LinFIR Documentation

Welcome to the official LinFIR documentation. LinFIR is a professional tool for designing FIR filters for loudspeaker crossovers and frequency response corrections.

Quick Navigation

• New to LinFIR? Start with the Overview
• Setting up audio? See Audio Setup & Measurements
• Designing filters? Check Driver Processing
• Need help? Visit Troubleshooting
• Updating LinFIR? See Updating LinFIR

About This Documentation

This documentation covers all aspects of LinFIR, from basic setup to advanced directivity analysis. Use the sidebar to navigate through different topics, or use the search function (top-left) to find specific information.

Accessing Documentation

You can access this documentation in several ways:

• From within LinFIR: Select Help > Documentation from the menu bar or press H from any window
• Online: Visit https://demaudio.com/doc/linfir/ for the latest version
• Offline: LinFIR automatically serves embedded documentation locally when no internet connection is available

When you access documentation from within LinFIR, the application will:

1. First attempt to open the online version
2. If the online version is unreachable (no internet or timeout after 3 seconds), automatically fall back to the embedded local documentation server
3. The local server runs on port 3030+ and serves the documentation directly from the application

This ensures you always have access to documentation, whether online or offline.

Safari Configuration (Mac OS)

⚠️ Important for Safari users on Mac OS: The local documentation server uses HTTP (not HTTPS) for localhost connections. To access the embedded local documentation in Safari, you need to:

1. Open Safari > Settings (or Safari > Preferences)
2. Go to the Security tab
3. Disable the “Non-secure site connections” option

This allows Safari to connect to the local HTTP server running on http://localhost:3030+.

Alternative: If you prefer to keep Safari’s security settings unchanged, the online documentation at https://demaudio.com/doc/linfir/ is always available and uses HTTPS.

LinFIR - Precision FIR Filter Design for Speaker Crossovers and Corrections
Visit https://demaudio.com/linfir/ for updates and support

Release Notes

This page is generated automatically from the project’s CHANGELOG.md and always reflects the latest released version.

[1.4.11] - 2026-10-03

Added

• Drag-and-drop reordering: drivers (and measurements in Room Calibration mode) can now be reordered by grabbing their header - the new ☰ grip or the name - and dropping it above or below another one. A line shows where it will land, and everything tied to the driver (its filter windows, IR management window, its position in the graphs, DI and exports) follows. Graphs can be reordered the same way: drag a graph’s title above or below another graph in the main view, or drag the rows of Settings → Graphs → Graph Order; hidden graphs keep their place in the order. The ⬆⬇ buttons still work, and clicking a driver’s name still folds and unfolds its section.

• Auto-level: a new button next to Auto-align delays that sets every active driver’s Gain so the summed response is as flat as possible across the whole system pass band, crossover regions included. The drivers are time-aligned internally for the calculation (without changing their delays), the natural roll-offs at both ends of the system are left out of the fit, and the fit uses a robust criterion so a driver’s own ripple or the start of a roll-off can’t skew it. No driver is ever made louder - only the relative levels change - so the headroom against clipping never shrinks. The confirmation shows the gain changes and how far the summed response is from flat before and after.

• Rotation center correction for off-axis measurements: when the turntable axis wasn’t at the front baffle - typically somewhere under the cabinet - the off-axis measurements were taken further from the speaker and at a wider angle than their nominal angle (with the axis 150 mm behind the baffle and the mic at 1 m, the 90° measurement is 0.47 ms late, 1.3 dB low and really 97° off axis), which skews the off-axis sums, the Listening Window and the directivity. Enter where the axis was (depth behind the baffle, sideways offset, mic distance - separately for the horizontal and vertical planes) and each driver’s position on the baffle under the measurement lists of the IR Management window: LinFIR rebuilds every off-axis response as if the speaker had turned about its baffle, correcting the time of flight, the level (1/r) and the angle (interpolating between neighboring measurements). The on-axis measurement is never changed, the measurement files aren’t modified, and the info icon shows a drawing of what to measure. Apply placement to all drivers copies the turntable placement to every driver, each keeping its own position on the baffle.

Changed

• A more modern look, in every window: the left column is now a set of cards - the target DSP settings in one, and each driver in its own card edged with its colour, with its filters, IIR filters and speaker settings grouped under small section titles - and every graph sits in its own card. Window, section and graph titles use the accent colour (red or blue, as set in Settings), tabs are segmented controls, buttons, toggles, status badges (such as “IR loaded”) and help icons look the same everywhere, and form labels no longer end with a colon. All the other windows follow the same style: IR management, FRD converter, LP/HP/correction filters, IIR and Auto-EQ, target curves, Hypex export, Settings (including the License tab), About, bug report, directivity, spectrogram, distortion IR and the near-field splicing wizard, in light and dark themes.
• Much lighter slider drags with several drivers: LinFIR recomputes live on every frame while a parameter is being dragged, and used to redo the whole filter chain and response analysis of every driver each time, even the ones the drag didn’t touch. Only the driver whose settings actually changed is recomputed now - the others are left exactly as they were - and the dashed raw (pre-filter) curves, which don’t depend on the filters, are no longer recomputed on a crossover or EQ edit.
• Clipping detection no longer slows down editing: the clipping test - the pink-noise test signal run through every driver’s filter chain and through the global filters - now runs in the background instead of on every recompute. With long filters it used to cost the interface tens of milliseconds on each frame of a slider drag; the graphs now follow the slider at full speed and the warnings update a moment later. Each filter chain is only re-tested when it actually changes (dragging one driver’s slider re-tests that driver alone), and a warning always stays on the right driver, even after reordering. The warnings also come in a stable order now, global first then drivers in order.
• Spectrogram CQT and Burst Decay up to 13× faster: each frequency band of the Morlet filter bank is now computed on its own bandwidth only, instead of a full-length transform of the whole spectrum for each of the thousands of bands. On the default 4096×4096 grid a 0.6 s loudspeaker IR drops from about 0.8 s to 0.12 s, and a 2.5 s room IR from 6.4 s to 0.5 s. The result is the same to within 0.01 dB.
• Faster IR windowing with many off-axis measurements: changing a driver’s time window used to re-window all its measurements one after another; they are now windowed in parallel, and the adaptive (frequency-dependent) window itself was streamlined - with the exact same result. Re-windowing 72 off-axis measurements is about 4× faster on a 4-core machine, more with more cores.
• Faster graph drawing, whatever the plot resolution: curves are now drawn at the graph’s actual pixel resolution instead of their full 16k-64k point spectrum, keeping the exact peak and dip of every pixel column, so the graphs look the same and auto-scaling is unchanged while drawing them costs a fraction of before. Zooming in reveals the full detail, as before.

Fixed

• Phase correction inverting a driver, and its guard tolerance not held: the phase correction brought the band’s mean phase to 0° whatever it was, so a driver - or the global sum - whose phase sat nearer 180° came out with its polarity inverted, flipping back and forth as the band, smoothing or guard settings changed. The correction filter now never inverts: its step response ends above where it starts and its main peak is positive (the whole filter is turned over when needed, which keeps the corrected impulse response symmetric). The magnitude guard also checked the filter on a grid too coarse to see the ripple between its points (over 1 dB at low frequency on a 4096-tap filter while 0.5 dB was asked): it now checks a finely sampled response, so the Guard Tolerance holds, on the drivers and on the global filter.
• Pololu Tic: Energize button out of step after a Go: sending the table to an angle from Settings energizes the motor, but the button kept offering Energize until clicked once; the move now energizes the motor explicitly and the button shows De-energize.
• Graphs overflowing the window, or an infinite graph height, with the HD graph: graph heights were computed by counting the HD graph and then subtracting it again, which went wrong whenever the two didn’t match - with HD detached in its own window the other graphs got a share too many and ran past the bottom of the panel, and with HD the only graph toggled on in the FIR, IIR or FIR+IIR views the count reached zero. Only the graphs actually drawn are counted now.
• Global correction filter not re-checked for clipping after an off-axis refresh: when the global FIR correction is redesigned from fresh off-axis data (Listening Window or Predicted In-Room reference), its new version was never tested for clipping, so a warning could be missing or out of date until some other change. It is checked like any other change now.
• FRD converter: driver offset delay had the wrong sign on one side of the axis: the time-of-flight difference added for a driver’s offset was always positive, so a woofer below the tweeter arrived later both above and below the axis - while below the axis, on its side, it is closer and must arrive earlier. Lobing predicted from FRD imports was therefore mirrored on one side. The delay is now signed (earlier on the driver’s side, later on the other side); X is counted toward positive horizontal angles and Y up. A ⚠ warns when the converter’s Delay is too short to leave room for the earliest angles.
• Saving and quitting could lose work: “Save” in the Unsaved Changes prompt now saves right away and only then quits, loads or creates a project (it used to queue a background save and go on, so a quit could kill the save mid-write, and a failed save was reported after the project was already gone). Quitting from that prompt now really closes LinFIR (it used to close only the prompt and leave the app half-quit), a second close request while the prompt is up no longer quits without saving, and queued background saves always finish before LinFIR exits. A failed auto-save is retried instead of being forgotten, and the final block of a project file is now checked and flushed to disk before it replaces the previous file. “Restart Now” after an update asks first when auto-save is off and there are unsaved changes. “Reset Application Settings” keeps the licence activation and closes the app as announced.
• Crashes: a driver, measurement, calibration or audio-device name with an accent or another non-ASCII character in the wrong place; two target-curve points at the same frequency (e.g. while typing 200 Hz); an Undo from an IR window that removes a driver; a Hypex project whose 15-biquad budget is full; a project file whose driver count does not match its drivers (it crashed at every launch); an impulse response or WAV file containing NaN or infinite samples (now refused with a message, instead of crashing or making the driver vanish from the graphs).
• Measurements and imports going to the wrong place: the angle and axis controls are locked while a capture, an auto-scan or the low-signal prompt is pending, so a measurement is stored where it was started; a single file with no angle in its name goes to the selected angle instead of replacing the on-axis measurement; closing the axis dialog cancels the import instead of importing every file at 0°; reordering drivers keeps every open IR window on its driver; in Room Calibration a stale off-axis angle no longer applies. A file, FRD conversion or capture at a sample rate different from the driver’s other measurements is refused instead of silently reinterpreting all of them (the TXT dialog no longer defaults to 8000 Hz after a restart).
• Measurement capture: clipping is detected on the real converter level, before the input gain compensation (no more false rejections with a positive compensation, or missed clipping with a negative one), and no longer on the start-up glitch the live check ignores; the last half-second of each capture is recorded again; the low-level check of the first sweep is no longer fooled by the loopback chirp; a timing reference that cannot be detected is reported as such instead of a misleading “h2 contaminates h1”; the sweep and timing-reference outputs on the same channel are refused; a negative manual offset no longer overflows; output-stream errors are reported as such. Closing the IR window - or deleting its driver, or opening or creating a project while it captures or scans - stops its capture and auto-scan (the table still returns to 0° and is de-energized), an auto-scan stops when a capture cannot start, Measure is disabled during a capture or scan, and starting a scan cancels the previous scan’s pending de-energize. In the near-field splice wizard, a second capture can no longer be started and saved into the wrong element, and auto-splicing new captures works again after reopening a project.
• Directivity, AutoEQ and exports: the Directivity Index is computed from the acoustic sum of the drivers, like the other directivity curves (it ignored phase and could drop a driver entirely); the AutoEQ optimizer actually optimizes now (it stopped after the first step and returned its initial guess); near-field splice levels follow the piston radius, not its area (vents and passive radiators came out several dB too quiet - re-run Match level to far-field); the global FIR correction in post-IIR mode follows the Auto-EQ tab when it is the active one; MiniDSP biquad exports have the sign miniDSP expects for a1/a2; the HFD export refuses filters it cannot carry instead of dropping them, and follows the device model set from the menu.
• Rotation table and audio devices: with GRBL 1.1, the angle now accounts for “Set 0°”, so the auto-scan no longer times out; plugging in an audio interface updates the device list again; waking an ASIO computer from sleep keeps the selected interface; choosing “None” as timing reference is remembered.
• Smaller fixes: the “What’s new” note no longer covers the update notification; dialogs open centred on the LinFIR window, also on a second screen; the update check no longer hides the download progress or “Restart Now”; the N shortcut recomputes like the toolbar toggle; “HD”, “(horizontal)/(vertical)” and “Plot” are translated; the bug-report window closes on Cancel; graph legends list the drivers in their order, top to bottom (they were sorted by name, and upside down in the bottom corners); the spectrogram explains when it cannot be computed instead of showing “Computing” forever; FIR export padding and the reports follow the per-driver taps setting.
• Performance: a connected turntable no longer redraws the whole interface every frame; renaming a driver no longer recomputes every IR; DI, Sound Power and Predicted In-Room integrate the sphere far faster; no per-frame copy of every driver in the driver list and the distortion IR window; faster phase alignment; the spectrogram runs one computation at a time instead of piling up workers during a drag; the directivity sonogram no longer recomputes forever with horizontal-only measurements; timing-reference detection only analyses the start of a long multi-average capture; stale off-axis results are dropped and the global FIR correction is only redesigned after an off-axis update when it depends on it; the Listening Window and Predicted In-Room curves (and Early Reflections, Sound Power, ER DI) are only computed while shown with a valid license, instead of after every edit (turning one on computes it); a spectrogram computation abandoned for a new setting stops at once instead of finishing its transform.

[1.4.10] - 2026-09-30

Fixed

• Sweep confirmation buttons cut off in French: the “Start sweep measurement?” window was too short for the French text, which wraps onto more lines, so its Start Measurement / Cancel buttons ran past the bottom of the window. The window is now tall enough.
• “+ Add Point” slightly cut off in the target curve window: in every language, the button below the control point list ran a few pixels past the bottom of the window. The list now leaves room for it, whatever the window size and the number of points.

[1.4.9] - 2026-09-26

Added

• LinFIR in French: the whole interface - menus, windows, settings, tooltips, graph titles and legends, notifications and error messages - is now available in French, using the usual French terms of signal processing and acoustics (réponse impulsionnelle, retard de groupe, passe-haut, fenêtrage, directivité…). Pick it in Settings → General → Appearance → Language: Automatic (the default) follows the language of your operating system, or force English or Français. The switch is immediate, no restart needed. The documentation, exported files and bug reports stay in English.

Changed

• Graph toolbar wraps when it needs to: the toolbar above the graphs moves its second half to a new row based on the actual width of its controls instead of a fixed window width, so it never runs off the edge - whatever the language - and stays on one row whenever it fits.
• Save as… moved to the File menu, right below Save: File → Export → LinFIR Project is gone, replaced by File → Save as… where you’d expect to find it. It works as before: you choose a new name and location, and LinFIR then keeps working on that new file (Save and Auto-save go to it from then on), leaving the original file as it was.
• Cmd/Ctrl+E now does Save as…: this shortcut used to write a copy of the project while you kept working on the original file - not the same thing as the menu entry it was paired with. It now does exactly what File → Save as… does, so saving under a new name always behaves the same way and you always know which file you are editing.

Fixed

• License deactivated on some networks: some networks (a home router, a hotel or office Wi-Fi) give the computer another name - a Mac called “MacBook-Air-de-…” becomes “mac.home” - and LinFIR took that for another machine and cleared the license. The license is now tied to the machine’s hardware identifier and its own name, whatever the network calls it. Existing activations keep working, including ones made while the network had renamed the computer, and nothing needs to be reactivated.
• Tooltips could run past the edge of narrow windows: long hover texts were wrapped at a fixed width, wider than some small windows (filter settings, IR management, export dialogs), so part of the text was cut off. Tooltips now wrap to fit the window they appear in.

[1.4.8] - 2026-09-24

Changed

• Raw pre-filter curves follow the focused filter window: the dashed raw (pre-filter) curves on the frequency, phase and group delay graphs are now drawn only for the driver whose filter window (LP, HP, FIR correction or IIR) is in the foreground - or for the sum when a global filter window is. With several filter windows open, you only see the curves of the one you are working in, and nothing when they are all in the background, the same way the IIR Auto EQ target overlay already behaved.
• FIR correction target overlay follows focus too: the green correction target overlay (per driver and global) is now shown only while its FIR correction window is in the foreground, like the IIR Auto EQ target.
• Bug Report now requires your email address: Help → Report a Bug has a new, required email field so we can follow up on your report. It’s remembered between sessions and used as the report email’s Reply-To, so a reply from support reaches you directly.
• Extended project backward compatibility to LinFIR 1.2.0: projects created with version 1.2.0 or later can now be opened directly without requiring a pre-conversion re-save.

Fixed

• Bug reports always claimed LinFIR v0.9.9: the emailed report’s “Application Version” line was a hardcoded placeholder instead of the actual running version - fixed to report the real version.
• “Save As…” and “Export Project” could produce a corrupted project file: unlike the regular Save path, these wrote a plain, uncompressed .lnf directly to the target path - an interruption partway through (crash, force-quit, power loss) could leave a truncated or empty file that fails to reopen (“Invalid JSON: expected value at line 1 column 1”). Both now use the same compressed, atomic (temp file + rename) write as every other save path, so an interruption leaves either the previous file intact or the fully-written new one.

[1.4.7] - 2026-09-23

Changed

• Target Curve Configuration window colored by filter type: the window opened via a filter’s Configure button now plots its curve in green for a FIR magnitude correction filter, matching that filter’s own target overlay on the frequency graph - Auto EQ keeps the existing blue.

Fixed

• Correction target overlay didn’t refresh on a Gain change: the FIR magnitude correction target overlay (new in 1.4.6) stayed at its last position after changing a driver’s Gain - it only caught up once some other correction parameter was touched, forcing a recompute. The overlay’s cache-invalidation check now also watches Gain.

[1.4.6] - 2026-09-23

Added

• Early Reflections, Sound Power and Early Reflections DI curves: the remaining pieces of the CTA-2034-A (“CEA-2034”) spinorama suite are now available as their own curves on the sum’s frequency graph, not just as internal ingredients of Predicted In-Room Response - Early Reflections (purple, dashed), Sound Power (teal, dashed) and Early Reflections DI (dark orange, the on-axis-minus-Early-Reflections ratio, offset alongside DI). Each has its own toggle in the Graphs dropdown’s Overlays section (Settings → Graphs → Default Overlays sets what a new project starts with) - all three are hidden by default, unlike the existing overlays.
• Listening Window / Predicted In-Room as a per-driver reference: the driver-level FIR correction filter and driver Auto EQ each gain the same Reference selector as their global counterparts - On-axis, Listening Window or Predicted In-Room - now computed from that driver’s own polar measurements alone instead of the summed system. Useful for voicing a driver’s own off-axis trends (tweeter beaming, baffle diffraction); a crossover region’s interference with a neighboring driver is still only visible to the global filters’ own Reference, which sees the summed system. Same license requirement and On-axis fallback behavior as the global version.
• Magnitude correction FIR filter: target curve overlay: same idea as the existing Auto EQ target overlay - while a driver or global correction filter window is open, a green overlay on the frequency response graph shows what the response converges to at 100% Strength (target curve shape, level-matched to the response’s own average - including the driver’s own Gain, for a driver correction filter - plus Correction Gain). Not shown with Use legacy FIR correction tuning enabled, since that algorithm has no equivalent fixed curve to converge to.

Changed

• Listening Window and Predicted In-Room no longer need a second driver or a global filter: they used to only appear once at least two drivers had measurement data, or a global IIR/FIR correction filter was active - even though DI already showed for a single, unfiltered driver. All three overlays now follow the same rule: available as soon as there’s enough off-axis angle coverage, regardless of driver count or filters. The Show Listening Window / Show Predicted In-Room Response toggles in the Graphs dropdown still let you hide them, unchanged.
• Magnitude correction FIR filter (driver and global): Max. Attenuation + Correction Gain replaced by a single Strength slider: instead of tuning “how much attenuation” and “how much compensating gain” as two separate, manually-balanced fields, a single Strength slider (0-100%, styled like the Phase section’s Intensity) now controls how far the correction goes, from untouched (0%) to fully matching the target within the configured frequency range (100%). The new algorithm references response and target to their own average level within that range (not their peak), so correction is symmetric - bumps get tamed, dips get filled in - instead of only ever boosting, and it’s no longer thrown off by a peak or dip outside the configured range: Strength now always has an immediate, visible effect the moment you move it. Correction Gain is kept, now as a plain, independent broadband offset on top of the Strength correction (default changed from 4 dB to 0 dB, since it’s no longer compensating for anything) rather than the sole gain control. A new Use legacy FIR correction tuning toggle (Settings → General → Filter Processing, off by default) brings back the exact original two-field algorithm, project-wide, when needed.

Breaking Changes

• Magnitude correction filters may sound different after this update: the algorithm change above is not backward-compatible at the numeric level - a project’s existing Max. Attenuation/Correction Gain values are reinterpreted under the new Strength model rather than converted, since the two algorithms don’t correspond via a simple formula. Check any correction filter you’re relying on after upgrading (the FIR display mode, press F, shows the actual result) and re-tune Strength/Correction Gain as needed, or enable Use legacy FIR correction tuning (Settings → General → Filter Processing) to recover the exact previous behavior.

[1.4.5] - 2026-09-20

Added

• Predicted In-Room Response (PIR) curve: a new PIR curve (cherry red, dashed) on the sum’s frequency graph, computed per the ANSI/CTA-2034-A (“CEA-2034”) spinorama standard - 0.12·Listening Window + 0.44·Early Reflections + 0.44·Sound Power. Early Reflections and Sound Power are new too: Early Reflections averages the standard Floor/Ceiling/Front Wall/Side Wall/Rear Wall bounce angle sets. Together with the existing DI and Listening Window curves, LinFIR now produces a complete, standard-compliant spinorama directly comparable to third-party ones built to the same standard. Gated the same way as DI/Listening Window: valid license, off-axis data, and a new Show Predicted In-Room Response curve toggle (default on) - see below, it now lives in the Graphs dropdown.
• Predicted In-Room as a correction/Auto-EQ reference: the global magnitude correction filter and global Auto EQ’s Reference selector gain a third option, Predicted In-Room, alongside the existing On-axis and Listening Window. Falls back to On-axis when there isn’t enough Early Reflections angle coverage, same as Listening Window falling back when there aren’t enough polar angles.
• Directivity Index visibility toggle: DI previously had no way to be hidden - it now has its own toggle alongside the other magnitude overlays (see below).
• Directivity Sonogram: configurable frequency/angle range: the Sonogram tab gains Frequency (min/max Hz) and Angle (± degrees, symmetric around 0°) drag-values with a Reset button, on the same row as the existing Colormap range/Normalization controls - shared by both the Horizontal and Vertical heatmaps, same idea as the Spectrogram window’s own frequency-range control. The Overlay Plot tab gains its own Frequency controls and Reset button too, kept independent per axis (Horizontal/Vertical), next to each plot’s existing Angle range controls.

Changed

• Listening Window curve color: the LW curve is now a fixed royal blue, dashed line (was sharing the Sum curve’s white/black), so it’s distinguishable from Sum, DI, PIR and the per-driver curves at a glance.
• Sum curve always drawn on top: the Sum/Average curve now paints over any driver, DI, LW or PIR curve it overlaps, instead of whichever was drawn last winning by coincidence.
• LW/PIR/DI hover tooltip spells out the full name: hovering the curve on the frequency graph now shows “Listening Window” / “Predicted In-Room Response” / “Directivity Index” instead of the abbreviation - the legend itself stays compact (“LW”/“PIR”/“DI”).
• Overlay toggles moved into the Graphs dropdown, and now save per project: Sum / Average, Listening Window, Predicted In-Room Response and Directivity Index visibility are no longer app-wide Settings toggles - they’re now in the graph toolbar’s Graphs dropdown, under a new Overlays section (separated from the existing graph-type toggles, now labelled Main plots, by a divider). Each project remembers its own choice, the same way it already does for which graph panels are shown. Sum/Average applies in both project modes (label follows the mode); Listening Window/PIR/DI stay Loudspeaker-Design-only and license-gated, as before. Settings → Graphs keeps a Default Overlays section (mirroring the existing Default Visibility one) to set what a new project starts with.
• Large projects load much faster: opening a project with a lot of measurement data (many drivers, many off-axis angles, distortion/harmonic data) is now several times quicker - the bigger the project, the bigger the difference. Project files created by this version are also somewhat smaller on disk.
• Dropped support for projects created before LinFIR 1.3.0: opening one now shows a clear “unsupported version” message instead of attempting a conversion. Open it with an older LinFIR version first and re-save to bring it forward.
• Saving is faster too: compressing a large project file on save now uses multiple CPU cores instead of one, on top of the smaller file size above. Quitting (or the auto-restart after an update) no longer re-saves a project that hasn’t changed since it was last auto-saved.

Fixed

• Auto-save before switching projects wrote an uncompressed file: opening a different project (New Project / Load Project / Open Recent) with Auto-save enabled and unsaved changes silently saved the current one first, but as plain, uncompressed JSON instead of the normal compressed format - multiple times the size on disk, and not crash-safe (an interruption mid-write could leave a corrupted file) unlike every other save path. It now goes through the same compressed, crash-safe save as everything else.

[1.4.4] - 2026-09-14

Added

• Low signal-to-noise ratio: keep anyway instead of automatic rejection: a sweep measurement or an imported impulse response with a peak level below −25 dB used to be discarded automatically. It’s now a confirmation dialog instead - an always-on-top popup that warns about the poor signal-to-noise ratio and offers Keep Anyway or Discard, so a marginal capture or import is no longer lost outright. Clipping still rejects immediately, unconditionally - that data is corrupted, not just noisy. An unattended auto-scan still aborts automatically on a low level rather than popping up a dialog nobody is there to answer. Settings → General → Measurement & Import Quality gains an Always keep low-SNR measurements toggle that skips the dialog and keeps low-level data silently from then on; its hover text is explicit that correction filter quality depends on measurement quality.
• One-click restart after an update: once an update finishes installing, the notification window (bottom-right corner) switches from Install to 🔄 Restart Now, which relaunches LinFIR with the update already applied and closes the current instance. No need to quit and reopen manually anymore.
• Soundcard calibration (Kirkeby correction): measurements can now be corrected for the audio interface’s own response, not only the microphone’s. Measure the interface once through an output-to-input loopback cable in Settings → Measurement Calibration → Soundcard, then select that calibration in a driver’s IR window: every measurement angle of the driver - near-field / far-field spliced responses included - is corrected in magnitude and phase by a regularized Kirkeby inverse filter, removing the interface’s phase rotation in the low bass and its roll-off near Nyquist. A calibration only corrects measurements captured at its own sample rate: calibrations measured at another rate are greyed out, and a warning is shown when the selected one doesn’t match. Soundcard calibrations are embedded in project files like microphone calibrations, and a default can be set for new drivers.

Changed

• Settings: Measurement Calibration tab: Mic Calibration becomes Measurement Calibration, grouping the microphone and soundcard calibrations in one tab. A driver’s IR window likewise selects both in a single Measurement Calibration section.

Fixed

• Harmonic distortion with a microphone calibration: the fundamental was calibration-corrected but its harmonics were not, skewing THD and the individual harmonic curves wherever the calibration isn’t flat - and changing the calibration didn’t update them. Each harmonic now gets the calibration at its own frequency, and the curves update when the calibration changes.
• Time-centered phase at another sample rate: the phase’s time centering has been reworked so the small fractional delay left over no longer changes how the phase looks when the project sample rate changes - the same measurement now shows the same phase curve at 44.1, 48, 96 or 192 kHz, instead of a slope appearing above about 1 kHz.
• Distortion IR window: driver dropdown ignored custom names: it always showed “Driver N”, never the name typed into the Driver IR window, unlike every other driver selector in the app. It now reads the same underlying name field, falling back to “Driver N” only when none is set.
• Distortion IR window: auto-fit didn’t reactivate on a new project: opening a different project (or restarting the app, which reloads the last one) could leave the graph zoomed to the previous project’s scale instead of re-fitting to the new data, matching what the Directivity Sonogram and Spectrogram windows already did.

[1.4.3] - 2026-09-11

Added

• Configurable frequency axis range: Settings → Graphs gains a Frequency Axis section with Min / Max frequency (Hz) fields (default 10 Hz – 24 kHz, with a Reset button) that now drive every frequency-axis graph - Magnitude, Phase, Group Delay, HD, the Directivity Sonogram (both the heatmap and its on-axis overlay plot), and the Spectrogram - instead of each hardcoding its own range. The existing Sync X-axis cross-graph linking, per-graph saved zoom, and double-click-to-reset all keep working unchanged; only the range they reset to is now configurable. Also fixes a pre-existing inconsistency where the Spectrogram privately floored at 5 Hz while every other graph floored at 10 Hz - they now all agree.
• Spectrogram: independent frequency range: the Spectrogram window gains its own Frequency drag-values (Min/Max Hz) with a Reset button that pulls in the current Settings → Graphs → Frequency Axis values. Unlike every other frequency-axis graph, once set it no longer tracks later Settings changes automatically - only that Reset button does.
• Spectrogram: manual time/period crop: the auto time crop (CQT/STFT/CSD, in ms) and Burst Decay’s Periods control (which used to fight the auto crop instead of matching it) are now one mechanism: a new Crop control (ms), independent for each of CQT/STFT/CSD, and Periods for Burst Decay - all four auto-detected from where the decay clears Min dB until dragged, then pinned exactly where left (through any other change, including switching modes) until that mode’s own Reset button brings it back to auto. Burst Decay now computes a generous ceiling internally while auto and reports back only the periods that actually carry signal, instead of showing a fixed 30-period window that could cut the decay short or leave most of the plot empty. The 3D waterfall now always shows the exact same crop as the 2D image - it used to run its own, independent (and occasionally wildly different) detection. The Crop/Periods controls now sit on their own row together with Frequency, separate from the transform/preset/processing-parameters row above them, and the crop’s Reset button is its own - decoupled from that row’s processing-parameters Reset, which no longer touches the crop at all, and only ever resets the mode currently on screen.
• Directivity Sonogram: view saved with the project: colormap range, normalization mode, active tab, and the Overlay Plot’s angle ranges/normalize/auto-Y are now stored in the .lnf and restored on reload, matching what the Spectrogram already did. Projects saved by an older LinFIR open at the app defaults.
• Settings → Graphs: Reset buttons: every section now has its own Reset button, restoring just that section’s settings to their app defaults - the same idea as the Frequency Axis section’s existing one.
• Auto EQ: take a locked filter into account when generating new ones: a locked filter (driver or global Auto EQ) now has a second button next to the padlock - when turned on, that filter is applied to the response before Auto EQ designs new filters, so the new ones correctly build around it instead of overlapping or ignoring it. Off by default, per filter, so nothing changes until you turn it on.
• Update download progress: installing an available update now shows a progress bar instead of freezing the app until the download finishes - handy on a slower connection. The app stays fully responsive while it downloads.

Changed

• Resampling: the sample-rate conversion used when changing the project’s working Fs (and when exporting IRs at a different Fs) has been rewritten as a self-contained, optimized implementation.
• FIR LP/HP: higher Kaiser β ceiling: the Kaiser β slider on the FIR Low-Pass and High-Pass filters now goes up to 500 (was 50), for users who need much gentler stopband attenuation than the old ceiling allowed.

Fixed

• IR Windowing: fade-in preview could fail to appear: with a Delay Compensation active, the transient window-envelope preview’s fade-in could land at a negative on-graph time and silently not draw at all instead of showing the ramp at its correct (negative) position.
• Undo/Redo could reset a graph’s zoom: undoing or redoing an unrelated change could silently snap a Magnitude/Phase/Group Delay/THD/Impulse/Step graph back to auto-fit, discarding a manual zoom or pan. Graph view state is no longer part of the undo history, matching how the Spectrogram’s view already behaved.
• Undo/Redo: one step per keystroke when typing a value: typing a number into a field (rather than dragging it) created a separate undo step for every digit, so undoing one edit meant pressing Undo repeatedly. Typing now debounces into a single undo step per edit, the same way dragging already did.
• Magnitude/Phase/Group Delay: raw pre-filter preview could distort the auto-fit range: opening a driver’s filter window shows its raw, pre-filter response as a dashed reference overlay - but that overlay’s own range was being folded into the graph’s auto-fit, which could shrink or shift the visible range away from the curves actually being worked on. The raw overlay no longer affects auto-fit.

[1.4.2] - 2026-09-09

Added

• Spectrogram: detail setting — Settings → Graphs → Spectrogram detail picks the compute-grid size (Standard / High / Ultra). High (4096²) is the new default; the transform now runs on a finer grid than before.

• Spectrogram: group-delay toggle — the grey group-delay comparison line can now be turned off.

• Spectrogram: default lower bound setting — Settings → Graphs → Spectrogram now has its own lower-bound (Min dB) default, −30 dB, matching the sonogram’s. It’s what the window opens with and what its Reset button restores.

• Spectrogram: CSD mode — Cumulative Spectral Decay: for a sweep of gate positions the impulse response from the gate to its end is FFT’d, showing the spectrum decaying over time in milliseconds. A Rise control sets the gate’s leading-edge taper.

• Spectrogram: Burst Decay mode — Morlet-wavelet burst decay on a period-based time axis: each frequency’s decay is read in cycles rather than milliseconds, so equal-Q resonances look identical everywhere. Period 0 is auto-locked to the arrival peak. Controls: Bandwidth (1/3 – 1/24 octave), Freq res (12–96 points/octave), Periods, Pts/period.

• Spectrogram: Reset button — a Reset at the end of the control bar returns every mode’s processing parameters (Q, Rise, Periods…) to their defaults.

• Spectrogram: 3D waterfall view — a 2D / 3D button renders the time–frequency grid as a shaded coloured perspective surface (same colormap and Min/Max dB as the 2D image), for all four transform modes. Left-drag rotates, right-drag pans, scroll zooms, double-click resets; the time, frequency and dB axes are labelled. Hovering the surface draws the time and frequency slices through the cursor with a value readout. A Show control switches between surface, surface + slice lines, and a slices-only line waterfall (slice count adjustable). The surface keeps narrow resonances and fast high-frequency detail that a plain average would have smoothed away.

• Spectrogram: frequency smoothing — a Smooth dropdown averages the spectrogram across frequency in fractions of an octave (None to 1/3 octave, or ERB), independently of the main graph’s smoothing. Applies to every transform and to both the 2D and 3D views; a free re-render.

• Spectrogram: analysis parameters are undoable — Undo / Redo now covers the spectrogram’s analysis controls (transform mode and its preset / window, CSD Rise, Burst-decay parameters, frequency smoothing). Pure view settings (colour range, Normalize, axis layout, 2D/3D, view angle) stay outside the history.

• Room Calibration: export a measurement — the Impulse Response Management window now has an ⬇ Export button (next to Import / Measure) that saves the selected measurement as WAV or TXT, matching what Loudspeaker Design mode already offered per angle.

Changed

• Spectrogram: automatic time crop — the time axis is now cropped to the part that carries signal: leading and trailing time slices where every frequency is at or below the Min dB level are dropped. It reads the rendered spectrogram directly, so the plot edges sit exactly where the colour reaches the floor, and it follows Min dB and Normalize instantly. Raise Min dB (or use STFT) for a tighter view; the CQT’s low-frequency smear keeps it wider.
• Spectrogram: view stays auto-fitted — after a double-click the plot keeps auto-fitting to the cropped image, so it re-frames itself when you change Min dB, the transform, or the axis layout, until you pan or zoom again.
• Spectrogram: view saved with the project — transform mode, resolution preset, normalization, group-delay toggle, axis layout and colormap range are stored in the .lnf and restored on reload. Projects saved by an older LinFIR open at the app defaults.
• Settings → Graphs — the Directivity Sonogram and Spectrogram options are now in separate sections.

Fixed

• Spectrogram: group-delay overlay drifted with smoothing — with smoothing active the grey group-delay line crept toward low frequencies as the smoothing widened. It is now smoothed symmetrically per octave and stays on the true frequency.
• Splicing wizard: mouse-wheel scroll panned the graphs instead of the page: scrolling while the cursor happened to be over one of the wizard’s graphs panned that graph instead of scrolling the wizard’s steps. The graphs no longer capture wheel/trackpad scroll; dragging still translates them as before.
• Off-axis harmonic distortion went blank, and off-axis recompute was slow: with the distortion graph visible, the background job that refreshes the off-axis responses (for the sonogram, directivity index and listening window) also re-ran the full harmonic-distortion analysis on every cycle - on data it couldn’t process - and then overwrote the good distortion data with the empty result. That job now leaves distortion alone entirely: it is captured data the main view already computes once. Off-axis refreshes are also noticeably quicker. Separately, the windowed impulse response is no longer stored in the project file (it is rebuilt from the raw capture on load), so a project saved with an incomplete one can’t carry that state back in, and project files are a little smaller.

[1.4.1] - 2026-09-06

Added

• Auto-align driver delays: a button below ➕ Add Driver sets every active driver’s Time Delay so their post-filter phase curves cross at the middle of each crossover overlap (the band where both drivers are within −20 dB of their own maximum). A driver that shares no band with the rest is left alone.
• Reorder drivers / measurements: each driver’s action bar gains ⬆ / ⬇ buttons (between paste and the remove button, which is now set slightly apart to avoid mis-clicks) that move it up or down. The new order applies everywhere — parameter list, graphs and legends, directivity index, and exports. The same buttons reorder measurement positions in Room Calibration mode.
• Directivity sonogram: clipping highlight: levels that rise above the colormap maximum (which per-band normalization can produce) now show in off-white instead of saturating to the same red as the loudest in-range energy, so it is obvious where the response goes over the reference level. It can be turned off in Settings → Graphs.
• Directivity sonogram: default normalization: Settings → Graphs now has a Sonogram normalization setting that picks which normalization mode the sonogram opens with. It can still be changed per-session from the sonogram’s own toolbar.
• Spectrogram: default axis orientation: Settings → Graphs now has a Spectrogram axes setting choosing whether the spectrogram opens with time on X (default) or frequency on X. The toolbar button still swaps it per-session.
• Splicing wizard: automatic level match: the Far-field window & predicted level step gains a Match level to far-field button next to the manual level offset. It shifts the predicted near-field curve onto the windowed far-field measurement across the valid blend range, using the median of the per-bin level difference so a diffraction ripple or a narrow notch near a band edge does not throw it off.
• Splicing wizard: automatic phase match: the Transition & blend step gains a Match phase button next to the near-field time offset. It smooths the predicted and far-field phase curves, then sets the offset to the delay that makes them cross at the middle of the transition band (where the blend is 50/50), leaving only a residual slope or a whole-turn wrap to sort out by hand.

Changed

• Settings → Graphs: the directivity-sonogram and spectrogram options moved out of “Magnitude Plot Bounds” into their own “Directivity Sonogram & Spectrogram” section. The General and Graphs settings documentation has been rewritten to match the actual tab layout and default values.
• Spectrogram: delay compensation: the spectrogram time axis and its grey group-delay overlay are now shifted by Delay Compensation, the same as the Phase and Group Delay graphs, so the acoustic arrival sits at t ≈ 0 and the views agree.

Fixed

• Aberrant driver phase / group delay with very long FIR filters: a driver whose filter chain included a very long FIR could show a rising phase and a negative group delay, even though the system sum stayed correct. The driver response is now computed the same way as the sum.
• Subtle phase rotation from delay compensation: enabling mic-distance delay compensation tilted the displayed phase slightly more than it should. Part of the cause was shared with the long-FIR fix above; the remainder, visible in time-centred and unwrapped phase views, was the compensation being rounded to a whole sample on the display side while applied exactly everywhere else. It is now applied consistently.

[1.4.0] - 2026-09-05

Added

• Near-field / far-field splicing wizard (license required): a new seven-step wizard that builds a full-range anechoic response for a driver by combining near-field captures of each radiating element (driver, vent, passive radiator) with its existing far-field polar measurements — the Don Keele method. The near-field captures cover the bass, where an indoor gated measurement can’t reach; the far-field measurements keep their real directivity and diffraction above the transition. The wizard models the enclosure (box or sphere, baffle edge treatment, per-element positions on any face), predicts each element’s directivity and edge diffraction per angle with a Distributed Edge Dipole model, level-matches the elements from their Sd and mic distance, guards against room reflections with the image-source method, and blends the two responses in a physics-bounded transition band with a magnitude and phase preview. Once applied, the spliced impulse response replaces the raw capture for every measured angle throughout the app (response, DI, sonogram, export), and newly captured angles are spliced automatically. Reached from IR Management → Near-field measurements (splicing); see the documentation for a step-by-step guide.
• Unsaved-work protection: closing LinFIR with unsaved changes now asks whether to save first. If a brand-new project that was never saved to disk is closed, it is offered back for recovery the next time you open the app.
• Configurable sweep End Frequency: the sweep’s upper limit can now be set in the IR Management window and the splicing wizard (defaults in Audio Settings, with a Reset button), instead of always being fixed just below Nyquist. The sweep stays flat at full amplitude up to the End Frequency then fades out over at most an octave. Lower it to stop the sweep below a loud cone-breakup resonance: previously such a resonance forced you to drop the sweep level to avoid clipping, sacrificing the signal-to-noise ratio of the band that matters. Harmonic distortion is still captured at full bandwidth — the deconvolution filter keeps its full range, so a harmonic landing above the End Frequency is not truncated. “Auto” keeps the previous behaviour exactly.
• Spectrogram: dB in the hover tooltip and a group delay overlay: hovering the spectrogram now shows the exact level in dB at the cursor alongside time and frequency, using the same reference as the colour map (global or per-band). A thin grey line also traces the group delay across frequency directly on top of the colour map — the same curve as the main Group Delay graph (same Smoothing setting), just not shifted by Delay Compensation — making it easy to check whether a resonance or smear in the spectrogram corresponds to a group-delay bump at the same frequency. The tooltip labels this reading explicitly (“Group Delay: … ms”) and only shows it when the cursor is actually hooking the grey line.
• Spectrogram: selectable STFT window shape: in STFT mode a new Shape control chooses the analysis window — Rectangular, Hann, Hamming, Blackman-Harris, Tukey, or Gaussian — to trade time localisation against spectral leakage. Gaussian (the previous fixed behaviour) stays the default. The transform is also now picked from a dropdown instead of separate buttons.

Fixed

• Interface freezing on projects with many off-axis measurements: editing any parameter (gain, delay, filters, target curve…) no longer freezes the interface on every change. The off-axis and directivity calculations now run in the background without copying the measurement data each time.
• Directivity sonogram not keeping up with edits: the sonogram now updates in place as you change parameters. It no longer updated only the on-axis response, and it no longer closed and reopened itself - jumping back to the center of the screen - on every change.
• Saved session could fail to reload: on large projects the file holding the application state could grow very large and then fail to load on startup, silently opening LinFIR with an empty project and forgetting the last-opened project. That state is now kept small and the previous project reloads reliably.
• Crash during impulse response capture: the distortion-zone calculation used during IR capture could underflow and crash the app when the main (H1) peak was close to the start of the recorded buffer (e.g. a short sweep or a small manual offset).
• “Missing measurement” warning on every angle change: selecting an off-axis angle where a driver has no measurement no longer raises a persistent warning toast. Browsing angles a driver was only measured on fewer axes is normal — its curve simply doesn’t appear.
• IR window Start / Stop crossing: setting the window start and stop within a taper of each other (e.g. 4.0 ms and 4.2 ms) used to make the fade-in and fade-out overlap and produce a malformed window. The window is now the product of the fade-in and fade-out envelopes, so overlapping edges compose into a smooth sub-unity bump — in both the applied window and the on-graph preview, simple and adaptive alike. Start and Stop are fully independent again: editing one no longer moves the other, which also fixes the value jumping while you typed a number into either field (including in the splicing wizard).
• Raw phase overlay misaligned from the filtered curve: with “Remove time of flight rotations” on, the dashed raw (pre-filter) phase curve shown while a driver’s filter window is open could sit a large linear trend or a full 360° turn away from the filtered curve it is meant to overlay. Two causes: (1) for a driver with only IIR filtering (no LP/HP/correction FIR) the raw curve was compensated for an FIR-chain delay its filtered response never carries; (2) the raw curve was given its own independently-rounded N×360° offset instead of the shared one. The raw phase now gets exactly the same time-of-flight removal and DC alignment as the filtered phase.
• Global phase correction ignored the global magnitude correction: unlike the per-driver FIR correction, the global FIR’s phase correction was designed from the summed response before the global magnitude correction FIR, so it did not linearise the excess phase a minimum-phase (Causality > 0) global magnitude correction introduces. The global phase correction now runs on the magnitude-corrected response, mirroring the per-driver behaviour; the fusion padding is also sized for the real magnitude-FIR causality.
• IR window fade-in applied even at Start = 0: Start = 0 means “keep the measurement from its very first sample” — nothing is being truncated there, so there is nothing to smooth. The window nonetheless faded the first 0.5 ms in from silence. The window (and its on-graph preview, simple and adaptive alike) now opens at full amplitude immediately when Start = 0. The Stop side needed no equivalent fix: it is already a no-op once Stop reaches the measurement’s own duration.

Improved

• IR window preview: the transient overlay shown while editing a driver’s time window now also draws that measurement’s raw, un-windowed impulse response on the Impulse Response graph, as a dashed line in the same colour as the driver’s curve. Placed with the same time shift as the processed curve underneath and fading out with the red envelope, it lets you see directly what the window keeps and what it removes. A new Show window button in the IR Windowing section brings the whole overlay up at any time without changing a setting. The documentation now also explains why the overlay’s start/stop do not sit at the entered values once the driver has filters or a time delay (the window times are relative to the raw IR; the graph shows the processed, delayed signal).
• Why the window preview doesn’t line up with Start/Stop: hovering Start or Stop in the IR Windowing section now shows exactly where that value shows up on the Impulse/Step Response graphs and why, broken down by FIR delay / Time Delay / Delay Compensation. Whenever the two positions actually differ, a new On Graph row appears right under Window Duration with both shifted values at a glance (e.g. 2.00 ➡ 12.00 ms). Start/Stop themselves are untouched — they stay defined on the raw measurement, this only translates where they currently land.
• Auto-save: no longer causes a brief periodic hitch while you are only viewing a project without editing it.
• Undo / Redo: the undo history now keeps the last 50 steps (was 10). Each step is stored as parameters plus shared references to the measurement data rather than a full copy, so the deeper history still uses only a few MB even on a large project.

[1.3.14] - 2026-08-28

Added

• Rotation table position dial during auto-scan: the driver IR window now shows the same angular position dial as Settings -> Rotation Table (graphical only, no controls) above the measurement buttons, visible while an auto-scan is running.
• Auto-scan end-of-scan behavior: whenever an auto-scan ends - whether it completes, is stopped, or is aborted due to a rejected capture - the rotation table is now sent back to 0°. For Pololu Tic controllers, the motor is also energized automatically at the start of the scan if it wasn’t already. A new De-energize motor after auto-scan toggle (Settings -> Rotation Table, off by default) de-energizes the motor once it’s confirmed to have settled back at 0°.

Improved

• Adaptive IR windowing: the adaptive window’s stop-side fade-out now scales its taper duration with each frequency band’s own wavelength, instead of using a fixed duration for every band. This avoids artifacts from cutting off low-frequency content mid-cycle.
  • This changes the exact shape of the adaptive window’s fade-out compared to earlier versions, so results computed with it (especially the low-frequency end of the response) may shift slightly for projects created with older versions of LinFIR. Re-check the bass estimation on existing projects using adaptive windowing if this matters for your work.
• Driver IR window layout: the growing list of already-captured off-axis measurements (“Measurements”) is now shown below the measurement controls (“Measurement Controls”: axis selector, angle field, position dial, Import/Measure/Scan/Stop buttons, capture progress) instead of above them, so those controls no longer get pushed further down - and out of view - as more angles are captured.

Fixed

• Undo/auto-save snapshots while dragging: dragging a value in an IIR (global or per-driver), target curve, or FIR correction window no longer triggers extra undo-history/auto-save snapshots mid-drag when another such window happens to be open at the same time.
• Directivity Analysis overlay plot: the overlay plot’s curves no longer get truncated at the bottom by the sonogram colorbar’s lower bound setting. They now use the “Drivers range” setting (Settings -> Graphs) instead, which isn’t clamped to a heatmap-friendly dynamic range.
• Auto-scan not stopping on a rejected capture: if a measurement was rejected during an auto-scan (clipping, signal too low, or h1/h2 harmonic contamination), the scan no longer silently continued as if it had succeeded - it now properly aborts with an error, since the underlying problem would likely affect every remaining angle the same way.

[1.3.13] - 2026-08-27

Added

• Native file opening on macOS: double-clicking a .lnf project file in Finder (or using “Open With”) now opens it directly in LinFIR - including reusing an already-running instance of the app.
• Release notes in the documentation: the full changelog is now available as a “Release Notes” page in the documentation (Help -> Release notes), always in sync with the latest version.

Improved

• UI library: updated to the latest version for overall stability and appearance improvements.
• Update notification: after updating, LinFIR now shows a small, dismissible “See what’s new” prompt instead of automatically opening a large release notes window.

[1.3.12] - 2026-08-25

Fixed

• Audio crackling on Windows (ASIO): boosted the priority of ASIO audio threads to reduce crackling that could occasionally occur during measurements.

Improved

• Audio library: updated to the latest version for overall stability improvements.

[1.3.11] - 2026-08-23

Added

• IR window envelope preview: adjusting IR windowing (start/stop time, drag, Auto Set Window, Reset Window, or toggling Adaptive Window) now briefly overlays the window’s shape in red on the Impulse Response and Step Response graphs, scaled to the curve’s own peak. The overlay stays visible for 3 seconds after the last change, then fades out over 2 seconds. With Adaptive Window enabled, it shows a gradient of fade-out curves - one per frequency band, darkest (and drawn on top) for the highest frequencies.
• Verbose logging setting: a new toggle in Settings -> General -> Diagnostics raises the application log to debug detail on demand, without restarting or a command line. Enable it when asked to, reproduce the issue, then use Help -> Report a Bug to send the log.

Improved

• Maximum number of drivers: increased from 10 to 50 per project.
• Documentation: clarified what the magnitude correction filter’s Correction Gain does - a single, uniform gain applied within the configured frequency range to compensate for the level loss introduced by the from-the-top equalization - and how to inspect the filter’s final response (gain included) in the graph bar’s FIR tab to avoid excessive gain on a given band.
• Log verbosity: routine tracing that used to fire continuously while dragging any parameter (LinFIR recomputes filters live), and audio ring-buffer overflow warnings that could log on every callback during an overrun, are now much quieter by default - the log file stays focused on events that matter. Full detail remains available via the new Verbose logging setting.

Fixed

• ASIO capture failures (loadDriver returned false): the automatic device hotplug detection added in 1.3.8 could race against a capture that was starting at the same time - both tried to load the ASIO driver concurrently, and ASIO only allows one loaded driver at a time. The hotplug check now defers to an in-progress capture instead of racing it.
• Undo/redo debounce with multiple driver/measurement windows open: dragging a value no longer records one undo step per intermediate value when a second driver or measurement window is open at the same time.

[1.3.10] - 2026-07-28

Improved

• Airplay / wireless device support: automatic detection of Airplay, HomePod, Sonos, and similar streaming devices; adjusts timing constants (extended pre-silence, post-silence, and chirp detection delay) to compensate for network latency
• Deconvolution noise reduction: minor improvements to limit deconvolution noise and artifacts via smooth tapering at the beginning and end of the averaged recording.

Added

• h2/h1 contamination guards: measurements are now rejected when the second harmonic (h2) falls inside the h1 zone (reduce Manual offset) or when h1 arrives before the truncation point (increase Manual offset or check acoustic distances). A permanent toast notifies the user with guidance for each case.

[1.3.9] - 2026-07-22

Improved

• Spectrogram image quality and performance: the spectrogram now uses bilinear interpolation for smoother, higher-quality rendering, and has been optimised with parallel processing for faster display updates.
• Directivity sonogram normalization: added a new “Normalize to ±10° avg (per freq)” mode that normalises each frequency column to the average magnitude between −10° and +10°, providing a more robust reference than the single 0° point when narrow on-axis peaks or dips are present. A 20 kHz cap on normalization references now applies to the Global max. and Normalize to 0° modes, preventing tweeter ultrasonic resonances from skewing the colour scale.

[1.3.8] - 2026-07-21

Added

• Audio device hotplug detection: the audio settings page now automatically detects when speakers or microphones are connected or disconnected, and refreshes the device list without requiring a manual click.
• Directivity Sonogram (-6 dB curves): thin grey reference lines now appear on the sonogram plots, marking the -6 dB beamwidth curves. A matching grey mark also appears on the colour bar for quick level reference.
• Sonogram Normalization dropdown: three normalization modes control how sonogram magnitudes are mapped to dB — Global max. (single reference from the strongest data point), Normalize to 0° (single reference from the strongest on-axis measurement across all frequencies), and Normalize to 0° (per freq) (each frequency column normalised to its own 0° value, revealing per-band directivity patterns).

Improved

• Plot axes: grid spacers and tick formatters replaced with dynamic frequency formatting — axes now show meaningful labels at every zoom level
• Documentation: faster fallback when online documentation is not available
• Audio library: updated to the latest version for overall stability improvements
• Harmonic distortion windowing: replaced fixed 5% tapers with an adaptive windowing algorithm for individual harmonic slices. The analysis now dynamically strips background noise before and after each impulse peak while automatically preserving long distortion tails, ensuring cleaner measurements without cutting off legitimate decay.

[1.3.7] - 2026-07-13

Added

• Rotation Table (Pololu Tic): new Go to 0°, Energize, and De-energize buttons in Settings -> Rotation Table

Improved

• Harmonic distortion: recomputing THD after a windowing or delay-compensation change is now much faster — the expensive per-harmonic spectra (H2-Hn) are cached and only recomputed after a new capture or a sample-rate change, instead of on every window adjustment

Fixed

• Rotation Table auto-scan: fixed a hang after the table’s first move when the Settings window was closed — the table position is now refreshed every frame regardless of which window is open

[1.3.6] - 2026-07-09

Fixed

• macOS/Windows crash at launch: libusb is now compiled statically into the binary (rusb vendored feature); the app no longer depends on an external libusb-1.0.0.dylib from Homebrew and will launch correctly on machines where Homebrew is not installed

[1.3.5] - 2026-07-08

Added

• Rotation Table support: connect and control a motorized turntable directly from LinFIR — supported controllers are Pololu Tic (T500, T834, T825, T249, 36v4, detected automatically via USB) and GRBL-based boards (Arduino, e.g. Audiomatica Medusa, connected via serial); the new Settings -> Rotation Table tab provides a live position dial, manual jog controls, and a zero-point reset button
• Auto-scan (requires a valid license): in the IR Management window, a ▶ Scan button drives the rotation table through a configurable angle range (min, max, step) and captures an impulse response at each position automatically — already-measured angles are skipped; a 30-second per-angle timeout guards against mechanical failures
• New C array export format options (f32 and f64) for FIR filters

Improved

• Documentation: the documentation printing experience is now fully supported — text is black on white, images use the light theme variant, and all UI chrome is hidden for a clean printout
• UI library updated to the latest version, bringing general bug fixes and stability improvements; note: a multi-monitor window positioning issue may still occur on Windows

[1.3.4] - 2026-06-08

Improved

• Harmonic distortion extraction: the HD slices are now cut at the minimum of the smoothed Hilbert envelope between each pair of adjacent theoretical peaks, maximising separation between harmonics
• Harmonic distortion windowing: replaced the h1-synchronised window with an independent automatic windowing; this is more robust for harmonics whose impulse response spreads significantly further in time than h1

[1.3.3] - 2026-06-07

Added

• Distortion IR window (shortcut J, View menu, requires a valid license): visualises the time-domain harmonic impulse responses H2..H5 extracted by the Farina ESS method; each curve is the same windowed slice used to compute the THD spectrum so the display is fully consistent with the distortion graphs; per-harmonic visibility toggles, independent per-harmonic normalisation, zoom/pan with double-click reset; the displayed angle follows the global angle selector

Improved

• Spectrogram: the window now follows the global angle selector — when a non-zero angle is selected the spectrogram shows the predicted sum IR for that angle (same data as the impulse graph in Drivers mode); the window title and in-window header update accordingly; on-axis behaviour is unchanged

Fixed

• THD formula corrected: harmonic distortion (total and individual) is now normalised by the fundamental level (H1) in accordance with the standard THD definition — THD = √(H2² + H3² + H4² + …) / H1; previously the denominator was the total RMS signal (fundamental + harmonics), which capped the displayed value at 100% and caused the curves to underestimate distortion at high harmonic levels (below resonance or at high SPL); values above 100% (0 dB in relative mode) are now correctly displayed
• Harmonic distortion windowing: H2–H5 are now windowed using the start and stop boundaries configured for H1 in Time Windowing, improving SNR and eliminating room echoes from the harmonic distortion estimates

[1.3.2] - 2026-04-29

Added

• Spectrogram window (shortcut W, requires a valid license): time-frequency colour map of the predicted on-axis sum IR (Loudspeaker) or averaged IR (Room Calibration); choose between CQT (Morlet filter bank, constant-Q) and STFT (Hann window); adjustable resolution preset, colormap dB range, and axis swap (time ↔ frequency)
• Directivity Analysis: new Overlay Plot tab showing all off-axis frequency response curves superimposed, one rainbow-coloured line per angle, with angle/frequency/magnitude tooltip on hover
• Directivity Analysis — Overlay Plot: angle range filter (All / ≥ 0° / ≤ 0° presets + drag-value min/max), defaults to 0°–180°
• Directivity Analysis — Overlay Plot: Normalize to 0° toggle — divides each curve by the on-axis response to display relative deviation
• Directivity Analysis — Sonogram: Colormap dB range controls — drag min/max dB values to saturate the colour scale and reveal low-amplitude detail; a Reset button restores the defaults

Improved

• Directivity Analysis: renamed from “Directivity Sonogram” to “Directivity Analysis”

[1.3.1] - 2026-04-19

Added

• Frequency graph (Drivers mode): the IIR Auto EQ target curve is now displayed as an overlay on the frequency response graph when the IIR window is in the foreground and the Auto EQ tab is selected; the overlay is automatically hidden when the window is sent to the background

Improved

• Auto EQ: improved filter initialization: filters are now placed using a score that trades off residual magnitude against proximity to already-placed filters, preventing wide low-Q filters from masking the residual error they leave behind and causing subsequent filters to be initialized in the wrong place
• Auto EQ (driver and global): Q Min and Q Max now use adaptive bounds: increasing Q Min automatically pushes Q Max up if needed (up to 20), and decreasing Q Max pulls Q Min down if needed (down to 0.1), so the two values can no longer cross; both parameters now share the full 0.1–20 range

Fixed

• HFD example in the documentation: the Tweeter was mapped to channel 1 which is inconsistent with Hypex plate amplifiers

[1.3.0] - 2026-04-15

Improved

• Auto EQ: improved algorithm to better respect the configured gain limits and boost cap
• Adaptive Window: upgraded internal wavelet transform from CDF 5/3 — the biorthogonal wavelet now used is CDF 9/7, which provides ~40 dB of stopband attenuation versus ~18 dB for CDF 5/3; this significantly reduces spectral leakage between frequency bands before temporal masking is applied, eliminating the frequency ripples that could let high-frequency echoes bleed through the window boundary
• UI framework updated to the latest version
• Documentation: microphone calibration phase column clarified — files with a third phase column are accepted but ignored; phase is reconstructed from magnitude via Hilbert transform, which is more robust given measurement microphones’ predominantly minimum-phase behavior
• Documentation: buffer size guidance corrected — buffer latency is compensated internally; the only criterion is avoiding underruns
• HFD export: the patch/create mode is now a dropdown (Create from template / Patch existing Config.xml) instead of a toggle button, making the choice more explicit
• HFD export: the channel mapping panel now uses the term “Channel” instead of “Group”; the description clarifies that each channel receives the FIR filters, IIR filters, gain, delay, and polarity of the assigned LinFIR driver; by default, drivers are assigned to channels sequentially (driver 1 -> channel 1, driver 2 -> channel 2, …); if there are more channels than drivers, the last driver is repeated for the remaining channels
• HFD export: the output directory field is removed; clicking Apply mapping and export in Create from template mode now opens a Save As dialog (pre-filled with Config.xml); in Patch mode the selected file is overwritten directly, with no extra dialog
• HFD export: a confirmation message is displayed after a successful export, showing the path of the written file
• Documentation: HFD section updated to clarify that the exported file is Config.xml, the native project format opened by Hypex Filter Design software
• Settings: Adaptive Crossover Smoothing moved from Settings -> General -> Default Values to Settings -> Graphs -> Display — it controls graph rendering for the current project and is not a new-project default value; documentation updated accordingly
• Graphs: “Adaptive Crossover Smoothing” renamed to “Steep Crossover Smoothing” — smoothing reduction is now proportional to the filter slope; no effect below 12 dB/oct (LR2), full reduction at 36 dB/oct (LR6+), proportional in between
• UI: boolean options have been replaced by an iOS-style toggle switch throughout the application for improved readability

Added

• Two new IIR filter types: Asym Low Shelf and Asym High Shelf, available in per-driver IIR filters and global IIR filters; both expose F1/Q1 (zeros) and F2/Q2 (poles) parameters; the High Shelf variant is normalised so that the DC gain is always 0 dB
• HFD export: driver alignment delays are now validated against the 19.2 ms hardware limit; an error is shown if any driver exceeds this value, preventing a silent export with incorrect timing
• HFD export: first-order high-pass and low-pass filters (Butterworth order 1) are now exported as ftHighPass1 / ftLowPass1 instead of the second-order variant; odd-order filters (e.g. Butterworth order 3) correctly export the first-order section as ftHighPass1 / ftLowPass1 and the remaining second-order sections as ftHighPass2 / ftLowPass2
• Auto EQ: new Auto-generate toggle in the IIR Auto EQ tab (per-driver and global) — when enabled, filters are automatically re-generated whenever any parameter in the Auto EQ tab changes; requires a valid license
• Sync X-axis: the button is now active by default for all new projects
• Settings -> General -> Default Values for New Projects: new “Sync X-axis” option to configure whether X-axis synchronisation is enabled or disabled by default for new projects
• IR Management window: each measurement in the angle list now shows an ℹ icon; hovering it displays the measurement source (measured with LinFIR or imported), sweep parameters (frequency range, duration, level, averages), and IR length
• Frequency, phase, and group delay graphs (Drivers mode): a dashed pre-filter overlay is now available for each driver and for the sum (requires a valid license); the overlay shows the unfiltered speaker response and group delay/phase when the corresponding filter window (LP, HP, correction FIR, IIR, or global filters) is open; toggled in Settings -> Graphs -> Response Overlays

Fixed

• HFD export: Bessel filters of order 1 are not supported by the HFD hardware; the export now raises an error if such a filter is detected on any driver; in FusionAmp mode the per-driver IIR filter UI enforces a minimum order of 2 for Bessel filters, preventing the configuration of an unsupported filter
• Frequency graphs (IIR Filters, FIR Filters, FIR+IIR Filters modes): the Y-axis lower bound no longer drops to the configured limit (~−100 dB) when only EQ-type filters are active (PEQ, shelves, all-pass, Asym Low/High Shelf); the deep lower bound is now applied only when at least one true high-pass or low-pass filter is present in the relevant filter set
• Frequency graph (Drivers mode): Listening Window curve is no longer displayed when only a single driver is active and no global filters are applied — it now follows the same visibility rules as the Sum curve (requires multiple drivers with IR data, or at least one driver with active global filters)
• About window (Windows): clicking “Read terms in browser” or “Open in browser” no longer causes a black console window to flash
• Directivity Sonogram: the sonogram no longer remains visible after measurements have been cleared — the cached images are now properly discarded when the async computation returns no data
• IR Management window (Hypex FusionAmp mode): the off-axis section (axis selector, angle list, measurement angle input, Export All) was missing — it now appears identically to Loudspeaker Design mode
• Phase graph (Drivers mode, off-axis angles): in time-centered phase and unwrapped mode, the linear phase reference was incorrectly taken from the on-axis sum instead of the sum at the selected angle; individual driver curves and the sum could appear misaligned when viewing off-axis measurements
• Phase graph (Drivers mode, unwrapped): individual driver curves could be offset by a multiple of 360° relative to the sum when steep high-pass crossover filters were active; the DC alignment now only uses in-band frequencies (within each driver’s passband) to avoid corruption by the out-of-band phase ramp introduced by the crossover
• IIR filters: setting a Bessel filter order to 0 via keyboard arrows caused a crash; the filter design now returns a pass-through (no filtering) for order 0

[1.2.21] - 2026-03-25

Fixed

• License activation: fixed a crash with “Invalid response: Failed to parse validation response: expected value at line 1 column 1” — the HTTP status was not checked before parsing the API response body; any non-success response (Cloudflare challenge, rate limiting, transient server error) now returns a clear error message instead of a JSON parse failure
• macOS full-screen: secondary windows (filters, IR management, etc.) no longer open as a tab inside the full-screen space — they now float above the main window as expected; this also fixes the black empty space that appeared when a resizable window was opened in full-screen mode

Added

• Several windows whose content is variable in length are now resizable

Added

• Frequency graph (Speakers mode): Listening Window curve (requires valid license)
  • Displayed alongside the Directivity Index when off-axis measurements are available
  • Defined as the spatial average of all measurement angles within ±30° horizontal and ±10° vertical (inclusive)
  • Shown only when at least 2 qualifying angles are available to average
• Global FIR magnitude correction and global Auto EQ: new “Reference” option (requires valid license)
  • Choose between On-axis (default) and Listening Window as the magnitude reference for correction
  • When set to Listening Window, the correction targets the spatial average of ±30°H / ±10°V angles instead of the on-axis response
  • Falls back to on-axis automatically if fewer than 2 qualifying angles are available

[1.2.20] - 2026-03-18

Fixed

• ASIO (Windows): measurements no longer abort with a spurious “timing reference signal too weak” warning despite the signal being correctly configured — note that ASIO stream startup can take significantly longer than on other platforms; this delay is now properly accounted for before any signal level check is performed
• Hypex mode: several features that were incorrectly restricted to Loudspeaker Design mode now work as expected — affected areas include phase display options, per-driver filter curve views, and the detailed report export
• Phase graph: auto-bounds vertical axis now enforces a minimum range of ±1° around the data centre, preventing near-flat phase curves (e.g. a Dirac) from zooming into numerical noise
• Phase / group delay graphs: toggling visibility via keyboard shortcuts (G / P) no longer leaves the graphs empty

[1.2.19] - 2026-03-12

Fixed

• Hypex mode: the directivity sonogram window is now accessible
• Hypex mode: exporting FIR coefficients now works correctly

[1.2.18] - 2026-03-10

Added

• Phase correction FIR: auto-correction guard for magnitude artifacts
  • LinFIR automatically iterates f_min and f_max in 1/3-octave steps (up to 100 iterations, f_min ceiling 15 kHz, f_max ceiling 18 kHz) until the FIR magnitude deviation falls within the per-filter tolerance
  • Guard Tolerance parameter added to each phase correction filter (driver and global): 0.1–15 dB, default 0.5 dB
  • A warning toast is shown only when the guard fails (exhausts all iterations without converging); warnings can be disabled in General Settings -> “Warn when phase FIR guard fails”

Improved

• Phase and group delay graphs: auto-bounds for the vertical axis now ignore data above 20 kHz, preventing ultrasonic content from distorting the scale
• Phase graph: auto-bounds no longer force the 0° value to be visible; the axis now fits tightly to the actual data range (± 10% margin)
• Graph toolbar: graph type toggles, time domain options, and phase options are now grouped into three dropdowns (“Graphs”, “Time Domain”, “Phase Options”), reducing clutter in the toolbar; dropdowns stay open when toggling checkboxes inside them
• Software update notification moved from the toolbar to a persistent floating card anchored to the bottom-right corner, with a dismiss button and an Install button; after a successful update a permanent closable toast confirms the installation

[1.2.17] - 2026-03-06

Added

• Hypex FusionAmp project modes with hardware-specific constraints (requires valid license)
  • Sample rate locked at 93.75 kHz for all FusionAmp modes
  • Fixed driver count matching hardware specifications (cannot add/remove drivers)
  • Two FIR processing modes:
    • FIR IN: Global FIR correction at DSP input (4500 taps maximum including padding)
    • FIR OUT: Per-driver or global FIR at DSP output (1500 taps maximum per driver including padding)
  • IIR filter constraints: 15 biquads maximum per channel
    • Real-time biquad counter with color-coded warnings in IIR filter UI
    • Filter types and orders dynamically limited to prevent exceeding biquad capacity
• Phase correction FIR magnitude artifact detection (per-driver and global)
  • After computing the phase correction FIR, LinFIR evaluates its magnitude response and warns if any bin deviates by more than 0.5 dB from 0 dBFS
  • A phase-only filter should leave magnitude unchanged; deviations indicate a filter design problem
  • Warning toast is shown with actionable advice: reduce Kaiser beta, increase tap count, or raise the minimum correction frequency
  • Disableable in General Settings
• Target curve import/export (all target curve windows: per-driver FIR, per-driver Auto EQ, global FIR, global Auto EQ)
  • Import: loads control points from a plain-text file (two columns: frequency Hz, gain dB; separator: tab, space or comma; comment lines starting with # are ignored; minimum 2 valid points required; points are sorted by frequency on import)
  • Export: saves current control points to a plain-text file with a two-column tab-separated format and commented header
• Microphone calibration plot preview: each calibration file in Settings -> Mic Calibration now has a Plot button that opens a dedicated window showing its frequency response on a log-frequency axis

Improved

• All secondary windows (filter windows, IR windows, export dialogs, HFD window, FRD converters, detached graphs…) are now automatically closed when loading or creating a project, preventing stale windows from a previous session from cluttering the workspace; the Settings window is additionally closed on application startup
• Clipping detection test signal replaced by a deterministic periodic pink-noise buffer
  • New signal: frequency-domain pink noise (−3 dB/octave ASD), generated once per sample rate at startup and cached in memory
  • Schroeder initial phases + Newman/CFR iterative crest-factor reduction
  • Band-shaped: 12 dB/oct high-pass at 20 Hz, 24 dB/oct low-pass at 20 kHz
  • RMS = −18 dBFS, peak ≤ 0 dBFS (crest factor ≈ 12 dB — representative of real programme material)
• Documentation: added warning against using phase correction with non-anechoic measurements in both driver-processing and system-processing sections

Fixed

• Multi-average capture: input gain compensation is now applied in the audio callback instead of post-capture, so level validation (signal-too-weak check on timing reference) correctly sees the compensated signal
• Multi-average capture: sweep segments are now aligned with GCC-PHAT before averaging, using their individual deconvolutions as alignment references; removes the naive arithmetic average that could blur transients in the presence of inter-sweep timing jitter
• Audio Settings: switching to an input device that doesn’t support the current sample rate now automatically selects the closest supported rate and shows an info toast; previously the sweep was generated at the old rate while the device ran at its own rate
• Documentation: breadcrumb bar hiding the search panel
• macOS: project files (.lnf) hidden in Finder after save

[1.2.16] - 2026-02-19

Added

• Driver copy/paste parameters
  • Copy button on each driver copies all its parameters to a clipboard (IRs and driver name are excluded)
  • Paste button applies the copied parameters to any other driver, preserving the target driver’s IRs and name
  • Paste button is grayed out when the clipboard is empty
  • Useful for quickly duplicating filter settings, crossover configuration, or DSP parameters across drivers
• Default crossover frequencies applied when adding a new driver
  • Driver 1: HP=1 Hz, LP=110 Hz (subwoofer)
  • Driver 2: HP=110 Hz, LP=600 Hz (bass/low-mid)
  • Driver 3: HP=600 Hz, LP=3000 Hz (midrange)
  • Driver 4+: HP=3000 Hz, LP=21000 Hz (tweeter)
• Automatic measurement cancellation on signal problems
  • Clipping on main channel: measurement stops immediately when any input sample hits digital full scale, without waiting for sweep completion
  • Clipping on timing reference channel (Electric and Acoustic): measurement stops if the chirp clips on the reference input channel
  • Timing reference too weak: if the peak on the timing reference channel never exceeded −25 dBFS during the chirp window, the sweep is cancelled before it begins (Electric and Acoustic modes, regardless of number of averages); toast reports the measured level
  • Low signal on first sweep (multi-average only, ≥ 2 averages): if the peak input level never exceeded −25 dBFS on the first sweep, the measurement is cancelled before the second average begins
  • Real-time input level (dBFS) displayed during capture with color coding: green (good), orange (near saturation), red (clipping)
• Timing reference chirp now has short Hann fade-in and fade-out (1% of chirp duration each)
  • Eliminates clicks caused by the abrupt onset and end of the chirp signal
• CLIO time-domain IR import (.txt files exported from CLIO in time domain)
  • 3-column format: Time[s], Real[Pa], Imag[Pa] — the Imag column is ignored
  • Sample rate is inferred from timestamps and snapped to the nearest standard rate (same algorithm as TMD)
  • Auto-detected by the presence of the Time[s] header line, requiring no manual configuration

Improved

• Channel selectors (mic input, timing reference output/input) replaced with dropdowns
• TMD file import now snaps the inferred sample rate to the nearest standard rate (44.1, 48, 88.2, 96, 176.4, 192, 384 kHz…) when within 1% tolerance
  • Corrects systematic sample rate errors caused by limited timestamp precision in TMD files
  • Prevents subtle pitch/speed deviations that would otherwise affect the imported impulse response

Fixed

• Phase curve alignment in Loudspeaker Design mode (time-centered phase and unwrap)
  • Unwrap + time-centered: DC offset (absolute phase reference) for all curves — drivers and the filtered Sum — now anchored to the pre-filter sum phase; previously the filtered Sum had a different DC origin, making it appear shifted vertically relative to the drivers
  • Unwrap without time-centered: all curves (drivers + Sum) are now consistently aligned to the filtered Sum as reference; previously activating the pre-filter reference unconditionally caused a vertical offset on the Sum curve in this mode
• Adaptive wavelet windowing now covers the same physical frequency bands regardless of sample rate
  • Number of levels is now derived from target minimum frequency (5 Hz) relative to Nyquist, ensuring coverage down to ~5 Hz at all sample rates (48, 96, 192 kHz…)
  • Previously, at higher sample rates (e.g. 96 kHz), the cap prevented the lowest frequency bands from being reached, resulting in inconsistent temporal extensions for bass frequencies
• Keyboard shortcuts now work correctly when Settings window is open in background
  • Main window shortcuts were incorrectly blocked whenever Settings window was open regardless of focus
  • Settings window now only handles shortcuts when it has focus
  • License tab disables shortcuts when Settings has focus to allow typing in text fields without triggering application shortcuts

[1.2.15] - 2026-02-17

Added

• Microphone calibration settings page
  • Dedicated settings tab for managing microphone calibration files
  • Scrollable list displaying all imported calibrations with sensitivity factor and frequency point count
  • Default calibration selector for automatic application to newly created drivers
  • Import/delete controls for each calibration file
• Spacebar keyboard shortcut to interrupt sweep measurements at any time
  • Provides instant emergency stop during capture
  • Works alongside Stop Capture button for multiple interruption methods

Improved

• Microphone calibration workflow enhancements
  • Default calibration automatically applied to new drivers and new projects
  • Deleting a calibration file automatically resets affected drivers to None
  • Per-driver calibration selection moved to IR Management window for better context
• Adaptive windowing now uses wavelet transform for frequency-dependent IR windowing
  • Replaced 10-band FIR filter bank approach with CDF 5/3 biorthogonal wavelet decomposition
  • ~10× performance improvement
  • Start windowing applied uniformly in time domain, stop windowing applied adaptively in wavelet domain
• Delay compensation with sync toggle for consistent measurement delay compensation
  • New Sync toggle (enabled by default) automatically applies delay adjustments to all drivers
• Stop Capture button visual design during measurements
  • Red button with white text matching Reset button styling
  • Increased size for better visibility and accessibility
  • Appears only during active capture for unambiguous emergency stop
• IR Management window behavior during measurements
  • Window remains on top of all other windows during capture
  • Window maintains keyboard focus throughout measurement for reliable spacebar interruption
  • Prevents accidental loss of control if user clicks elsewhere during sweep
  • Automatically returns to normal window level when measurement completes or is stopped
• Documentation navigation enhancement with sticky breadcrumb indicator
  • Added breadcrumb banner below menu that tracks current position during scroll
  • Helps users orient themselves in long documentation pages
  • Improves navigation experience for complex multi-section pages

[1.2.14] - 2026-02-13

Added

• Individual harmonic distortion curves (H2, H3, H4) displayed on HD graph
  • HD graph now shows detailed breakdown of harmonic components in addition to total harmonic distortion
  • Individual harmonic curves require a valid license (total harmonic distortion remains available to all users)
  • Provides deeper insight into distortion characteristics and system non-linearity
• HD display mode selector with three visualization options
  • Percent (%) mode: Traditional percentage representation relative to total RMS
  • dB (relative) mode: Harmonic distortion relative to total RMS in decibels
  • dB (absolute) mode: Absolute level of harmonics without normalization
  • Dropdown control added to graph toolbar before smoothing selector
  • Mode applies to both total HD curve and individual harmonics (H2, H3, H4)
  • Useful for comparing distortion levels across different SPL measurements and evaluating signal-to-distortion ratio
• Driver offset parameters in FRD file import window
  • New X (horizontal) and Y (vertical) offset fields allow specifying driver position relative to baffle plane
  • Enables reconstruction of relative time-of-arrival delays that may have been removed during FRD file creation
  • Critical for accurate off-axis interference simulation around crossover frequencies
  • Automatically calculates geometric delays based on listening angles and driver positions
  • Improves accuracy of multi-driver system simulation from imported frequency response data
• TMD file format support for impulse response import

Improved

• Exponential sine sweep measurement system completely redesigned
  • Single-pass averaging: Timing reference eliminates need for multi-pass alignment, enabling more efficient averaging workflow
  • Reduced deconvolution artifacts: Improved signal processing prevents artifacts during impulse response extraction
  • Automatic sweep end frequency: f_end now automatically calculated to optimize harmonic distortion SNR (1/96 octave before Nyquist)
  • Simplified measurement workflow: End frequency parameter removed from UI, managed automatically based on sample rate
• Measurement capture stop behavior improved for graceful interruption
  • Stop button now immediately interrupts sweep in progress instead of waiting for completion
  • Sweep signal automatically attenuated with fade-out to prevent audio clicks and pops
  • Enables quick measurement cancellation without disruptive audio artifacts
• HD graph replaces THD nomenclature
  • Graph renamed from “THD” to “HD” (Harmonic Distortion) for clarity
  • Keyboard shortcut (K) remains unchanged
  • Updated throughout UI and documentation
• Documentation dark theme (coal) with darker background and lighter text for improved readability

[1.2.13] - 2026-01-29

Added

• Batch import for multiple measurement files with automatic angle parsing
  • Import multiple IR files simultaneously with multi-file selection
  • FRD converter parameters (sample rate, delay, extrapolation modes) automatically applied to all files in batch
• Manual FIR delay compensation visibility control (advanced setting)
  • New “Show manual FIR delay compensation” option in Settings -> General -> Filter Processing
  • Hidden by default to prevent accidental filter misalignment
  • When disabled, all FIR compensation delays are automatically forced to 0
  • Marked as advanced setting (⚙) with warnings about signal processing knowledge requirements
  • Affects both per-driver “FIR Offset Delay” and global “FIR compensation delay” fields

Improved

• Driver control interface clarity with organized section headers
  • Added colored section labels: “FIR Filters”, “IIR Filters”, and “Driver Adjustments”
  • Replaced ambiguous “Visualization Controls” with clear “Driver Adjustments” label
  • Improved visual hierarchy with consistent spacing and separators
  • Better organization distinguishes signal processing filters from gain/delay/polarity adjustments

Fixed

• FIR correction frequency range controls now use proper undo debouncing
  • Fixed issue where dragging frequency min/max sliders created separate undo history entries for each incremental value change
• Off-axis sum calculation now correctly handles missing measurements per driver
  • Fixed bug where drivers without measurements at a specific angle would incorrectly use their on-axis (0°) measurement as fallback
  • Fixes issues with directivity sonogram displaying incorrect data at extreme angles
• Directivity sonogram normalization now excludes Nyquist frequency artifacts
  • Sonogram color scale normalization now limited to frequencies up to 24 kHz
  • Prevents measurement artifacts near Nyquist frequency from affecting visualization scale
  • Default frequency axis bounds now limited to 24 kHz for cleaner visualization

Breaking Changes

• Distortion measurement storage structure has been changed
  • Internal data structure for storing distortion harmonics has been completely redesigned
  • Backward compatibility is NOT maintained: Projects with distortion measurements from versions < 1.2.13 are not compatible
  • Re-measurement is required: Importing old IRs will not restore distortion data - you must re-measure to capture distortion with the new structure

[1.2.12] - 2026-01-25

Fixed

• Off-axis measurements now correctly calculated after FRD file import
  • Fixed issue where importing FRD files (frequency response data) would not trigger off-axis response calculations
• Reduced log verbosity for Directivity Sonogram window
• Sum curve now displayed when some enabled drivers have no measurements
  • Previously, sum curve was hidden if any enabled driver lacked a measurement
  • Now, sum is calculated and displayed from all enabled drivers that have measurements
  • Drivers without measurements are treated as if they were disabled

Improved

• Impulse response level validation now uses frequency domain analysis
  • Previously checked instantaneous peak level in time domain
  • Now computes FFT and checks maximum magnitude across all frequencies
  • More accurate detection of low signal-to-noise ratio issues
  • Warning triggered if no frequency has magnitude above -25 dB

[1.2.11] - 2026-01-24

Added

• Custom file icon for .lnf project files
  • LinFIR project files (.lnf) now display a custom icon in file explorer
  • Works on Mac OS and Windows
  • Improves visual identification of project files

[1.2.10] - 2026-01-23

Added

• Adaptive crossover smoothing option for magnitude curves near crossover frequencies
  • Preserves crossover slope visibility by reducing smoothing within ±0.5 octave of cutoff frequencies
  • Controlled via Settings -> General -> Adaptive Crossover Smoothing (enabled by default)
• Phase smoothing
  • Phase curves can now be smoothed while preserving driver alignment
  • Magnitude-based filtering applied before smoothing to eliminate out-of-band artifacts
  • Provides cleaner phase visualization without compromising timing accuracy
• Butterworth 4th-order sweep tapering for reduced pre/post-ringing
  • New sweep windowing mode with smooth frequency-domain tapering (24dB/octave rolloff)
  • Significantly reduces pre-ringing and post-ringing artifacts in measured impulse responses
  • Eliminates Gibbs oscillations at spectrum edges caused by abrupt frequency cutoffs
  • Sweep extends to f_start/4 and f_end×4 with Butterworth filters centered at user frequencies
  • Recommended for cleaner measurements (enabled by default)
  • Available in Settings -> Audio -> Sweep Windowing: None (no tapering) or Butterworth 4th (24dB/oct)
• ERB (Equivalent Rectangular Bandwidth) smoothing mode
  • Perceptually-motivated variable smoothing that follows the ear’s frequency resolution
  • Available in the Smoothing dropdown menu (Drivers mode only)

Improved

• Off-axis response calculation performance significantly optimized
  • Unified threading architecture eliminates race conditions between off-axis and directivity index calculations
  • Substantially faster computation time through improved parallelization
  • Directivity Sonogram updates automatically without requiring mouse movement
• Automatic impulse response noise truncation now limits IR length to 2 seconds after peak
  • IRs are first truncated to 2 seconds maximum after the main impulse peak (maximum absolute value)
  • Then trailing noise below 0.1% of peak amplitude is removed
  • Minimum duration of 1.5 seconds is always preserved
  • Improves processing performance and reduces file sizes for long IRs with extended decay
• Plot data resampling optimization using continuous logarithmic function
  • Replaced fixed frequency bands with smooth exponential downsampling function
  • Eliminates hard frequency band transitions for visually smoother plots
  • More efficient computation
  • Maintains perceptually constant point density across logarithmic frequency axis
• ASIO device compatibility now includes 24-bit sample format (I24) support
  • Added I24 format to preferred and fallback format detection
  • Improves compatibility with professional audio interfaces like Dante Virtual Soundcard
  • Resolves “no compatible sample format found” errors with 24-bit ASIO devices
• Enhanced measurement safety warnings in confirmation dialog
  • Added explicit low frequency warning about high energy bass content
  • Warns against measuring unprotected tweeters or midrange drivers
  • Clarifies potential damage risks from high excursion at low frequencies
• Sweep measurement parameters are now remembered per-driver
  • Duration, level, frequency range, and averages are automatically saved to each driver
  • Settings persist when closing and reopening the IR window
  • New projects start with application default values, customizations are preserved in existing projects
• Default sweep parameters relocated to Audio Settings
  • Moved from Settings -> General to Settings -> Audio Settings -> Measurement Configuration
  • Now grouped with other measurement-related settings (sweep channels, calibration, timing reference)
  • More logical organization: all measurement parameters in one place

Fixed

• Directivity Index not updating with latest off-axis data
  • Race condition eliminated where DI could be calculated from stale measurements
  • DI now guaranteed to reflect current off-axis responses

[1.2.9] - 2025-12-30

Fixed

• FIR filter export filenames now display correct tap count
  • Fixed inconsistency where exported FIR filter filenames showed incorrect number of taps
  • Tap count in filename now always matches the actual length of the exported filter
  • Affects all export formats (CSV, WAV, bin, txt, Armonia) and both individual and batch exports
  • Correctly reflects “per driver tap length” setting when enabled
• Log file rotation now properly enforces 4 MB size limit
  • Previous implementation allowed log files to grow indefinitely (up to 188 MB observed)
  • Oversized logs are automatically rotated and renamed with timestamp
  • Only the 2 most recent log files are retained to manage disk space

Improved

• Bug report notifications now use toast messages
  • Success notifications display for 3 seconds before auto-dismissing
  • Error notifications remain visible until manually dismissed
  • Error messages include manual submission instructions and support email

[1.2.8] - 2025-12-30

Improved

• Filter Export behavior: disabled drivers are now skipped
  • FIR and IIR filter export now only exports filters from enabled drivers
  • Disabled drivers (checkbox toggled off in Drivers toolbar) are automatically skipped during export
  • Applies to all export formats and methods (individual FIR files, IIR files, and HFD export)
  • Improves workflow efficiency by preventing accidental export of inactive driver configurations
• FIR Magnitude Correction taper standardized to quarter-octave
  • Changed from 5% bandwidth taper to quarter-octave (1/4 octave) transition zones
  • Matches the taper behavior of Auto EQ for consistency
  • Provides smoother frequency transitions at the edges of the correction range
  • Applies to both per-driver and global FIR magnitude corrections

[1.2.7] - 2025-12-27

Added

• FIR Filter Export: Stereo WAV formats
  • New stereo export options: WAV i16 (stereo), WAV i24 (stereo), WAV f32 (stereo)
  • Mono formats renamed to clarify: WAV i16 (mono), WAV i24 (mono), WAV f32 (mono)
  • Filter is duplicated identically in both left and right channels for stereo files
• IIR Filter Export: MiniDSP format
  • New export format option for MiniDSP processors
  • Exports biquad coefficients in compact text format: biquad1, b0=…, b1=…, b2=…, a1=…, a2=…
  • Available alongside existing FQG, CamillaDSP, and Q-SYS export formats in File > Export > IIR Filters

Fixed

• Directivity Sonogram flickering during parameter changes
  • When modifying crossover or filter parameters, the sonogram would briefly disappear before reloading
  • Sonogram remains visible and smoothly updates to new parameters without flickering

[1.2.6] - 2025-12-22

Added

• IIR Filter Export: Q-SYS CSV format
  • New export format option for Q-SYS IIR Custom Filter components
  • Exports biquad coefficients in CSV format: b0, b1, b2, a0, a1, a2 (one biquad per line)
  • Supports all IIR filter types: Butterworth, Linkwitz-Riley, Bessel, PEQ, Low-shelf, High-shelf, Allpass (1st and 2nd order)
  • Coefficients exported with 16 decimal places precision for maximum accuracy
  • Available alongside existing FQG and CamillaDSP export formats in File > Export > IIR Filters
• Documentation enhancements
  • New “Design Philosophy” section in Overview explaining LinFIR’s DSP-faithful, time-domain approach
  • New “Impulse Response Timing Requirements” section in IR Management explaining 200ms peak delay limit and best practices

Fixed

• In same cases a deleted driver/measurement instance could lead to a crash
  • Internal lists linked to drivers/measurements/ are now checked regularly to preserve coherence
• Long impulse responses with significant delays (>500ms) were being truncated during spectrum calculation
  • freqz() and freqz_complex() now dynamically adjust FFT size to accommodate full IR length
  • Prevents artifacts and incorrect frequency response for IRs longer than the requested spectrum points
• Import rejection for excessively delayed impulse responses
  • LinFIR now rejects impulse response imports where the main peak is located beyond 200ms
  • 200ms corresponds to ~68 meters acoustic distance, indicating missing timing reference in measurement
  • Prevents performance degradation from excessive zero-padding before useful signal
  • Error message explains computational constraints and recommends proper timing reference methods

Improved

• Phase Correction Filter stability
  • Improved algorithm robustness when correction band extends near Nyquist frequency
  • Better noise handling in high-frequency regions
  • Eliminates phase offsets and artifacts for cleaner, more stable impulse responses
• Auto Causal Alignment algorithm
  • Now uses logarithmic energy distribution instead of simple first-significant-sample detection
  • Positions FIR impulse peak based on energy ratio: if 30% of log-energy is before peak, places peak at 30% of FIR taps
  • Prevents truncation of both pre-ringing and post-ringing artifacts in combined causal filters
  • More robust with asymmetric impulse responses and complex filter combinations
• Clearer logging for debug
• Audio input channel defaults
  • When selecting an input device with at least 2 channels available, LinFIR now automatically sets input channels to 2 instead of 1
  • Applies both when manually changing devices and during initial device detection
  • Electric loopback timing reference now defaults to channel 2 for both input and output (avoids collision with microphone on channel 1)
  • Improves out-of-box experience for electric timing reference without manual channel configuration
• Documentation enhancements
  • User Interface section has been moved to the beginning of the documentation (after Project Modes) for better pedagogical flow
  • New comprehensive Graph Toolbar section in Graph Interaction page, now positioned at the top
  • Added interactive zoom-on-hover for graph toolbar images (hover to zoom in on specific areas)
  • Clarified Time-Centered Phase description (linear regression to remove linear phase component)
  • Clarified Smoothing behavior (explicitly states which graphs are affected)
• Time-Centered Phase algorithm improvements
  • Replaced phase regression method with IR peak position detection for more accurate bulk delay removal
  • Provides more robust phase centering, especially for impulse responses with significant delays or complex phase behavior
  • Applied to both graph display (time-centered phase toggle) and resampling algorithm (upsampling phase correction)
• Resampling algorithm optimizations
  • Improved stability with long-delay impulse responses
  • Reduced sinc filter length from 256 to 128 to minimize edge artifacts
  • Upgraded interpolation from Linear to Cubic for better quality
  • Reduced oversampling factor from 256 to 128 for improved numerical stability
• Spectrum calculation performance for long IRs
  • Significantly reduces FFT size and computation time for IRs with long initial delays while maintaining accuracy

[1.2.5] - 2025-12-19

Added

• Documentation enhancements
  • Added comprehensive “Performance Considerations” section in global-settings.md
    • Explains LinFIR’s real-time filter processing architecture
    • Documents filter application methods (FIR: fast convolution, IIR: difference equations)
    • Performance impact factors: filter length thresholds, off-axis measurement considerations
    • Background processing architecture for UI responsiveness
    • Optimization recommendations for slower systems and complex projects
  • Enhanced clipping detection documentation
    • Clarified that warnings are purely about filter processing, not measurement levels
    • Explained why phase processing can cause clipping even with flat magnitude responses
    • Added technical details: 32-tone test signal (20 Hz - 20 kHz), 1.1 detection threshold
    • Documented common scenarios: phase correction, all-pass filters, multiple FIR filters

Fixed

• Clipping detection test signal
  • Corrected multitone test signal from 31 to 32 frequencies
• Documentation accuracy improvements
  • Corrected Tukey window formula in ir-management.md to match actual implementation (was using undefined “\text{prog}” variable)
• Undo/Redo bug causing empty driver graphs
  • Fixed issue where undo/redo operations would sometimes cause driver graphs to become empty
  • Root cause: Restored measurements weren’t being resampled/recalculated due to cached state
  • Solution: Invalidate resampling cache on snapshot restore to force complete recalculation
  • Affects: Driver frequency response, phase, group delay, and impulse graphs after undo/redo
• Documentation refering to the View menu instead of Edit for the settings
• Buttons not showing in Sweep confirmation dialog

[1.2.4] - 2025-12-18

Added

• Documentation screenshots
• Documentation reorganization
  • Added comprehensive “Target DSP Settings” section in driver-processing.md
  • Consolidated sample rate, FIR length, per-driver tap lengths, and export padding documentation
  • Added per-driver tap lengths screenshot showing UI changes when enabled
  • Clarified that Filter length and Export padding controls relocate to individual driver sections when per-driver mode is active
  • Removed redundant “Filter Length Configuration” section

Fixed

• Documentation accuracy improvements
  • Removed Auto EQ workflow references from Manual Tab section in driver-processing.md
  • Clarified locked filters usage in manual IIR filter design context
  • Added comprehensive TXT file format documentation in Supported Formats section
  • Documented TXT file metadata support (sample rate, distortion data, ESS parameters)
  • Clarified that TXT files can contain either impulse samples or FRD format data
  • Corrected sweep-measurements.md to reference 🎤 Measure button (was incorrectly called “Capture”)
  • Added documentation for measurement confirmation dialog and skip option in settings
• License validation offline behavior
  • Fixed infinite validation loop when machine is offline
• IR import validation improvements
  • Added file extension validation (only .wav, .txt, .frd supported)
  • Added data validation to reject empty files, single-sample files, and files containing NaN values
  • Improved error messages with reference to documentation for unsupported formats

[1.2.3] - 2025-12-17

Fixed

• Documentation theme customization
  • Fixed theme selector to show only Auto, Light, and Coal themes
  • Corrected inconsistent styling between auto mode and manual theme selection
  • Ensured proper LinFIR branding colors across all themes

[1.2.2] - 2025-12-17

Improved

• Auto Detect Delay now applies minimum delay across all drivers
  • Previous behavior: Applied delay compensation only to the current driver
  • New behavior: Detects delay for all drivers and applies the minimum delay to all simultaneously
  • Benefits:
    • Preserves temporal alignment between drivers automatically
    • Removes common propagation delay from all measurements
    • Prevents manual alignment errors when setting up crossovers
  • Detection method: Uses peak position of on-axis measurements for precise alignment at zero
  • Manual adjustment still available for individual drivers if needed-
• Documentation access prioritizes online version
  • When accessing documentation (Help > Documentation / H key), LinFIR now attempts to open the online documentation at https://demaudio.com/doc/linfir/ first
  • If no internet connection is available or the online version is unreachable, automatically falls back to the embedded local documentation server
  • Benefits:
    • Users always access the latest documentation updates when online
    • Seamless offline access with no additional configuration required
• FRD Converter: Configurable magnitude extrapolation modes
  • Added separate extrapolation control for DC and high-frequency regions
  • Two modes available:
    • Constant: Flat extrapolation using endpoint value
    • Roll-off: Applies smooth roll-off with adaptive curvature
  • DC roll-off: 12 dB/octave high-pass characteristic for realistic low-frequency behavior
  • HF roll-off: 12 dB/octave low-pass characteristic, compensates for upward slopes
  • Default settings: Roll-off for DC, Constant for Nyquist
  • Benefits:
    • More realistic impulse responses from incomplete FRD data
    • Prevents artifacts from abrupt extrapolation
    • Maintains adequate signal level up to Nyquist frequency

[1.2.1] - 2025-12-13

Added

• Import validation warning for excessive delays
  • When importing impulse responses from TXT or WAV files (excluding FRD files), LinFIR now detects the main peak position
  • Displays a warning toast if the main peak is located beyond 20 ms
• Import validation warning for low signal levels
  • When importing impulse responses from TXT or WAV files, LinFIR now checks peak signal level
  • Displays a warning toast if peak level is below -25 dB
  • Low levels often indicate poor signal-to-noise ratio, which can compromise correction filter quality and phase visualization accuracy
• Web-based documentation system
  • Replaced in-app egui documentation with comprehensive mdBook-based documentation
  • Embedded directly into the application binary for offline access
  • Served locally via built-in HTTP server on port 3030+
  • Features:
    • 17 comprehensive documentation pages covering all LinFIR features
    • Theme-aware screenshots with automatic dark/light mode switching
    • Integrated search functionality
  • Access via Help > Documentation (H key) or Menu > Documentation

Fixed

• HFD export driver naming bug
  • Fixed issue where all drivers displayed as “Driver 1” in group-to-driver mapping dropdown
  • Driver names now correctly show as “Driver 1”, “Driver 2”, etc. when not renamed
  • Custom driver names are properly displayed when set by user

[1.2.0] - 2025-12-11

Added

• Room Calibration Mode
  • New dual-mode project system: Loudspeaker Design (default) and Room Calibration
  • Mode selection available in File menu
  • Room Calibration features:
    • Measurements are temporally aligned using GCC-PHAT (Generalized Cross-Correlation with Phase Transform) algorithm
    • Subsample precision via parabolic interpolation of the correlation peak
    • Measurements automatically averaged after alignment to create system impulse response
    • UI adapted: “Drivers” renamed to “Measurements”, directivity features disabled
    • Sweep output channel selector in IR window (excludes timing reference channel)
    • Graph display shows only global filters (FIR and IIR), individual measurement curves hidden
    • Export restrictions: Only global filters exported (FIR/IIR), HFD config and reports disabled
  • Designed for in-room acoustic measurements and correction filter generation
  • Preserves all loudspeaker design capabilities in Loudspeaker Design mode
• FRD to IR Converter utility window
  • Integrated advanced FRD-to-impulse response converter directly into LinFIR
  • Opens automatically when importing .frd or FRD-formatted .txt files
  • Features:
    • PCHIP interpolation with adjustable sampling frequency and low/high-frequency extrapolation modes
    • Real-time visualization: Original + Interpolated FRD, Impulse Response, Reconstructed Response
    • Quality metrics: data coverage percentage, reconstruction error analysis
    • Minimum phase transformation option to eliminate pre-ringing
    • Configurable delay compensation
    • Phase display options: wrap/unwrap, with/without linear phase removal
  • Replaces previous basic FRD import with full-featured converter from standalone frd-to-ir tool
  • Direct integration: converted impulses automatically update the target driver measurement
• Directivity analysis tools (requires license)
  • Polar measurement capabilities on both vertical and horizontal axes
  • Off-axis curve visualization with selector in toolbar
  • Directivity Index (DI) prediction curve
  • New window displaying predicted directivity sonograms (vertical and horizontal)
  • DI and sonograms reflect filtering choices between drivers
  • Requires IR measurements preserving time of flight to accurately represent interference patterns between drivers
  • Note: Directivity tools require a purchased license; other features remain free

Changed

• IR naming behavior
  • Removed automatic renaming of impulse responses during import and measurement
  • IR name field now preserves user-entered names across operations
  • Users must manually name their IRs via the text field in the Driver IR window
  • Previous behavior: IRs were automatically renamed to filename/timestamp on import/capture
  • New behavior: Manual naming only, preventing unwanted name changes

Improved

• File saving optimizations
  • Implemented compression to reduce file sizes
  • Faster save operations with smaller project files
• Memory optimizations
  • Reduced RAM usage through internal memory management improvements
  • More efficient memory allocation and deallocation
• Performance optimizations
  • Optimized calculations for improved app responsiveness
  • Faster UI updates and graph rendering

Fixed

• Phase correction in upsampling
  • Fixed phase artifacts in interpolated spectrum (above original Nyquist frequency)
  • Algorithm now: calculates phase slope before resampling, removes it, enforces constant phase above old Nyquist, then restores temporal coherence
  • Applies only to upsampling operations (e.g., resampling from 48 kHz to 96 kHz)
• Time-centered phase plot
  • Limited phase slope calculation to 20 kHz maximum
  • Prevents phase noise at high frequencies from affecting the centering calculation
  • Applies to both Speakers mode and FIR/IIR filter modes

[1.1.20] - 2025-11-23

Fixed

• Graph interaction improvements

  • Fixed axis drag zoom not disabling auto-bounds
    • Dragging on X or Y axes to zoom now correctly disables auto-bounds for the affected axis
  • THD graph now only displays in Speakers/Drivers mode
    • Previously shown in all display modes, now correctly restricted to Speakers mode only
    • Maintains proper display in detached windows

• Configuration report export (TXT/PDF) fixes

  • Fixed AutoEQ filters not being exported when AutoEQ tab is selected
    • Filter type counting now correctly uses the active tab’s filters (Manual or AutoEQ)
    • Correction Filters (PEQ) section now properly displays AutoEQ-generated filters
  • Added missing Phase Correction FIR parameters to reports
    • Now includes Phase Offset (degrees) and Kaiser β parameters
    • Provides complete documentation of phase correction settings
  • PDF report now includes biquad references for IIR filters
    • Added “Biquads” column to IIR filter tables in PDF reports
    • Displays biquad section numbers for each filter (e.g., “1” for single biquad, “1,2” for cascaded)
    • Matches the biquad numbering shown in TXT reports and appendices
    • Enables cross-referencing between filter definitions and biquad coefficient listings

[1.1.19] - 2025-11-20

Fixed

• Per-driver tap lengths toggle now correctly applies individual tap values to each driver
  • Fixed bug where enabling “Enable per-driver tap lengths” would still use the global tap count

[1.1.18] - 2025-11-20

Improved

• Phase correction algorithm enhancement
  • Upgraded from linear interpolation to PCHIP (Piecewise Cubic Hermite Interpolating Polynomial)
  • Provides smoother, more accurate phase interpolation during correction process
• Graph state management refactoring
  • Complete architectural cleanup of graph bounds and auto-bounds management
  • Each graph now maintains its own saved bounds and auto-bounds flags independently

[1.1.17] - 2025-11-19

Added

• New direct phase correction algorithm
  • Works directly with unwrapped phase instead of group delay integration
  • Removes linear phase component before smoothing for stable, predictable corrections
  • Better impulse symmetrization without arbitrary phase offsets
  • More reproducible results across different correction parameters
  • Band-limited correction with constant phase continuation outside the working band
  • Preserves phase coherence and avoids group delay artifacts at band edges
• Phase offset control for phase correction
  • Replaces binary polarity inversion with continuous phase offset slider
  • Range: -180° to +180°
  • Allows fine-tuning of phase alignment and acoustic polarity
  • Controls impulse response symmetry to adjust pre-ringing introduced by phase correction FIR
  • Applied after phase correction, before spectral synthesis
• Detached graph windows
  • Right-click any plot -> “Open in separate window” to spawn a native window
  • Multiple instances per plot type supported (e.g., several frequency plots)
  • Windows are resizable and can be closed with Cmd/Ctrl+W
  • Live rendering: detached plots update in real time and stay synchronized with the current project state
  • Display-mode capture: each window renders using the mode captured at detachment (FIR/IIR/FIR+IIR/Speakers) without altering the main window’s mode
• Automatic phase alignment in unwrapped mode
  • Phase curves are now automatically aligned when displaying unwrapped phase
  • Removes phse offset differences between drivers/filters for easier comparison
  • Works with or without time-centered phase enabled
  • Speakers mode: aligns all drivers to the Sum curve
  • FIR/IIR modes: aligns all filters to the first active filter
  • Uses robust median-based offset calculation on passband (middle 50% of data)
  • Eliminates random offsets caused by out-of-band noise and measurement artifacts
  • Makes crossover phase continuity immediately visible

Improved

• Smart directional truncation for group delay and phase plots
  • Graph truncation now respects the type of active filters (high-pass, low-pass, or both)
  • High-pass filters only: truncates low frequencies below threshold
  • Low-pass filters only: truncates high frequencies below threshold
  • Both filters active: truncates both sides as before
  • No LP/HP filters active: no truncation (e.g., when only using PEQ, shelves, or all-pass filters)
  • Applies to all contexts: FIR, IIR, FIR+IIR combined, speaker responses, and global filters
  • Provides cleaner visualization by showing only relevant frequency ranges

Fixed

• Undo/Redo no longer affects visualization controls (graph toggles, display mode, phase unwrapping, smoothing, etc.)
• FRD file import crash
  • Fixed panic when importing FRD files with excessive group delay
  • Added proper bounds checking for impulse response rotation to prevent array index overflow
  • Import now handles edge cases gracefully without crashing
• FRD phase extrapolation artifacts
  • Improved phase extrapolation beyond measured frequency range
  • Now uses average group delay to maintain constant GD in extrapolated regions
  • Prevents phase from diverging and creating negative group delay
  • Eliminates impulse response artifacts in high/low frequency extrapolation zones

[1.1.16] - 2025-11-15

Improved

• Interface optimization for small windows
  • Graph toolbar now adapts to window width
  • Below 1300px width, toolbar splits into two rows
  • First row: Display mode selectors (FIR/IIR/Drivers), graph toggles (Mag./IR/Step/G. Delay/Phase), THD
  • Second row: Normalize, Sum Focus, Sync X-axis, Unwrap Phase, Time-centered Phase, Smoothing

Fixed

• Phase control buttons behavior
  • “Unwrap Phase” and “Time-centered Phase” buttons now disabled (greyed out) instead of hidden when phase graph is not displayed
  • Improves UI consistency and predictability

[1.1.15] - 2025-11-14

Improved

• Interface optimization for small screens
  • Shortened button labels in graph toolbar for more compact display
  • Reduced driver parameter column width for better space utilization
  • Improved overall layout efficiency for laptop and smaller displays

[1.1.14] - 2025-11-10

Added

• Theme selector in Settings
  • Moved theme selection (Light/Dark/System) from “View” menu to Settings > General > Appearance
  • Now grouped with other appearance settings (UI Color Scheme)
  • More intuitive access to visual preferences in a single location
• Time-centered phase display mode
  • New “Time-centered Phase” toggle button in phase display controls (next to “Unwrap Phase”)
  • Removes the linear phase component (constant group delay) from all driver phase responses
  • Makes phase alignment between drivers easier to visualize and compare
  • Keyboard shortcut: C key
  • Available only in Speakers display mode
  • Applies correction to all visible curves while preserving phase differences
  • New default setting in Settings > General > Display: “Time-centered phase” dropdown (Enabled/Disabled)
  • Controls whether time-centered phase is enabled by default for new projects
• Customizable magnitude plot bounds
  • New “Magnitude Plot Bounds” section in Settings > General > Graphs
  • “Drivers range (dB)”: Controls Y-axis range for speaker magnitude plots
    • Default: 60 dB (from max_response - 60 to max_response + 5)
    • Adjustable from 20 dB to 200 dB
    • Allows fine-tuning the vertical zoom level for better resolution or overview
  • “Filters lower bound (dB)”: Controls lower Y-axis bound for filter magnitude plots
    • Default: -100 dB
    • Adjustable from -20 dB to -200 dB
    • Applies to FIR, IIR, and FIR+IIR filter magnitude plots
    • Useful for examining filter attenuation characteristics

Improved

• Project data structure refactoring
  • Consolidated all project-related data into a unified struct
  • Backward compatibility maintained through legacy migration system
  • Cleaner separation of concerns for future development
• Magnitude-threshold truncation replaces octave margins
  • Group Delay and Phase truncation now use a magnitude threshold in dB below the maximum instead of octave-based margins
  • New per-context settings: thresholds for Speakers, IIR, FIR, and FIR+IIR
  • Default threshold set to -50 dB (configurable in Settings)
  • UI updated: controls display values in dB below maximum

Removed

• “View” menu from menu bar
  • Theme selection moved to Settings > General > Appearance
  • Simplifies menu bar by consolidating appearance settings in one location

Fixed

• Phase display accuracy
  • Removed fractional octave smoothing from phase curves
  • Smoothing was introducing visual artifacts and local slope changes
  • Could bias visual assessment of driver phase alignment
  • Phase now displays raw calculated values without smoothing
  • Group delay continues to be smoothed as before (not affected by this change)
  • Results in more accurate phase representation for alignment analysis
• IIR and FIR+IIR phase/GD correctness
  • Group delay is now computed from unwrapped phase for IIR and combined FIR+IIR, eliminating artifacts at phase wrap boundaries
  • “Unwrap Phase” toggle now affects display only; underlying GD is always derived from unwrapped phase

[1.1.13] - 2025-11-07

Added

• Auto causal alignment setting
  • New “Auto causal alignment” option in Settings > General > Filter Processing
  • Automatically optimizes impulse response positioning when multiple filters with causality are active
  • Detects first significant sample (>1/10000 of peak) and aligns it with crop window start
  • Prevents pre-ringing truncation and maximizes useful signal capture
  • Activates when: ≥2 FIR filters active AND at least one has non-zero causality
  • Can be disabled for manual control over impulse positioning
  • Enabled by default

Improved

• FIR filter convolution stability
  • Improved impulse response alignment algorithm for causal filter combinations
• FIR correction gain application
  • Gain adjustment for FIR correction now applied before frequency band limitation in signal processing
  • Only corrects gain within the active correction band (f_min to f_max)
  • Preserves magnitude outside correction range, avoiding unintended global gain changes

[1.1.12] - 2025-11-06

Added

• Independent calculation modes for magnitude and phase FIR correction
  • Separate “Correction calculation” dropdowns for magnitude and phase correction
  • Each can independently choose pre-IIR or post-IIR signal calculation
  • Available for both per-driver and global FIR correction filters
• Pre-ringing analysis in impulse response plots
  • Automatic weighted RMS calculation of pre-ringing using Hilbert envelope
  • Perceptually-weighted metric using sigmoid function (masking near peak, penalty far from peak)
  • Displayed in hover tooltip with level in dB and qualitative risk assessment
  • Risk labels: “Likely masked” (≤-36 dB), “Low risk” (-36 to -24 dB), “Moderate risk” (-24 to -12 dB), “High risk” (>-12 dB)
  • Non-intrusive: only shown when hovering over impulse curves
• UI color scheme selection
  • New “UI Color Scheme” option in Settings > General > Appearance
  • Choose between “Red” (custom red theme) or “Blue” (default egui theme)
  • Red scheme: custom red accent colors for headers and active buttons
  • Blue scheme: standard egui colors with blue accents
  • Changes apply immediately without restart

Improved

• Correction filter UI reorganization
  • Removed pre/post-LP and HP calculation modes as it could cause issues
  • All FIR corrections are now computed without FIR LP or HP filters

[1.1.11] - 2025-11-05

Added

• X-axis synchronization for time-domain plots
  • New “Sync X-axis” toggle now also synchronizes impulse and step response plots
  • Double-click any synchronized plot to reset all time axes to full range
  • Independent Y-axis control per plot for flexible visualization
  • Auto-bounds system: plots automatically adjust to data range, disabled by user zoom/pan, re-enabled by double-click
• “All” button for driver management
  • Quickly enable all drivers with a single click
  • Complements existing “Solo” button for easy workflow: Solo -> test -> All -> compare
  • Located next to “Solo” button in each driver panel
• Peak time information in IR Information section
  • Displays temporal position of IR peak for manual window definition reference
• Maximum audio inputs and outputs increased to 256
  • Enables use of professional audio interfaces with numerous channels
• Measurement notes field for each driver
  • Add notes about hardware configuration, measurement setup, etc. (up to 1000 characters)
  • Accessible in the IR management window under “Measurement Notes”
• Default sweep parameters in settings
  • Configure default sweep duration, level, start/end frequency
  • New windows automatically use these defaults
• Graph display configuration in settings
  • New “Graphs” tab in Settings window
  • Reorder graphs with up/down (\u{2B06}\u{2B07}) buttons to customize display order
  • Configure which graphs are shown by default in new projects

Improved

• Y-axis alignment across all plots
  • Fixed Y-axis width ensures vertical alignment of plot areas
  • Better readability when comparing multiple graphs with different label lengths
• Clipping warnings now use a dedicated notification group
  • More reliable notification system with less visual clutter

Fixed

• Audio settings preservation across all project operations
  • All audio configuration (devices, sample rate, channels, calibration, measurement channels, timing reference, etc.) now persists when:
    • Loading existing projects
    • Creating new projects
    • Resetting to new project with defaults
  • AudioManager is properly reinitialized with user’s saved settings instead of reverting to defaults
  • Prevents accidental overwrite of carefully configured audio settings

[1.1.10] - 2025-11-02

Added

• THD truncation margin parameter for drivers in settings
• Adaptive impulse/step response display with configurable detection
  • Automatically detects signal boundaries at 1/1000th of peak amplitude
  • Configurable margins before (default: 5ms) and after (default: 10ms) signal
  • Can be disabled to show full impulse response at all times
• “Focus on Summed IR” button for impulse and step response displays
  • When enabled, all curves use the same time boundaries as the summed IR

Fixed

• Auto-save now forces save on application shutdown
  • Ensures pending changes are saved before closing
  • Resolves issue where rapid edits could be lost if closing within 1.5s of last change
  • Project state now consistently preserved across sessions
• Phase compensation now correctly accounts for driver IR delay compensation
  • Previously only affected group delay, now phase is also compensated
  • Ensures phase and group delay remain consistent when using IR delay compensation

[1.1.9] - 2025-11-01

Added

• FIR compensation delay parameter for global FIR correction filters
  • Fine-tune FIR alignment when filter ringing permits
  • Rotates the impulse response in time domain (±100 ms range)

Fixed

• Phase delay compensation now correctly applied to total response (Sum)
  • Affects display when “Compensate plot delay” is enabled in Global FIR settings
• “Open recent projects” menu now persists correctly across sessions
• IR delay compensation now resets to zero on every new measurement or import

[1.1.8] - 2025-10-31

Improved

• Phase and group delay calculations significantly improved when working with FRD files
  • More accurate phase extraction from frequency response data
  • Better group delay derivation from phase information
  • Improved unwrapping and numerical stability
• IIR and FIR+IIR graph tabs simplified to reduce visual clutter
  • Removed aggregated sum curves that were causing confusion
  • Individual driver responses remain available and accurate
  • Cleaner display focusing on relevant per-driver information

Added

• “Open recent projects” menu for quick access to previously used projects
  • Tracks up to 20 most recent project files
  • Provides faster workflow for switching between projects
• Measurement confirmation dialog before starting sweep
  • Displays safety warnings and reminders (mic position, gain levels, audible sweep)
  • Can be disabled in audio settings with toggle (“Skip confirmation”)
  • Helps prevent accidental measurements with improper setup

[1.1.7] - 2025-10-29

Added

• Added an input gain compensation setting in the audio preferences.
  • This allows adjusting the recording level directly in software
  • Useful for devices like the UMIK-1 on MacOS, where the system enforces a very low fixed gain and no hardware control is available.
• Added microphone calibration file import in audio preferences.
  • Supports standard calibration file format (text files with “Sens Factor” header and frequency/error columns)
  • Applies minimum-phase correction to impulse response measurements
  • Correction is baked into the captured impulse response, not just applied visually to magnitude plots

[1.1.6] - 2025-10-28

Fixed

• Wrong total phase with some combination of parameters

Improved

• IIR magnitude plot auto-bounds now use -100 dB lower limit when high-pass or low-pass filters are active
  • Matches FIR+IIR display behavior for better filter slope visualization
  • Dynamically adapts based on active filter types (HP/LP vs. PEQ/shelving)

[1.1.5] - 2025-10-27

Added

• HFD Export for Hypex FA Series Amplifiers
  • Direct export to Hypex FA amplifier configuration files (FA122, FA123, FA251, FA252, FA253, FA501, FA502, FA503)
  • Support for both Input FIR and Output FIR positions
  • Automatic FIR padding to required lengths (1500 taps for Output, 4500 taps for Input)
  • Per-driver IIR filter export (up to 15 biquads per channel)
  • Patch mode to update existing HFD configuration files
    • Automatically detects FIR position from existing files
    • Allows selective preset updates (Preset 1, 2, or 3)
  • Comprehensive validation:
    • Sample rate check (93.75 kHz required)
    • FIR tap length validation per position
    • Biquad count verification (max 15 per channel)
  • Driver-to-group mapping interface for multi-channel systems
• Configurable Group Delay & Phase Truncation
  • Optional truncation of GD and phase curves to active frequency range
  • Per-filter-type margin settings (Speakers, IIR, FIR, FIR+IIR)
  • Margin values configurable in octaves (0.0-5.0)
  • Reduces visual clutter by hiding curves outside useful frequency range

Fixed

• Fixed FRD file import phase reconstruction
  • Automatic compensation for negative group delay to prevent impulse wrapping
  • Imported impulses may appear time-shifted based on the group delay characteristics of the FRD data
• Global FIR delay compensation on graphs inactive after restarting the app

[1.1.4] - 2025-10-23

Fixed

• Fixed keyboard shortcuts not working in driver IR management windows
  • Undo/Redo (Cmd/Ctrl+Z, Shift+Cmd/Ctrl+Z) now work in all IR windows
  • Save (Cmd/Ctrl+S) and Export (Cmd/Ctrl+E) shortcuts now available
  • Settings shortcut (Cmd/Ctrl+,) now functional in IR windows
• Fixed impulse export dialog behavior
  • “Include distortion” option now unchecked by default (previously checked)
  • TXT export button now enabled by default (previously disabled)
  • TXT export now available even when distortion metadata is not present
• Fixed impulse clipping during WAV export
  • WAV export now automatically normalizes impulses with amplitude exceeding ±1.0
  • Prevents clipping while preserving the impulse shape
  • Particularly important for FRD-imported impulses which may have unnormalized amplitudes
• Fixed undo/redo creating multiple snapshots when dragging numeric values
  • Drag operations (DragValue widgets) now create a single snapshot on release
  • Works across all windows (main window and all viewport windows)
  • Single Cmd+Z now correctly reverts an entire drag operation to its initial state
  • Prevents undo history pollution from intermediate drag values

Improved

• Improved impulse response display layout in driver IR management window
  • Long filenames are now truncated to 30 characters with ellipsis to prevent overflow
  • Duration and sample rate information moved to a separate line for better readability
  • Prevents UI elements from extending beyond window boundaries
• Improved IR Delay Compensation System
  • IR delay compensation no longer truncates impulse responses
  • Impulses are now preserved intact in memory with delay applied as display offset
  • Delay compensation now affects:
    • Time axis display (visual shift without data loss)
    • Group delay calculations (compensated by subtracting delay)
  • Prevents magnitude and phase artifacts from truncation
  • Maintains full impulse information including pre-ringing

[1.1.3] - 2025-10-21

Added

• Option to disable clipping warnings in General Settings
• EQ Boost Cap parameter in AutoEQ (both driver and global IIR)
  • Prevents AutoEQ from stacking multiple filters at the same frequency
  • Configurable from 0 to 20 dB
  • Set lower values (e.g., 10 dB) for more conservative corrections
  • Helps achieve more natural-sounding frequency responses
• Gain Offset parameter in AutoEQ (both driver and global IIR)
  • Adjusts the target level before EQ optimization
  • Range: -20 to +20 dB (default: 0 dB)
  • Useful for compensating overall system level shifts
  • Applied to target curve before computing corrections

Fixed

• FIR+IIR impulse responses no longer truncated, ensuring accurate plots
• Clipping warnings now display correctly for global IIR filters
• FIR+IIR frequency response curves now render properly with active global IIR filters
• Keyboard shortcuts now work correctly in all native windows (Settings, Documentation, Filter windows, etc.)

Improved

• “Compensate FIR delay on graphs” option now preserves pre-ringing instead of truncating it, preventing magnitude and phase artifacts
• Magnitude response graph toggle changed from ‘R’ to ‘M’ key for better ergonomics
• AutoEQ parameter layout reorganized into 3 columns for better space efficiency
• All AutoEQ parameters now use consistent 6-column grid layout (3 visual parameter pairs)
• Parameter grouping: Target/Resolution/Gain Offset, Gain limits/EQ Boost Cap, Q limits/Max Filters, Frequency limits/Shelves

[1.1.2] - 2025-10-21

Added

• Added “Compensate FIR delay on graphs” toggle in Global FIR Correction window
  • Optionally shifts impulse response left to remove visual delay introduced by the global FIR filter
  • Does not affect exported filters, only improves plot alignment
  • Useful for comparing time-aligned driver responses in the presence of global FIR correction
• Added Reset Application Settings feature in Settings
  • Completely resets all application settings and preferences to factory defaults
  • Clears application cache and UI state
  • Accessible via Settings > General Settings > Reset section
  • Useful for troubleshooting or starting fresh with default configuration

Fixed

• Fixed phase response not reflecting polarity inversion
  • Phase now correctly shifts by 180° when driver polarity is inverted
  • Phase correction filters now properly display 180° phase shift when their polarity is inverted
• Fixed global IIR and FIR+IIR graphs not clearing when using “Clear Filters” button
  • IIR response curves now properly reset to flat when all filters are removed
  • Combined FIR+IIR graphs now correctly update when IIR filters are cleared
• Fixed manually added filters in AutoEQ tab not being applied to driver response before running optimization
  • AutoEQ filters are now automatically enabled when adding the first filter
  • Locked filters (e.g., high-pass or low-pass) are now properly integrated into the AutoEQ target curve instead of being compensated by the optimizer
• Fixed AutoEQ tab requiring manual IIR toggle to be enabled (AutoEQ and Manual IIR tabs are now fully independent)
• Fixed IIR filter type changes not triggering UI updates
  • Changing filter types (e.g., LR low-pass -> high-shelf) now properly invalidates the filter cache
  • Filter response is immediately recalculated and displayed when switching types

Improved

• Reorganized documentation to enhance readability and better align with the app’s workflow
• All windows now use the native OS window system, improving multi-monitor usability
• Lockable filters now available in the manual tab
• Locked filters now stay in place when clearing the filter list
• AutoEQ now respects locked filters by incorporating their response into the optimization target
  • Only high-pass and low-pass filters (Butterworth, Linkwitz-Riley, Bessel) are included in the target curve
  • PEQ, shelf, and allpass locked filters are preserved but not integrated into the optimization target
  • Example: Setting a high-pass or low-pass filter will preserve that behavior instead of trying to correct it
  • Locked filters’ frequency response is calculated and multiplied with the target curve for optimization

[1.1.1] - 2025-10-19

Added

• Preferences for default values.
• Configurable sweep frequency range: Users can now set custom lower and upper limits for measurement sweeps.
  • Default: 0.1 Hz to Nyquist (full bandwidth).
  • Manual controls: start frequency adjustable from 0.1 Hz to 20 kHz; end frequency up to Nyquist.
  • “Reset Range” button to quickly return to full bandwidth.
  • Automatic Nyquist limiting based on current sample rate.
  • Typical use cases:
    • Subwoofers: 20 Hz-200 Hz
    • Tweeters: 2 kHz-20 kHz
    • Full-range: 20 Hz-20 kHz
• Frequency values are automatically rounded to maintain optimal phase coherence whenever possible.

Improved

• FIR delay now shown directly in graph legends for a cleaner UI.
• Maximum of 10 speakers supported for complex setups.
• Improved layout of settings.
• Labels refined for a clearer interface.

[1.1.0] - 2025-10-18

Added

• Added “Save” menu item in File menu for easier project saving
• Added adaptive display window for impulse and step responses with long filters (>10000 taps)
• Added progress toast with progress bar during IR measurements showing estimated completion time

Fixed

• Fixed window positions jumping when renaming impulse responses
• Fixed filter configuration windows (LP, HP, Correction, IIR) disappearing when changing IR names
• Fixed collapsing headers closing unexpectedly when renaming drivers
• Fixed autosave not triggering when changing IR names
• Fixed crash when n_taps exceeds FFT size in impulse correction filter generation

Improved

• Simplified impulse response management window interface
• Impulse and step response plots now automatically extend display window for very long filters based on peak position
• Enhanced toast notifications with shadow effects for better visibility

[1.0.9] - 2025-10-16

Added

• Added lockable IIR filters in AutoEQ mode

• Added timing reference via loopback (Electric) or pre-chirp reference (Acoustic)

• Added phase graph with wrapped/unwrapped display

• Added adjustable delay compensation for raw impulse response (in IR Window management)

• IIR filter export functionality in File -> Export menu- FQG format (Frequency, Q, Gain): Simple text format for filter parameters

  • CamillaDSP format: YAML configuration for direct import into CamillaDSP
  • Exports global IIR filters and per-speaker IIR filters separately
  • Supports all filter types: Butterworth, Linkwitz-Riley, Bessel, PEQ, Shelving, Allpass
  • Only active filters are exported
  • Respects Manual/Auto-EQ tab selection

[1.0.8] - 2025-10-15

[1.0.7] - 2025-10-13

Added

• Added FRD and txt impulses support

Improved

• Improved latency compensation for sweep measurements
• Improved sweep error messages with notifications
• Improved audio support for Windows

Overview

LinFIR is a professional audio software tool designed for precise FIR (Finite Impulse Response) and IIR (Infinite Impulse Reponse) filter design, tailored specifically for loudspeaker crossovers and frequency response correction. The application provides comprehensive control over both magnitude and phase response characteristics, enabling engineers and enthusiasts to craft high-quality audio filter solutions.

Design Philosophy

LinFIR is designed to be both a filter creation tool and a DSP-faithful system simulator. It shows what the filters will actually do when applied to your measurements, not an idealized version.

What You See Is What You Get: LinFIR operates with real DSP constraints—impulse timing, causality, delay, phase rotation, and resampling effects all matter. The application does not silently recenter impulses or adjust phase for prettier graphs. What is displayed is what a real DSP would produce.

This approach requires some understanding of signal processing fundamentals, but LinFIR automates most technical decisions under the hood to keep the workflow practical. The trade-off is that results may sometimes look less “clean” than in tools that abstract away these constraints, but they are physically accurate and predictable.

Different by Design: LinFIR does not replicate the workflows or interface conventions of other tools. The internal DSP-like architecture drives both the workflow and presentation. This is intentional, the goal is to bring a time-domain perspective into DIY loudspeaker and room-correction work, not to replace existing approaches.

Understanding what LinFIR shows and why helps avoid confusion and makes the tool more effective for its intended purpose.

Dual Operating Modes

LinFIR offers two distinct modes of operation, each optimized for different workflows:

Loudspeaker Design Mode

This mode is dedicated to designing and optimizing multi-way loudspeaker systems. Key applications include:

• Crossover design with multiple filter types
• Individual driver frequency response correction
• Phase and time alignment between drivers
• Directivity pattern analysis and optimization
• Complete system integration and tuning

Room Calibration Mode

Focused on in-room acoustic measurements and correction:

• Multiple measurement position capture and averaging
• Automatic temporal alignment using GCC-PHAT algorithm
• Room response correction filter generation
• Spatially-averaged frequency response optimization
• Integration with existing room correction workflows

The mode is selected when creating a new project and remains fixed throughout the project’s lifetime. To switch modes, simply create a new project with the desired configuration.

Core Features

FIR Filter Design

LinFIR implements sophisticated FIR filtering with:

• Configurable filter length: From 32 to 65536 taps for precise frequency resolution
• Causality control: Continuous adjustment from pure linear-phase (0.0) to minimum-phase (1.0)
• Multiple filter types: Brickwall (Sinc), Linkwitz-Riley, Butterworth, and Bessel characteristics
• Kaiser window shaping: Fine-tune transition band characteristics

Frequency Response Correction

Advanced magnitude and phase correction capabilities:

• Target curve options: Flat response, Harman in-room curve, or custom user-defined targets
• Independent correction modes: Separate control for magnitude and phase correction
• Frequency range limiting: Apply corrections only where needed
• Maximum attenuation control: Prevent over-correction of deep nulls

IIR Filtering

Complement FIR filters with cascaded IIR sections:

• Parametric EQ: Precise peak/notch filters for resonance control
• Shelving filters: Low-shelf and high-shelf for tonal balance
• Crossover filters: Butterworth, Linkwitz-Riley, and Bessel topologies
• All-pass filters: Dedicated phase shaping without magnitude changes
• Auto-EQ: Automatic optimization to target curves with configurable parameters

Measurement System

Built-in exponential sine sweep (ESS, Farina method) for high-quality impulse response capture:

• Harmonic distortion separation: Automatic extraction of H2-H4 through deterministic time positioning
• Driver protection: 4th-order Butterworth high-pass filter with configurable cutoff (-3 dB at fc, -24 dB/octave)
• Configurable sweep parameters: Duration, amplitude, frequency range (starting from 1 Hz)
• Quality validation: Automatic rejection of poor captures (clipping, low SNR)
• Microphone calibration: Import manufacturer calibration files for accurate measurements

Workflow Integration

LinFIR seamlessly integrates into professional audio workflows:

• Import/export WAV, txt, and specialized formats
• Direct export to Powersoft Armonia processors
• Export FIR coefficients in binary, text, WAV or CSV format
• Hypex FA Series amplifier configuration (HFD format)
• CamillaDSP filter export
• Comprehensive project management with auto-save

Platform Support

• Mac OS: Full feature support with low-latency audio
• Windows: Complete functionality with platform-specific audio optimizations
• Cross-platform projects: Projects are fully compatible across platforms

Project Modes

LinFIR offers three specialized operating modes, each designed for specific audio engineering workflows. The mode is selected during project creation and determines the available features and interface elements throughout the project’s lifetime.

Note on Hypex FusionAmp mode: The Hypex FusionAmp mode is a derivative of Loudspeaker Design mode. Everything documented for Loudspeaker Design — driver processing, crossover design, FIR/IIR filtering, measurement import, directivity analysis — applies equally to Hypex FusionAmp projects. The mode adds hardware-specific UI constraints (locked sample rate, fixed channel count, FIR tap limits, IIR biquad limit) that prevent invalid configurations and eliminate the need to manually track hardware limits during the design process.

Loudspeaker Design Mode

Purpose

This mode is optimized for designing and analyzing multi-way loudspeaker systems. It provides comprehensive tools for crossover design, driver integration, and directivity analysis.

Key Features

• Multi-driver support: Design systems with up to 50 drivers (subwoofers, woofers, midranges, tweeters)
• Crossover: FIR and IIR crossover filters with various types
• Driver correction: Individual magnitude and phase correction for each driver
• Directivity tools: Analyze off-axis response and polar patterns (license required)
• Complete export: Export individual driver filters, global filters, and HFD configurations

Typical Workflow

1. Create a new Loudspeaker Design project
2. Import or measure impulse responses for each driver
3. Design crossover filters (low-pass, high-pass)
4. Apply frequency response corrections
5. Analyze summed system response
6. Export filters for DSP implementation

When to Use

• Designing passive loudspeaker conversions to active DSP
• Optimizing existing multi-way systems
• Analyzing driver interactions and phase relationships
• Creating custom crossover
• Performing anechoic or quasi-anechoic measurements
• Aligning a complete sound reinforcement system (see below)

Sound Reinforcement System Alignment

The 50-driver capacity is not only meant for multi-way cabinets. It also covers sound reinforcement system alignment, where a driver slot represents a cabinet, an array section, or a zone rather than a transducer:

• Line array: each element, or each group of elements, as its own driver
• Subwoofers: cardioid or end-fire arrays, left/right stacks, sub zones
• Front fills / out fills / delays: one driver per fill position

Working this way, the same tools used for crossover design apply to system alignment: relative gain, delay and polarity per element, per-zone FIR or IIR correction, and the summed response of the complete system.

⚠️ Performance: with many drivers, the processing load scales with the drivers count and filter length. Long FIR filters (>8192 taps) applied across dozens of drivers make interactive adjustments sluggish — keep taps moderate during alignment.

This kind of project normally uses on-axis measurements only, so that’s the whole story. The one thing to avoid, even though it’s not something you’d run into by accident, is capturing full polar measurement sets across many drivers at once — that multiplies the cost rather than adding to it and can slow the UI down noticeably. See Large Multi-Driver Projects for details (not a concern in Room Calibration mode, where polar measurements aren’t available).

Room Calibration Mode

Purpose

Dedicated to in-room acoustic measurements and correction filter generation. This mode focuses on capturing multiple measurement positions, aligning them temporally, and creating averaged correction filters.

Key Features

• Multiple measurement positions: Capture IRs at different listening locations (up to 50 positions); reorder them by dragging their header (☰ grip or name) or with the ⬆ / ⬇ buttons on each position
• Automatic temporal alignment: GCC-PHAT algorithm aligns measurements
• Spatial averaging: Creates averaged response across measurement positions
• Global correction only: Simplified interface focused on room correction
• Streamlined export: Export only global correction filters

Typical Workflow

1. Create a new Room Calibration project
2. Configure sweep output channel
3. Capture measurements at 3-5 different listening positions
4. Measurements are automatically aligned using GCC-PHAT
5. Design global correction filters (FIR and/or IIR)
6. Export correction filters for room EQ implementation

UI Adaptations

When in Room Calibration mode, the interface adapts to focus on relevant features:

• Disabled: Directivity analysis tools (not applicable to room measurements)
• Simplified: Filter graphs show only global filters
• Restricted: Export options limited to global correction filters
• Hidden: Individual driver processing controls

Export Restrictions

Room Calibration projects export only:

• Global FIR correction filter (if enabled)
• Global IIR filters (Manual or Auto-EQ)

The following exports are disabled:

• HFD config export
• Detailed reports (TXT/PDF)
• Individual driver filters

This ensures clean, focused output for room correction workflows.

When to Use

• Correcting in-room frequency response
• Creating stereo-linked or mono room correction
• Working with existing loudspeaker systems
• Integrating with convolution engines

Hypex FusionAmp Mode

⚠️ License required: Creating Hypex FusionAmp projects requires a valid LinFIR license. See License for activation details.

Purpose

Hypex FusionAmp mode is a derivative of Loudspeaker Design mode tailored specifically for Hypex FusionAmp series amplifiers (FA122, FA123, FA251, FA252, FA253, FA501, FA502, FA503). All Loudspeaker Design features are available — driver processing, crossover design, FIR/IIR filtering, measurement import, directivity analysis — but several parameters are locked to match the DSP capabilities of the target hardware.

The goal is to eliminate manual bookkeeping: instead of counting biquads, tracking tap budgets, or checking compatibility after the fact, the UI enforces hardware limits in real time so the resulting configuration is always valid and ready to export.

Available Models

LinFIR supports all FusionAmp models:

• FA122: 2-channel amplifier (2 × 125W @ 4Ω)
• FA123: 3-channel amplifier (2 × 125W + 100W @ 4Ω)
• FA251: 1-channel amplifier (1 × 250W @ 4Ω)
• FA252: 2-channel amplifier (2 × 250W @ 4Ω)
• FA253: 2-channel amplifier (2 × 250W + 100W @ 4Ω)
• FA501: 1-channel amplifier (1 × 500W @ 4Ω)
• FA502: 2-channel amplifier (2 × 500W @ 4Ω)
• FA503: 3-channel amplifier (2 × 500W + 100W @ 4Ω)

Hardware Constraints

When in Hypex FusionAmp mode, several parameters are locked to match hardware specifications:

• Sample Rate: Fixed at 93.75 kHz (cannot be changed)
• Channel Count: Fixed by model (1, 2 or 3 channels, cannot add/remove drivers)
• IIR Filters: Maximum 15 biquads per channel
• FIR Filters: Fixed total tap count depends on FIR position (see below)

FIR Processing Modes

FusionAmp mode offers two mutually exclusive FIR processing configurations:

FIR IN (Input FIR)

Global FIR correction applied at the DSP input stage before channel processing.

• Location: Before IIR filters and channel routing
• Total taps: Fixed at 4500 (filter taps + padding)
• Use case: Global room correction, global speaker compensation
• Constraint: FIR Taps + Export padding = 4500 (always)
• UI behavior: Per-driver FIR controls are hidden; adjusting taps automatically adjusts padding to maintain 4500 total

FIR OUT (Output FIR)

FIR correction applied at the output stage, after IIR processing.

• Location: After IIR filters, per-channel or global
• Total taps: Fixed at 1500 per channel (filter taps + padding)
• Use case: Individual driver correction, per-channel equalization
• Constraint: Filter length (taps) + Export padding = 1500 (always, per driver or global)
• UI behavior: Adjusting taps automatically adjusts padding to maintain 1500 total

IIR Filter Constraints

FusionAmp DSP limits each channel to 15 biquads maximum. LinFIR enforces this through:

• Add Filter button: Automatically disabled when at 15 biquads
• Filter type restrictions: Types exceeding the limit are grayed out with tooltips
• Order limitations: Filter order sliders dynamically limited based on remaining capacity
• Auto EQ constraints: “Max Filters” parameter accounts for locked filters’ biquad usage

Note: For details on how biquad counts are calculated for different filter types, see the IIR Filtering section.

Key Features

• Hardware-matched UI: Interface adapts to show only applicable controls
• Automatic validation: Prevents configurations exceeding hardware limits
• Export compatibility: Direct HFD export for FusionAmp amplifiers
• Constraint warnings: Visual indicators when approaching limits

Typical Workflow

1. Create new project: File → New Project → Hypex FusionAmp
2. Select model (FA122, FA123, FA251, FA252, FA253, FA501, FA502, FA503)
3. Choose FIR position (Input or Output)
4. Import/measure impulse responses for each channel
5. Design IIR filters (monitor biquad counter to stay within 15 biquad limit)
6. Configure FIR correction (taps + padding always equals 4500 for IN or 1500 for OUT)
7. Export HFD configuration file for amplifier

UI Adaptations

The interface automatically adjusts based on FIR position:

FIR IN mode:

• Global FIR section visible and active
• Per-driver FIR controls hidden
• Taps and padding controls linked to maintain 4500 total
• Adjusting taps automatically recalculates padding, and vice versa

FIR OUT mode with per-driver FIR:

• Per-driver FIR controls visible
• Each driver has linked taps/padding controls maintaining 1500 total
• Global FIR section hidden or disabled

FIR OUT mode without per-driver FIR:

• Taps and padding controls linked to maintain 1500 total
• Per-driver FIR controls hidden

Export Configuration

Hypex FusionAmp projects support:

• HFD export: Native configuration format for FusionAmp amplifiers
• FIR filter export: Individual channel FIR filters
• IIR filter export: Biquad coefficients per channel

When to Use

• Configuring Hypex FusionAmp series amplifiers
• Ensuring DSP configuration fits hardware constraints
• Exporting ready-to-use HFD configuration files
• Working within strict real-time processing limits

Limitations

⚠️ Hardware constraints cannot be bypassed:

• Sample rate is locked at 93.75 kHz
• Channel count is fixed by model (1, 2, or 3 channels)
• FIR total tap count is fixed (4500 for IN, 1500 for OUT) - taps and padding sum must always equal this value
• IIR biquad limit (15 per channel) is strictly enforced
• Cannot use both FIR IN and FIR OUT simultaneously

Mode Selection

Creating a New Project

Project mode is selected via File → New Project:

1. Click “New Project”
2. Choose between “Loudspeaker Design”, “Room Calibration”, or “Hypex FusionAmp”
3. For Hypex FusionAmp: Select model (FA122/FA123/FA251/FA252/FA253/FA501/FA502/FA503) and FIR position (IN/OUT)
4. For other modes: Configure initial project settings (sample rate, filter length, etc.)
5. Begin working in the selected mode

Mode Permanence

⚠️ Important: Once a project is created, its mode cannot be changed. The mode is permanently associated with the project file.

To work in a different mode:

1. Save your current project (if needed)
2. Create a new project with the desired mode
3. Import measurements or data as required

Choosing the Right Mode

Use Loudspeaker Design Mode when:

• Designing crossovers for multi-driver systems
• Analyzing individual driver characteristics
• Performing directivity analysis
• Working with anechoic or quasi-anechoic data
• Aligning a sound reinforcement system (line array + subs + front fills)
• Need flexible configuration options

Use Room Calibration Mode when:

• Correcting in-room frequency response
• Creating averaged room correction filters
• Working with existing complete loudspeaker systems
• Focusing on global system correction only

Use Hypex FusionAmp Mode when: (license required)

• Configuring Hypex FusionAmp series amplifiers
• Exporting HFD configuration files
• Need to ensure configurations match hardware limits (93.75 kHz, fixed taps, 15 biquads)

Best Practices

Room Calibration Mode

• Measurement count: Capture 3-5 measurements at different positions (up to 50 are supported, but more positions rarely improve a domestic correction)
• Position spacing: Keep positions within 30-50 cm of main listening area
• Height consistency: Use consistent microphone height across measurements
• Reference position: First measurement should be at primary listening position
• Correction philosophy: Apply gentle correction, avoid over-equalization
• Deep nulls: Don’t attempt to fill room mode nulls below 300 Hz
• Phase type: Consider minimum-phase FIR for reduced latency
• Acoustic treatment: Combine with room treatment for best results

Loudspeaker Design Mode

• Measurement quality: Use anechoic or quasi-anechoic measurements when possible
• Windowing: Gate reflections using IR time windowing
• Alignment: Align drivers using time delay controls, not FIR compensation delay
• Crossover design: Start with appropriate crossover frequencies and filter slopes
• Phase analysis: Monitor phase relationships between drivers
• Directivity: Capture multiple angles for comprehensive analysis (license required)

Sound Reinforcement Systems (Loudspeaker Design Mode)

• Slot mapping: Decide upfront what a driver slot represents (cabinet, array section, zone) and name the drivers accordingly
• Filter length: Stay at or below 4096 taps during alignment; raise it only at the end if low-frequency resolution requires it
• Polar measurements: not part of a typical alignment workflow (on-axis is normally enough); if you do capture them, restrict full polar sets to a few representative cabinets rather than every driver
• Section by section: Use Solo / Enable-Disable to align subs, array and fills separately before evaluating the full sum
• Copy / Paste parameters: Use the 📋 / 📥 buttons to replicate identical settings across identical cabinets instead of configuring each one manually

Mode Comparison Table

Hypex FusionAmp is a superset of Loudspeaker Design: all ✅ features from Loudspeaker Design are available, with the hardware constraints in the last rows added on top.

Graph Interaction & Display

This section covers all interactive features, display controls, and analysis tools available when working with LinFIR’s graphs.

Graph Toolbar

The graph toolbar provides comprehensive controls for selecting which data to display and how to visualize it. The toolbar adapts based on the project mode (Loudspeaker Design vs Room Calibration).

Loudspeaker Design Mode:

Room Calibration Mode:

Display Mode Selection

Purpose: Choose which stage of the signal processing chain to visualize.

Available modes:

• FIR - FIR filter responses only (crossovers, magnitude/phase correction)
• IIR - IIR filter responses only (parametric EQ, shelves, crossovers)
• FIR+IIR - Combined FIR and IIR responses
• Drivers (Loudspeaker Design) / Measurements (Room Calibration) - Full signal chain applied to driver or room measurements (theoretical system response)

Keyboard shortcuts:

• Press F to cycle through filter modes (FIR → IIR → FIR+IIR)
• Press D to switch to Drivers/Measurements mode

Mode behavior:

• Loudspeaker Design mode: Shows individual driver filters plus global filters
• Room Calibration mode: Shows only global filters in FIR/IIR/FIR+IIR modes

See Display Modes section below for detailed explanations of each mode.

Graph Visibility Toggles — “Graphs” Dropdown

Purpose: Show or hide specific graph types and magnitude overlay curves.

The Graphs dropdown in the toolbar has two sections, separated by a divider.

Main plots - which graph panels are shown:

• Magnitude (M) - Frequency response in dB
• Impulse Response (I) - Impulse response in time domain
• Step Response (T) - Step response in time domain
• Group Delay (G) - Group delay in milliseconds
• Phase (P) - Phase response in degrees
• Harmonic Distortion (K) - (Loudspeaker Design / Hypex mode only)

Letters in parentheses indicate keyboard shortcuts. The dropdown stays open when toggling items, allowing multiple graphs to be shown or hidden without reopening it.

Note: Harmonic Distortion is only available in Loudspeaker Design / Hypex mode with Drivers display mode active, as harmonic distortion analysis is not applicable to in-room measurements.

License Feature: The HD graph displays the total harmonic distortion for all users. Individual harmonic curves (\(h_2\), \(h_3\), \(h_4\), \(h_5\)) are only available with a valid license.

Overlays - which curves are drawn on the frequency graph itself:

• Sum / Average - the summed (Loudspeaker Design) or averaged (Room Calibration) response; labelled to match the mode. Available in both project modes, no license required.
• Listening Window - (Loudspeaker Design mode only, license required) the ±30°H / ±10°V spatial average, shown when off-axis measurements exist.
• Predicted In-Room Response - (Loudspeaker Design mode only, license required) the CEA-2034-A curve, shown when there’s enough Early Reflections angle coverage - see Directivity Analysis.
• Directivity Index - (Loudspeaker Design mode only, license required) shown when off-axis measurements exist.

Each project remembers its own choice for these four toggles. Settings → Graphs → Default Overlays sets what a new project starts with (see Graphs Settings).

Angle Selector (Loudspeaker Design / Hypex mode only)

Purpose: Select which off-axis measurement angle to display when drivers have directivity data.

• Horizontal angles: Typically ±15°, ±30°, ±45°, ±60°, ±75°, ±90°
• Vertical angles: Same as horizontal
• Toggle H/V: Switch between horizontal and vertical axis selection

Only available when at least one driver has off-axis measurements. The selector is hidden in Room Calibration mode as directivity analysis is specific to anechoic/quasi-anechoic loudspeaker measurements.

See Directivity Analysis for complete details on polar measurements and analysis.

Time Domain Options — “Time Domain” Dropdown

The Time Domain dropdown groups options that affect impulse and step response display.

Normalize

Purpose: Scale impulse and step responses to ±1 range for easier comparison.

When enabled:

• Impulse responses normalized to peak amplitude = 1
• Step responses normalized to maximum value = 1
• Useful for comparing relative timing and shape without amplitude differences

Focus on Summed Response

Purpose: Use the time boundaries of the summed impulse response for all IR and step response plots.

• Loudspeaker Design: All time-domain plots are framed around the summed driver output
• Individual curves remain visible within those boundaries
• Useful for evaluating overall system timing alignment

Phase Options — “Phase Options” Dropdown

The Phase Options dropdown is enabled only when the Phase graph is visible. It contains two options:

Unwrap Phase

Purpose: Display continuous phase response without ±180° wrapping discontinuities.

• Disabled: Phase wraps at ±180° (sawtooth pattern)
• Enabled: Phase continues beyond ±180° showing true accumulated phase shift
• Essential for analyzing linear phase filters and group delay
• Makes phase response easier to interpret across wide frequency ranges

Remove time of flight rotations

What is a time of flight? A pure propagation delay — the time it takes for sound to travel from the driver to the measurement microphone — appears in the phase response as a constant linear slope: the further the impulse response peak is from t = 0, the steeper the slope, and the more the phase accumulates rotations across the frequency range. This linear ramp carries no useful information about the filter or driver behaviour; it just reflects the physical distance between source and mic.

This option removes that linear component, leaving only the non-linear phase rotations introduced by the filters and the acoustics of each driver. The result is a much flatter, easier-to-read phase curve, and — when multiple drivers are shown — it becomes straightforward to compare their phase alignment without the bulk delay dominating the picture.

• Disabled: Phase includes the full linear delay component (steep slope proportional to time of flight)
• Enabled: Linear component subtracted, phase flattens around 0° — non-linear deviations stand out clearly

HD Display Mode (Loudspeaker Design mode only)

Purpose: Select how harmonic distortion values are displayed on the HD plot.

Available display modes:

• Percent (%) - Traditional percentage representation
• dB (relative) - Relative level compared to fundamental
• dB (absolute) - Absolute RMS level of harmonics

Percent (%) Mode:

Displays harmonic distortion as a percentage relative to the fundamental (\(h_1\)). Values above 100% are possible when harmonic components exceed the fundamental level (e.g., below the driver’s resonance frequency).

dB (relative) Mode:

Expresses harmonic distortion relative to the fundamental (\(h_1\)) in decibels. Negative values indicate harmonics are quieter than the fundamental; positive values indicate harmonics exceed the fundamental. For example:

• -40 dB ≈ 1% distortion
• -60 dB ≈ 0.1% distortion
• -80 dB ≈ 0.01% distortion
• 0 dB = 100% distortion (harmonics equal the fundamental)

Useful for evaluating signal-to-distortion ratio.

dB (absolute) Mode:

Shows the absolute level of the harmonics alone, without normalization. This mode is useful for comparing distortion levels across different SPL measurements, as it shows the actual acoustic power in harmonic components regardless of the fundamental level.

Individual Harmonics:

All three modes also display individual harmonic curves (\(h_2\), \(h_3\), \(h_4\), \(h_5\)) with dashed/dotted line styles. The selected mode applies to both the total HD curve and individual harmonics.

License Feature: Individual harmonic curves (\(h_2\)-\(h_5\)) are only available with a valid license.

Smoothing

Purpose: Apply smoothing to magnitude, HD, directivity, group delay and phase plots.

Smoothing options:

• None - No smoothing (raw response)
• 1/48 octave - Very fine smoothing
• 1/24 octave - Fine smoothing
• 1/12 octave - Moderate smoothing
• 1/6 octave - Coarse smoothing
• 1/3 octave - Very coarse smoothing
• ERB - Perceptually-motivated variable smoothing (see below)

Applies to:

• Magnitude (frequency response)
• Group delay
• Phase (preserves driver alignment via trend removal method)
• HD (Harmonic Distortion)
• Directivity Analysis (sonogram)

Does not affect: Time-domain plots (impulse, step response)

ERB Smoothing:

ERB (Equivalent Rectangular Bandwidth) smoothing uses a variable bandwidth that corresponds to the frequency resolution of the human auditory system. The smoothing bandwidth follows the formula: (107.77f + 24.673) Hz, where f is frequency in kHz.

This results in:

• Heavy smoothing at low frequencies (~1 octave at 50Hz, 1/2 octave at 100Hz, 1/3 octave at 200Hz)
• Moderate smoothing at mid frequencies (approximately 1/6 octave above 1kHz)
• Natural perceptual weighting that reflects how the ear integrates frequency information

ERB smoothing is particularly useful for:

• Understanding what a listener actually perceives
• Comparing measurements to subjective listening impressions
• Evaluating whether fine response details are audible

Note: Smoothing is cosmetic and does not affect filter calculations or exports. See Fractional Octave Smoothing section below for detailed behavior.

Mouse & Trackpad Controls

Zoom Controls

Zoom In - Multiple methods:

• Box zoom: Right-click and drag to select a region
• Trackpad pinch: Pinch gesture to zoom in/out
• Mouse wheel: Scroll to move the view within a graph

Zoom Out:

• Double-click anywhere on a graph to reset to auto-calculated bounds
• Double-click re-enables auto-bounds mode (graphs automatically adjust to data range)

Pan

Left-click and drag to move the view within a zoomed graph.

Legend Interaction

Legends list the curves in plotting order - the drivers in their order in the left column, then the sum and the other curves - read from top to bottom whatever corner the legend sits in.

Click on color dots in the legend to show/hide individual curves:

• Hidden curves are grayed out in the legend
• Click again to show the curve
• Useful for isolating specific drivers or filters

Reorder Graphs

Drag a graph’s title (the ☰ grip or the title text) and drop it above or below another graph to change the order of the graph stack:

• While dragging, the title follows the pointer and a line shows where the graph will land
• Releasing away from the graphs, or pressing Esc, cancels the move
• Only the graphs currently shown take part; hidden graphs keep their place in the order, so they reappear where they were when toggled back on
• The new order is the same one as Settings → Graphs → Graph Order (see Graph Order): it applies to every display mode and is kept for new projects
• Detached graph windows have a plain title and are not part of the stack

Detach Graphs

Right-click on any graph to open it in a separate window:

• All graph types supported (Frequency, Phase, Group Delay, Impulse, Step, HD)
• Detached windows render in real-time following the active display mode
• Multiple graphs can be opened simultaneously for side-by-side comparison
• Ideal for multi-monitor setups and detailed analysis workflows
• Close detached windows: Cmd+W (macOS) / Ctrl+W (Windows)

Display Modes

Project Mode Impact

LinFIR supports two project modes that affect display terminology:

• Loudspeaker Design - Speaker design with drivers
• Room Calibration - Multi-position measurements for room correction

In Room Calibration mode:

• “Drivers” mode is renamed to “Measurements”
• Individual measurement curves are shown in Measurements mode
• Filter modes (FIR/IIR/FIR+IIR) display only global filters

In Loudspeaker Design mode:

• “Drivers” mode displays individual drivers
• Filter modes show both per-driver filters and global filters

Filters vs. Drivers/Measurements Mode

Filters Mode (Keyboard: F):

• Shows individual filter responses (FIR, IIR, or FIR+IIR combined)
• Press F multiple times to cycle through filter submodes:
  • FIR Filters - Shows only FIR filter responses
  • IIR Filters - Shows only IIR filter responses
  • FIR+IIR Filters - Shows combined FIR and IIR responses
• Loudspeaker Design: Per-driver crossover and correction filters displayed separately
• Room Calibration: Only global filters are shown
• Global filters shown as additional curves
• Useful for analyzing filter design and frequency response shaping

Drivers/Measurements Mode (Keyboard: D):

• Loudspeaker Design: Shows combined driver + filter responses (acoustic output)
  • Each curve represents the complete signal chain for that driver
  • Sum curve shows the total system response at listening position
• Room Calibration: Shows individual measurement positions and averaged response
  • Each curve represents a raw measurement location (no filters applied)
  • Sum curve shows the averaged response with global filters applied
• Useful for analyzing final acoustic performance

Graph Visibility Toggles

Show/hide individual graph types using keyboard shortcuts:

• M - Toggle Magnitude (Frequency Response) plot
• P - Toggle Phase Response plot
• G - Toggle Group Delay plot
• I - Toggle Impulse Response plot
• T - Toggle Step Response plot
• K - Toggle HD (Harmonic Distortion) plot

X-Axis Synchronization

Toggle: X key or toolbar button

When enabled, Sync X-axis synchronizes the horizontal axis across related plots:

Frequency Plots (Magnitude, Phase, Group Delay, Directivity Analysis):

• All plots share the same frequency range
• Zooming any plot updates all frequency plots simultaneously
• Double-click resets all frequency plots to full range — set by Settings → Graphs → Frequency Axis (default 10 Hz – 24 kHz)

Time Plots (Impulse, Step Response):

• All time-domain plots share the same time range
• Zooming any plot updates all time plots simultaneously
• Double-click resets all time plots to full range

Independent Y-Axes:

• Each plot maintains its own vertical scale
• Y-axis auto-bounds independently for each graph type

Use Cases:

• Compare magnitude and phase behavior at the same frequency
• Analyze impulse and step response in the same time window
• Maintain consistent zoom across multiple graph types

Impulse & Step Response Normalization

Toggle: N key or toolbar button

Normalizes impulse and step responses to peak amplitude = 1.0:

Purpose:

• Allows visual comparison between responses with different amplitudes
• Focuses on filter shape and time-domain characteristics
• Eliminates gain differences for easier visual analysis

Effect:

• All IR and SR curves are scaled to the same peak height
• Does not affect magnitude or phase plots
• Purely visual - does not modify underlying data

Use Cases:

• Compare filter shapes between drivers with different sensitivity
• Analyze time-domain behavior without gain differences obscuring details
• Identify pre-ringing or timing issues across multiple drivers

Focus on Summed Response

Toggle: S key or Time Domain dropdown (Drivers/Measurements mode only)

Synchronizes the time window of all temporal plots (Impulse & Step Response) to the Sum/Average curve boundaries:

How it Works:

1. Analyzes the Sum/Average impulse response to find its energy boundaries
2. Detects first and last samples above -60 dB threshold (1/1000th of peak)
3. Adds configurable margins before and after the detected energy region
4. Applies the same time window to all temporal plots (both Impulse and Step Response)
5. Enables auto-bounds for all time plots

Behavior:

• Requires at least 2 active curves to activate (otherwise ignored)
• Focuses all temporal plots on the main acoustic energy of the Sum/Average response
• Individual driver/measurement curves are zoomed to the same time window as the sum
• Easier to compare arrival times and transient behavior across all curves
• Works with adaptive display mode margins (configurable in Settings)

Use Cases:

• Loudspeaker Design: Quickly identify timing alignment issues across drivers
  • See if all drivers arrive within the sum’s main energy window
  • Detect delays or misalignment relative to system response
• Room Calibration: Focus on the averaged room response energy region
  • Compare individual measurement positions to the average response timing
  • Identify room reflections within the main energy window
• Eliminate clutter from pre-ringing or tail noise outside the main energy region

Only available in Drivers/Measurements Mode (not applicable to individual filters).

Note: Works in both Loudspeaker Design and Room Calibration project modes when display mode is set to Drivers/Measurements.

Phase Display Options

Unwrap Phase

Toggle: U key or toolbar button

Controls phase display format:

Wrapped Phase (default):

• Phase values constrained to ±180°
• Shows discontinuities (phase wraps) at ±180° boundaries
• Easier to read for simple filters
• Standard display format for most audio applications

Unwrapped Phase:

• Continuous phase values extending beyond ±180°
• Shows total accumulated phase shift
• Useful for analyzing group delay trends and overall phase shift

Note: Group delay is always computed from unwrapped phase internally. The Unwrap toggle only affects phase display, not group delay accuracy.

Remove time of flight rotations

Toggle: C key or Phase Options dropdown (next to Unwrap Phase)

A pure propagation delay translates into a linear slope in the phase response — the further the impulse response peak is from t = 0, the steeper the slope, and the more rotations accumulate across the frequency range. This linear ramp is entirely determined by the physical distance between the driver and the microphone and does not reveal anything about filter behaviour or driver characteristics.

This option removes that linear component via a linear regression on the IR peak position, leaving only the non-linear phase introduced by the filters and acoustics. Phase curves become much flatter, easier to read, and directly comparable across drivers regardless of their physical placement.

Removes the linear phase component (constant group delay) from phase responses:

Algorithm:

• Computes linear phase slope from IR peak position for all curves (drivers and filters)
• Subtracts the constant group delay (linear phase shift)
• Flattens phase around 0° while preserving relative phase differences
• Works with both wrapped and unwrapped phase display modes

In Loudspeaker Design Mode (Drivers):

When multiple speakers are active:

• Removes the linear delay of the Sum curve from all responses
• Centers all driver phases around 0° relative to the system response
• Highlights phase alignment issues between drivers
• Makes phase differences visible without bulk delay obscuring them

When single driver is active:

• Removes the linear component of that driver’s phase
• Flattens phase around 0° to show only non-linear phase shifts

In Room Calibration Mode (Measurements):

• Independent correction: Removes the linear phase component of each individual measurement independently
• No shared reference - each measurement position is treated separately
• Flattens each measurement’s phase around 0°
• Useful for comparing phase behavior across different room positions without timing offsets
• The averaged curve also has its own linear component removed

In Filters Mode (FIR, IIR, FIR+IIR):

• Per-filter correction: Removes the linear phase component of each individual filter
• Flattens each filter’s phase around 0°
• Reveals only non-linear phase rotations introduced by the filter
• Useful for analyzing filter phase behavior independent of bulk delay

Use Cases:

• Identify phase alignment issues between drivers without delay obscuring differences
• Visualize phase rotations introduced by filters (independent of linear delay)
• Compare filter phase characteristics across different designs
• Verify minimum-phase vs. linear-phase behavior

Fractional Octave Smoothing

Control: Dropdown menu in toolbar (None, 1/48 oct to 1/3 oct, or ERB)

Apply smoothing to reduce measurement noise and reveal trends:

Fixed Fractional-Octave Smoothing:

• No smoothing - Show raw response (no averaging)
• 1/48 octave - Very fine smoothing (highest detail)
• 1/24 octave - Fine smoothing
• 1/12 octave - Moderate smoothing (default)
• 1/6 octave - Coarse smoothing
• 1/3 octave - Very coarse smoothing (lowest detail)

ERB (Equivalent Rectangular Bandwidth) Smoothing:

ERB smoothing uses a frequency-dependent bandwidth that models the human auditory system’s frequency resolution. Unlike fixed fractional-octave smoothing, ERB adapts its bandwidth at each frequency according to psychoacoustic research.

Bandwidth formula: (107.77f + 24.673) Hz, where f is in kHz

Frequency-dependent behavior:

• 50 Hz → ~1 octave bandwidth (heavy smoothing)
• 100 Hz → ~1/2 octave bandwidth
• 200 Hz → ~1/3 octave bandwidth
• 1 kHz and above → ~1/6 octave bandwidth (leveling out)

When to use ERB smoothing:

• To see what the ear actually perceives (perceptually accurate)
• When evaluating audibility of response features
• To match listening test results with measured data
• For publication-quality graphs that reflect human hearing

When to use fixed fractional-octave smoothing:

• For equalization and correction work (1/6 or 1/12 octave recommended)
• When analyzing crossover behavior in detail
• To maintain consistent resolution across all frequencies
• For technical comparisons between drivers

Effect:

• Wider smoothing = smoother curves, less detail
• Narrower smoothing = more detail, more noise visible
• ERB = perceptually-weighted smoothing (variable with frequency)

Applies to:

• Frequency response (magnitude), HD, Directivity Analysis (sonogram), group delay and phase plots
• Does not affect time-domain plots (impulse responses)

Note: Smoothing is applied after all processing (filters, crossovers, correction). It affects display only, not exported data. It can also alter crossover slopes on the graphs.

Impulse Response Analysis

Automatic Pre-Ringing Detection

When hovering over impulse response curves, LinFIR displays additional analysis information:

Pre-ring Level (in dB relative to peak):

• Weighted RMS measurement of signal energy before the main peak
• Analysis stops at the last zero-crossing before peak (excludes rising edge)
• Uses Hilbert envelope to capture perceptible energy regardless of phase
• Perceptually weighted: energy closer to the peak is less penalized (temporal masking)
• Displayed as: Pre-ring: -50.0 dB (Likely masked)

Algorithm Details:

1. Find the impulse peak position and amplitude
2. Find the last zero-crossing before the peak (to exclude the rising edge)
3. Extract the pre-ringing region (from start to zero-crossing)
4. Calculate the Hilbert envelope of the pre-ringing region
5. Apply sigmoid weighting (closer to peak = less perceptual penalty)
   • Masking time at 50% penalty: 0.6 ms (default)
   • Weight floor: 0.2 (minimum weight for distant pre-ring)
   • Weight ceiling: 1.0 (maximum weight for near-peak pre-ring)
6. Calculate weighted RMS of the envelope
7. Convert to dB relative to peak amplitude

Weighting Rationale:

• Pre-ringing close to the peak is masked by the main transient (temporal masking)
• Pre-ringing far from the peak is more audible (less masking)
• Sigmoid function models gradual transition from masked to unmasked

Risk Assessment Labels

Pre-ringing levels are classified based on typical perceptual thresholds (indicative):

Important Notes:

• These are indicative estimates, not absolute rules
• Actual audibility depends on:
  • Program material (transient-rich vs. sustained tones)
  • Listening level
  • Individual hearing sensitivity
  • Room acoustics and background noise
• Use as a guideline for filter design, not strict limits
• Shorter FIR filters or increased causality reduce pre-ringing at the cost of magnitude and phase accuracy

Trade-offs:

• Longer FIR filters → More pre-ringing, better magnitude/phase accuracy
• Shorter FIR filters → Less pre-ringing, reduced correction capability
• Higher causality (e.g., minimum-phase) → Less pre-ringing, potential phase distortion
• Linear-phase (causality = 0) → Maximum pre-ringing, linear phase

Auto-Bounds Behavior

LinFIR automatically adjusts graph bounds to fit data:

Auto-Bounds Enabled (default):

• Graphs automatically scale to show all data
• Y-axis adjusts when switching between drivers or display modes
• X-axis adjusts based on data range (frequency or time)

Auto-Bounds Disabled (after user interaction):

• Manual zoom/pan disables auto-bounds for that axis
• Graph retains user-specified bounds
• Switching modes preserves manual bounds
• Double-click re-enables auto-bounds and resets to calculated range

Per-Axis Independence:

• X-axis and Y-axis auto-bounds are independent
• Dragging vertically disables Y auto-bounds, preserves X auto-bounds
• Dragging horizontally disables X auto-bounds, preserves Y auto-bounds
• Box zoom (right-click drag) disables both X and Y auto-bounds

Per-Graph Independence:

• Each graph type has independent auto-bounds state
• Zooming Frequency plot does not affect Impulse plot
• Switching between display modes preserves per-graph bounds

Keyboard Shortcuts Summary

Display Modes

• F - Switch to Filters display mode
• D - Switch to Drivers/Measurements display mode

Display Controls

• X - Toggle X-axis synchronization
• N - Toggle normalize impulse and step responses
• U - Toggle unwrap/wrap phase response
• C - Toggle time of flight rotation removal
• S - Toggle focus on summed impulse response (Drivers/Measurements mode only)
• R - Toggle Directivity Analysis window (Loudspeaker Design mode only, requires license)

Graph Visibility

• M - Toggle Magnitude Response plot
• P - Toggle Phase Response plot
• G - Toggle Group Delay plot
• I - Toggle Impulse Response plot
• T - Toggle Step Response plot
• K - Toggle HD (Harmonic Distortion) plot

Related Documentation

• Settings - Graph resolution, titles, colors
• Keyboard Shortcuts - Complete shortcut reference
• Driver Processing - Filters affecting graph display
• System Processing - Global filters and sum calculation

Keyboard Shortcuts

This is a comprehensive reference of all keyboard shortcuts available in LinFIR.

Note: On macOS, use Cmd (⌘). On Windows/Linux, use Ctrl.

File Operations

Context:

• Works in main window and all detached windows
• Disabled when text input fields have focus (to avoid conflicts)
• Auto-save runs every 60 seconds in the background (if enabled)

Window Operations

Context:

• Cmd+W works in detached graph windows, driver IR windows, HFD export window, etc.
• Cmd+, is a system-wide shortcut to access application settings
• H toggles documentation visibility globally

Display Modes

Behavior:

• D: Shows acoustic output (driver impulse + all filters applied)
• F: Shows individual filter responses
  • First press: FIR filters only
  • Second press: IIR filters only
  • Third press: FIR+IIR combined
  • Fourth press: Back to FIR filters

Context: Only active when no text input has focus.

Display Controls

Details:

• X (Sync X-Axis):

  • Frequency plots (Magnitude, Phase, Group Delay) share X-axis
  • Time plots (Impulse, Step) share X-axis
  • Y-axes remain independent

• N (Normalize):

  • Visual only - does not affect exported data
  • Useful for comparing filter shapes with different gains

• U (Unwrap Phase):

  • Wrapped: phase constrained to ±180° (discontinuities)
  • Unwrapped: continuous phase beyond ±180° (smooth)

• C (Remove time of flight rotations):

  • Removes linear phase (constant group delay) via linear regression
  • Flattens phase around 0° to show only non-linear behavior
  • In Drivers mode: removes Sum curve delay from all responses
  • In Filters mode: per-filter linear phase removal

• S (Focus on Summed Response):

  • Only available in Drivers mode
  • Calculates time boundaries from summed impulse
  • Auto-enables auto-bounds for all time plots

Context: Only active when no text input has focus.

Graph Visibility Toggles

Behavior:

• Each toggle is independent
• State persists across sessions (saved in project)

Context: Only active when no text input has focus.

Advanced Features

Requirements:

• R and J only work in Loudspeaker Design and Hypex FusionAmp modes (not Room Calibration)
• R, W, and J require a valid license

Context: Only active when no text input has focus.

Undo/Redo

Behavior:

• Full project state snapshots (parameters + shared references to the measurement data, so each step costs little memory)
• Undo history keeps the last 50 steps
• Works in main window and all viewport windows (Driver IR, HFD export, etc.)

Context: Disabled when text input fields have focus.

Dialog-Specific Shortcuts

Shortcut Contexts

LinFIR disables shortcuts based on context to avoid conflicts:

When Text Input Has Focus

The following shortcuts are disabled when typing in text fields (IR names, filter frequencies, etc.):

• All single-letter shortcuts (D, F, J, M, K, G, I, T, P, U, C, X, S, N, R, W, H)
• Undo/Redo (Cmd+Z, Cmd+Shift+Z)
• Save/Save As (Cmd+S, Cmd+E)
• Settings (Cmd+,)

Still active:

• Close window (Cmd+W)

When Any Input Window is Open

Certain shortcuts are disabled when modal dialogs or input windows are open:

• Save (Cmd+S)
• Save As (Cmd+E)

This prevents accidental saves while configuring settings or importing files.

Platform-Specific Notes

macOS

• Use Cmd (⌘) for all modifier shortcuts
• Cmd+W closes windows but not the main application window
• Cmd+Q quits the application (system shortcut)
• Cmd+, is the standard macOS shortcut for Preferences/Settings

Windows/Linux

• Use Ctrl for all modifier shortcuts
• Ctrl+W closes windows
• Alt+F4 quits the application (system shortcut)

Quick Reference Card

File

• Cmd+S - Save
• Cmd+E - Save As

Window

• Cmd+W - Close Window
• Cmd+, - Settings
• H - Documentation

Display

• D - Drivers Mode
• F - Cycle Filters Mode
• X - Sync X-Axis
• N - Normalize IR/SR
• U - Unwrap Phase
• C - Remove time of flight rotations
• S - Focus on Summed Response

Graphs

• M - Magnitude
• P - Phase
• G - Group Delay
• I - Impulse
• T - Step
• K - Harmonic Distortion

Advanced

• R - Directivity Analysis (license required)
• W - Spectrogram (license required)

Edit

• Cmd+Z - Undo
• Cmd+Shift+Z - Redo

Related Documentation

• Graph Interaction - Mouse controls, zoom, pan, detach graphs
• Settings - Configure default display modes and preferences
• Driver Processing - Filter controls affected by shortcuts
• System Processing - Global filters and display modes
• Directivity Analysis - Directivity Analysis window (R key)
• Spectrogram - Spectrogram window (W key)

General Settings

The General tab holds application-wide preferences, filter-processing behaviour, and the defaults applied to new projects. Changing a default here does not touch projects that already exist.

The tab is organised into: Project, Appearance, Filter Processing, Measurement & Import Quality, Default Values for New Projects, Diagnostics, and Reset.

Project

Auto-save

Automatically saves the project after a committed change (a capture, a filter edit, an undo/redo…).

• Requires a project path — auto-save only activates once the project has been saved to disk at least once. Until then a hint reminds you that “Auto-save will activate after first manual save”.
• When off, a “(disabled)” label appears next to the toggle and you save manually with File → Save / Ctrl / Cmd + S.

Appearance

UI Color Scheme

• Red (default) — custom red accent for section headers, graph titles, active buttons and selection highlights.
• Blue — the standard egui theme with blue accents.

Applies immediately, no restart.

Theme

The standard light / dark / follow-system buttons. “Follow system” tracks your OS appearance setting.

Language

The language of the user interface: Automatic (default), English or Français.

• Automatic follows the language of your operating system - French when the OS is set to French, English otherwise. The entry shows which language it currently resolves to, e.g. Automatic (Français).
• The change applies immediately, without restarting, and is remembered between sessions.
• Only the interface is translated: this documentation, exported files (reports, filter files), log files and bug reports stay in English.

Filter Processing

Settings that change how filter chains are computed and validated. These apply to the current session, not just new projects.

Clipping detection

Default: enabled  ·  Threshold: 0 dB (range −5 to +2 dB)

Applies every active filter to a deterministic pink-noise test signal and warns when the output exceeds the threshold, suggesting a gain reduction.

The test runs in the background: it never slows down the graphs, even with long filters, and its warnings appear a moment after a change. Each filter chain is only re-tested when it actually changes — dragging one driver’s slider re-tests that driver alone — and a warning always stays attached to the right driver, even after reordering drivers.

Test signal: periodic pink noise, 20 Hz – 20 kHz, −3 dB/oct slope, band-shaped (12 dB/oct high-pass at 20 Hz, 24 dB/oct low-pass at 20 kHz), RMS ≈ −24 dBFS, peak = −6 dBFS.

Threshold (relative to 0 dBFS):

⚠️ The threshold field is only editable while clipping detection is on. Above 0 dB, a “⚠ Threshold above 0 dBFS” warning is shown. With detection off, “⚠ Safety warnings disabled” is shown.

Notes

• Measurement levels are not tested — the warning is purely about filter processing; the test signal is synthetic and independent of your measurement amplitude.
• Phase processing can clip even with a flat magnitude response — phase-correction FIRs, all-pass filters, and multiple FIRs with different causality can align components in time and push the temporal peak past 0 dBFS.
• Disabling removes the per-update test-signal processing (a small performance gain) and all safety warnings — always test with real content afterwards.

Warn when phase FIR guard fails

Default: enabled

The phase-correction auto-guard always runs: it raises f_min / f_max in 1/3-octave steps (up to 100 iterations) until the FIR’s magnitude deviation falls within the per-filter Guard Tolerance (0.1–15 dB, default 0.5 dB, set in each Phase Correction / Global FIR Phase Correction window).

When this toggle is on and the guard still gives up, a warning toast is shown. Turning it off silences the toast only — the guard keeps running. To fix the underlying failure: reduce the phase-correction amount, lower the Kaiser β, increase the tap count, or raise the Guard Tolerance.

Show manual FIR delay compensation   ⚙ Advanced

Default: disabled (hidden)

Shows the FIR Offset Delay field in each driver’s FIR section and the FIR compensation delay field in the Global FIR Correction window. When disabled, every FIR compensation delay is forced to 0.

Keep this off unless you need manual control over FIR impulse alignment — incorrect values cause filter truncation and frequency-response ripples.

Auto causal alignment   ⚙ Advanced

Default: enabled  ·  Forced on whenever Show manual FIR delay compensation is off (there would otherwise be no way to correct a crop misalignment).

Optimises the crop-window position of the combined FIR impulse response when several filters with non-zero causality are active. The log-energy distribution around the IR peak is used to place the window proportionally, so causal pre-ringing or post-ringing isn’t disproportionately truncated at the tap boundary.

Active only when at least 2 FIR filters are enabled (LP / HP / correction / phase) and at least one has non-zero causality.

Leave it enabled unless you have solid DSP experience and want to position the impulse manually with the FIR delay-compensation fields.

Measurement & Import Quality

Always keep low-SNR measurements

Default: disabled

A sweep measurement or an imported impulse response with a peak level below −25 dB has a poor signal-to-noise ratio. By default, LinFIR pops up an always-on-top Low Signal Level Warning dialog when this happens, asking whether to Keep Anyway or Discard it — see Quality Validation and Import Signal Level Check for the full behaviour.

Turning this on skips that dialog: low-level captures and imports are kept automatically, with only a toast as a reminder.

⚠️ Correction filter quality strongly depends on measurement quality. A low signal-to-noise ratio means noisier data, which can produce inaccurate correction filters and skew phase visualization. Leave this disabled unless you specifically want to bypass the check — for example when re-importing a known-good but intentionally low-level measurement.

This setting does not affect clipping, which is always rejected outright regardless of this toggle — clipped data is corrupted, not just noisy, so there’s nothing to usefully keep. It also doesn’t affect an unattended auto-scan (off-axis sweep automation): a low level there still aborts the scan automatically rather than popping up a dialog nobody is there to answer, unless this setting is enabled, in which case the scan keeps going.

Default Values for New Projects

A two-column grid. Every row seeds a new project only.

Sample rate

Match your target DSP.

• 44.1 kHz — compact DSP platforms with limited processing power
• 48 kHz — professional-audio standard, most DSP platforms
• 88.2 / 96 kHz — high-resolution audio
• 93.75 kHz — Hypex FA-series amplifiers (FA123, FA253, FA502…)
• 176.4 / 192 kHz — only when the DSP platform requires it (higher CPU load, larger filters)

FIR length (taps)  ·  Global FIR taps

Default tap count for per-driver FIR filters (LP / HP / correction) and for the global correction FIR.

• 512 – 2048 — enough for most crossover and correction work
• 4096 — typical maximum for standard loudspeaker design
• 8192 – 16384 — sub-bass phase / magnitude correction at very low frequencies
• More taps → finer resolution and steeper slopes, but higher latency (taps/2 samples for a linear-phase filter) and more DSP load

Curve smoothing

Fractional-octave smoothing applied to frequency-response curves. None by default — the raw response with all detail. 1/12 octave is a good balance of detail and readability; 1/3 octave shows overall trends only.

Graph display mode

Which curves a new project shows first. Drivers/Measurements (the fully-processed driver or measurement responses) by default. The other modes show filter responses only. Press F / D to switch modes at any time; the label reads “Drivers” in Loudspeaker Design mode and “Measurements” in Room Calibration mode.

Phase display

Wrapped (values in [−180°, 180°]) or Unwrapped (continuous). Group delay is always computed from unwrapped phase regardless of this setting. Press U to toggle.

Time-centered phase

Seeds the phase graph’s “Remove time of flight rotations” toggle (the toolbar and the C shortcut use that name; Settings calls the default “Time-centered phase”). When on, the linear-phase component (a constant group delay) is removed so the phase sits around 0°:

• Loudspeaker Design — removes the Sum curve’s linear delay from every driver (or the driver’s own delay for a single driver), so you can see phase alignment between drivers relative to the system
• Room Calibration — removes each measurement’s own linear phase independently (no shared reference)
• Filters mode — removes each filter’s own linear phase

Sync X-axis

Whether zooming one plot zooms all related plots. When on, every frequency plot (Magnitude, Phase, Group Delay, Directivity) shares one horizontal range and every time-domain plot (Impulse, Step) shares one time range. Enabled by default; toggle with X or the toolbar button.

Diagnostics

Verbose logging

Default: disabled  ·  not persisted — always starts off.

Raises the log filter to debug at runtime, no restart or terminal needed. Enable it only when support asks, reproduce the issue, then submit Help → Report a Bug (the log file is attached automatically). Turn it back off afterwards — verbose logs are larger and rotate more often. See Bug Reports.

Reset

Reset Application Settings

⚠️ Cannot be undone.

Deletes every application setting and preference, clears the app cache, resets all new-project defaults to factory values, and closes LinFIR immediately (relaunch manually). The license activation is kept. Existing project files are untouched — save your current project first.

Graphs Settings

The Graphs tab controls how the plots are computed and displayed. Some rows act on the current project (crossover smoothing, truncation, plot bounds); others are defaults for new projects (graph order and visibility).

Sections, in tab order: Display, Group Delay & Phase Truncation, Frequency Axis, Magnitude Plot Bounds, Directivity Sonogram, Spectrogram, Impulse/Step Response Display, Response Overlays, Graph Order, Default Visibility, Default Overlays.

Appearance, filter-processing and new-project value defaults live on the General tab, not here.

Display

Plot resolution

Options: 16k points (faster), 32k points (balanced), 64k points (maximum quality) Default: 64k points

Number of frequency points used to compute filter and driver responses.

• Higher resolution → finer low-frequency detail in the magnitude, phase and group-delay plots
• Affects computation time for filter and measurement responses, not filter quality
• Drop to 16k or 32k on slower systems or with very long filters

Steep crossover smoothing

Default: enabled  ·  acts on the current project (adaptive_crossover_smoothing).

Reduces the frequency-response smoothing near crossover frequencies so a steep filter’s true slope stays visible instead of being rounded off.

Reduction vs. effective acoustic slope — linear from 12 to 36 dB/oct, then held at 100 %:

At full reduction the smoothing drops to 1/96 octave at the cutoff. The reduction fades in over the 0.5 octave of passband next to the cutoff (smoothstep), reaching its minimum at the cutoff itself. Only the magnitude display curves are affected — phase, group delay and filter generation are untouched.

When disabled: uniform smoothing everywhere, matching external tools that use constant smoothing (which then show gentler crossover slopes than LinFIR’s adaptive mode).

Group Delay & Phase Truncation

Hide the group-delay and phase curves outside the active magnitude region.

Enable truncation

Default: enabled

When on, a curve is drawn only where its reference magnitude exceeds max + threshold. When off, the full curves are shown (“(showing full curves)” appears next to the toggle). Purpose: keep phase / GD out of stop-bands and cut visual clutter.

Truncation thresholds

A grid of DragValues, dB below the maximum, range −120 to 0 dB, all defaulting to −20 dB:

• −20 dB means “show the curve only where magnitude > (max − 20 dB)”
• The reference magnitude matches the context — FIR curves use the FIR magnitude, drivers use the driver magnitude, etc.
• HD (harmonic distortion) truncation exists for the Drivers context only

Frequency Axis

The frequency range shown on every frequency-axis graph in the app (current project).

Min frequency (Hz)

Range: 1 Hz to (Max frequency − 1 Hz)  ·  Default: 10 Hz

Max frequency (Hz)

Range: (Min frequency + 1 Hz) to 96 000 Hz  ·  Default: 24 000 Hz

Reset

Restores both fields to 10 Hz / 24 000 Hz.

These two values replace what used to be a fixed range hardcoded per graph. They now drive:

• Magnitude, Phase, Group Delay, HD — the default/reset X-axis range, and which data points are drawn at all (a point outside [Min, Max] is dropped, not just scrolled off-screen)
• Directivity Sonogram — both the frequency range the heatmap image is sampled over and its on-axis overlay plot’s default range; a change regenerates the sonogram image
• Spectrogram — only the range its own window opens with. It has an independent Frequency Range control with its own Reset button — once opened, it no longer tracks changes made here except when that Reset is clicked.

Narrowing the range does not change how filters or the FIR/IIR chain are computed — it only affects what the graphs draw. It is independent of Sync X-axis: synced or not, “full range” now means this setting instead of a fixed 10 Hz – 24 kHz.

Magnitude Plot Bounds

Default Y-axis bounds for magnitude plots (current project).

Drivers range (dB)

Range: 20 – 200 dB  ·  Default: 60 dB

Y-axis span for driver / measurement magnitude plots, drawn from max_response − range to max_response + 5 dB.

• 20–40 dB — zoomed, emphasises small variations
• 60–80 dB — good for most designs
• 100–200 dB — wide view of the full range

Filters lower bound (dB)

Range: −200 to −20 dB  ·  Default: −100 dB

Lower Y-axis bound for filter magnitude plots (FIR, IIR, FIR+IIR).

• −40 dB — focus on the pass-band
• −100 to −200 dB — show deep stop-band attenuation

Directivity Sonogram

License-gated analysis window. These rows sit just below the magnitude-bound rows in the tab. The sonogram’s frequency range itself is set by Frequency Axis above, not here.

Sonogram lower bound (dB)

Range: −100 to −10 dB  ·  Default: −30 dB

Bottom of the directivity-sonogram colour scale. Narrowing it exposes low-level decay / room modes; a change regenerates the sonogram image.

Highlight sonogram clipping

Default: enabled

Draws sonogram levels above the colormap maximum in off-white (fading from red over about 1 dB) instead of letting them saturate to the same red as the loudest in-range energy — so it is obvious where per-band normalization pushes the response over the reference level. Turn it off to have over-range values saturate to red.

Sonogram normalization

Options: Global max., Normalize to 0°, Normalize to 0° (per freq), Normalize to ±10° avg (per freq) Default: Global max.

The normalization mode the directivity sonogram opens with. It can still be changed per-session from the sonogram’s own toolbar. See Directivity Analysis.

Spectrogram

License-gated analysis window. See Spectrogram. Its frequency range is seeded by Frequency Axis above but adjustable independently from its own window — see Frequency Range.

Spectrogram lower bound (dB)

Range: −120 to −10 dB  ·  Default: −30 dB

The Min dB value the spectrogram opens with, and what its Reset button restores. It also drives the time crop. It can still be changed per-session from the spectrogram’s own toolbar.

Spectrogram axes

Options: Time × Freq (time on X), Freq × Time (frequency on X) Default: Time × Freq

The axis layout the spectrogram opens with. The spectrogram’s toolbar button still swaps it per-session. See Spectrogram.

Spectrogram detail

Options: Standard (2048), High (4096), Ultra (8192)  ·  Default: High (4096)

The size of the spectrogram compute grid — the number of frequency bins and time points the transform runs on before it is sampled onto the image. Higher is sharper but slower and more memory (Ultra ≈ 4× the compute time and ~256 MB per recompute). See Spectrogram → Detail.

Impulse/Step Response Display

Enable adaptive display

Default: enabled

Detects the signal boundaries (threshold 1/1000 of peak amplitude) and shows only that portion of the impulse and step responses, plus the margins below. When off, the complete impulse response (full tap length) is always shown — useful to inspect zero-padding, awkward with long IRs.

Margin before (ms)

Range: 0 – 50 ms  ·  Default: 5 ms — keeps pre-ringing and early arrivals visible.

Margin after (ms)

Range: 0 – 100 ms  ·  Default: 10 ms — captures the decay tail and late reflections.

Response Overlays

The Listening Window, Predicted In-Room Response and Directivity Index overlay toggles moved to the Graphs dropdown in the graph toolbar, alongside a new Sum / Average toggle - see Graph Visibility Toggles. Each project now remembers its own choice for these four; this tab only sets what a new project starts with - see Default Overlays below.

Show raw response when filter windows are open

Requires a valid license — without one the toggle is greyed out and the overlay isn’t drawn.

Default: enabled

When a filter window is open and in the foreground, a dashed overlay of that driver’s pre-filter response (gain + windowing only, no FIR/IIR) is drawn on the frequency, phase and group-delay graphs.

• Triggered by any per-driver LP / HP / correction / IIR window, or the Global FIR Correction / Global IIR window (the latter show the unfiltered sum / average)
• Only the window in the foreground counts: with several filter windows open, only the driver (or sum) of the window you are working in is overlaid, and nothing is shown when all filter windows are in the background
• Coloured to match the driver’s curve; the global overlay is white (dark mode) / black (light mode)

Use it to see how much correction a filter applies without toggling it off.

Graph Order

Drag a graph by its ☰ grip or its name and drop it above or below another one, or use the ⬆ ⬇ buttons to move it one step. The order applies to the graph stack immediately and is saved as the default layout for new projects.

The graphs can also be reordered directly in the main window, by dragging their title — see Reorder Graphs.

Default Visibility

A toggle per graph type (Frequency, THD, Group Delay, Phase, Impulse, Step) choosing which graphs are shown by default in new projects.

Default Overlays

A toggle per magnitude overlay curve (Sum / Average, Listening Window, Predicted In-Room Response, Directivity Index) choosing which are shown by default in new projects. Each project then remembers its own choice from the Graphs dropdown in the graph toolbar - see Graph Visibility Toggles.

Performance Considerations

Real-Time Filter Processing Architecture

LinFIR differs from traditional FIR design software in how it processes and displays results.

Real-time (“online”) filter generation:

• Filters are generated and applied to driver impulse responses in real time as you interact with the UI
• The complete signal-processing chain is computed immediately as you adjust parameters
• You see the true response of the entire chain instantly, including every windowing artifact

Why this matters:

• Immediate feedback — no “generate” / “apply” button
• Transparency — shows exactly what the filters can and cannot do, with all artifacts visible
• Accuracy — no approximation or preview mode

Filter Application Methods

FIR filters — applied via fast convolution (multiplication in the Fourier domain); cost scales with filter length and measurement count.

IIR filters — applied via recursive difference equations (time domain); extremely efficient regardless of complexity.

Performance Impact Factors

Filter length:

• Up to 4096 taps — negligible impact on modern CPUs
• 4096 – 8192 taps — slight delays on slower systems
• > 8192 taps — can noticeably slow interactive adjustments

Off-axis measurements: each angle re-runs the full filter chain — many angles + long filters means longer updates.

Driver count: every enabled driver adds a filter chain that must be summed; cost grows roughly linearly with driver count and multiplies with angles per driver.

Large Multi-Driver Projects (Sound Reinforcement)

Loudspeaker Design mode accepts up to 50 drivers, which makes it suitable for aligning a complete sound-reinforcement system — a line array, its subwoofer arrays, and the front fills, each cabinet or zone captured as a separate driver.

This is a supported workflow. Drivers are normally measured on-axis only (one measurement per cabinet or zone), so the load stays close to driver count × filter length:

• A 50-driver project recomputes 50 full convolution chains on every parameter change
• With filters of 8192 taps or more, dragging a crossover or sweeping a gain can become sluggish
• Recommendation: keep filter lengths at or below 4096 taps while setting up and aligning a large system, and only raise the tap count at the end if low-frequency resolution genuinely requires it

Edge case — polar measurements on many drivers at once

Capturing full polar (off-axis) sets for many drivers at the same time multiplies the cost rather than adding to it: 30 drivers × 19 angles is 570 complete filter chains per update. If you need directivity data alongside a large alignment project, capture polar sets for a few representative cabinets only, or keep that analysis in a separate, smaller project.

Room Calibration mode is unaffected. Polar measurements are not available there, so a 50-measurement room project only carries the linear cost of the measurement count.

Additional measures for large systems:

• Reduce plot resolution to 16k or 32k points
• Clipping detection runs in the background and doesn’t slow the graphs down; disabling it (Settings → General) only frees some CPU time on very large projects

Background Processing for Responsiveness

Priority (synchronous, immediate): the on-axis (0°, 0°) measurement, the currently displayed angle, and filter parameters.

Background (asynchronous, non-blocking): the remaining off-axis measurements, the directivity index, multi-driver sum calculations, and clipping detection.

The main graph always updates immediately for the selected angle, and the UI stays responsive during background work.

Optimization Recommendations

For slower systems or very long filters (> 8192 taps):

1. Reduce plot resolution — Settings → Graphs → Plot resolution (16k or 32k)
2. Limit off-axis measurements — reasonable angle increments (10 or 15°); on a large multi-driver project, capture polar sets only for the drivers that need directivity analysis

Performance is always good for: projects with standard filter lengths (≤ 4096 taps), IIR-only filtering, single on-axis measurements, and modern multi-core CPUs.

Best Practices

Performance

• Slower systems — reduce plot resolution to 16k / 32k points
• Large systems — keep FIR lengths moderate and work section by section with Solo / Disable

Filter design

• Keep Auto causal alignment enabled unless you need precise manual IR positioning
• Keep Show manual FIR delay compensation disabled unless you are experienced with FIR alignment
• Keep Clipping detection on during design; test with real content before deploying

Graph display

• Enable truncation for cleaner phase / GD plots
• Enable Steep crossover smoothing to read true crossover slopes

Audio Configuration

Before performing measurements or analyzing audio, LinFIR needs to be configured with the correct audio devices and parameters. This section covers the audio settings available in the application.

Accessing Audio Settings

Audio settings are accessed via:

• Menu: View → Settings → Audio Settings
• Keyboard: Cmd+, (macOS) or Ctrl+, (Windows), then select Audio Settings tab

Device Selection

Input Device

The input device captures audio from microphones or measurement equipment:

• Click the dropdown to see available input devices
• Select your measurement microphone or audio interface input
• Verify the device shows the correct channel count
• Ensure the device isn’t in use by another application

Output Device

The output device plays back sweep signals or test tones:

• Select the audio interface connected to your loudspeaker system
• Verify the device supports the desired sample rate
• Check that the device has sufficient channels for your application

Refreshing Devices

If your audio device doesn’t appear:

1. Click “Refresh Devices” to rescan available hardware
2. Ensure the device drivers are properly installed
3. Check that the device is powered on and connected
4. Try restarting the application if the device remains unavailable

Sample Rate Configuration

Choose a sample rate matching your needs:

• 44.1 kHz: CD quality, most common for consumer audio
• 48 kHz: Professional audio standard, video production
• 88.2/96 kHz: High-resolution audio, improved time-domain precision
• 176.4/192 kHz: Ultra-high resolution (use only when necessary for THD analysis)

⚠️ Note: The sample rate must be supported by both your input and output devices. If a rate is unavailable, try a different value.

Sample Format Detection

LinFIR automatically detects and selects the best compatible audio sample format between your input and output devices:

• 16-bit integer (I16): Standard for most consumer audio devices
• 24-bit integer (I24): Professional audio interfaces (e.g., Dante Virtual Soundcard, RME, MOTU)
• 32-bit float (F32): High-end interfaces, ASIO and Core Audio devices
• 32-bit integer (I32): Some professional hardware

The format is automatically negotiated during measurement setup - no manual configuration required. If LinFIR cannot find a compatible format, try:

• Using ASIO drivers (Windows) for better professional interface support
• Updating your audio interface drivers
• Checking device sample rate compatibility

Buffer Size

Buffer latency is compensated internally and has no effect on measurement accuracy. The only criterion for choosing a buffer size is avoiding underruns (audio dropouts caused by the buffer emptying before it can be refilled).

• Smaller buffers (128-256 samples): More demanding on the CPU — increase the risk of dropouts on slower systems
• Recommended (512-1024 samples): Reliable on most systems
• Larger buffers (2048+ samples): Maximum stability, useful if dropouts persist

If you experience audio dropouts during measurements, increase the buffer size. Otherwise, any value works.

Input Gain Compensation

Software-based gain adjustment applied to the input signal:

• Range: -32 dB to +32 dB
• Purpose: Compensate for fixed or unavailable hardware gain controls
• Use case: Particularly useful for USB microphones like UMIK-1 on Mac OS
• Clipping: always detected on the raw converter level, before this compensation - a positive compensation no longer triggers false clipping alarms, and a negative one no longer hides real clipping

Why Use Input Gain Compensation?

Some measurement microphones have fixed gain or no hardware gain control, particularly on certain platforms:

• Mac OS USB microphones: Mac OS enforces very low fixed gain for USB microphones
• UMIK-1 example: No hardware control available in Mac OS system settings
• Solution: Use positive gain compensation (+18 to +24 dB) to achieve optimal recording levels

Important Notes

• Input gain compensation does not improve signal-to-noise ratio
• It’s a convenience feature for level matching, not a replacement for proper gain staging
• Optimal SNR still requires appropriate acoustic levels and microphone positioning
• Use hardware gain controls when available for best results

Default Sweep Parameters

Configure default settings for the built-in sweep measurement system. These apply to new measurement captures and can be accessed in the Audio Settings tab under Measurement Configuration.

Sweep Duration

Range: 1.0 to 10 seconds Default: 5 seconds

Default length of exponential sine sweep (ESS) for impulse response measurements.

• Shorter (1.0-2s): Faster measurements, lower frequency resolution
• Longer (5-10s): Better low-frequency resolution, improved signal-to-noise ratio, better harmonic distortion measurement

Benefits of longer sweeps:

• Improved signal-to-noise ratio for harmonic distortion analysis
• Better frequency resolution at low frequencies
• More reliable detection and separation of weak harmonics

⚠️ Important: Longer sweeps deliver more energy to the driver at low frequencies (exponential sweeps spend more time in the bass region). For drivers with limited excursion capability, this may require raising the start frequency or reducing the sweep level to avoid over-excursion.

Recommendation: 5-7 seconds for full-range measurements, 2-3 seconds for quick driver checks.

Sweep Level

Range: -40 to 0 dBFS Default: -6 dBFS

Default output level for sweep signals.

Adjust this to match your amplifier’s input sensitivity and desired acoustic level. Lower values provide headroom for systems with high gain.

Start Frequency

Range: 1 to 5000 Hz Default: 1 Hz

Default cutoff frequency for the 4th-order Butterworth high-pass filter applied to the sweep signal for driver protection.

Filter Characteristics:

• Attenuation at Start Frequency: -3 dB (Butterworth characteristic)
• Roll-off below cutoff: -24 dB/octave (deterministic, 4th order)

Frequency Guidelines:

• 1 Hz: Lowest configurable frequency, captures near-DC response
• 20 Hz: Typical for full-range loudspeaker measurements
• 60-70 Hz: Example for 80 Hz passband (accounts for -3 dB at cutoff)
• >1000 Hz: Tweeter protection (above resonance frequency)

⚠️ Critical: Position Start Frequency slightly below your desired passband start to compensate for the -3 dB attenuation at cutoff, but always above driver safety limits. Setting too low can cause mechanical or thermal damage.

See Start Frequency: Driver Protection for detailed guidance on driver-specific settings.

End Frequency

Range: Auto, or a value from half an octave above Start Frequency up to the Auto value Default: Auto

End of the sweep’s full-amplitude band.

• Auto: 1/48 octave below Nyquist (~23.4 kHz at 48 kHz, ~47.3 kHz at 96 kHz) — the historical fixed behaviour. Keep it here for normal measurements.
• Lower value: the sweep is flat up to this frequency, then a sin² fade-out to min(Nyquist, ×2), and nothing above. There is no filter and no roll-off — the response stays flat right up to the End Frequency (unlike the Start Frequency, which is a −3 dB filter cutoff).

When to lower it: a driver with severe cone breakup can have a resonance so loud that you must reduce the sweep level to avoid clipping, which ruins the SNR of the useful band. Setting End Frequency below the resonance keeps it out of the capture entirely.

See End Frequency for the full explanation, including why the top is not symmetric with the bottom.

Using Default Values

These default values are applied when:

• Opening a new measurement window for the first time
• Clicking the Reset Start Freq button in the measurement window
• Creating a new driver without saved sweep parameters

Changing these defaults affects only future measurements - existing measurements retain their configured parameters.

Platform-Specific Considerations

Mac OS

• Timing stability: Excellent (< 0.01ms typical jitter)
• Timing reference: Optional when the same device is used as input and output but recommended for best accuracy
• Multi-interface: Highly reliable even with different input/output devices
• USB microphones: May require input gain compensation due to OS-enforced low gain with usb devices

Windows

• Timing stability: Variable (1-150ms jitter due to scheduler instability)
• Timing reference: Strongly Recommended (Electric or Acoustic mode) for absolute timing accuracy
• Timing inconsistency: IR arrival time may vary without timing reference
• ASIO drivers: Provide improved stability compared to WASAPI
• Measurement averaging: Works with or without timing reference (continuous capture method)

A timing reference on Windows improves absolute timing consistency when comparing measurements across drivers.

Configuration Checklist

Before starting measurements, verify:

• Input device selected and recognized
• Output device selected and functional
• Sample rate matches project settings
• Buffer size set to 256 or 512 samples
• Input gain compensation configured (if needed)
• Microphone calibration loaded (if available)
• Timing reference configured (see Reference Timing below)
• Devices not in use by other applications

Reference Timing for Sweep Measurements

Accurate timing is critical for consistent impulse response measurements, especially when capturing multiple measurements for averaging or multi-position analysis. LinFIR provides three timing reference modes to accommodate different hardware configurations and platform requirements.

Why Timing Reference Matters

Audio playback and recording systems introduce latency from:

• Digital-to-analog conversion (output)
• Analog-to-digital conversion (input)
• Buffer processing delays
• Operating system scheduler variability

Without a stable timing reference, each measurement may have slightly different latency, making:

• Driver alignment unreliable
• Directivity analysis inaccurate (time-of-flight variations)

Timing Reference Modes

None

No timing reference - relies solely on system scheduler timestamps.

How it works:

• Uses operating system audio clock
• Timestamps based on buffer callback timing
• No additional hardware or connections required

When to use:

• Mac OS systems (low scheduler jitter, typically < 0.01 ms)
• Single-shot measurements without averaging
• Simple setups where high precision isn’t critical

Limitations:

• Windows: Not recommended due to significant scheduler jitter (1-150 ms)
• IR arrival time may vary between captures

Mac OS: Reliable and accurate for most applications when using the same device as input and output. Windows: Avoid for professional work - use Electric or Acoustic mode instead.

Electric (Loopback)

Uses a physical cable connection between output and input to provide a stable timing reference.

How it works:

• Connect line output to line input with a cable
• Loopback signal captures exact OS and electrical latency
• Combined electrical and conversion latency measured
• Provides sample-accurate timing reference

Setup:

1. Connect output channel to spare input channel with audio cable
2. Select “Electric” timing reference mode in Audio Settings
3. Ensure loopback channel is not the measurement input channel, and that the sweep and the timing reference use different output channels (a measurement with both on the same output is refused)
4. Run sweep - loopback captures timing automatically

Advantages:

• Most accurate timing reference
• Platform-independent reliability
• Compensates for interface-specific latency
• Eliminates scheduler jitter completely

When to use:

• Windows systems (strongly recommended)
• Designing crossovers
• Professional applications requiring highest accuracy
• Different input/output devices (aggregate devices)

Limitations:

• Requires spare input channel for loopback
• Requires physical cable connection
• May not be available on simple USB audio interfaces

Acoustic

Uses another driver (reference driver) as a timing microphone.

How it works:

• Reference driver captures sweep output acoustically
• Provides stable timing reference based on acoustic arrival
• Measures relative timing between reference and measured driver
• Does not capture true absolute acoustic + electrical travel time

Setup:

1. Select a reference driver (must reproduce high frequencies > 5 kHz)
2. Position reference driver microphone to capture sweep clearly
3. Select “Acoustic” timing reference mode in Audio Settings
4. Reference driver captures timing on each sweep

Advantages:

• No loopback cable required
• Works when electrical loopback unavailable
• Good alternative for Windows users without spare input channels
• Suitable for multi-driver directivity measurements

When to use:

• Loopback hardware not available
• Windows systems (recommended alternative to None)

Limitations:

• Only measures relative timing (not absolute acoustic delay)
• Reference driver must reproduce high frequencies cleanly
• Acoustic path must be clear and consistent
• Less accurate than Electric mode

Important: Reference driver should be full-range or at least capable of reproducing 5-10 kHz cleanly for reliable timing detection.

Manual Offset (Preferences)

Adjustable time offset to shift the impulse response position.

• Range: ±4000 ms
• Applies to: All timing reference modes (None, Electric, Acoustic)
• Purpose: Shift the IR forward or backward in time

Use in all modes (convenient)

The offset can be used in any timing mode to position the IR at a convenient location in the window — for example, to center it or add pre-delay before a gate. This is optional and has no impact on accuracy.

⚠️ Caution: A positive offset shifts the IR earlier in the window. If the offset is too large (positive), the impulse peak may move before t=0 and be clipped. Keep the offset small enough to leave sufficient pre-delay in front of the impulse.

Use in Acoustic mode (required in certain cases)

In Acoustic mode, the timing reference is based on the acoustic arrival at the reference driver. If the reference driver is at a similar or shorter distance from the microphone than the measured driver, the IR of the measured driver will appear at time zero or even before it — making it impossible to capture the full impulse response.

In this case, a positive offset must be applied to shift the IR forward and restore the expected relative delay.

Example — reference closer than measured driver:

• Reference driver is 150 cm from the microphone
• Measured driver is 200 cm from the microphone
• Acoustic delay difference ≈ 1.5 ms (50 cm / 343 m/s)
• Apply a positive offset just above ~1.5 ms — large enough to place the impulse response after t=0 with a safety margin
• Use the same offset, mic distance, and reference driver for all measurements in the project — since the offset is constant, the relative acoustic delay between drivers is preserved, enabling accurate temporal alignment when designing crossovers

Fine-tuning:

1. Run test measurement
2. Check impulse response time alignment
3. Adjust offset in small increments
4. Re-measure to verify alignment

Platform Requirements

Windows

• Averaging: Works with or without timing reference (continuous capture eliminates jitter issues)
• Scheduler jitter: Significant (1-150 ms variable latency) for initial timing only
• Recommendation: Timing reference still recommended for absolute timing accuracy across sessions
• Single measurements: Work with “None” mode but may have timing variation

Without timing reference on Windows:

• Each measurement has unpredictable latency
• Directivity analysis compromised
• Time and phase alignment between drivers will lead to wrong results

Mac OS

• Scheduler jitter: Minimal (< 0.01 ms typical)
• Recommendation: Timing reference optional but still beneficial for best accuracy
• Multi-interface: More stable than Windows even without reference

Choosing the Right Mode

Configuration Workflow

1. Assess requirements:

   • Single measurement or averaging?
   • Mac OS or Windows?
   • Spare input channel available?

2. Select mode:

   • Windows averaging → Electric or Acoustic
   • Mac OS → None acceptable, Electric for best results
   • No loopback → Acoustic (if reference driver available)

3. Setup hardware (if Electric):

   • Connect output to spare input channel
   • Verify loopback signal path

4. Setup reference driver (if Acoustic):

   • Select driver with good high-frequency response
   • Position reference microphone to capture sweep clearly

5. Configure in Audio Settings:

   • Select timing reference mode
   • Adjust timing offset if needed (Acoustic mode)

6. Test:

   • Run test measurement
   • Verify timing consistency
   • Adjust offset if required

Timing Reference Troubleshooting

Inconsistent Measurements

Symptom: Each measurement has different arrival time Cause: No timing reference active Solution: Enable Electric or Acoustic mode

Reference Signal Not Detected

Symptom: Error message saying the timing reference could not be detected (the measurement is not aligned on a guessed timestamp instead) Cause: Loopback not connected or reference driver not capturing sweep Solution:

• Electric: Verify cable connection and input channel selection
• Acoustic: Ensure reference driver can reproduce sweep frequencies

Timing Offset Not Working

Symptom: Impulse still misaligned after offset adjustment Cause: Incorrect offset direction or magnitude Solution:

• Try opposite sign (positive ↔ negative)
• Increase offset magnitude
• Verify reference driver is capturing sweep correctly

Best Practices

• Windows users: Always use Electric or Acoustic mode for professional work
• Loopback preferred: Most accurate, eliminates all timing variables
• Acoustic alternative: Good fallback when loopback unavailable
• Test before session: Verify timing consistency with test measurements
• Document settings: Record timing mode and offset in project notes
• Consistent hardware: Use same interface for all measurements in a session

Troubleshooting

Device Not Listed

• Click “Refresh Devices” to rescan
• Ensure device drivers are installed
• Check that device isn’t in use by another application
• Try different sample rates or buffer sizes
• Restart LinFIR

Configuration Invalid

• Selected devices may have been disconnected
• Sample rate may not be supported by device
• Channel count may exceed device capabilities
• Try selecting default devices and refreshing

Audio Dropouts

• Increase buffer size
• Close other applications using audio
• Check for system resource constraints
• Update audio interface drivers

No Sound During Sweep

• Verify output device is selected and powered on
• Check system audio settings
• Ensure exclusive mode isn’t blocking access (Windows)
• Try different buffer sizes
• Verify output channel selection in IR Management window

Measurement Calibration

Correct measurements for the measurement chain itself, in two independent parts:

• Microphone: the measurement microphone’s frequency response and sensitivity, from its calibration file
• Soundcard: the audio interface’s own response, measured through a loopback cable

Each part keeps a library of calibrations in application settings, with a default for newly created drivers. Each driver then selects its own microphone and soundcard calibration (or none) in its IR window - see IR Management → Measurement Calibration.

Microphone

Manage calibration files for measurement microphones and define default calibrations for new drivers.

Overview

Microphone calibration compensates for the frequency response and sensitivity of your measurement microphone. LinFIR supports:

• Multiple calibration files stored in application settings
• Per-driver calibration selection in the IR Management window (see IR Management)
• Default calibration for newly created drivers
• Automatic application to new projects when a default is set
• Minimum-phase correction computed in real-time from calibration data

Calibration scope:

• Calibration data is embedded in project files (.lnf) for portability
• Calibrations from projects are also stored in application settings for reuse
• Each driver can use a different calibration file or none at all

Calibration File List

The main section displays all imported calibration files in a scrollable grid:

Automatic driver updates:

• When you delete a calibration file that’s currently assigned to one or more drivers, those drivers are automatically reset to “None”
• A warning toast will notify you which drivers were affected
• This prevents broken references and ensures project stability

Default Calibration

Select a calibration file to automatically apply to newly created drivers.

Dropdown selector:

• Shows “None” when no default is set
• Lists all imported calibration files by name
• Changes apply immediately to new drivers and projects

When the default is applied:

• Manual driver creation: When you add a new driver via the Drivers toolbar
• New Loudspeaker Design project: All initial drivers receive the default calibration
• New Room Calibration project: The initial measurement receives the default calibration

Note: Changing the default does not affect existing drivers in the current project or previously saved projects. Only new drivers created after the change receive the new default.

Import Calibration File

Click “Import Calibration File” to load a microphone calibration file.

Supported file format:

• Plain text files with two or three columns: Frequency (Hz), Magnitude (dB), and optionally Phase (degrees)
• Whitespace or comma-separated
• Lines starting with non-numeric characters are treated as comments/headers
• Optional header line with sensitivity factor and serial number

Note on phase column: If a third column (phase) is present, it is intentionally ignored. Phase data exported by calibration software is typically smoothed, wrapped, or otherwise processed in ways that make it unsuitable for direct use in transfer function reconstruction. Since measurement microphones exhibit predominantly minimum-phase behavior, reconstructing the phase from the magnitude alone (via Hilbert transform) is both accurate and more robust than relying on the exported phase values.

Example file format:

# Sens Factor = -2.5dB, SERNO: 41486651
# Frequency (Hz)  Magnitude (dB)
20.0              -0.5
50.0              -0.3
100.0             -0.1
200.0              0.0
500.0              0.1
1000.0             0.0
2000.0            -0.2
5000.0            -0.5
10000.0           -1.0
20000.0           -2.0

Sensitivity factor and serial number:

• Sensitivity factor (in dB) is automatically extracted from the file header if present (line containing “Sens Factor”)
• Serial number is automatically extracted if present (SERNO: field in header)
• If no sensitivity factor is found in the file, it defaults to 0 dB
• Serial number is used to differentiate calibrations with identical filenames when sharing projects

Storage: Calibration data (frequency response, sensitivity, and serial number) is embedded in project files (.lnf) to ensure portability:

• Projects can be shared without requiring recipients to have the calibration files
• When loading a project, calibrations are automatically added to application settings if not already present
• If a calibration with the same name exists:
  • With serial number: differentiated using serial number (e.g., filename (SN:7016078))
  • Without serial number: differentiated using numeric suffix (e.g., filename (1), filename (2))
• You can also manually import calibrations to your settings library for reuse across projects

How Calibration is Applied

Minimum-phase correction: LinFIR converts the imported frequency response correction into a minimum-phase FIR filter applied during IR processing.

Processing steps:

1. Frequency filtering: Bins below 5 Hz are removed to prevent phase distortion at very low frequencies
2. Interpolation: Calibration data is interpolated to match your IR’s sample rate and FFT size
3. Minimum-phase conversion: Magnitude correction is converted to minimum-phase using Hilbert transform
4. Sensitivity compensation: The sensitivity factor (dB) is applied as a gain adjustment
5. IR application: The resulting correction filter is convolved with the driver’s impulse response

Phase behavior: LinFIR does not use phase data from calibration files, even when present. Instead, the correction filter’s phase is reconstructed from the magnitude response via the Hilbert transform (cepstral minimum-phase lifting). This is intentional:

• Phase columns in calibration files are typically smoothed, wrapped, or quantized by the calibration software, and directly applying such data would introduce artifacts.
• Measurement microphones exhibit predominantly minimum-phase behavior in their frequency response deviations, meaning the phase is largely determined by the magnitude — making reconstruction accurate without measured phase.

The resulting correction therefore affects both magnitude and phase of the measurement, with the phase derived from minimum-phase theory rather than read from the file.

Per-Driver Calibration Selection

Each driver can use a different calibration file (or none). To assign a calibration to a specific driver, use the Microphone dropdown of the Measurement Calibration section in the IR Management window.

See: IR Management → Measurement Calibration

Best Practices

File organization:

• Use descriptive filenames that include microphone model and serial number
• Example: UMIK-1_SN123456.txt or Earthworks_M30_SN789.txt

Calibration maintenance:

• Import calibration files for all your measurement microphones
• Set your most frequently used microphone as the default

Project workflow:

• Assign calibrations in the IR Management window after importing measurements
• Use “None” for measurements taken with already-flat reference microphones
• Different drivers in the same project can use different calibrations (useful when combining measurements from multiple sessions)

Frequency range:

• Ensure calibration files cover the full measurement bandwidth
• LinFIR interpolates between provided points, but extrapolation beyond the calibration range may produce inaccurate results
• Most measurement microphones provide calibration data from 20 Hz to 20 kHz

Sensitivity factor:

• Use the value from your microphone’s calibration certificate
• This is typically a negative dB value (e.g., -0.5 dB, -2.4 dB)
• If unsure, use 0 dB and rely solely on the frequency response correction

Soundcard

Measure your audio interface’s own response once, and remove it from every measurement taken through it.

Overview

An audio interface is not perfectly transparent: its converters and their filters shape the magnitude and phase of everything you measure, mostly at the frequency extremes (phase rotation in the low bass, roll-off and phase shift near Nyquist). A soundcard calibration measures that response through an output-to-input loopback cable and corrects it with a regularized (Kirkeby) inverse filter:

• Magnitude and phase are both corrected, from the measured response itself
• Timing is preserved: the corrected impulse response stays where it was, so window positions and delays are unaffected
• Overall level is unchanged: only the response’s shape is corrected
• Correction range follows the sweep the calibration was measured with; outside it, measurements are left untouched
• Independent of the microphone calibration: a driver can use either, both, or neither

Default Calibration

Select a calibration to automatically assign to newly created drivers. As for microphones, changing the default does not affect existing drivers.

Measure a New Calibration

1. Connect the Output Channel directly to the Input Channel with a cable - no amplifier, no microphone
2. Select both channels, then click Start Measurement
3. When the measurement completes, give it a Name and click Save (or Discard)

The measurement uses a 5 s sweep at the sample rate currently set in Audio Settings.

Sample rate matters. A soundcard calibration only corrects measurements captured at the sample rate it was measured at: an interface’s response changes from one sample rate to another, so a calibration is never resampled. Measure one calibration per sample rate you work at. In a driver’s IR window, calibrations measured at another sample rate than the driver’s measurements are greyed out, and a warning is shown if the selected one doesn’t match.

Calibration Library

Storage: like microphone calibrations, soundcard calibrations are embedded in project files and added to application settings when a project is loaded.

How the Correction is Applied

• Applied to every measurement of the driver, on its impulse response (or its near-field / far-field spliced response) before the time window
• Computed once per measurement and reused: changing the window or other settings doesn’t recompute it
• Harmonic distortion is analyzed on the uncorrected capture, since a correction would shift the harmonics’ positions; the calibration is then applied as a gain to the fundamental and to each harmonic at its own frequency

Best Practices

• Measure through the same interface, channels, and sample rate as your driver measurements
• Measure again after changing the interface, its routing, or the sample rate you work at
• On a loopback, the interface’s own distortion is part of both the measurement and the calibration: interpret the THD of a loopback measurement with care

License

The License tab allows you to activate and manage your LinFIR license.

License Features

A valid LinFIR license unlocks advanced analysis features:

• Off-axis measurements: Import and analyze measurements at angles other than 0° (horizontal and vertical)
• Directivity Index (DI): Calculate and visualize directivity patterns across frequency
• Directivity Analysis: Access the Directivity Analysis window (sonogram and overlay plot) for detailed directivity visualization
• Spectrogram: Access the Spectrogram window (shortcut W) for time-frequency visualisation of the impulse response
• Multi-angle visualization: View multiple measurement angles simultaneously in graphs
• Pre-filter response overlays: When a FIR or IIR filter configuration window is open and in the foreground, a dashed overlay of that driver’s (or the sum’s) unfiltered response (magnitude, phase, and group delay) is displayed on the corresponding graph, making it easy to compare the raw and filtered curves without toggling filters on and off
• Auto EQ — Auto-generate: The Auto-generate toggle in the IIR Auto EQ tab (per-driver and global) automatically re-runs the EQ optimization whenever any parameter in the Auto EQ tab changes, removing the need to manually trigger a recalculation
• Harmonic distortion details: View individual harmonic levels (THD breakdown by harmonic)
• Near-field / far-field splicing: The splicing wizard, which synthesizes a full-range anechoic response from near-field element captures and far-field polar measurements (see Near-Field / Far-Field Splicing)
• Hypex FusionAmp mode: Create projects for Hypex FA series amplifiers with hardware-locked parameters and HFD export

Without a license, LinFIR operates with full functionality for on-axis measurements only (0°, 0°). The following features are unavailable without a license: off-axis analysis, directivity tools (DI, sonogram and overlay plot), spectrogram, pre-filter response overlays, Auto EQ auto-generate, harmonic distortion breakdown by harmonic, near-field / far-field splicing, and Hypex FusionAmp project creation.

Activating Your License

If you don’t have an active license, the License tab will display an activation form.

Activation Process

1. Enter your license key

   • Format: xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx
   • Provided in your purchase confirmation email

2. Enter your email address

   • Must match the email used for purchase
   • Used for license validation

3. Click “Activate License”

   • Creates a new activation tied to this specific machine
   • Each activation is identified by a unique instance name and ID

Important:

• Each license allows one activation at a time
• Each activation is machine-specific
• You must deactivate one machine before activating another
• You can manage activations from the License tab

License Status

When a license is active, the License tab displays:

Current License Information

• Customer name: License holder’s name
• Email: Email address associated with the license
• Product: LinFIR license type
• Status: Current license status (active, expired, etc.)
• Activations: Number of active machines (e.g., “1 / 1”)
• Last checked: Date and time of last validation

This Machine

• Instance Name: Automatically generated name for this activation
• Instance ID: Unique identifier for this machine’s activation

Managing License Activations

Validate License

Click “Validate License” to check the current status of your license with the licensing server.

Use cases:

• Verify license is still active
• Update activation count
• Check for license changes

Deactivate This Machine

Click “Deactivate This Machine” to free up this activation slot.

When to use:

• Switching to a different computer
• Freeing up an activation for another machine
• System reinstall or hardware change

Effect:

• Removes this machine’s activation
• Frees one activation slot
• License features will no longer be available on this machine
• You can reactivate on this or another machine later

Managing Other Activations

The “All License Activations” section displays all machines where the license is currently active.

For each activation:

• Name: Instance name
• Instance ID: Unique identifier
• Created: Activation date

Actions:

• This Machine: Marked with ✓ (cannot be deactivated from this view)
• Other machines: Click “Deactivate” to remove that activation remotely

Use case: If you activated LinFIR on a machine you no longer have access to, you can deactivate it from here to free up that slot.

Troubleshooting

Unable to reach the license server

Symptoms: Activation fails with a message about a network proxy, firewall, or “temporarily unavailable”.

Causes: The license server is protected by Cloudflare. A corporate firewall, VPN, proxy, or content filter may intercept the request and return an HTML page instead of a valid response.

Solutions:

1. Disable VPN or proxy temporarily and try again
2. Check firewall rules: ensure outgoing HTTPS connections to api.lemonsqueezy.com are not blocked
3. Try a different network: switch to a mobile hotspot or home network to rule out corporate filtering
4. Wait and retry: Cloudflare rate limiting is temporary — try again after a few minutes
5. Check your internet connection: ensure basic web browsing works before activating

“No Active License” but I activated

Symptoms: The License tab shows no active license despite a previous activation.

Solutions:

• Click “Validate License” to refresh status from the server
• Check internet connection
• The activation may have been lost (e.g., after a system reinstall) — activate again using your license key and purchase email

Too many activation attempts (rate limited)

Symptoms: Activation fails with a message about too many requests or rate limiting.

Solution: Wait a few minutes and try again. Repeated failed attempts (e.g., due to a network issue) can temporarily trigger rate limiting on the server.

Activation limit reached

Symptoms: Activation fails with “Activation limit reached”.

Cause: Your license is already activated on the maximum number of machines allowed.

Solutions:

1. Scroll down to the “All License Activations” section in the License tab
2. Identify machines you no longer use and click “Deactivate” next to them
3. Then try activating this machine again
4. If you no longer have access to the machine listed, use “Deactivate” to free up that slot remotely
5. Contact support if you need additional activation slots

Wrong email or license key

Symptoms: Activation fails with an error about an invalid key or email mismatch.

Solutions:

• Use the exact email address from your purchase confirmation (check for typos, extra spaces)
• Copy and paste the license key directly from the confirmation email to avoid transcription errors
• If you cannot find your purchase email, contact support

License deactivated on this machine

Symptoms: A toast appears saying “License has been deactivated for this machine”.

Cause: The license was revoked, expired, or deactivated remotely (e.g., from another device).

Solutions:

• Click “Validate License” to confirm the current status
• If the license is still valid, re-enter your license key and email to reactivate

License not validating

Symptoms: Background validation silently fails (no toast) or shows a warning about license status.

Solutions:

• Ensure internet connection is active
• Check firewall settings (HTTPS access to api.lemonsqueezy.com required)
• Verify your license has not expired by checking the status in the License tab
• Click “Validate License” to trigger a manual check

Privacy & Data

What is stored:

• License key (encrypted)
• Email address
• Instance name and ID (machine identifier)
• Activation timestamps

What is sent to licensing server:

• License key
• Email address
• Instance information (for activation/validation)

No project data, measurements, or filter designs are transmitted during license operations.

Rotation Table

The Rotation Table settings panel lets you connect and control a motorized turntable for automated polar measurements. Once connected, the table integrates with the IR Management window to drive the auto-scan feature, which rotates the speaker and captures impulse responses at each angle without any manual intervention.

Requires a valid LinFIR license. All rotation table features — connection, manual jog controls, position dial, and auto-scan — are available to license holders only. Without a valid license the Settings → Rotation Table tab is displayed grayed out.

Supported Hardware

LinFIR supports two families of rotation table controllers:

Pololu Tic controllers are identified automatically by their USB vendor ID. GRBL boards appear as serial ports; LinFIR probes them with a status query to confirm the firmware.

Accessing the Settings

Open Edit → Settings (or Cmd+, / Ctrl+,) and select the Rotation Table tab.

Connecting a Table

1. Port and Baud Rate

LinFIR scans available ports each time the settings panel is drawn. Use the port combobox to select the device; the detected controller type is shown next to each entry (e.g. COM3 — GRBL or /dev/cu.usbmodem… — Pololu Tic).

For GRBL boards, choose the baud rate that matches your firmware (typically 115200). Pololu Tic controllers connected over USB do not use a baud rate.

A small label below the combobox confirms the auto-detected type:

Detected: GRBL

or

Detected: Pololu Tic

2. Connect / Disconnect

Click Connect to open the connection. LinFIR probes the controller and updates the detected type if needed. A status indicator appears to the right of the Disconnect button:

Click Disconnect to cleanly close the serial/USB connection.

Position Dial

After connecting, a dial shows the current table angle:

• Red needle points to the current position (−180° to +180°)
• Cardinal labels: 0°, 90°, 180°, −90°
• ↻ button — sends a position query and refreshes the display immediately

The angle is updated automatically whenever the table moves or reports its position.

Manual Controls

GRBL Controllers

LinFIR reads the GRBL angle as the work position (machine position minus the G92 offset), so after Set 0° the displayed angle and the auto-scan targets count from the new zero, with GRBL 1.1 as with older firmware.

The Go button is disabled while the table is moving.

Pololu Tic Controllers

Tic range calibration: Set the range so that the min and max step values span a full 360° rotation. For example, if your Tic is configured for 200 steps/rev × 16 microsteps = 3200 steps/rev, use −1600 to +1600 for ±180°.

Workflow: Manual Polar Measurement

If you prefer to position the table by hand and measure angle-by-angle without auto-scan:

1. Connect the table in Settings → Rotation Table.
2. Open the IR Management window for a driver.
3. Use Go to angle to move the table to the desired off-axis angle.
4. Set the Measurement angle field in the IR window to match.
5. Click 🎤 Measure to capture.
6. Repeat for each angle.

For a fully automated workflow, see Auto-Scan with a Rotation Table in the Sweep Measurements documentation.

Tips and Best Practices

• Set zero before scanning — use “Set 0° at current position” with the speaker pointing directly at the microphone, so 0° = on-axis.
• Always capture on-axis (0°) first — LinFIR requires a 0° reference before allowing off-axis measurements. If 0° is in the scan range it will be measured first automatically.
• Use a timing reference — for accurate phase relationships between angles, configure an Electric (loopback) or Acoustic timing reference in Audio Settings. This is critical for Directivity Analysis.
• Keep the mic at constant distance and height for all angles.
• Note where the rotation axis is relative to the front baffle (depth, and any sideways offset): if it isn’t at the baffle, enter those offsets under the measurement lists of the IR Management window - see Rotation Center Correction.
• 10° step or finer gives good sonogram and DI resolution. 5° is ideal.
• For the Pololu Tic, calibrate the step range carefully: a wrong range shifts all angle positions.

Troubleshooting

Related Documentation

• Directivity Analysis — using polar measurements for DI and sonograms
• Sweep Measurements — sweep parameters and capture workflow
• Audio Setup — timing reference configuration

IR Management

The IR (Impulse Response) Management window provides comprehensive tools for obtaining, processing, and managing impulse responses for each driver or measurement position in your system.

Accessing IR Management

Loudspeaker Design Mode:

• To open: Driver → Manage IR
• Purpose: Import, capture, and process driver impulse responses at multiple angles

Room Calibration Mode:

• To open: Measurements → Manage IR
• Purpose: Capture IR at a given in room position

Window Layout

The top of the window holds the controls you use on every capture — the name field, the axis / angle selectors (Loudspeaker Design mode), and the 📁 Import, 🎤 Measure and ▶ Scan buttons.

Everything below is organized into collapsible sections, in this order:

1. Measurement Settings — sweep parameters (duration, level, start / end frequency, …). Open by default. These configure the capture buttons above, not the list below. See Sweep Measurements.
2. Measurements — the captured / imported angles, in two columns (horizontal, vertical), with per-measurement actions. Open by default. Not shown in Room Calibration mode.
3. Near-field measurements (splicing) 🔒 — element status and the Open splicing wizard… button. Collapsed by default. Not shown in Room Calibration mode. See below.
4. Measurement Calibration — per-driver microphone and soundcard calibration selection. Collapsed by default.
5. IR Windowing — time window, adaptive window, delay compensation.
6. Measurement Notes — free-text notes saved with the project. Collapsed by default.
7. IR Information — read-out of the current measurement’s properties.

Driver/Measurement Name

Manually enter a descriptive name for the impulse response.

• Label: “Driver Name” (Loudspeaker Design) or “Measurement Name” (Room Calibration)
• Purpose: Identify IRs in project and exports
• Persistence: Names are saved with the project
• Optional: Empty names default to “Driver X” or “Measurement X”
• Best practice: Use descriptive names like “Woofer_Left”, “Tweeter_RefXYZ”, “Position_Center”

Capturing IR

Button: 🎤 Measure

Triggers a measurement sweep with the configured parameters to capture the impulse response.

Features:

• Exponential sine sweep generation with configurable parameters
• Automatic harmonic distortion analysis (THD)
• Timing reference support for multi-driver measurements

For complete documentation on sweep parameters, driver protection, measurement workflow, and best practices, see Sweep Measurements.

Off-Axis Measurements 🔒

Available in: Loudspeaker Design mode only License required: Valid LinFIR license needed for off-axis measurements

Overview

Off-axis measurements enable directivity analysis by capturing impulse responses at different horizontal and vertical angles. These measurements are essential for understanding speaker radiation patterns and optimizing crossover design for consistent off-axis response.

Measurement Organization

Measurements are displayed in two columns:

• Horizontal column: Measurements with vertical angle = 0°, varying horizontal angle

  • Example: -90°, -60°, -30°, 0°, +30°, +60°, +90°

• Vertical column: Measurements with horizontal angle = 0°, varying vertical angle

  • Example: -60°, -40°, -20°, 0°, +20°, +40°, +60°

On-axis reference: The (0°, 0°) measurement appears in both columns.

Axis Selection

Before importing or capturing an off-axis measurement, select the measurement axis:

• Horizontal button: Measurements along horizontal axis (left/right)
• Vertical button: Measurements along vertical axis (up/down)

The selected axis determines whether the angle applies to horizontal or vertical position.

Measurement Angle

Field: “Measurement angle” Range: -180° to +180° Indicates: Angle for next import or capture operation

How it works:

• If Horizontal axis selected: angle applies to horizontal position (vertical = 0°)
• If Vertical axis selected: angle applies to vertical position (horizontal = 0°)
• The field shows which axis is active: “(horizontal)” or “(vertical)”

Restrictions:

• Disabled until on-axis (0°, 0°) measurement exists
• Requires valid LinFIR license
• Must import on-axis measurement first before adding off-axis angles

Per-Measurement Actions

Each measurement in the table has two action buttons:

Export Button (⬇)

Click to export this specific measurement:

1. Opens export format dialog
2. Choose WAV or TXT format
3. Option to include distortion metadata
4. File is saved with angle suffix (e.g., Driver1_v0h30.wav, Driver1_v-15h0.wav)

Delete Button (❌)

Remove a measurement from the driver:

• Off-axis measurements: Can be deleted freely
• On-axis (0°, 0°) measurement: Can only be deleted if all off-axis measurements are removed first
  • This restriction exists because the on-axis measurement serves as the reference for all filter calculations

Tooltip guidance:

• Enabled: “Delete measurement”
• Disabled: “Delete all off-axis measurements first”

Export All Button

Button: ⬇ Export All Location: Above “Measurement angle” field Purpose: Export all measurements in a single operation

How it works:

1. Click “⬇ Export All”
2. Select output directory in file dialog
3. All measurements are exported as WAV files
4. Files use consistent naming: DriverName_v{vertical}h{horizontal}.wav

File naming examples:

• On-axis: Driver 1_v0h0.wav
• Horizontal +30°: Driver 1_v0h30.wav
• Vertical -15°: Driver 1_v-15h0.wav
• Horizontal -45°: Driver 1_v0h-45.wav

Format:

• All files exported as WAV (32-bit float, mono)
• Sample rate matches original measurement or importation sample rate
• Distortion metadata option applies to all files
• No individual format selection (always WAV)

Benefits:

• Fast export of complete measurement sets
• Consistent naming for re-import or external processing
• Files can be re-imported via batch import with automatic angle detection

Use cases:

• Archiving complete measurement sets
• Sharing measurements with collaborators
• Processing measurements in external tools (FIR Designer, Matlab, etc.)

Symmetry Tool

Button: ⚖ Symmetry Location: Below each column Purpose: Automatically duplicate measurements to opposite angles

How it works:

• Horizontal Symmetry: If you have +30°, creates -30° by mirroring the measurement
• Vertical Symmetry: If you have +20°, creates -20° by mirroring the measurement

Use case: Save measurement time by capturing one side and mirroring to the other side, assuming the speaker has symmetric radiation.

Rotation Center Correction

Location: Below the measurement lists (shown once the driver has off-axis measurements) Purpose: Rebuild the off-axis measurements as if the speaker had turned about its front baffle, when the turntable axis was behind the baffle or off to the side

A polar measurement is meant to turn the speaker about a reference point on its front baffle, so the microphone stays at the same distance from it at every angle. On most turntables the rotation axis is somewhere under the cabinet instead, so off axis the microphone is actually further away and sees the driver at a wider angle than the nominal one. With the baffle 150 mm in front of the axis and the microphone at 1 m, the 90° measurement is really taken 16 cm further away - 0.47 ms late and 1.3 dB low - and 97° off the driver’s axis; at 180° it arrives 0.88 ms late. These timing errors go straight into the off-axis sums of a multi-way speaker (and into the Listening Window, the predicted in-room response and the directivity index), since the drivers’ relative timing is what makes the lobes.

Settings (in millimeters, one column per plane - the ℹ icon shows the drawing above):

The + side is the side the microphone faces at +90°. Check it once on your setup: if the microphone faces the other side at +90°, flip the signs of Lateral and Driver offset for that plane. The horizontal column applies to the measurements with a 0° vertical angle, the vertical column to those with a 0° horizontal angle - for the vertical plane, the speaker is laid on its side on the turntable, so its height becomes the “lateral” direction. Measurements off both planes are left as they are.

Apply placement to all drivers copies the depth, lateral offset and mic distance of both planes (and the on/off switch) to every driver; each driver keeps its own driver offset, which is the one value that differs between drivers. Copy/paste of driver parameters does the same.

How it works:

For each nominal angle, LinFIR rebuilds the response the microphone would have picked up on the ideal circle around the reference point:

1. Geometry, per driver. From the offsets, the real microphone positions (the 0° position turned about the turntable axis) and the ideal ones are known, so is the direction and distance at which each measurement saw the driver, and the direction and distance each nominal angle needs. The driver offset matters: correcting a woofer 20 cm below the tweeter as if it were at the reference point would still leave 20-55 µs of timing error, about 20-56° of phase at 3 kHz.
2. Distance: time of flight and level. Around the driver, changing the distance only delays and scales the sound: each measurement is moved to the target distance with a fractional delay of (target distance − measured distance) / 343 m/s and a gain of measured distance / target distance (1/r law).
3. Angle: interpolation. The response in the needed direction is interpolated between the two measurements whose directions bracket it - after both were moved to the target distance, so they arrive together and blend without comb filtering. Beyond the measured range the nearest measurement is used; a full 360° set wraps around.

The on-axis measurement is the reference and is never changed. The correction applies after the IR time window and before resampling, so everything downstream (filters, sums, directivity graphs, Listening Window, DI) uses the corrected responses; the measurement files themselves are not modified, and turning the correction off restores the original responses. Symmetry copies are recognized and treated as the mirror image of the measurement they were copied from.

Requirements and limits:

• The measurements must keep their relative timing (a fixed acoustic or loopback timing reference, the same for every angle - see Impulse Response Timing Requirements). Measurements whose delay was removed one by one can’t be corrected.
• The driver is treated as a point source on the baffle: the acoustic center’s recess behind the baffle and the driver’s own near field are ignored. At 1 m and beyond this is accurate for typical drivers; very close measurements of large drivers are less so.
• The correction moves the microphone, not the diffraction: when the microphone distance is small, the baffle diffraction seen at the real and ideal positions differ slightly, and that difference is not corrected.

Workflow for Off-Axis Measurements

1. Import or capture on-axis measurement (0°, 0°) first
2. Select axis (Horizontal or Vertical)
3. Set measurement angle (e.g., +30°)
4. Import file or capture measurement
5. Repeat for additional angles
6. Use Symmetry tool to fill opposite angles (optional)
7. Enter the turntable offsets under the measurement lists if the rotation axis wasn’t at the front baffle (see Rotation Center Correction)
8. Analyze directivity patterns in main application graphs

🔄 Automated workflow: If you have a motorized turntable (Pololu Tic or GRBL/Arduino), use the ▶ Scan button to rotate and measure all angles automatically. See Auto-Scan with a Rotation Table and Rotation Table Settings.

Near-Field / Far-Field Splicing 🔒

Available in: every mode except Room Calibration (Loudspeaker Design, Hypex FusionAmp) License required: Valid LinFIR license

Section: Collapsible header Near-field measurements (splicing), directly under the measurement list. Collapsed by default.

Splicing synthesizes a full-range anechoic response by combining near-field captures of each radiating element (driver, vent, passive radiator) with this driver’s far-field polar measurements — the Don Keele method. The near field covers the bass, where a gated far-field measurement can’t; the far field covers everything above the transition, with its real diffraction and directivity. Once applied, the spliced IR replaces the raw capture for every measured angle throughout LinFIR.

The section shows each near-field element’s capture status and the Open splicing wizard… button (enabled once an on-axis far-field measurement exists).

The wizard is documented on its own page: Near-Field / Far-Field Splicing — prerequisites, the seven steps, and step-by-step advice.

When a spliced IR is active, the IR Windowing section below applies the driver’s time window on top of the spliced (anechoic) response rather than the raw capture — a note in that section flags this.

Import from File

Load impulse responses from external files.

LinFIR works internally with impulse responses (IRs). All imported files are ultimately converted to an IR before being used for filter design. Two import paths exist:

• Direct IR formats (WAV, TXT, TMD): Files already contain time-domain impulse response samples. They are resampled to the project sample rate and loaded directly.
• Frequency-domain format (FRD): Files contain frequency response data (magnitude + phase). LinFIR reconstructs an impulse response via the FRD to IR Converter window before proceeding.

All the measurements of a driver share one sample rate. A file, an FRD conversion or a capture at a different rate than the driver’s existing measurements is refused with a message: import or re-measure it at the driver’s rate, or start a new driver. Files containing NaN or infinite samples are refused as well.

Supported Formats

WAV Files

• Channels: Mono or stereo (left channel used for stereo files)
• Bit depth: 16-bit, 24-bit, or 32-bit PCM
• Automatic resampling: The IR is resampled to match the project sample rate for processing. The original samples are preserved internally at their native rate, so changing the project sample rate later does not cause quality loss from successive resamplings. This is handled transparently in the background.

With distortion metadata:

• If WAV contains custom ‘dist’ chunk (exported from LinFIR with “Include dist.” option)
• Enables THD (Total Harmonic Distortion) computation

Without distortion metadata:

• Standard WAV import
• Only practical/cropped IR available
• THD computation not available

TXT Files

Plain text files containing impulse response samples, one value per line or space/comma-separated. LinFIR auto-detects this format.

Note: .txt files containing FRD-formatted data (3 columns: frequency, magnitude, phase) are automatically detected and processed as FRD files (see FRD Files section below).

TMD Files

MATAA (Mat’s Audio Analyzer) time-domain data format. Each line encodes a sample with its timestamp, allowing the sample rate to be derived from the data itself.

• Format: Two-column text file (time in seconds, amplitude)
• Comments: Lines starting with * are metadata/comments
• Sample rate: Inferred from the time intervals between samples. If the result is within 1% of a standard rate (44.1, 48, 88.2, 96, 176.4, 192, 384 kHz, etc.), LinFIR snaps to that exact rate to correct for the systematic sample rate error introduced by limited timestamp precision.
• Timestamp precision: Timestamps must have enough decimal places for accurate sample rate inference (e.g. 9 digits for 96 kHz). Insufficient precision can produce a slightly off sample rate that will still be corrected if it falls within the 1% snapping tolerance.
• Precision: Full floating-point precision (no bit-depth limitations)

Example TMD file:

* TMD data written by MATAA
* Measurement description
0.000000000	0.000000
0.000020833	0.125000
0.000041667	0.850000
...

CLIO Files

Time-domain impulse response exports from CLIO (Audiomatica) measurement software. These are plain .txt files with three columns.

• Format: Text file with a header row followed by data rows
  • Column 1 — Time[s]: sample timestamp in seconds
  • Column 2 — Real[Pa]: signal amplitude (used as the IR)
  • Column 3 — Imag[Pa]: imaginary part (ignored by LinFIR)
• Header line: the first non-empty line must start with Time (case-insensitive)
• Sample rate: inferred from the time intervals between samples, then snapped to the nearest standard rate within 1% tolerance (same algorithm as TMD)
• Precision: full floating-point precision

Example CLIO file:

Time[s]         Real[Pa]        Imag[Pa]
0.000000E+00    -4.790139E-03   -4.583853E-10
2.083333E-05     4.526122E-03   -8.076313E-10
4.166667E-05     3.454284E-04    1.007265E-10
...

FRD Files

Frequency response data files (commonly exported by REW, VituixCAD, or provided by manufacturers). These files do not contain an impulse response directly—LinFIR reconstructs one via the FRD to IR Converter window (see dedicated section below).

Format:

• 3 columns per line: Frequency (Hz), Magnitude (dB), Phase (degrees)
• Text-based, space or tab delimited
• Can use either .frd extension or .txt extension (auto-detected)

Processing:

• Automatic interpolation to required frequency points using PCHIP (Piecewise Cubic Hermite Interpolating Polynomial)
• Configurable magnitude extrapolation for DC and high-frequency regions
• Phase handling with group delay compensation

Extrapolation Modes:

When importing FRD files, LinFIR opens the FRD to IR Converter window with the following extrapolation controls:

• DC Extrapolation: Controls how magnitude is extended below the lowest measured frequency

  • Constant: Flat extrapolation (holds the lowest measured value)
  • Roll-off: Applies a 12 dB/octave high-pass roll-off for realistic loudspeaker behavior
  • Default: Roll-off

• High Frequency Extrapolation: Controls how magnitude is extended above the highest measured frequency toward Nyquist

  • Constant: Flat extrapolation (holds the highest measured value)
  • Roll-off: Applies a 12 dB/octave low-pass roll-off, automatically compensates for upward slopes
  • Default: Constant

Benefits of Roll-off mode:

• More realistic impulse responses from incomplete FRD data
• DC roll-off mimics natural loudspeaker low-frequency limitations
• HF roll-off prevents artifacts from measurements that don’t extend to Nyquist
• Adaptive: compensates for incorrect slope trends in measured data

Important notes:

• FRD files are sparse (often logarithmically spaced frequency points)
• Requires complex interpolation, potentially introducing artifacts
• Generally do not contain the time of flight required to properly time align drivers
• Recommendation: Prefer importing impulse responses (WAV) whenever possible

Automatic Noise Truncation

Once an impulse response has been imported or reconstructed (from any format), LinFIR automatically truncates it to optimize file length and processing performance. This is an internal post-processing step applied uniformly to all IR data before it is stored in the project.

Truncation Algorithm:

1. Peak detection: Locates the main impulse peak (maximum absolute value in the IR)
2. 5-second limit: Truncates the IR to 5 seconds maximum after the peak
   • Corresponds to approximately 1.7 km acoustic distance
   • Sufficient for capturing full decay in most practical scenarios
3. Noise threshold: Removes trailing samples below 0.1% of peak amplitude
   • Adaptive threshold based on the actual signal level
   • Eliminates noise floor without affecting useful signal
4. Minimum duration: Always preserves at least 1.0 seconds of the IR
   • Ensures sufficient data for low-frequency analysis

Benefits:

• Reduces processing load for FFT operations and convolutions
• Decreases file sizes for project saving and export
• Removes unnecessary noise floor that doesn’t contribute to filter design

What gets removed:

• Excessive silence beyond 5 seconds after the main impulse
• Noise floor below the 0.1% threshold (≈ -60 dB relative to peak)
• Applies to both anechoic and in-room measurements

What is preserved:

• The complete main impulse and initial decay
• At least 1.0 seconds of IR data regardless of noise level
• All frequency content relevant for crossover design
• Sufficient time-domain information for phase analysis

Impulse Response Timing Requirements

Peak delay limit: LinFIR rejects impulse responses where the main peak is located more than 200 ms after the start of the file.

Why this limit exists:

LinFIR performs extensive signal processing operations—FFT transformations, convolutions, resampling, and real-time filter preview—that must remain responsive. These operations scale with impulse response length, and excessive zero padding before the actual signal significantly impacts performance without adding useful information.

Physical context:

A 200 ms delay corresponds to approximately 68 meters of acoustic travel distance. In practical loudspeaker measurement and room calibration scenarios, such excessive delays indicate that no timing reference method was used during capture (loopback trigger, electrical reference, or pre-alignment in the measurement software).

Impact on workflow:

Without proper timing reference, relative time alignment between drivers becomes unreliable, as the arbitrary delays vary unpredictably between measurements. This compromises crossover phase coherence and summation accuracy—core functions of LinFIR.

Best practice:

When capturing impulse responses, use measurement software that provides:

• Loopback triggering (audio interface output → input reference)
• Electrical reference triggering (synchronized start)
• Pre-alignment features that position the IR peak near the beginning of the recording

These methods ensure minimal file length while preserving the precise time-of-flight information needed for accurate driver alignment and filter design.

If your measurement contains a valid signal buried in initial silence, pre-process the file to remove leading zeros before importing into LinFIR.

Import Signal Level Check

When importing a WAV/TXT impulse response that wasn’t captured by LinFIR’s own sweep engine, the frequency response level is checked: if its peak magnitude is below -25 dB, that’s usually a sign of a poor signal-to-noise ratio in the original measurement.

What happens:

• An always-on-top Low Signal Level Warning dialog appears, offering Keep Anyway or Discard
• Keep Anyway imports the file as usual
• Discard cancels the import — nothing is stored, and you can re-import a different file
• The file is not imported until you answer the dialog

Why this matters: correction filter quality strongly depends on measurement quality. A low signal-to-noise ratio means noisier data, which can produce inaccurate correction filters and skew phase visualization — see Quality Validation for the same check applied to LinFIR’s own captures.

💡 Settings → General → Always keep low-SNR measurements skips this dialog and imports low-level files automatically, with only a toast as a reminder. See General Settings.

This check only applies to files not captured via LinFIR’s own sweep engine — an IR that already carries LinFIR’s own capture metadata skips it, since its quality was already validated at capture time. It’s independent of the peak delay check above: a file can fail either, both, or neither.

Import Workflow

1. Click “📁 Import” button
2. Select WAV, TXT or FRD file
3. File is automatically detected and processed:
   • WAV: Resampled to project sample rate, metadata extracted
   • TXT (impulse samples): Either auto-imported (with metadata) or user prompted for sample rate (the prompt proposes the driver’s sample rate)
   • TXT/FRD (frequency data): Opens FRD to IR Converter window
4. IR appears in impulse response graph
5. Manually enter IR name in text field
6. Apply windowing if necessary (see Time Windowing section)

Special case for FRD files: When importing .frd or FRD-formatted .txt files, the FRD to IR Converter window opens automatically. See the dedicated section below for details.

⚠️ Display Mode Check: If the imported impulse response does not appear in the graphs, verify that Drivers/Measurements display mode is selected in the graph toolbar (not FIR, IIR, or FIR+IIR modes which only show filter responses). The graph title and legend clearly indicate which mode is active.

Batch Import of Multiple Measurements

Available in: Loudspeaker Design mode (off-axis measurements) License required: Valid LinFIR license needed for off-axis measurements

Overview

Batch import enables importing multiple measurement files simultaneously with automatic angle detection from filenames. This significantly speeds up the workflow when importing multi-angle measurement sets from tools like VituixCAD, ARTA, or REW.

Filename Angle Parsing

LinFIR automatically detects measurement angles from filenames using conventions from other measurement software.

Important: Angle patterns must be at the end of the filename (before extension). If the axis marker (h/v) has a prefix format (like h30), it must not be preceded by a letter. This prevents false detection in cases like “Driverh30” while allowing “SIG270-4h30” or “Driver 2_h30”.

Format with single axis specification (H = Horizontal, V = Vertical):

• H30, h30, H-20 → Horizontal axis, angles: +30°, +30°, -20°
• V15, v15, V-45 → Vertical axis, angles: +15°, +15°, -45°
• 30H, 30h, -20H → Horizontal axis, angles: +30°, +30°, -20°
• 15V, 15v, -45V → Vertical axis, angles: +15°, +15°, -45°

Format without axis (@ marker):

• @30, @-20, Tweeter@15 → Angle detected, axis must be specified by user
• When files use @ notation, a dialog prompts you to select Horizontal or Vertical axis

MF 2D format (both H and V in filename):

• v0h30, V0H30, v000h030 → Horizontal +30° (horizontal plane: V=0)
• v30h0, V30H0, v030h000 → Vertical +30° (vertical plane: H=0)
• h30v0, H30V0, h030v000 → Horizontal +30° (order-independent)
• h0v30, H0V30, h00v030 → Vertical +30° (order-independent)
• v0h0, V0H0, v000h000 → On-axis (0°, 0°)
• v30h50 → Rejected (out-of-plane measurement)

Note: Leading zeros are supported (e.g., 030 = 30)

Note on MF 2D format: Only measurements in a plane are accepted (one angle must be 0). Files with both H and V non-zero are ignored as they represent oblique measurements that can’t be represented in LinFIR’s horizontal/vertical coordinate system. The order of h/v doesn’t matter.

Examples of valid filenames:

• Tweeter_h0.wav → On-axis (0°, 0°)
• Tweeter_h30.wav → Horizontal +30°
• Tweeter_v-20.frd → Vertical -20°
• Woofer@15.txt → Angle +15° (user selects axis)
• measurement_30h.wav → Horizontal +30°
• Left_-45v.frd → Vertical -45°
• Driver_v0h45.frd → Horizontal +45° (MF 2D format)
• Tweeter_v20h0.txt → Vertical +20° (MF 2D format)
• SIG270-4h30.frd → Horizontal +30° (number before axis letter is OK)
• SIG270-4v0h30.frd → Horizontal +30° (MF 2D with number prefix)

Examples of rejected filenames:

• Driverh30.wav → Rejected (letter “r” directly before “h”)
• Testh30v0.frd → Rejected (letter “t” directly before “h”)

Files without angle markers:

• A single file goes to the angle currently selected in the window’s angle controls, or to the on-axis (0°, 0°) slot while the driver has no on-axis measurement yet
• In a multi-file batch, it is imported as the on-axis (0°, 0°) measurement

Batch Import Workflow

1. Multi-File Selection

1. Click “📁 Import” button
2. Select multiple files in the file dialog (Ctrl+Click or Shift+Click)
3. Click Open

2. Automatic Processing

LinFIR automatically analyzes all selected files:

If all files have explicit axis (H/V in filename):

• Files are processed immediately
• Each file imported at the angle specified in its filename
• No user interaction needed

If files use @ notation (ambiguous axis):

• Axis Selection Dialog appears
• Select either Horizontal or Vertical axis
• Click Continue
• Selected axis applies to all @ files in the batch
• Closing the dialog cancels the whole batch import

3. FRD Converter Integration

If the batch includes FRD files:

1. LinFIR opens the FRD Converter for the on-axis FRD file (if present)
2. Configure conversion parameters (sample rate, delay, extrapolation modes)
3. Click “Use This Impulse Response”
4. All remaining FRD files automatically use the same parameters
5. No need to configure each file individually

How it works:

• Parameters are configured using the on-axis measurement (0°, 0°)
• Same delay and extrapolation settings applied to all off-axis FRD files
• Ensures consistent conversion across the measurement set

⚠️ Important Limitation - Missing Relative Time Delays:

FRD files contain frequency response data only (magnitude and phase) without encoding the relative time delays between different measurement angles. When importing multiple FRD files at different angles, all files use the same delay parameter configured on the on-axis measurement.

Why this matters for off-axis measurements:

In real loudspeakers, off-axis measurements have different acoustic path lengths and interference patterns compared to on-axis:

• Sound from tweeter and woofer arrives at different relative times depending on measurement angle
• These time differences create frequency-dependent interference (comb filtering)
• Proper time alignment between drivers varies with angle

Consequence:

• Batch-imported FRD files will have identical relative timing at all angles
• Phase relationships between drivers won’t reflect actual acoustic behavior
• Directivity analysis and crossover optimization will be inaccurate
• Simulated off-axis response won’t match real-world interference patterns

Recommendation:

• Prefer WAV impulse response files for multi-angle measurements whenever possible
• WAV files preserve the actual time-of-flight at each angle
• Only use FRD batch import for preliminary analysis or when WAV files aren’t available
• If using FRD files, be aware that off-axis crossover behavior won’t be accurately modeled

Valid use cases for batch FRD import:

• Quick magnitude response comparison across angles
• Preliminary directivity visualization
• Single-driver systems where relative timing isn’t critical
• When no impulse response measurements are available

Practical Example

Scenario: You have a set of horizontal measurements exported from REW:

Tweeter_h0.frd
Tweeter_h15.frd
Tweeter_h30.frd
Tweeter_h45.frd
Tweeter_h60.frd

Workflow:

1. Select all 5 files in file dialog
2. FRD Converter opens for Tweeter_H0.frd (on-axis)
3. Set sample rate to 96000 Hz, delay to 5 ms
4. Choose DC extrapolation: Roll-off
5. Click “Use This Impulse Response”
6. Files H15, H30, H45, H60 automatically converted with same settings
7. All 5 measurements appear in the horizontal column

Result:

• All angles have same delay (5 ms) configured on on-axis measurement
• ⚠️ Relative time-of-flight differences between angles are lost
• For accurate crossover modeling, prefer WAV impulse response files

Limitations

• Batch import only works with files having consistent format (all WAV, all FRD, or mixed)
• Files must follow supported angle naming conventions
• On-axis measurement (0°, 0°) should be included when importing FRD files for optimal parameter configuration
• If mixing WAV and FRD formats, FRD files must include on-axis or parameters will use defaults
• FRD files don’t preserve relative time delays between angles—prefer WAV files for multi-angle measurements

Tips

Consistent naming convention:

• Use descriptive base names: Tweeter_h30.wav instead of just h30.wav
• Include driver name in filename for easier identification
• Use lowercase h/v for consistency with LinFIR exports
• MF 2D format is order-independent: v0h30 = h30v0

Large measurement sets:

• Import horizontal and vertical measurements separately if you have both
• Use Symmetry tool after batch import to fill missing opposite angles
• Verify measurement table after import to ensure correct angle assignment

FRD files:

• ⚠️ Not recommended for off-axis measurements (missing relative time delays)
• Always include on-axis (H0 or V0) measurement in batch for parameter reference
• Configure extrapolation modes carefully for first file—settings propagate to all
• Verify impulse response quality after batch conversion
• Consider this a preliminary workflow—prefer WAV files for accurate directivity analysis

Recommended format priority:

1. WAV files: Best option—preserves all timing information
2. TXT with metadata: Acceptable if exported from LinFIR
3. FRD files: Only for magnitude analysis—timing not preserved

FRD to IR Converter

When importing FRD files, LinFIR opens a dedicated conversion window with advanced tools for interpolation, visualization, and quality assessment.

Overview

The FRD to IR Converter transforms sparse frequency-domain data into time-domain impulse responses using sophisticated interpolation and extrapolation techniques. It provides real-time visualization and quality metrics to help you assess the conversion quality.

Conversion Parameters

Sample Rate

Options: 44.1 kHz, 48 kHz, 88.2 kHz, 96 kHz, 176.4 kHz, 192 kHz Default: 96 kHz

Target sampling frequency for the resulting impulse response.

Considerations:

• Higher sample rates provide better frequency resolution
• Must match or exceed your project’s sample rate
• FRD data is interpolated to fill all frequency bins from DC up to Nyquist

Delay Compensation

Range: 0.0 to 100.0 ms Default: 5.0 ms

Time offset applied to the impulse response to allow for pre-ringing visualization and processing.

Purpose:

• Provides time buffer before the main impulse
• Essential when FRD data contains non-minimum phase characteristics
• Allows visualization of pre-ringing artifacts

Warning indicator: If pre-ringing duration exceeds the delay setting, a warning appears suggesting either:

• Increase delay to accommodate pre-ringing
• Enable Minimum Phase transformation

Minimum Phase Transformation

Default: Disabled

Applies Hilbert transform to convert the impulse response to minimum phase, eliminating all pre-ringing.

When to use:

• FRD data shows significant pre-ringing (> 5% energy before peak)
• Phase data is corrupted, smoothed, or contains wrapping errors
• You need a causal, physically realizable impulse response
• Time-domain alignment is critical

Effects:

• Removes all energy before the main peak
• Preserves magnitude response
• Changes phase response to minimum phase equivalent
• Results in a causal system (no pre-ringing)

Note: Pre-ringing in FRD-derived impulses is often an artifact of measurement or file creation, not a physical characteristic of the device.

DC Extrapolation

Modes: Constant, Roll-off Default: Roll-off

Controls how magnitude is extended below the lowest measured frequency toward DC (0 Hz).

Constant mode:

• Flat extrapolation using the lowest measured value
• Use when FRD data already reaches very low frequencies

Roll-off mode (recommended):

• Applies 12 dB/octave high-pass characteristic
• Mimics natural loudspeaker low-frequency roll-off
• More physically realistic for incomplete low-frequency data

High Frequency Extrapolation

Modes: Constant, Roll-off Default: Constant

Controls how magnitude is extended above the highest measured frequency toward Nyquist.

Constant mode (recommended):

• Flat extrapolation using the highest measured value
• Maintains adequate signal level to Nyquist
• Minimizes noise amplification

Roll-off mode:

• Applies 12 dB/octave low-pass characteristic
• Automatically compensates for upward slope trends
• Use when FRD data shows unrealistic high-frequency behavior

Driver Offset Compensation

The FRD to IR Converter provides Driver Offset controls to compensate for the physical displacement of the driver relative to the reference point (typically the front baffle center or another driver).

Parameters:

• X (Horizontal): Driver offset in the horizontal direction, in millimeters, positive toward positive horizontal angles (to the right when facing the baffle, as in the diagram below)
• Y (Vertical): Driver offset in the vertical direction, in millimeters, positive up, toward positive vertical angles
• Located below the Extrapolation controls in the converter window

If your measurements count positive horizontal angles toward the left when facing the baffle, enter X with the opposite sign: what matters is that a positive X is on the side the microphone is on at positive angles.

Coordinate System and Geometry:

Reference Point (Origin):

The reference point (X=0, Y=0) represents the desired acoustic axis of the loudspeaker system. In the diagram example:

• The tweeter is positioned at the origin (0, 0) — it defines the acoustic axis
• The woofer is offset by -200 mm vertically (0, -200)

Note for designers: The choice of origin is not necessarily the center of a driver — it’s an engineering decision. You could choose:

• The acoustic center of a specific driver (as shown in the example)
• A point between drivers (e.g., midway between tweeter and woofer)
• Any other reference point that makes sense for your design goals

What matters is consistency: all driver offsets are measured relative to this chosen origin.

Important: For off-axis measurements, only a driver with offset (woofer in the example) will have a geometric delay added or subtracted. A driver at the origin (tweeter in the example) has no delay added or subtracted for off-axis angles — it remains the reference. This delay difference is then used to calculate interference patterns between drivers at off-axis angles.

How it works:

The driver offset introduces a geometric time-of-flight difference for off-axis measurements, calculated with the far-field approximation (parallel rays): the offset’s component along the direction of the microphone, divided by the speed of sound.

$$\Delta t = -\frac{x \cdot \sin\theta \cdot \cos\varphi + y \cdot \sin\varphi}{c}$$

Where:

• \(x\) = horizontal driver offset (mm), positive toward positive horizontal angles
• \(y\) = vertical driver offset (mm), positive up
• \(\theta\) = horizontal measurement angle (degrees)
• \(\varphi\) = vertical measurement angle (degrees), positive up
• \(c\) = speed of sound ≈ 343,000 mm/s

The sign matters: a driver on the same side of the reference point as the microphone is closer to it, so its sound arrives earlier (negative \(\Delta t\)); a driver on the opposite side is farther and arrives later. This is what makes the off-axis lobing of a multi-way speaker asymmetric between, say, above and below the axis.

In practice (measurements in a single plane):

Since LinFIR only supports measurements in the horizontal plane (\(\theta \neq 0\), \(\varphi = 0\)) or vertical plane (\(\theta = 0\), \(\varphi \neq 0\)), the formula simplifies:

• Horizontal plane: \(\Delta t = -\frac{x \cdot \sin\theta}{c}\) (only x offset matters)
• Vertical plane: \(\Delta t = -\frac{y \cdot \sin\varphi}{c}\) (only y offset matters)

This geometric delay is added to the user-configured Delay before converting the FRD data to an impulse response. Since it can be negative, the Delay must be at least the largest advance the offset can cause, \(\sqrt{x^2 + y^2} / c\) (0.58 ms for a 200 mm offset): otherwise the angles where the driver is closest would need to start before 0 ms: their delay is clamped at 0 ms and their timing is then wrong. A ⚠ icon next to the offsets warns when the Delay is too short. The default 5 ms Delay covers offsets up to 1.7 m.

Example:

• Tweeter at origin (0, 0): no geometric delay applied at any angle
• Woofer offset: X = 0 mm, Y = -200 mm (200 mm below the tweeter)
• Measurement at vertical -30° (microphone 30° below the axis, on the woofer’s side): the woofer is closer, \(\Delta t = -200 \times \sin(30°)/343000 \approx -0.291\) ms - it arrives 0.291 ms earlier
• Measurement at vertical +30° (microphone above the axis): the woofer is farther, \(\Delta t \approx +0.291\) ms - it arrives 0.291 ms later
• This delay difference creates the correct interference pattern between tweeter and woofer on both sides of the axis

Purpose and Far-Field Approximation:

⚠️ Purpose: Driver offsets are used to predict far-field interference patterns between drivers when the listener is in the far field (listening distance >> driver spacing).

Key insight: Since typical FRD data has its phase recalculated (constant delay removed), the original measurement conditions don’t matter. What matters is modeling the acoustic geometry at the listening position.

When to use driver offsets:

• Listening distance ≥ 10× driver spacing (far-field condition for the listener)
• Example: For drivers spaced 200 mm apart, listener at ≥ 2 meters
• Typical home listening (2-4m) and studio monitoring (2-3m) scenarios
• Predicting directivity and off-axis response in real-world use

Geometric assumption:

The formula assumes acoustic rays are parallel at the listening position:

• Valid when listener distance >> driver spacing
• Path difference depends only on driver offset and angle

Near-field listening considerations:

For close listening distances (< 1m), the parallel ray approximation becomes less accurate for the direct sound, and the predicted response will deviate more from reality as you get closer to the drivers.

However, the approximation remains valid for the reflected sound field since walls, floor, and ceiling are typically far enough that the acoustic path length is much larger than driver spacing. This is important because off-axis directivity affects how sound interacts with room boundaries.

Best practice:

• Use driver offsets in most scenarios, including near-field listening
• The approximation is excellent for far-field listening (≥ 2m)
• For near-field use (< 1m), understand that direct sound predictions will be approximate, but reflected field modeling remains valid
• Leaving offsets at 0 mm corresponds to a coaxial driver configuration (all drivers at the same point)

Display Parameters

These parameters control visualization only and do not affect the conversion.

IR Start / Stop Time

Range: 0.0 ms to impulse duration Purpose: Zoom into specific time region of the impulse response graph

Useful for examining:

• Pre-ringing characteristics
• Main peak detail

Frequency Min / Max

Range: 1 Hz to 48000 Hz Purpose: Set frequency range for magnitude and phase graphs

Allows focusing on:

• Specific frequency bands of interest
• Problem areas in the response
• Comparison between original and interpolated data

Wrap Phase

Default: Disabled

Toggles between wrapped (-180° to +180°) and unwrapped phase display.

Wrapped:

• Phase limited to -180° to +180° range
• Shows discontinuities at ±180° boundaries
• Familiar format for some users

Unwrapped:

• Continuous phase (can exceed ±180°)
• Shows true phase accumulation
• Better for assessing group delay trends

Remove Delay Phase

Default: Disabled

Removes the linear phase component corresponding to the delay setting.

When enabled:

• Subtracts linear phase = -360° × f × delay
• Reveals underlying phase structure
• Useful for comparing final phase response with the original one

Data Quality Assessment

The converter displays comprehensive quality metrics to help you evaluate the FRD data.

Coverage Metric

Format: “X / Y frequency points (Z% coverage)”

• X: Number of frequency points in FRD file (within 0 Hz to Nyquist)
• Y: Expected number of points for complete FFT coverage
• Z: Coverage percentage = (X / Y) × 100

Quality indicators:

• ✅ Good (≥ 50%): Green - Sufficient data density
• ⚠️ Moderate (20-50%): Yellow - Acceptable but sparse
• ❌ Low (< 20%): Red - Very sparse, heavy interpolation required

FFT bin calculation: Based on sample rate and FFT size (4× next power of 2)

Info Button - Detailed Diagnostics

Click the ℹ Info button to open the Data Quality Information window with detailed analysis.

Missing DC (0 Hz):

• Impact: Low-frequency extrapolation required
• Consequence: Uncertainty in bass response below lowest measured frequency
• Solution: Use Roll-off extrapolation for DC

Missing Nyquist Data:

• Impact: High-frequency extrapolation required
• Consequence: Uncertainty in high-frequency response
• Recommended: FRD should extend to at least Nyquist frequency

Sparse Frequency Data:

• Threshold: < 50% coverage
• Impact: Heavy interpolation between measured points
• Risk: Artifacts from PCHIP interpolation of widely-spaced points
• Note: Shows exact percentage of missing data points

Pre-ringing Detection:

Analyzed automatically after conversion. Three severity levels:

• Minor (0.1% - 0.5% energy before peak):

  • Blue indicator
  • May indicate slight phase smoothing or measurement artifacts
  • Usually acceptable for most applications

• Moderate (0.5% - 1.0% energy before peak):

  • Yellow/Orange warning
  • Phase data likely smoothed or contains wrapping errors
  • Consider minimum phase transformation

• Significant (≥ 1.0% energy before peak):

  • Red warning
  • Severe phase corruption or non-physical characteristics
  • Strongly recommended: Enable Minimum Phase transformation
  • Will affect transient response and phase correction accuracy

Pre-ring duration: Time span from first significant energy to main peak

Physical interpretation: Real loudspeakers cannot produce energy before the stimulus arrives. Pre-ringing in FRD files indicates:

• Phase data smoothing in measurement software
• Phase wrapping/unwrapping errors
• File export artifacts
• Non-minimum phase processing applied to data

Impact Warnings

The Data Quality Information window provides context on how issues affect the conversion:

Missing Data Effects:

• Artifacts in reconstructed impulse response
• Estimated information through interpolation/extrapolation
• Final impulse may not fully represent original system

Best Practices for FRD Files:

• Include DC (0 Hz) measurement point
• Extend data to at least Nyquist frequency
• Aim for > 50% frequency coverage
• Prefer linear frequency spacing over logarithmic
• Avoid phase smoothing in export settings

Time-Domain Preference:

The converter emphasizes: When possible, import impulse responses (WAV files) instead of FRD.

Advantages of time-domain data:

• No interpolation artifacts
• Preserves time-of-flight information (with proper timing reference)
• Essential for proper driver alignment in crossover design
• Natural representation of system behavior

FRD limitations for crossover design:

• Often lacks time-of-flight data
• Driver motors may not be coplanar
• Diaphragm geometry affects acoustic center
• Horn loading adds group delay
• Relative delays between drivers are essential

Visualization Graphs

The converter displays three synchronized graphs:

1. FRD Data (Original + Interpolated):

• Left: Magnitude response
• Right: Phase response
• Red curve: Original FRD measurements
• Dashed blue curve: PCHIP interpolated curve

2. Impulse Response:

• Reconstructed time-domain impulse
• Zoom using IR Start/Stop controls
• Shows pre-ringing and overall IR structure

3. Reconstructed Frequency Response:

• Left: Magnitude computed from impulse via FFT
• Right: Phase computed from impulse via FFT
• Validation: Should closely match interpolated FRD (with Remove Delay Phase button enabled)
• Differences indicate reconstruction issues

Workflow

1. Import FRD file - Converter window opens automatically
2. Review quality metrics - Check coverage and warnings
3. Adjust extrapolation - Set DC and HF modes as needed
4. Set sample rate - Match or exceed project sample rate
5. Configure delay - Ensure adequate pre-ringing buffer
6. Enable minimum phase - If pre-ringing > 1% or phase corrupted
7. Inspect graphs - Verify interpolation and reconstruction quality
8. Click “Use This Impulse Response” - Apply to selected measurement angle

The converted impulse is automatically stored in the driver’s measurement at the specified horizontal/vertical angle.

Measurement Calibration

Correct this driver’s measurements for the measurement chain itself: a microphone calibration and a soundcard calibration, selected independently (either, both, or none).

Calibration management:

• Calibration libraries are managed in Settings → Measurement Calibration (see Measurement Calibration)
• A default calibration of each kind can be set for new drivers
• Each driver/measurement can use different calibrations or none at all

Microphone

Compensates for:

• Frequency response deviation from ideal flat response
• Sensitivity factor (dB conversion from microphone output to SPL)

Options:

• None: No calibration applied (use for reference microphones or pre-calibrated measurements)
• [Calibration file names]: All imported calibration files from Settings → Measurement Calibration

Soundcard

Compensates for the audio interface’s own response - magnitude and phase - measured through a loopback cable (see Measurement Calibration → Soundcard).

Options:

• None: No soundcard correction
• [Calibration names]: All calibrations measured in Settings → Measurement Calibration. Calibrations measured at another sample rate than this driver’s measurements are greyed out: a soundcard calibration only corrects measurements captured at its own sample rate. If the selected one doesn’t match, a warning is shown and it is not applied.

Behavior

• Selection is saved with the project file, calibration data included
• Changing the selection immediately recomputes the driver’s measurements and all filter responses

How Calibration is Applied

Processing pipeline:

1. Calibration corrections applied (if selected) to each measurement’s impulse response, or to its spliced response - soundcard first, then microphone:
   • Soundcard: regularized inverse of the loopback response, timing preserved
   • Microphone: minimum-phase FIR filter created from frequency response data, sensitivity factor (dB) applied as gain adjustment, frequencies below 5 Hz filtered out before processing
2. Measurement windowed (see Time Windowing)
3. Corrected IR used for all subsequent filtering (crossovers, correction filters, etc.)

Phase behavior:

• Microphone correction uses minimum-phase reconstruction, affecting both magnitude and phase - physically accurate for measurement microphones, which exhibit mainly minimum-phase behavior
• Soundcard correction uses the interface’s measured phase

Real-time updates:

• Corrections are computed once per measurement and reused when you change the windowing parameters
• Filter responses update automatically when you change the calibration selection

When to Use Calibrations

Apply a microphone calibration when:

• Using consumer or measurement microphones with non-flat response
• You have calibration data from the manufacturer or third-party calibration service
• Measurements were taken with the same microphone used for the calibration file

Apply a soundcard calibration when:

• Measurements were taken through the interface, channels, and sample rate the calibration was measured with

Use “None” when:

• Measurements were taken with an already-flat reference microphone
• Importing measurements already compensated externally

Per-Driver vs. Default Calibration

Per-driver assignment (this window):

• Overrides any default set in Settings
• Useful when combining measurements from different microphones or interfaces in one project
• Saved with the project file

Default calibration (Settings → Measurement Calibration):

• Automatically applied to newly created drivers/measurements
• Does not affect existing drivers
• See Measurement Calibration for configuration

Time Windowing

Isolate the direct sound from an impulse response and exclude reflections or noise.

Note: Time windowing is not available in Room Calibration mode. This is intentional: Room Calibration mode is designed to capture the full speaker-plus-room response, including reflections and room decay. Applying a time window would produce an anechoic-like impulse response — corrections derived from it would be wrong for the actual listening environment — what you hear is the direct sound from the speakers combined with all the reflections that follow, and the correction filters must account for that full response.

Parameters

Start Time

Beginning of the time window in milliseconds.

• Purpose: Remove pre-ringing or early noise
• Typical: 0.0 to 5.0 ms before main peak
• Effect: Sets the beginning of the extracted portion

Stop Time

End of the time window in milliseconds.

• Purpose: Exclude late reflections or room decay
• Typical: 5.0 to 100.0 ms after main peak
• Effect: Sets the end of the extracted portion

Duration Display

Window duration (Stop Time - Start Time).

Start and Stop are independent — neither one drags the other. If you bring them within a taper of each other (or cross them), the fade-in and the fade-out simply multiply: the window becomes a smooth sub-unity bump rather than a malformed shape. With the default 0.5 ms taper, 4.0 ms / 4.5 ms already gives a clean triangular window (the fade-in reaches full amplitude exactly where the fade-out starts).

Time Reference

Start Time and Stop Time are defined on the raw impulse response — the measurement exactly as captured or imported, before any gain, IIR/FIR filtering, physical time delay, or delay compensation. A Start of 0 means “keep the measurement from its very first sample”; the values never change meaning as you build up filters and delays on the driver.

The Impulse Response and Step Response graphs, on the other hand, show the processed driver signal, and the red window-envelope preview (see Preview) is drawn on top of that processed curve. Since processing shifts the signal along the time axis, the window’s position on the graph is offset from the number you entered:

For example, with a 2 ms Time Delay on the driver and no delay compensation, a window Start of 0 ms shows its fade-in at the 2 ms mark: the window has not moved — the displayed impulse has been delayed by 2 ms, and the overlay follows it so the two stay aligned. Remove every filter and set both delays to zero and the overlay lines up exactly with the entered value.

The overlay is deliberately drawn relative to the curve beneath it, so it always shows what the window does to the response you are actually looking at rather than to an abstract time axis. Note also that with Adaptive IR display on (Graphs settings), the graph zooms to the signal and may crop the region before the impulse — if your Start sits in that cropped area, the overlay simply enters from the left edge already open.

No need to do the arithmetic yourself: hovering Start or Stop shows exactly where that value lands on the graph, with the shift broken down by cause (FIR delay / Time Delay / Delay Compensation). Whenever that shift is non-zero, an On Graph row also appears right under Window Duration, showing both shifted values at a glance — e.g. 2.00 ➡ 12.00 ms for a Start/Stop of 0 / 10 ms with a 2 ms Time Delay. This readout always uses the raw Start/Stop values as the source of truth (they never change meaning), it only translates where they currently show up.

Auto Set Window

Click “Auto Set Window” to automatically detect and configure optimal window boundaries.

How It Works:

The algorithm analyzes the on-axis measurement (0°, 0°) and automatically sets the start and stop times:

1. Detect main peak: Locates the largest absolute amplitude in the impulse response
2. Compute Hilbert envelope: Calculates the amplitude envelope for robust reflection detection
3. Adaptive threshold: Adjusts detection sensitivity based on signal-to-noise ratio (3-8% of peak)
4. Set start time: Places window start 3 ms before the main peak (clamped to 0.0 minimum)
   • This margin accounts for off-axis measurements where the sound path may be shorter (gives a margin of ~1m)
   • Prevents truncation of the direct sound in off-axis measurements
5. Search for reflections: Scans the envelope for the first significant reflection/echo after the main peak:
   • Must occur at least 3 ms after the main peak
   • Must exceed the adaptive threshold
   • Must be a local maximum in the envelope
6. Set stop time: Places window stop 0.5 ms before the first detected reflection

If no significant reflection is detected, the stop time is set to the end of the search window or IR length.

Detection Method:

The algorithm uses Hilbert envelope analysis to detect reflections. This provides more robust detection than simple amplitude analysis by following the overall energy contour of the signal, making it less sensitive to oscillations within the waveform.

Window Margins:

• Start margin (3 ms): Ensures off-axis measurements are not truncated when rotating large loudspeakers. At the speed of sound (~343 m/s), 3 ms corresponds to approximately 1 meter of path length difference. During rotation, drivers may move closer to or further from the microphone—this margin provides sufficient safety for rotations of large cabinet.
• Stop margin (0.5 ms): Small margin before the detected reflection to avoid including the beginning of the echo.

Best Use Cases:

• ✅ Quasi-anechoic measurements: Ideal for measurements with clear, isolated reflections (e.g., outdoor, treated rooms, gated measurements)
• ✅ Controlled environments: Works well when the direct sound is clearly separated from reflections
• ✅ Multi-angle measurements: The 3ms start margin prevents truncation of off-axis measurements
• ✅ Time-saving: Quick initial window setup that can be manually refined if needed

Limitations:

• ❌ Reverberant environments: Not suitable for highly reflective rooms where reflections arrive continuously
• ❌ Dense reflection patterns: May not detect the first reflection correctly if multiple reflections arrive in quick succession
• ❌ Room measurements: For full-room measurements including reflections, manual windowing is more appropriate

Recommendation:

Use “Auto Set Window” as a starting point in quasi-anechoic conditions, then verify and adjust the window boundaries manually while observing the impulse response graph. In reverberant environments, set window parameters manually based on your measurement conditions.

Windowing Purpose

• Anechoic approximation: Simulate anechoic conditions from room measurements
• Reflection gating: Remove floor, ceiling, and wall reflections
• Crossover and correction design: Essential for accurate filter design using measured data
• Clean response: Isolate direct driver response

Window Function

LinFIR uses a Tukey window (tapered cosine/raised cosine) for smooth fade-in and fade-out transitions. This prevents spectral artifacts and high-frequency ringing caused by abrupt signal truncation.

Windowing Convention:

The fade-out starts at the stop parameter:

• Fade-in: [start → start + taper]
• Full signal: [start + taper → stop]
• Fade-out: [stop → stop + taper]

The applied window is the product of the fade-in and the fade-out envelopes, so if start and stop are within a taper of each other the two overlap gracefully into a smooth sub-unity bump instead of a broken shape — no minimum gap is enforced.

No fade at the extremes: a fade exists only to smooth a discontinuity created by truncating real signal — a start of 0 truncates nothing before it (see Time Reference), so there is no fade-in at all in that case: the window opens at full amplitude from the very first sample. The stop side needs no equivalent special case: once stop reaches the measurement’s own duration (its maximum allowed value), every real sample already falls before it, so the fade-out ramp naturally has nothing left to attenuate.

Taper Duration: 0.5 ms fixed for simple windowing. For adaptive windowing, the taper is frequency-adaptive: each band gets a half-period taper centered on its center frequency (half of 1/frequency), with a floor at 0.5 ms. This means high frequencies use ~0.5 ms tapers while low frequencies use much longer tapers (e.g., 5 ms at 100 Hz, 10 ms at 50 Hz) to match the wavelength and avoid introducing ripples in the magnitude.

Preview

The impulse response graph shows the windowed result in real-time as you adjust start and stop times.

Whenever you change a windowing parameter (start/stop time, drag, “Auto Set Window”, “Reset Window”, or toggling Adaptive Window), the window envelope itself is briefly overlaid in red on both the Impulse Response and Step Response graphs, scaled to line up with the peak of the curve underneath. Alongside it, on the Impulse Response graph, the raw (un-windowed) impulse response of the measurement currently shown is drawn as a dashed line in the same colour as that driver’s impulse curve, so you can see directly what the window keeps and what it removes. Both fade together: fully visible for 3 seconds, then fading out over 2 seconds - changing a parameter again resets the timer. The overlay is only shown for the driver you’re currently editing, and only while its window is enabled.

You can also bring the overlay up at any time without changing anything by clicking the Show window button in the IR Windowing section.

The overlay (and the dashed reference) is positioned to match the processed curve on the graph, so its start/stop will not sit at the raw Start/Stop values you entered once the driver has filters or a time delay — see Time Reference above for why.

With Adaptive Window enabled, the overlay instead draws a telescoping sequence of fade-out curves, one per frequency band: the shortest (darkest, treble) band’s complete window is drawn first, and each following, longer band picks up exactly where the previous one’s plateau ends and extends the shape further out, growing lighter as frequency decreases. Because bass bands extend their fade-out much further in time than treble bands, you’ll often only see the first few (darkest) steps within the graph’s visible range - the lower-frequency ones continue further right, beyond the current time axis.

Adaptive Window

Advanced frequency-dependent windowing that preserves low-frequency response while gating high-frequency reflections.

How It Works

When enabled, LinFIR uses a wavelet transform to apply different window lengths for different frequencies:

• Start parameter: Applied uniformly to all frequencies
• Stop parameter: Applied with frequency-dependent extensions
  • Lower frequencies get longer decay times
  • Higher frequencies use shorter windows

Examples of automatic extensions:

• 30 Hz: +33 ms beyond stop time
• 480 Hz: +2 ms beyond stop time
• 7.5 kHz: +0.1 ms beyond stop time
• 18 kHz: +0.06 ms beyond stop time

This reflects the physical reality that longer wavelengths make it difficult to temporally separate direct sound from reflections.

Benefits

• Preserves bass response: Low frequencies automatically get the time they need
• Avoids bass roll-off: Short windows don’t unnaturally cut off bass
• Frequency-aware gating: Removes high-frequency reflections while keeping bass intact
• No phase distortion: Maintains natural phase relationships

When to Use

Use Adaptive Window when:

• Gating reflections but need to preserve bass response
• Using short measurement windows where bass would be compromised
• Simple windowing causes unnatural bass attenuation
• You want frequency-dependent decay times matching physical acoustics

Don’t use Adaptive Window when:

• Working with anechoic or quasi-anechoic measurements with very clean direct sound (simple window is cleaner)
• You have sufficient window length for all frequencies uniformly
• You prefer uniform windowing for consistency

Enabling

Check the “Adaptive Window” checkbox in the IR Windowing section.

Delay Compensation

Shift the impulse response earlier in time without truncating data.

How It Works

• Non-destructive: IR data remains intact in memory
• Display offset: Delay applied as metadata for display and calculations
• Purpose: Bring acoustic peak closer to time zero to flatten the phase

Auto Detect Delay

Click “Auto Detect Delay” to automatically detect and apply optimal delay compensation across all drivers:

Behavior:

1. Detects the peak position for each driver (based on on-axis measurements)
2. Finds the minimum detected delay across all drivers
3. Applies this minimum delay to all drivers simultaneously

Advantages:

• Preserves temporal alignment: All drivers maintain their relative timing relationships
• Removes common delay: Eliminates the shared propagation delay from all measurements
• Prevents errors: Ensures consistent delay compensation across all drivers automatically

Detection Method:

• Locates the position of the main peak in the impulse response
• Uses on-axis measurement (0°, 0°) as reference for each driver
• Aligns all driver peaks precisely at time zero

Manual Adjustment

Manually enter delay value in milliseconds to fine-tune compensation for individual drivers.

Sync Toggle (enabled by default):

When the Sync toggle is enabled, any manual delay adjustment automatically applies to all drivers in the project. This ensures consistent compensation across all drivers without needing to adjust each one individually.

• Enabled: Changes to the delay parameter propagate to all drivers automatically
• Disabled: Changes only affect the current driver, allowing independent delay values
• Activating Sync: If Sync is currently disabled and you enable it, an automatic delay detection is triggered to ensure all drivers have coherent delay values

When to use independently (Sync disabled):

• Only when drivers were measured with significantly different microphone positions
• Even then, prefer fixing microphone position and re-measuring rather than relying on delay compensation

Purpose and Limitations

What delay compensation is FOR:

• Removing the time of flight of sound from the measurement (optional, depends on personal preference)
• Correcting for different microphone positions between measurements

What delay compensation is NOT for:

• ❌ Time-aligning drivers during crossover design (this is handled separately by the Time Delay parameter)
• ❌ A tool for precise multi-driver coherence (proper crossover phase design is required instead)

Important: Microphone Position Corrections

Delay compensation can correct for microphone position differences, but this method has significant limitations:

⚠️ High-frequency measurement errors scale with wavelength:

• 10 cm microphone position difference → Errors above ~3400 Hz
• 5 cm microphone position difference → Errors above ~6900 Hz
• Ultra-short wavelengths (high frequencies) amplify measurement uncertainties

Best practice:

• ✅ Prefer re-measuring with fixed microphone placement rather than compensating for position differences
• ✅ Maintain consistent microphone position across all measurements when possible
• ✅ Delay compensation cannot replace careful measurement technique

Off-Axis Measurements:

• The selected delay compensation applies uniformly to all off-axis measurements (cannot vary by angle)
• This affects the relative timing between different measurement angles consistently

Measurement Notes

Section: Collapsible header (default closed) Purpose: Document measurement conditions and hardware setup

Overview

The Measurement Notes field provides a space to record important information about your measurement setup, hardware configuration, environmental conditions, and any other relevant details.

Features

• Multi-line text editor: Supports up to 1000 characters
• Auto-save: Notes are saved with the project
• Hint text: Helpful placeholder text suggests what information to include
• Character counter: Live display of character count (e.g., “127/1000 characters”)

What to Document

Hardware configuration:

• Microphone model and serial number
• Audio interface model and settings
• Amplifier configuration
• Cable connections and routing

Measurement conditions:

• Date and time of measurement
• Room conditions (temperature, humidity if relevant)
• Microphone position and distance
• Driver orientation and mounting

Reference information:

• Calibration file version used
• Sample rate and sweep settings
• Any unusual conditions or observations
• Version of measurements for tracking

Example Notes

Microphone: UMIK-1 (S/N: 123456) with cal file v2.3
Interface: Focusrite Scarlett 2i2 (48kHz, 512 buffer)
Distance: 1m on-axis, driver mounted in baffle
Room: Semi-anechoic (windows gated at 10ms)
Date: 2025-12-14
Notes: First measurement after driver break-in period

Best Practices

• Be specific: Include model numbers and serial numbers when relevant
• Date your measurements: Track when measurements were taken
• Note calibration: Record which calibration file was used
• Document changes: If re-measuring, note what changed
• Reference conditions: Record anything that might affect comparisons

Exporting Impulse Responses

Impulse responses can be exported individually or in batch using two methods:

Spliced measurements: if a near-field / far-field splice is active (see Near-Field / Far-Field Splicing), the export dialog adds a Raw capture / Spliced (anechoic) choice. “Spliced” writes the blended anechoic IR (and adds a _spliced marker to the filename). “Include distortion” can be combined with either choice — it prepends the distortion tail measured at capture time to whichever main response you export (marker becomes _spliced_dist when both apply). The choice defaults to “Spliced” when the measurement being exported has a spliced version.

Individual Export (Per-Measurement)

For each measurement (horizontal and vertical angles):

1. Click the ⬇ (Export) button next to the measurement angle
2. Export dialog appears with format options
3. Choose export format (WAV or TXT) and options
4. File is saved with automatic angle suffix

File naming format: