Resonalyze

Measurement-Driven Tuning for Car Audio and Multi-Way Loudspeaker Systems

Automatic time alignment with honest verdicts, a virtual DSP crossover designer, and engineering-grade acoustic analysis — impulse response, frequency response, phase, loopback-referenced timing, and live transfer functions — on Windows.

Measure each driver once, then leave the car: align, combine, and optimize the whole system from your desk — and only then type the result into the DSP.

Download latest release · Your first measurement · Build from source

Resonalyze is an open-source desktop application for measuring and tuning multi-way loudspeaker systems — with a special focus on the hardest room of all: the car cabin. It generates test signals, records the response through a Windows audio device, and turns the captured data into engineering-focused plots and concrete DSP settings: crossover corners, per-driver delays, polarity, and PEQ. The same toolset measures rooms, home loudspeakers, headphones, microphones, and complete signal paths.

Resonalyze's center of gravity is the step most measurement workflows leave to you: turning a set of per-driver measurements into one coherent system. Auto delay and Auto crossover search the actual settings against the phase-aware predicted sum, and every automatic result comes with an honest verdict — the engine reports why it trusts an arrival, and refuses loudly instead of fabricating a number when the measurement cannot support one.

Resonalyze is under active development. Treat its results as diagnostic measurements, not as certified laboratory data.

Project Showcase

Resonalyze is built around a practical loudspeaker workflow: measure real drivers, inspect timing, design EQ and DSP settings virtually, then apply the result with fewer blind tuning passes.

Virtual DSP — combine measured drivers through gain, delay, polarity, crossover filters, an all-pass stage, and PEQ before touching the hardware DSP.

Does the automation actually help? Sum loss — how many dB the real phase-aware sum falls short of a phase-blind magnitude addition at each crossover junction (average / worst dip). Left: a three-way system tuned by ear over years. Right: the same system after one Auto crossover + Auto delay pass — the worst junction dip shrinks from −8.0 to −2.3 dB.

EQ Wizard equalizes any measured response — an impulse response, a captured overlay curve, or a moving-microphone RTA in dB SPL — designing an up-to-32-band PEQ toward its own target curve, with microphone calibration, Auto Tune, a vertical fader bank, import/export, and printable tuning sheets.

Time Alignment estimates loopback-referenced delay from the transfer impulse response and shows confidence, levels, distance, and the arrival envelope.

Why Resonalyze?

If you already use tools like REW, OpenSoundMeter, or Smaart, the obvious question is: why install another analyzer?

The short answer: those are broad measurement toolboxes; Resonalyze is a focused, end-to-end tuning workflow for active multi-way systems. It is not that the classics lack EQ or alignment features — REW's alignment tool sums a pair of measurements, and its EQ module corrects a response. Resonalyze operates one level up: separate per-driver measurements on one absolute time base, complete virtual DSP chains, a phase-aware sum of the whole system, and optimizers that work every crossover junction and both stereo sides at once. Its home turf is the car, where every driver sits at a different distance, doors leak and resonate, and the "room" fights back — and its output is not just a plot but the DSP settings themselves:

• Built for multi-way active systems Measure each driver separately, then design the whole system virtually: crossover corners, slopes and filter families, per-driver delay and polarity, all-pass stages, and PEQ — tuned against the phase-aware predicted sum, and only then written into the hardware DSP. Auto crossover and Auto delay search these settings automatically, across both stereo sides in one run.
• Honest automation An automatic tuner that guesses is worse than none. Resonalyze's engines certify their evidence: arrival estimates carry confidence and verdicts, modal build-up latches are detected and flagged instead of aligned to, playback-crosstalk contamination is detected and removed from the analysis, and when a measurement cannot support a decision the engine says so out loud instead of returning a plausible-looking number.
• Loopback-referenced timing Measurements can use a recorded loopback channel as the time reference, so delay and transfer-function analysis are tied to the actual playback path instead of to guesswork.
• Repeatable, confidence-scored measurements Average up to 64 sweeps into one cross-spectrum transfer estimate to pull the response out of the noise, and read a per-frequency coherence (γ²) curve that flags exactly which bands are trustworthy. An optional confirm-between-runs pause turns the same path into spatial averaging across microphone positions.
• Absolute, calibrated levels Calibrate the microphone against an acoustic 1 kHz calibrator and read the Frequency Response and the live RTA in real dB SPL, not just relative dB. The anchor is stored as its ingredients and re-validated against every measurement, so a reading is shown as absolute only when it is genuinely backed — otherwise the plot falls back to relative dB and says so.
• Crossover summation prediction Measure each driver once, then virtually align, combine, and optimize your loudspeaker system before applying a single change to the DSP. Because every measurement carries a loopback transfer IR, Resonalyze can compute the true complex (vector) sum of two measurements — Main ⊕ Compare — summing their impulse responses sample-by-sample so relative delay, polarity, and phase are all accounted for. That predicts how two drivers (or the two sides of a crossover) actually combine, which arithmetic on dB curves cannot. Compare-side delay and polarity controls let you tune the alignment live, and a companion sum-loss curve shows exactly how many dB the real phase-aware sum falls short of a phase-blind magnitude addition — a direct read-out of the summation loss you are dialing out. The Virtual DSP tool takes this to its conclusion: complete virtual DSP chains (gain, delay, polarity, crossover filters, all-pass, PEQ) per driver, tuned against the live predicted sum before a single setting is applied to the hardware. Two auto-fit modes do the tedious part: an Auto crossover optimizer searches the crossover frequencies, filter families, slopes, and cut-only gains that flatten the summed magnitude (favoring tight, minimally overlapping splits), and Auto delay aligns each junction's delay and polarity against the phase-aware sum — across both stereo sides in one run, holding a configurable L/R scene offset between the sides. The search level-matches each junction internally, so the result does not follow the channel gains; still, set the gains at least approximately first — past ±30 dB of in-band imbalance the correction saturates and the log will ask you to level the gains and re-run.
• Practical loudspeaker alignment Time Alignment reports first arrival and strongest peak, each refined to sub-sample precision by a GCC-PHAT cross-correlation, plus distance at 20 °C, confidence, signal levels, and a visible envelope around the detected arrival.
• Fast compare-and-adjust work Persistent overlays, calculated overlays, target curves, and on-plot labels make it quick to compare measurements, tuning passes, channels, listening positions, or before/after changes.
• Live transfer-function analysis Live Spectrum drives the system with a selectable excitation signal — including a leakage-free periodic pink noise — and uses a loopback reference, coherence, overlap, averaging, peak hold, and overload detection to show the driven response rather than only the raw microphone spectrum.
• Measurement history as a working shelf Recent captures stay available in memory, saved files are remembered across launches, and each entry has a frequency-response preview. Entries also remember their full working state — active mode, per-mode settings, and shown overlays — so switching between measurements restores the whole context, and a one-click reset starts a fresh session from defaults.
• Developer-friendly, inspectable data IR files, overlays, settings, and history metadata are stored as readable JSON where practical, making measurements easy to archive, diff, and debug.

Resonalyze does not try to be every acoustic tool at once. Its sweet spot is measurement-driven multi-way tuning — above all in the car — where timing, repeatability, quick comparison, and transparent data matter more than a large legacy feature set. For room EQ at home, REW remains excellent; when the question is "what delays, crossovers, and polarities do I put into this six-channel DSP", that is what Resonalyze is for.

Demo

A one-minute tour of the main features:

Download

Download the latest ready-to-run build from GitHub Releases:

• Resonalyze-Setup-vX.Y.Z-win-x64.exe — the recommended installed build
• Resonalyze-vX.Y.Z-win-x64.zip — for most Windows computers
• Resonalyze-vX.Y.Z-win-arm64.zip — for Windows on ARM

The .zip builds are self-contained and do not require a separate .NET installation. The installer adds shortcuts, uninstall support, and automatic in-app updates for the installed x64 build. A SHA-256 checksum file is provided with every release.

Note that "self-contained" refers to the runtime, not to your data: by default every build keeps settings, history, overlays, Virtual DSP state and logs in %LocalAppData%\Resonalyze. To make a .zip build fully portable — data beside the executable, nothing left on the machine — create an empty file named portable.flag next to Resonalyze.exe.

Windows SmartScreen note: the builds are not code-signed (signing certificates are expensive for a free open-source project), so the first launch may show a "Windows protected your PC" dialog. Click More info → Run anyway, or verify the download first against the published SHA-256 checksum. The full source code is right here if you prefer to build it yourself.

Highlights

• Exponential sine sweep measurement with impulse-response JSON save/load
• Band-defined sweep: set the low and high frequency the sweep must cover (20 Hz – 20 kHz) and a per-octave pace instead of an octave count pinned to Nyquist. The panel reports the range it can actually deliver, and the transfer estimate is gated to the excited band, so a band-limited sweep no longer shows noise spikes just below its low edge
• Mandatory loopback-referenced sweep processing: every measurement captures a loopback reference, and all analysis is derived from the resulting transfer function (harmonics and THD+N stay on the sweep deconvolution)
• Multi-sweep averaging (1–64 runs) combined as a cross-spectrum transfer estimate to lift the signal-to-noise ratio, with a per-frequency coherence (γ²) curve in the Frequency Response, Phase, and Group Delay views and an optional confirm-between-runs pause for spatial averaging
• Noise-robust reliability-anchored phase unwrapping: deep nulls and low-coherence bands are bridged by a slope prediction instead of anchoring the unwrap, so one noisy bin can no longer throw the whole phase tail off by a multiple of 360°
• Selectable phase windowing: the original fixed Tukey gate or a frequency-dependent window (FDW, 4 / 6 / 8 cycles) that progressively removes late reflections at mid and high frequencies while retaining the fixed gate and useful resolution at low frequencies
• Selectable microphone calibration profiles (0° / 90°) applied per view, with lenient parsing of common .txt / .cal / .frd / .csv correction files
• Absolute dB SPL: calibrate against an acoustic 1 kHz calibrator (94 / 104 / 114 dB) and read the Frequency Response and the Live Spectrum RTA on a true dB SPL axis. The anchor is stored as its ingredients (reference and measured levels, tone frequency, capture identity) and re-validated against each measurement, quietly falling back to relative dB with a note when it does not apply
• Time Alignment with sub-sample delay estimation from the transfer IR, refined by a GCC-PHAT cross-correlation
• Crossover summation prediction: the true complex (vector) sum of two measurements (Main ⊕ Compare) with Compare delay/polarity controls, plus a sum-loss curve — accounts for delay, polarity, and phase the way dB-curve math cannot
• Virtual DSP tool: run up to eight measured L/R driver pairs (with mono channels for a shared subwoofer) through virtual DSP chains — gain, delay, polarity, Butterworth / Linkwitz-Riley / Bessel / Chebyshev crossovers, an all-pass stage, and imported PEQ — and see their complex sum, sum loss, the opposite side's sum, phase tracking, a per-junction phase read-out (phase at the crossover, the coherence-maximizing delay fix and its lobe margin), per-pair Δ L−R timing, auto crossover proposals, a stereo-aware auto delay with a scene offset, gated phase view, overlay capture, a headphone track audition (your own music rendered through the tune to a WAV), sessions, and tuning-sheet export
• Live Spectrum: real-time loopback transfer function with selectable excitation (leakage-free periodic pink, pink, brown/red, white noise) and coherence, plus a Silent mode that measures the ambient room with no excitation at all
• Frequency response, phase, group delay, waterfall, Burst Decay, and autocorrelation
• Compare a second measurement (from a file or History) against the current one across Time Alignment, Phase, Group Delay, Frequency Response, and Impulse Response, with matching analysis settings and per-metric deltas
• Minimum-phase / excess-phase decomposition from the same selected Fixed/FDW spectrum, with Off / Auto / Manual τ detrending and one shared Auto reference when Main and Compare are shown together
• Per-curve visibility toggles in every analysis view; curves redraw on the fly with no separate draw/clear step
• Harmonic distortion, THD, and THD+N analysis
• Persistent comparison overlays with labels, styling, curve math over captured or live plot curves, targets, import/export, and saved per-mode state
• Live overlay preview: captured, calculated, and target overlay dialogs redraw the candidate curve on the plot as you edit, and revert on Cancel
• EQ Wizard: equalize any measured frequency response — an impulse response from a file or History, a captured overlay slot, or a curve imported from text — toward its own target curve, designing an up-to-32-band parametric EQ (with selectable microphone calibration), using a vertical fader bank, Auto Tune, a live results read-out, cross-tool PEQ import/export, and a printable tuning-sheet PDF. The car case it was built for is a moving-microphone RTA in dB SPL: no coherence, no impulse response, an absolute datum
• Signal Generator: play pink (periodic and continuous), brown/red, white noise, or a sine tone through the configured playback device for level setting and channel checks
• Measurement History with in-memory snapshots, saved-file recall, FR previews, per-entry working state (mode, settings, active overlays), and a one-click new-session reset
• Four audio backends — MME Compatibility, ASIO, WASAPI Shared, and WASAPI Exclusive — with device-aware sample-rate selection and backend-specific channel routing
• Compact Mic/Loop input level meter with Peak, RMS, Peak Hold, and stored measurement levels
• Docked, non-modal settings panels with live previews and instant graph updates
• Auto-update support for installed builds through a signed NetSparkle appcast

Gallery

Virtual DSP Measure each driver once, then design gain, delay, polarity, crossover filters, an all-pass stage, PEQ, and the complex acoustic sum before applying settings to the real DSP.	EQ Wizard Load an impulse response and design a parametric EQ toward its own target curve with microphone calibration, Auto Tune, a 32-band fader bank, per-band curves, a live results read-out, cross-tool import/export, and a tuning-sheet PDF.
Time Alignment Sub-sample delay estimation from a loopback-referenced transfer impulse response, with confidence, levels, distance, and envelope view.	Frequency Response One-click loudspeaker response measurement with smoothing, calibration, an optional dB SPL scale, distortion curves, overlays, target comparison, and an optional coherence curve from averaged sweeps.
Live Spectrum Real-time loopback transfer-function analyzer with selectable excitation noise, coherence, averaging, overlap, peak hold, unreliable-band marking, and an absolute dB SPL RTA from an acoustic calibrator.	Impulse Response Inspect the measured impulse response, save it as readable JSON, load it later, and reuse it across analysis modes without re-measuring.
Group Delay Analyze timing behavior from the loopback-referenced transfer IR, with a millisecond gate, gate offset, and a live impulse-window preview.	Waterfall and Burst Decay Visualize frequency decay and stored energy with the Fourier waterfall and Burst Decay views.

More plots

Phase response

Burst Decay

Overlays

Requirements

To run a release build:

• Windows 10 or later
• Working Windows playback and recording devices
• An optional ASIO driver for low-latency audio interfaces
• A suitable loopback, microphone, or other measurement connection

The self-contained release archives include the required .NET runtime.

To build Resonalyze from source:

• Windows 10 or later
• .NET 10 SDK — global.json pins the exact version (rollForward: latestPatch), so an older feature band fails restore even though it is also .NET 10
• Visual Studio 2026 with the .NET desktop development workload, or the .NET CLI

Use conservative playback levels when connecting physical equipment. Start with the output turned down and verify the signal path before running a measurement.

Your First Measurement

If you are starting from zero, this is the minimal hardware path — roughly €100–200 total:

• A USB audio interface with at least two inputs (any entry-level two-channel interface with phantom power works). Two inputs matter because every Resonalyze measurement records a loopback reference alongside the microphone — that is what makes the timing analysis absolute. Community-verified so far: Focusrite Scarlett Solo (the developer's own rig). Confirmed another interface working? Open an issue and it will be listed here.
• An analog measurement microphone (an inexpensive electret measurement mic with an individual calibration file is ideal; Resonalyze imports the calibration text file). A USB measurement mic such as the UMIK-1 will not work — see the FAQ for why.
• Two cables: one to feed the system under test from the interface's output 1, and one short cable from the interface's output 2 straight back into input 2 — that is the loopback.

Then, in about ten minutes:

1. Wire it up: mic → input 1, output 2 → input 2 (loopback), output 1 → the system's input (in a car: the DSP's aux/optical input, with only the driver under test unmuted).
2. Start Resonalyze, open the measurement settings, select the interface, and assign the input and loopback channels. The measurement will not start without a loopback — that is by design. Set Measurements to at least 4: the sweeps are averaged into one cross-spectrum estimate, which lifts the response out of the cabin's noise floor and produces the per-frequency coherence curve that tells you which bands to trust.
3. Turn the playback level well down, place the mic at the listening position, and run the sweeps. Watch the input level meter for microphone level, loopback presence, and headroom.
4. Explore the views: Frequency Response, Time Alignment (arrival, delay, distance), Phase, Impulse. Save the impulse response — saved measurements are the raw material for everything else.
5. Measure each driver the same way, then open Virtual DSP and let Auto crossover and Auto delay design the tune against the phase-aware predicted sum before you touch the hardware.

The full path with averaging, coherence, spatial averaging, and comparison is described in Measurement Workflow.

FAQ

Can I use a UMIK-1 or another USB microphone?

No — and it is physics, not stubbornness. Every Resonalyze measurement records a loopback reference next to the microphone signal, and the two streams must share one hardware clock to stay sample-accurate. A USB microphone is its own audio device with its own free-running clock; pairing it with a separate playback/loopback device gives two streams with an unknown run-to-run start offset plus continuous drift — which silently corrupts every timing-sensitive result: phase, group delay, and above all automatic delay alignment. This is also why the settings deliberately do not offer a separate loopback device. Use an analog measurement microphone through a two-input interface instead (see Your First Measurement).

Why is the loopback mandatory? REW works without one.

The loopback records what actually left the playback chain and exactly when it left, so every analysis is derived from the mic-vs-loopback transfer function — timing becomes absolute rather than relative to an arbitrary trigger. That absolute time base is what allows separate measurements, taken minutes apart, to be combined later: it is the foundation of the whole measure-once-tune-at-your-desk workflow, of the complex (vector) sum prediction, and of automatic delay alignment. Without a shared reference none of those operations would be honest.

Is one microphone position enough to tune a whole car?

At that one point, yes, and exactly: sound pressure sums linearly, so the predicted combination of individually measured drivers is the physics of what the microphone would record — not an approximation. The honest boundaries are the ones any single-point method has: the prediction holds at the microphone position (put it where your head is), in the linear non-clipping regime, with the same playback chain and mic position for every measurement in the set, and at a roughly stable cabin temperature. For frequency-response work you can go further with spatial averaging (Confirm each run pauses between sweeps so you can move the microphone). And the final judge of a tune is still your ears — the tool's job is to make the version you audition worth auditioning.

Building from Source

Clone the repository:

git clone https://github.com/DIMOSUS/Resonalyze.git

Then open:

source/Resonalyze.sln

Or build and run it from the command line:

dotnet restore source/Resonalyze.sln
dotnet build source/Resonalyze.sln --configuration Release
dotnet run --project source/Resonalyze.csproj

Run all application and deterministic DSP tests with:

dotnet test source/Resonalyze.sln -c Release --filter "Category!=Hardware"

That covers three test projects: Resonalyze.Dsp.Tests (deterministic and synthetic), Resonalyze.App.Tests (file formats and non-UI application logic against a fake audio factory) and Resonalyze.Audio.Tests (audio internals — PCM decoding, capture sessions, WASAPI configuration). The filter drops the hardware smoke tests, which need real WASAPI endpoints named through the RESONALYZE_WASAPI_CAPTURE_ENDPOINT_ID and RESONALYZE_WASAPI_RENDER_ENDPOINT_ID environment variables; without those they report as skipped rather than passed.

For local performance profiling, build the dedicated Tracy configuration:

dotnet run --project source/Resonalyze.csproj -c Tracy

This configuration defines TRACY_ENABLE and references Tracy-CSharp; normal Debug and Release builds do not load Tracy. Add instrumentation through AppProfiler.Zone(...), AppProfiler.FrameMark(...), and AppProfiler.SetThreadName(...) so profiling code stays isolated behind the build flag. Tracy zones are thread-bound and strictly LIFO: never let a zone span an await (Tracy terminates the session with "Invalid order of zone begin and end events") — zone the synchronous sections, including inside Task.Run bodies.

The Release executable is produced at:

source/bin/Release/net10.0-windows/Resonalyze.exe

Tagged GitHub releases also produce:

• portable self-contained .zip packages for win-x64 and win-arm64
• an x64 Setup.exe installer with uninstall support
• NetSparkle appcast files that the installed build uses for automatic updates

The build.yml workflow runs on every push to main and every pull request. It builds the solution, runs all three test projects — application, audio and DSP — each with --filter "Category!=Hardware", then proves the release path still works by producing the single-file self-contained publish and compiling installer/Resonalyze.iss. Warnings are errors, so a new one fails the build.

Measurement Workflow

This workflow covers impulse-response (IR) based analysis: a swept-sine measurement is captured once and then inspected across the frequency-response, phase, group-delay, impulse, waterfall, and burst-decay views. For continuous, real-time analysis without capturing an IR, use the additional Live Spectrum mode instead.

1. Connect the output of the device under test to the selected input, either directly or through a microphone and a suitable interface.
2. Start Resonalyze and open the measurement settings.
3. Select the audio backend, sample rate, devices or backend-specific input and loopback channels, the sweep band and pace, playback channel, and analysis parameters. A loopback reference channel is required — all analysis is derived from the transfer IR it produces, so the settings panel flags an unset loopback and the measurement will not start without one. To average several sweeps, set Measurements above 1; enable Confirm each run to pause before each sweep so you can reposition the microphone for spatial averaging.
4. Start a recording to generate and capture the exponential sine sweep. With averaging enabled the runs are combined into one transfer IR and a coherence (γ²) curve, debiased by the number of runs: the raw estimate over K averages reads 1/K even for pure noise (0.5 at two runs — estimator bias, not information), so the stored figure maps that null expectation to 0 and stays comparable across run counts.
5. Watch the compact input level meter to confirm microphone level, loopback presence, and headroom before trusting the measurement.
6. Select the analysis view you need.
7. Adjust smoothing, windows, offsets, and display options as needed. Mode settings open in a docked, non-modal panel attached to the plot, so the main window stays usable while the settings are visible. Changes apply on the fly and update the active graph without closing the panel or resetting the current zoom/pan.
8. Capture and compare with overlays: store the current curve in an overlay slot, import a reference from text, or combine slots with curve math. When tuning home or car systems, add a target curve overlay (a parametric house/Harman-style target with presets) and switch its deviation readout to EQ correction to see how much to dial into an equalizer.
9. Pin a second measurement with Compare to overlay a reference from a file or a History entry across Time Alignment, Phase, Group Delay, Frequency Response, and Impulse Response.
10. Use Save to keep the captured impulse response for later analysis or comparison.
11. Use History to review recent measurements, preview their frequency response, reload an older snapshot, or save an in-memory capture to disk.

For acoustic measurements, microphone placement and room conditions strongly affect the result. For electrical loopback measurements, make sure the signal levels and impedances are safe for both devices.

Mode Settings

The Mode Settings... button opens the settings for the current analysis mode in a docked panel aligned to the plot area. The panel has no title bar, can stay open while the main window has focus, and switches automatically to the matching panel when you change modes.

Settings apply on the fly: changing a value immediately redraws the current analysis while preserving the visible plot range. This makes it easier to tune smoothing, FFT windows, Tukey fades, offsets, and display options without losing the area you were inspecting.

Each curve-based view groups its plotted curves under a Curves: heading with one checkbox per curve — for example Primary / HD2–HD4 / THD+N in Frequency Response, or measured / minimum / excess in Phase. Toggling a curve redraws immediately; there is no separate draw or clear step. Numeric and dropdown settings carry a small R button that resets them to the built-in default, and double-clicking a plot axis restores its default scale.

The Frequency Response, Phase, Group Delay, Waterfall, and Burst settings include a compact impulse-window preview where applicable. The preview shows the impulse response used by that mode together with the selected Tukey window. Phase and Group Delay analyze the loopback transfer IR and are only drawn when the active record provides one; their preview marks the gate position used for the analysis.

Live Spectrum has its own docked settings panel. It lets you choose the Signal Type (excitation noise), a Scale (relative dB or dB SPL, when an SPL anchor applies), a microphone calibration profile (Off / 0° / 90°), and a Sequence Length from a power-of-two list. The sequence length is the FFT block size used by the live analyzer and is preserved between sessions.

Phase and Group Delay

Phase and group-delay analysis run on the loopback transfer impulse response: both views need the common timing reference it provides, so they are only drawn when the active record contains a transfer IR. Without one, the plot says that loopback is required instead of showing a misleading curve.

Both modes use a millisecond-based fixed gate built from a left Tukey fade, a flat plateau, and a right Tukey fade. A Gate offset positions the end of the left fade inside the analysis frame, and the Auto checkbox beside it keeps that offset snapped to the detected start of the impulse response — the band-limited first-arrival front, not the peak — re-snapping on every new measurement and making the field read-only while it is checked. It is on by default; release it to place the gate by hand. (When the detector cannot get a trustworthy reading from a record it falls back to the transfer-IR peak, which is what older releases always used.) The docked preview draws the impulse response, the fixed gate, and a marker at the gate offset. A read-only readout shows its lowest reliable frequency (≈ 1 / gate length), so it is clear where the gated curve stops being trustworthy.

Phase additionally offers Window: Fixed / FDW. Fixed reproduces the original single Tukey gate across the whole spectrum. FDW builds a small bank of time-aligned spectra whose effective right-side duration follows cycles / frequency: the fixed gate is the maximum, so low frequencies retain the long window, while mid and high frequencies progressively reject the late reflection tail. FDW cycles selects 4, 6, or 8 periods: 4 gives the strongest reflection suppression, 6 is the recommended balance, and 8 retains more late detail. Every bank spectrum is referenced to the same absolute sample origin before interpolation, so changing cycle count does not create an artificial time shift.

The Phase view can show four independently toggled curves:

• Measured phase — the raw response, including delay and reflections.
• Minimum phase — the part tied to the magnitude (correctable with EQ), reconstructed with a real-cepstrum method.
• Excess phase — measured minus minimum: the all-pass part (pure delay and reflections) that an equalizer cannot fix.
• Coherence (γ²) — drawn when the impulse response was captured with two or more averaged runs, to show which frequencies the phase traces can be trusted at.

Detrend removes one constant delay before phase unwrapping:

• Auto estimates the slope-based excess delay from the same Fixed/FDW spectrum being displayed. The resolved read-only value appears directly in τ (ms).
• Manual uses the editable τ value. Switching through Auto does not overwrite the stored Manual value.
• Off applies no additional detrend and keeps the absolute phase slope.

When Main and Compare are displayed together, Auto is resolved once from Main and that common reference is applied to both measurements. Their real relative delay therefore remains visible as a linear phase difference instead of each curve being flattened independently. The same common reference is used for measured and excess phase.

Unwrapped phase uses a reliability-anchored algorithm instead of naive bin-to-bin accumulation: each bin takes the 360° branch closest to a phase predicted from the last trustworthy bin and the running phase slope. Bins well below the local magnitude envelope (so one tall resonance cannot disqualify a quieter but repeatable band) — or with low coherence, when the measurement carries a γ² estimate from averaged runs — are still displayed but never trusted as anchors, so deep nulls, reflection notches, and masked bands are bridged cleanly and a single bad bin can no longer shift the entire remaining curve by a multiple of 360°. A dead stretch too long to bridge honestly (the turn count inside it is genuinely unknowable) is blanked instead of guessed, and the curve restarts as a fresh segment after it. On clean data the result is identical to the classic unwrap.

Measured, minimum, and excess phase are internally consistent: minimum phase is derived from the magnitude of the selected Fixed/FDW analysis spectrum, and excess phase subtracts that curve from measured phase without smoothing across NaN segment breaks.

Group Delay reads absolute delay referenced to the start of the transfer IR, so a peak well into the impulse response reports its true arrival time. The curve is computed energy-weighted: the numerator and the energy of the per-bin group-delay ratio are smoothed separately and divided afterwards, so near-null bins — where the raw ratio legitimately spikes to tens of milliseconds while carrying almost no energy — follow the delay of the dominant energy instead of the singularity. The smoothing window never narrows below the gate's own spectral resolution (features finer than 1/T cannot be resolved by a gate of duration T anyway), so interference-null spikes stay suppressed even with display smoothing off. FDW is deliberately not applied to Group Delay in this version: Group Delay continues to use the fixed Tukey gate. Consequently an FDW phase curve is a direct-sound-oriented representation and is not the exact integral of the currently displayed fixed-gate Group Delay. Selecting Fixed phase restores the mathematically compatible phase/group-delay pair.

Audio Backends

Resonalyze can run measurements through four backends, chosen in the measurement settings dialog:

Backend	Use it for
MME Compatibility	Ordinary Windows playback and recording devices. The most compatible option and the fallback when nothing else works.
ASIO	Audio interfaces with a native ASIO driver: lowest latency and arbitrary multi-channel routing.
WASAPI Shared	Windows endpoints without an ASIO driver, while other applications keep using the device. The endpoint's own mix format applies, so Windows may resample.
WASAPI Exclusive	The same endpoints taken exclusively: no Windows mixer in the path, and the requested sample rate and bit depth reach the hardware unresampled. Nothing else can play while a measurement runs.

Both WASAPI modes address devices by endpoint id rather than by index, so a chosen device survives reboots and device reordering. The two subsections below cover the MME and ASIO settings in detail; the WASAPI modes use the same microphone and loopback channel selection as MME.

The microphone input is the primary measurement channel, and a loopback reference channel is required for every measurement. Resonalyze records both channels simultaneously and derives the main impulse response as a transfer function from the loopback reference to the microphone response, which removes the playback path (DAC, amplifier, output routing) from the analysis. All IR-based views — frequency response, phase, group delay, impulse response, waterfall, Burst Decay, and autocorrelation — are computed from this transfer IR. Harmonic distortion, THD, and THD+N curves use the ordinary sweep-deconvolution response instead, because the harmonic separation belongs to the sweep analysis itself. The HD2–HD4 curves draw at the excitation frequency (a second-harmonic hump caused by a 1 kHz drive appears at 1 kHz, not at its 2 kHz product; microphone calibration is applied at the product frequency first), so each curve ends at Nyquist/n. Note the harmonic curves are on the sweep-deconvolution scale while the primary curve is loopback-normalized — their vertical distance is not yet a calibrated distortion percentage.

The group-delay reference is the start of the transfer IR, so reported delay is absolute rather than relative to a response peak.

Because the loopback is mandatory, a measurement will not start until a loopback channel is selected for the active backend; the settings panel flags an unset loopback in place. Records loaded from older files that were captured without a transfer IR still open, but their transfer-IR views show a "requires loopback transfer IR" note instead of a misleading curve.

Sweep band and duration

The exponential sweep is described by the band it must cover: a Low frequency (Hz) and a High frequency (Hz) anywhere between 20 Hz and 20 kHz, plus a Per octave (ms) pace that sets the duration. Measuring a tweeter through a 2 kHz crossover no longer means sweeping from 20 Hz and pinning the top to Nyquist — sweep the band the driver actually plays, and spend the whole excitation there.

Phase alignment is preserved by rounding the band outward to whole start and end cycles, so the achieved range always encloses the one you asked for, and the fade-in and fade-out live in the guard bands outside it rather than eating into the band under test. The Actual range line reports what the current settings really deliver: where a short sweep cannot honour the request — one cycle at 20 Hz alone takes 50 ms — it says so instead of quietly shortening the sweep at run time. The duration is capped by one length limit that the preview, the generator and the stored settings all resolve through, so the panel cannot promise a sweep the measurement then fails to play.

The transfer estimate is gated to the excited band: full weight between the points where the sweep envelope is actually open, raised-cosine ramps across the fades, zero outside. Outside the swept band the transfer function is microphone noise divided by the reference's leakage skirt, which used to surface as +15 dB spikes just below the start of a band-limited sweep; those bins are now excluded rather than half-passed.

Measurement options apply as you edit them — the band, the pace, the playback channel, the averaging — and touch the audio session only when its identity actually changed. The audio backend, the format the device is opened with and its device panel are the exception: they sit in their own bordered panel and commit together with Apply settings.

MME Compatibility

Use MME Compatibility for ordinary Windows playback and recording devices. (Older releases and the settings file call this backend Wave; the dropdown entry is "MME Compatibility".) The measurement settings dialog lets you choose:

• playback device
• recording device (microphone)
• sample rate from the values supported by the current configuration
• playback channel
• microphone input channel (Left or Right)
• loopback input channel (Left or Right) — required

The loopback is captured from a second channel of the same recording device as the microphone, so both signals share one hardware clock and stay sample-accurate — the timing every phase, group-delay and time-alignment result relies on. This is deliberate: capturing the loopback from a second input device would put the two streams on independent clocks with an unknown, run-to-run start offset plus drift, silently degrading every timing-sensitive result. Resonalyze therefore does not offer a separate loopback device at all; loopback is mandatory, so the recording device must expose a stereo input (or use ASIO).

ASIO

Use ASIO for audio interfaces that provide a native ASIO driver. The measurement settings dialog lets you choose:

• ASIO driver
• sample rate from the values supported by the selected driver
• ASIO input channel used for the microphone
• ASIO loopback input channel — required
• ASIO output channel pair used for playback
• playback routing within the selected output pair

ASIO output routing works as follows:

• Mono sends the same signal to both channels of the selected output pair
• Left sends the signal only to the first channel of the pair
• Right sends the signal only to the second channel of the pair
• Stereo sends the signal to both channels of the pair

Before applying ASIO settings, Resonalyze checks whether the selected driver supports the current sample rate: the dialog shows the driver's playback latency in samples and a "supported / not supported" line for the chosen rate.

Click ASIO Control Panel to open the driver's native control panel. Use it to configure driver-level settings such as buffer size, clock source, or sample rate when the driver requires those to be set outside the application.

Click Test ASIO Inputs to capture a short diagnostic snapshot of the available ASIO inputs. This helps verify that the microphone and loopback channels are truly separate and are not being mono-summed by the driver or the audio-interface control software.

ASIO support depends on the installed driver. If a driver is already in use by another application or refuses the selected sample rate, Resonalyze reports the driver error before starting the measurement.

Input Level Meter

The right-side control column includes a compact two-channel input meter for Mic and Loop. It is designed to stay useful — while routing, checking loopback, or validating a completed measurement — without opening extra dialogs.

• the bar shows a filtered RMS level
• the bright vertical marker shows Peak Hold
• the text shows Peak / RMS in dBFS
• after a sweep or time-alignment measurement completes, the meter retains the final levels from the last valid capture instead of dropping back to idle

This makes it easy to spot missing loopback, a weak microphone level, overload, or an unexpectedly hot reference path before you start analyzing the curves.

Live Spectrum

The Live Spectrum mode is a live, dual-FFT transfer-function analyzer. It plays a continuous excitation signal, uses the configured loopback channel as a reference, and shows the real-time frequency-domain relationship from loopback to microphone. Because the estimate is referenced to loopback rather than to the microphone alone, it suppresses noise and other input-side content that is not correlated with the playback signal.

Alongside the transfer function, Resonalyze draws a coherence curve (γ²) on a secondary right-hand axis scaled from 0 to 1. Coherence shows how much of the measured response is linearly correlated with the loopback reference: values near 1 mark frequencies where the transfer-function estimate is trustworthy, while low values flag bands dominated by noise, reflections, or non-linear behavior.

The live estimate averages in the power domain, which avoids the downward bias that magnitude averaging introduces on noise-like signals. The level is calibrated for tones rather than as a power spectral density. On-screen smoothing is referenced to wall-clock time, so the response stays consistent regardless of the chosen overlap and sequence length, and the display refreshes at roughly 30 frames per second.

Live Spectrum works with any backend. Configure a loopback input in Record Settings, on a separate channel from the microphone, to get the full loopback-referenced transfer function with coherence.

Without a loopback it still runs, as a single-channel RTA: the microphone's own spectrum, with no transfer function and no coherence. That is the mode the car workflow actually wants — a moving-microphone average in dB SPL, where there is no reference signal to divide by.

In practice, referencing to loopback is far more stable than viewing the raw microphone spectrum when the room or measurement chain contains unrelated noise. It is not a magic denoiser, but it lets you focus on the driven response rather than on whatever the microphone happens to hear.

A Scale control switches the plot between relative dB and dB SPL (offered only while a matching SPL anchor applies to the live input). In dB SPL the transfer function is hidden — a dimensionless ratio has no absolute level under noise excitation — and the whole plot becomes the microphone RTA on a true dB SPL axis. That RTA level is integrated as power per fractional-octave band, so the absolute reading is independent of the FFT size above the analysis resolution limit. Below it — where a single FFT can no longer resolve that band — the integration band is floored to the window's main lobe, which does narrow as the FFT lengthens, so a broadband level there reads roughly 3 dB lower per doubling of the FFT size. That is the unavoidable resolution limit of a single FFT, shared by every FFT RTA; use a longer FFT to push it lower. Because dB SPL needs no excitation, it also unlocks a Silent signal type — an ambient RTA that plays nothing and simply measures what the microphone hears — while the periodic pink noise, which exists only to converge the transfer function, is dropped from the list.

Signal Type selects the excitation noise, ordered by usefulness:

• Pink noise (periodic) — the default. One FFT-length period of exactly pink noise, synthesised in the frequency domain and looped. Because it is periodic with the analysis block, every frame captures a whole period, so it is measured leakage-free with a rectangular window and perfect bin resolution, and the average converges almost instantly. When it is selected, Window is forced to Rectangular and Overlap to Off (both would only add correlated frames that do not improve the estimate); your own picks are restored for other signals.
• Pink noise — continuous random pink noise, −3 dB/octave.
• Brown / red noise — −6 dB/octave, with more low-frequency drive for subwoofer and room-mode work.
• White noise — flat energy per hertz.

The Live Spectrum settings panel also exposes Sequence Length, the FFT block size used by the live analyzer. Only power-of-two values are offered, to keep the live FFT path efficient and predictable.

It also exposes Overlap (Off, 50%, or 75%), which slides the analysis window by a fraction of its size instead of advancing in non-overlapping blocks. With a tapering window (the usual case) overlap reclaims the samples the window attenuates at the block edges, giving faster, smoother averaging and a more responsive display at the cost of more FFTs per second. It is disabled for periodic pink noise, where overlapped frames are correlated and add no averaging.

Smoothing applies fractional-octave smoothing (Off, 1/1 … 1/48) to the displayed curve, using the same presets as the Frequency Response mode.

Every magnitude-smoothing selector (Frequency Response, Live Spectrum, Fourier Waterfall, Virtual DSP, EQ Wizard, magnitude overlays) also offers Psychoacoustic: smoothing whose width follows frequency instead of staying constant — 1/3 octave at and below 100 Hz, narrowing smoothly to 1/6 octave from 1 kHz upward. That is roughly how hearing resolution varies: broad in the bass, where room modes dominate and fine structure is neither audible nor fixable, and finer in the midrange and treble, where real resonances live and are worth seeing. It shapes magnitude curves only: phase, group-delay and coherence traces (including captured overlays of them) and the harmonic widths fall back to the plain 1/6-octave width, Burst Decay does not offer the mode at all (its per-band envelope pipeline has no magnitude grid to floor), and the Auto delay engine never reads the display smoothing — the drawn curve and the panel's Sum-loss read-outs follow the selection, the proposed delays do not. Files that store a smoothing choice (overlays, Virtual DSP sessions) save the mode as a plain 1/6-octave width plus a separate flag, so older builds open them with plain smoothing instead of rejecting them.

Window selects the analysis window applied before the FFT: Hann (a good general default), Flat Top (maximum amplitude accuracy for tones), Blackman-Harris (strong spectral-leakage suppression), or Rectangular (unwindowed). It is forced to Rectangular for periodic pink noise, which is already leakage-free.

Averaging sets how quickly the trace responds: Fast, Medium, and Slow select exponential time constants (referenced to wall-clock time, so they are independent of overlap and sequence length), while Infinite integrates a cumulative average indefinitely. Reset Average clears the running average and peak-hold envelope without restarting the measurement.

Main curve (on by default) shows the primary live trace itself; turning it off leaves only the optional RTA, peak-hold, and coherence curves.

RTA (input) (off by default) overlays a reference-free real-time analyzer curve: the plain magnitude spectrum of the microphone input alone, with no division by the loopback reference. It is what a classic RTA shows — the actual spectral content the microphone hears — and is drawn on the same dB axis as the transfer function. Coherence does not apply to it, so it is never dimmed by the Coherence Limit, and its level is normalized by the analysis window's coherent gain, so switching windows does not shift it. In relative dB it is a single-channel level whose vertical position floats with input gain; with the Scale set to dB SPL it becomes calibrated absolute sound pressure (and the only curve — see Sound Pressure Level). Captured into an overlay slot, an RTA trace — typically a microphone moved slowly around the listening area — can be equalized directly in the EQ Wizard.

Peak Hold overlays a second curve that retains the maximum level seen on the trace until it is reset. Coherence (on by default) toggles the γ² curve shown on a secondary 0-to-1 axis.

Coherence Limit marks unreliable parts of the transfer-function curve: any frequency whose coherence falls below the chosen percentage (default 25%) is drawn dimmed and dashed, so it is immediately clear which portions of the trace should not be trusted. Set it to Off to draw the whole curve uniformly.

If the CPU cannot keep up with the chosen settings, captured blocks are dropped rather than allowed to stall the measurement, and a processing overload warning appears at the top of the plot, making the cause of a stuttering display clear.

Switching to another analysis mode and back restores the last Live Spectrum curve, its peak-hold envelope, and any active overlays, so a captured trace is not lost when you step away to inspect a different view. Press Start to resume live capture; starting a new capture replaces the remembered trace automatically.

Measurement History

The History button opens a docked measurement-history panel with:

• a list of recent measurement snapshots
• a compact frequency-response preview for the selected row
• row tooltips with capture metadata such as time, mode, sample rate, duration, channel, peak index, and stored mic/loopback meter levels

History entries come in two kinds:

• RAM for in-memory snapshots from the current session
• FILE for saved IR files remembered across launches, as long as the files still exist on disk

The newest entries appear at the top of the list. Column-header sorting is intentionally disabled, so the history keeps a stable chronological order and the row actions always match the visible item.

The currently active loaded snapshot stays highlighted in the list, even when you click another row only to inspect its preview. This makes it easier to compare entries without losing track of which measurement is actually driving the main plots.

Double-click a row to load it into the main workspace. Use:

• Save to turn an in-memory snapshot into a regular IR JSON file
• Delete to remove an item from history without deleting the underlying file from disk
• New session (reset to defaults) to start with a clean slate: all per-mode settings return to their defaults, and the current measurement and overlays are cleared. Audio device and routing settings are kept, and the history list and saved files are left intact. The active entry's working state is saved first, so nothing is lost.

Working state remembered per entry

Each history entry remembers the working state it was last used with: the active mode, every per-mode setting (frequency response, phase, group delay, impulse response, waterfall, burst decay, live spectrum, time alignment), and which overlay slots were shown. Switching to another entry and back therefore restores not just the impulse response but the whole working context.

This state is kept current as you work — it is written back into the active entry whenever you switch to another entry and when you close the app — so it reflects what you were actually doing, not just the moment of capture. Audio device and routing settings are never changed by switching entries. Overlays keep their own separate on-disk storage; the history records only which slots were active and reloads their contents from there.

Saving an in-memory snapshot turns that row into a file-backed history entry and updates the visible name to the chosen file name. Loaded IR files appear in the same list as fresh captures, so History works as a practical short-term measurement shelf rather than as a separate file browser.

To keep memory use predictable, Resonalyze retains only a small rolling set of unsaved in-memory snapshots. Saved file-backed entries are persisted separately.

Compare

The Compare button in the action panel overlays a second measurement on top of the current one, so two responses can be read side by side with the same analysis settings. Choose the reference from a file (Choose file…) or from a History entry; the button then shows its name, and Clear removes it.

Compare is applied everywhere it is meaningful, always recomputed with the current mode's settings so the two curves stay directly comparable:

• Time Alignment — the reference envelope is overlaid on the peak preview with its own first-arrival and strongest-peak markers, and the delay table gains a second block whose every value shows the delta against the source in parentheses (for example 1.006 (+0.010)).
• Phase and Group Delay — the reference curves are computed with the identical gate/window and smoothing and drawn dashed and dimmed. Phase Auto detrend is resolved once from Main and reused for Compare, preserving their relative delay; the gated IR preview shows the reference impulse as well.
• Frequency Response — the reference magnitude is overlaid (harmonics stay source-only to keep the plot readable).
• Impulse Response — the reference impulse is drawn alongside the source with the axis fitting both curves. When the record contains a transfer IR, the plot uses an absolute sample timeline from sample 0 to the IR peak plus the configured Length, so the arrival positions of the two impulses can be compared directly.

A reference is only drawn when its sample rate matches the current measurement: Time Alignment states the mismatch explicitly, the other modes simply omit the curve. Because Compare applies the same settings to both measurements, adjusting a window or gate updates both curves at once — the intended way to compare two phase or group-delay responses fairly. The source and Compare curves are also selectable as live operands in a calculated overlay, so their difference can be plotted and watched live while you tune the analysis window.

Time Alignment

The Time Alignment mode analyzes acoustic delay from the currently active measurement record. It is designed for practical loudspeaker, microphone, and channel alignment work, where the result has to be more precise than a single audio sample.

Time Alignment no longer runs its own separate capture path. Instead, it reads the active transfer impulse response already stored in the current record. That means it works immediately after:

• a new sweep measurement captured with loopback enabled
• loading an IR JSON file that contains transfer-response data

If the active record does not contain a transfer IR, the mode clearly reports that the measurement was captured without loopback and does not attempt to estimate delay from the ordinary sweep-deconvolution response.

The transfer IR itself comes from the same loopback-based sweep measurement pipeline used elsewhere in the app: Resonalyze plays the exponential sweep, records the microphone and loopback simultaneously, and computes the microphone response relative to the loopback reference path. This removes unknown playback latency from the response used for timing analysis. With ASIO, both recorded channels also stay locked to the same hardware clock, which gives the most repeatable result.

The delay estimator uses a deliberately robust two-stage chain:

• the active transfer impulse response from the current record
• an optional raised-cosine bandpass window around the frequency range of interest
• the analytic-signal envelope of that impulse response, whose first arrival and strongest peak are detected robustly — this is the coarse, polarity-blind anchor
• a GCC-PHAT (phase-transform) correlation, computed from the transfer IR's own spectrum, that refines each anchor to sub-sample precision

The first-arrival search rejects pre-ringing sidelobes: the zero-phase stages of the chain (the bandpass window and the Hilbert envelope itself) ring exactly symmetrically around each arrival, and the stronger of those early lobes clear the arrival threshold — on a clean measurement they used to read as an arrival up to several milliseconds before the true wavefront, and the better the SNR, the more of them survived the noise gate. The kernel that makes the ringing is known, so each candidate is tested against physics rather than heuristics: an arrival can pre-ring at a given distance no louder than the analysis kernel's own envelope allows there. A candidate above that ceiling is a genuine arrival no matter how the surroundings look — which is what keeps weak direct sound alive in reverberant bass, where everything around a reflection cluster is energized. A candidate at or below the ceiling is confirmed as pre-ring by its mirror twin: an exactly even kernel puts an equal lobe at the mirrored position after the peak, and room decay only adds energy on the late side, so the mirror cannot hide a lobe.

That second stage is what makes the numbers trustworthy. The transfer IR's spectrum already carries the microphone-to-loopback cross-phase, so whitening it to unit magnitude over a soft band mask (built from where the response actually has energy) collapses the correlation to a sharp peak at the true broadband delay — independent of the driver's own magnitude shape, which would otherwise pull an envelope peak off the real arrival. A short search window keeps the refinement on the arrival the envelope found, a windowed-sinc plus parabolic interpolation reads the peak between samples, and the search runs on peak magnitude so a polarity-inverted arrival (a trough) is located just as reliably as a normal one. When the whitened peak is weak or pinned to the window edge, the estimate falls back to the envelope's own fractional peak, so the result is never worse than the plain envelope.

When the record was captured with averaging and carries a coherence (γ²) curve, the whitening is additionally weighted by it: bands whose phase does not repeat across the averages — noise, level- or drift-varying distortion, non-averaging reflections — get less say in the correlation than clean, repeatable bands, while every in-band bin keeps at least a quarter of its weight so the occupied bandwidth (and the peak sharpness that follows from it) is preserved. Repeatable content, including harmonic distortion, reads as coherent and is not suppressed.

The payoff is delay estimates such as 87.0 samples or 1.972 ms resolved to a hundredth of a sample instead of a coarse integer, and refined against the true acoustic arrival rather than the driver-tinted envelope shape. For time alignment, this is a serious practical upgrade: smaller timing adjustments become visible, repeatable, and easier to trust.

The first arrival — the one the panel aligns by — also shows a GCC-PHAT alignment confidence: the normalized height of the whitened correlation peak, displayed as Alignment: NN%, together with whether the sub-sample position came from the correlation (GCC-PHAT) or fell back to the envelope. It is separate from the meter-based signal-quality readout: a high signal level with a low alignment confidence means the level was fine but the delay itself is only coarsely located.

When the strongest peak lands well after the first arrival — the classic narrowband-subwoofer case, where room modes ring louder than the direct sound long after it — Time Alignment flags it and points you at the first arrival, so a modal or reflected peak is not mistaken for the driver's real timing. The flag requires a real valley (6 dB) between the two peaks: a low-frequency driver's direct sound can keep rising for milliseconds, and a shoulder of that one wave packet peaking later is its rise time, not a reflection.

The mode recalculates immediately when you switch into Time Alignment, and also updates live as soon as you change the bandpass settings.

It reports signal quality using the analysis envelope and the stored meter snapshot from the same measurement record:

• a color-coded Excellent, Good, Fair, or Poor signal grade from the recording's SNR — the strongest envelope peak against the record's noise floor (the RMS of its quietest quarter, so reflections and modal decay do not count as noise the way an average over the whole record would)
• the first-arrival prominence — the first arrival's envelope level relative to the strongest peak. A low value means the pick sits on a broad leading edge (normal physics for band-limited low-frequency drivers), so its exact position is less sharply defined; it says nothing bad about the recording itself, which is why it is reported separately instead of being folded into the signal grade
• microphone peak and RMS levels in dBFS
• loopback peak and RMS levels in dBFS
• a CLIP warning for an overloaded microphone input
• a FULL SCALE marker for a digital loopback reference running at 0 dBFS

The compact input level meter remains useful here too: it preserves the final captured levels from the last valid sweep measurement or loaded file, so the Time Alignment readout still has the signal context that produced the current transfer IR.

The measured time, distance, and sample count are clickable. Click one of those result lines to copy just the numeric value to the clipboard, which is convenient when pasting delay values into another tool or a spreadsheet.

When the bandpass window is enabled, Resonalyze shows a small frequency-domain preview of the selected pass band. It also shows the envelope around the detected peak, making it easy to see whether the reported delay comes from a clean dominant arrival or from a noisy or ambiguous response.

Selecting a Compare reference overlays its envelope on the same preview and adds a second delay-table block whose values carry the delta against the source, which turns Time Alignment into a direct A/B of two arrivals.

Time Alignment therefore depends on how the underlying sweep record was captured. To produce a usable transfer IR, the sweep measurement itself must be run with a configured loopback input channel. That loopback-enabled sweep can be captured through:

• ASIO, the recommended path for the best timing accuracy
• Wave, if the selected recording device exposes a stereo input and one side can be dedicated to loopback

Wave loopback is supported for convenience, but ASIO remains the preferred path for serious timing work because the capture chain is more tightly controlled.

Installer and Updates

Resonalyze supports two release styles:

• Portable .zip builds Extract and run Resonalyze.exe directly. This is convenient for quick testing or for keeping multiple versions side by side.
• Installed Setup.exe build Installs to the current user's local Programs folder, creates shortcuts, and registers an uninstaller.

When the application detects a newer GitHub release, the version label in the custom title bar changes to Update available. Clicking it opens a focused update dialog:

• installed builds can either start an Automatic Update or open the GitHub releases page for a manual download
• portable builds offer a manual download only, because they are not tied to a managed install location

Automatic update currently targets the installed x64 build distributed through Setup.exe. Portable .zip builds remain fully supported, but they are updated manually by downloading a newer archive.

Saving and Loading Impulse Responses

After a sweep measurement completes, click Save to store the measured impulse-response data. Resonalyze proposes a timestamped file name such as:

Resonalyze-IR-2026-06-15_14-30-00.json

Files are saved as indented, human-readable JSON. Each file contains:

• format and schema version
• save time in UTC
• sample rate and bit depth
• the requested sweep band and the band the sweep achieved, plus its duration and sample count (files written before the band settings carry an octave count instead and are migrated on load)
• playback channel
• measurement mode (SweepDeconvolution or LoopbackTransfer)
• sweep-deconvolution impulse-response samples and peak index
• optional loopback transfer-function impulse-response samples and peak index when loopback was enabled
• optional transfer-function coherence (γ²) data when two or more sweeps were averaged, plus the requested and accepted run counts
• stored microphone and loopback Peak/RMS meter values from the measurement
• an optional SPL calibration anchor when one was configured, so the measurement can be placed on a dB SPL axis after reloading
• embedded preview frequency-response data for the Measurement History panel

Click Load to open a previously saved response. Resonalyze validates the file before using it, rejects files below 44100 Hz, restores the associated measurement metadata into the active record, stays in the current analysis view, and redraws it from the loaded data. Loading an IR does not rewrite the current audio-device configuration in Record Settings. All analyses derived from an impulse response — frequency response, phase, group delay, waterfall, Burst Decay, autocorrelation, and loopback-based Time Alignment — can then be generated without repeating the measurement.

Saving and loading are disabled while a measurement is running. The current file format identifier is resonalyze-impulse-response, version 7. Files are meant to stay human-readable, but editing the sample arrays by hand may make a file invalid or produce misleading analysis results. Files that do not yet contain the embedded preview-frequency-response section can still be loaded; Resonalyze rebuilds the preview when needed.

Plot Overlays

Each supported overlay view provides twelve universal overlay slots. Every slot can hold one of three kinds, chosen from the numbered capture-button menu (or from the settings dialog):

• Captured — a snapshot of a curve currently on the plot.
• Operation — a calculation between two operands, each either a live plot curve or a captured slot (A - B, B - A, A + B, (A + B) / 2, |A - B|, or a frequency blend), plus a complex (vector) sum of the Main and Compare transfer IRs in Frequency Response (see below).
• Target — a parametric target curve compared against a source.

Overlay slots are stored automatically as human-readable JSON under the application data directory — %LocalAppData%\Resonalyze for an installed build, or beside the executable in a portable one (see Download):

overlays/<AnalysisMode>/overlay-01.json

The numbered button opens a menu to Capture curve, Import from text, Export to text, or switch the slot to a Calculated overlay or Target. The checkbox shows or hides the slot and the numeric control applies a vertical offset; the slot's settings dialog opens from ⚙ Settings… in the same numbered-button menu. A Target slot references a captured slot or the current measurement; an Operation slot references two operands that are each a live plot curve or a captured slot. A live-curve operand re-reads the plot on every rebuild, so a calculation over it — for example the difference between the source and a Compare curve — updates live as the analysis settings change.

Visible overlay names are drawn directly over the plot as a compact legend. Each entry uses the same color and line style as the curve itself, so exported screenshots and day-to-day comparisons stay readable even when many curves are active.

Target curves

A Target overlay compares a source against a parametric target shape and draws two curves from the one slot: the target itself and the deviation (source minus target), plus an optional shaded tolerance band (±dB). The source is either a captured slot or the current measurement (the main Frequency Response curve, or the Live Spectrum main trace — including the running trace, which the target and deviation follow frame by frame).

The target shape and its tolerance band are parametric over frequency, so they are drawn whenever the slot is enabled even if no measurement is on the plot yet; only the deviation curve waits for an incoming source.

The target shape is built from four editable terms: an overall tilt around a 1 kHz pivot, a bass shelf, a treble shelf, and a presence bump/dip. This covers room, car, home-theater, and voicing targets. Presets fill in the parameters and remain fully editable:

• Flat, Room (gentle), Harman room, Warm
• Car, Car (mild), House / bass boost
• X-curve (cinema), Smiley, BBC dip, Custom

The deviation curve has selectable modes: Deviation (measurement − target, how far the response sits from the target), EQ correction (target − measurement, the gain to dial into an equalizer to reach the target), or None to hide it.

The settings dialog shows a live preview of the target shape, and the shared slot offset moves the target up or down (the deviation follows automatically). Target overlays are available in Frequency Response and Live Spectrum (both the running and the paused trace).

Import and export

Import from text loads a captured overlay from a plain-text file of X Y pairs (for example, 123.4 -5.5), one per line. Parsing is lenient: values may be separated by spaces, tabs, commas, or semicolons; extra columns are ignored; and any line that is not a valid number pair (comments, headers, blanks) is skipped. Export to text writes the slot's current curve in the same format. For a Target slot, Export deviation writes the deviation or EQ-correction curve, which is handy for transferring corrections into an equalizer or another tool.

Exported files open with a commented # resonalyze-curve metadata header that records what the curve is — the analysis it came from, its role (a measured response, a deviation, a target), and the sample rate where one applies. Foreign files without the header still import exactly as before; the header only lets Resonalyze recognize its own curves on the way back in, so that, for example, the EQ Wizard can tell a measured response from an EQ-correction curve that must never be equalized as if it were one.

Captured overlay settings include:

• a user-defined name
• line color, thickness, style, and opacity
• optional 1/48, 1/24, 1/12, 1/6, or 1/3 octave smoothing in frequency-based views
• a Clear action that removes only that slot in the current analysis mode

Calculated overlay settings additionally include:

• two operands (Curve A and Curve B), each selected from the live plot curves or the captured overlay slots
• operations A - B, B - A, A + B, (A + B) / 2, and |A - B|
• a blend operation with a user-defined crossover frequency and transition width
• a complex (vector) sum (Main ⊕ Compare), available in Frequency Response when a Compare measurement is set and both records have a transfer IR (see below)
• a sum loss curve (complex − magnitude), the companion to the complex sum, showing how many dB the real phase-aware sum falls short of a phase-blind magnitude addition (see below)
• optional amplitude-space math for dB-based views, which converts both curves to linear amplitude before the operation and back to dB afterward
• independent octave smoothing applied after the selected operation

In Phase Response, the difference operations (A - B, B - A, |A - B|) are phase-aware. Each phase curve — captured or live — remembers whether it is an unwrapped (continuous) or wrapped (-180..180) representation: measured phase follows the Unwrap option in effect when it was captured, while minimum and excess phase are always continuous. When either operand is wrapped, the difference uses the shortest angular distance (atan2(sin Δ, cos Δ)) so it never jumps by ±360° across the branch cut. When both are unwrapped it stays a plain subtraction, preserving the accumulated slope (and therefore the delay) of the two curves. Imported text curves have no wrap hint and are treated as unwrapped.

Octave smoothing is available only for Frequency Response, Phase Response, Group Delay, and paused Live Spectrum. Impulse Response and Autocorrelation keep their original time-domain samples. Overlay JSON always stores the unsmoothed source points, so changing or disabling smoothing is lossless.

Calculated results use the same axes, units, zoom, and vertical pan as the ordinary overlays. Operations are applied to the displayed Y values after source offsets. As a result, addition and averaging on a decibel plot are arithmetic operations on dB coordinates, not physical summation of acoustic power.

Complex (vector) sum

In Frequency Response, a calculated overlay can compute the complex (vector) sum of the Main and Compare transfer impulse responses (Main ⊕ Compare). Both transfer IRs share the same sample-0 time reference, so summing them sample-by-sample and taking the magnitude gives the physically correct summed response of two sources — it accounts for their relative delay, polarity, and phase, unlike arithmetic on dB magnitudes. This predicts how two drivers, or the two sides of a crossover, actually combine.

Two Compare-side controls make it a DSP-style alignment tool:

• Time offset — a fractional-sample delay applied to the Compare IR (interpolated between samples), mirroring a delay you would dial into a DSP channel.
• Invert polarity — flips the Compare IR, to check a polarity swap at the crossover.

Both update the summed curve live as you turn the field or toggle the checkbox. The operation needs a Compare measurement with a transfer IR at the same sample rate; it stays armed and redraws automatically once that data is available.

A companion sum loss operation (complex − magnitude) plots the difference between the complex sum and a phase-blind magnitude addition (|Main| + |Compare|). By the triangle inequality it is always ≤ 0 dB: it shows how many decibels the real, phase-aware sum falls short of what naive dB/amplitude addition would predict — zero where the two sources are perfectly in phase, and dropping into deep negatives toward cancellation. Because the magnitude sum ignores delay and polarity, only the complex side moves as you tune Time offset and Invert polarity, so the curve rises back toward 0 dB as you bring the sources into phase — a direct read-out of the summation loss you are dialing out. It shares the same requirements and live behavior as the complex sum.

Live overlay preview

The captured, calculated, and target overlay settings dialogs preview their result on the plot while you edit. Every control change — name, color, style, opacity, smoothing, operands, the complex-sum delay/polarity, and the whole target shape, tolerance band, and deviation curve — redraws the candidate curve immediately, using the same rendering path as the committed overlay, so you see exactly what Save will keep. Closing with Cancel (or Esc) restores the previous state.

Overlay files are separated by analysis mode and restored automatically when the application starts or the active view changes. Changes to any source overlay immediately update the visible calculated overlays.

All overlay slots use a single file format, resonalyze-overlay, version 5, with a kind field selecting captured / operation / target. Older overlay schema versions are intentionally not loaded.

Overlays are available in the Impulse Response, Frequency Response, Phase Response, Group Delay, paused Live Spectrum, and Autocorrelation views. A Show all / Hide all pair above the overlay panel toggles every active overlay for the current mode at once, without deleting any saved JSON file.

EQ Wizard

The EQ Wizard (under the Tools tab) designs a parametric equalizer — up to 32 peaking (PK) bands plus a preamp — that moves a measured response toward a target. Open it from the Tools tab. It owns its own target curve, edited through the same dialog the Target overlays use but stored with the wizard's own settings — so tuning here never disturbs your overlay slots, and changing an overlay never moves the wizard's target.

Choosing what to equalize

The Source… button picks the curve to tune, and it does not have to be an impulse response:

• Impulse response from file… or from history — the loopback-referenced measurement, the same source the analysis views use
• Curve from overlay slot — any captured frequency-response slot, imported as a snapshot with no live link back to the slot
• Curve from text file… — a response exported from Resonalyze or produced elsewhere

The case this was built for is a moving-microphone RTA in dB SPL: park the Live Spectrum RTA on a car's listening area, capture it into an overlay slot, and equalize that. Such a curve has no impulse response and no coherence behind it, and its datum is absolute rather than relative — so the wizard treats the RTA as a first-class analysis curve rather than something to be reconstructed from an IR. Only measured responses can enter: a harmonic, THD, phase, deviation, EQ-correction, target or calculated curve is refused, including through a text round trip.

Imported curves carry their own Calibration choice, independent of the microphone-calibration preference used for impulse responses, because a curve captured through a calibrated RTA must not be calibrated a second time. The filter response is drawn against its own right-hand dB axis, scaled to the summed filter curve.

The plot shows, on shared frequency/dB axes:

• Source — the captured reference measurement (with optional extra smoothing)
• Target — the parametric target shape (colour, thickness and line style are yours to set in the target dialog)
• Source + EQ — the source with the current EQ applied
• EQ — the filter response itself (all bands, without the preamp) in white
• a shaded error fill between Source + EQ and Target, so the remaining deviation is visible at a glance

Click a band card (or any of its fields) to overlay that band's individual contribution as a dashed curve relative to the target; click empty space to clear it. Each band card carries its frequency, Q, and gain, and the panel adds a Target Level (target offset), a Gain (preamp), a Bands count, source Smoothing (1/N octave), and a Bypass toggle that draws the curves without the EQ. An overlay-settings shortcut reopens the underlying target.

Auto Tune

Auto Tune fits the whole EQ automatically. It works on the error between the target and the (smoothed) source, sets a preamp for the broadband level, then adds peaking bands greedily where the residual error is largest — choosing each band's frequency, gain, and the Q that reduces the error the most. It chooses the band count itself, up to the Max Filters limit (4–32). A cumulative-boost cap and minimum band spacing keep it from stacking many maxed-out bands where the response simply cannot be corrected (for example a deep-bass roll-off).

Cuts only (on by default) is the safe choice for a car tune: Auto Tune places only cut bands and never boosts. A boost cannot fill a reflective cabin's interference null — it just burns amplifier headroom on a dip that shifts the moment the microphone moves. Unticking it lets Auto Tune boost too, but only where boosting is trustworthy: high measured coherence (when the source is a loopback-transfer measurement) and not inside a narrow, deep null. Every boost still obeys the Max Gain limit and the total-gain ceiling.

A From / To frequency window limits where bands are placed; it is drawn on the plot as a shaded band between dashed guides, and the same window bounds the error metrics in the results panel.

Results

Replacing the overlay panel in this mode, a colour-coded Tuning results panel reports the fit quality and the EQ's own extents:

• RMS error and Max error between Source + EQ and Target, measured inside the From / To window
• Filters used, the number of active bands
• Peak boost and Peak cut of the combined EQ
• Headroom — the margin to 0 dB (red when the EQ nets a boost that could clip)

Import, export, and tuning sheet

The wizard imports and exports PEQ profiles in several formats, so tunings move between tools and DSPs:

• Import + export: Equalizer APO, REW filter settings, Generic CSV, EasyEffects (JSON), CamillaDSP (YAML)
• Export only: miniDSP biquads (RBJ coefficients, at 44.1 / 48 / 96 kHz to match the DSP's internal rate), GraphicEQ (Wavelet / JamesDSP)

Import is deliberately lenient: comments, blank lines, disabled (OFF) filters, non-peaking filter types, and malformed entries are skipped rather than rejected.

Export as tuning sheet produces a phone-friendly PDF for reading next to the car or speaker: the product banner, a title from the file name, the date and fit range, a small EQ preview graph with the fit window shaded, the tuning statistics, the preamp, and one large card per filter.

DSP Q convention

Processors do not agree on what the Q of a peaking band means, and the disagreement is invisible until you cut deep. Every convention states the bandwidth between the half-gain points as BW = m · Fc / Q and differs only in the multiplier, so the same three numbers produce different filters:

Convention	Bandwidth at half gain	Behaviour	Seen on
RBJ	Fc / Q	Independent of gain	Equalizer APO, CamillaDSP, REW Generic/Extended, Audiotec Fischer (HELIX / MATCH / BRAX), Audison/Hertz, Mosconi, miniDSP
Symmetric (Zölzer/DAFX)	sqrt(|gain|) · Fc / Q	Widens as the band deepens, boost and cut alike	AMP Panacea, Behringer DCX2496, Rockford Fosgate 3Sixty.3, Hypex Input EQ, rePhase, Crown USM810
Classic	sqrt(gain) · Fc / Q	Asymmetric — boost wider, cut narrower	JL Audio TwK-88

Resonalyze fits, plots and exports RBJ filters throughout. Hand a Q of 5.8 at −15 dB to a Symmetric processor and it realizes a band over twice as wide — an EQ that measures much broader than it was designed, over-correcting its neighbours.

The DSP Q selector in the EQ Wizard states which convention the processor being tuned uses. It moves the Q printed on the tuning sheets — the EQ Wizard's PDF and the Virtual DSP sheet, both of which name the convention they were written for — and nothing else: the fit, the curve on screen and the exported profile files stay RBJ. The conventions are exactly reconcilable, being one filter family reached through different Q scales, so a restated band reproduces the designed response on the device rather than approximating it:

Q_symmetric = Q_rbj × 10^( |gain| / 40)     ±3 dB ×1.19   ±12 dB ×2.00   ±15 dB ×2.37
Q_classic   = Q_rbj × 10^(  gain  / 40)     +12 dB ×2.00   −12 dB ×0.50

The lists above follow REW's equaliser reference, which states each supported processor's half-gain bandwidth formula. Note that these are conventions of a model, not of a manufacturer or a chip: JL Audio's TwK-88 and VXi disagree with each other behind the same tuning software, and a SHARC or SigmaDSP part is handed finished coefficients, so it implies nothing about Q. If your processor is not listed above, measure it: set one band to Fc 1 kHz and Q 4, at +12 dB and then at −12 dB, and read the bandwidth between the ±6 dB points off a sweep. RBJ gives ~250 Hz both times, Symmetric ~499 Hz both times, Classic ~499 Hz and ~125 Hz. The spread is far too large to confuse with measurement error.

Signal Generator

The Signal Generator (under the Tools tab) plays a continuous test signal through the current playback device, independent of any measurement. It is handy for setting output levels, checking channel routing and polarity, exercising a loudspeaker, or feeding an external analyzer.

Signal type offers the same excitation options as Live Spectrum plus a tone:

• Pink noise (periodic) — the default; one period of exactly pink noise looped seamlessly.
• Pink noise — continuous random pink noise, −3 dB/octave.
• Brown / red noise — −6 dB/octave, weighted toward low frequencies.
• White noise — flat energy per hertz.
• Sine — a pure tone; the Frequency, Hz field is only enabled for this type.

Duration, s sets how long the signal plays, and Level, % scales its amplitude. Play starts playback and Stop ends it; a status line reports whether the generator is Ready, Playing, or shows any playback error.

The generator reuses the audio configuration from Record Settings — backend (MME, ASIO, WASAPI Shared or WASAPI Exclusive), sample rate, bit depth, playback channel, and the output device, WASAPI render endpoint or ASIO output channel pair — and displays the resolved settings so you can confirm where the signal is going before pressing Play.

Virtual DSP

The Virtual DSP (under the Tools tab) is the summation-prediction workflow taken to its conclusion: measure each driver once, then design the whole DSP setup virtually. Channels (A, B, C, …) are stereo L/R pairs: each side picks its own measurement — from a file or from History — and runs its own virtual DSP chain. Tight L / R selector radios switch which side the block's controls edit (the plots follow), L→R / R→L buttons copy chain settings across sides for the channels you tick in a small dialog, and a Mono checkbox turns a pair into a single shared driver — the typical one-subwoofer car layout — that feeds both sides' sums. Add channel / Remove channel grow the setup from two up to eight pairs; the blocks live in a scrolling list, so a many-way system stays in one window without crowding the plots. The +/− button in a block folds it down to its header — source, gain, delay, polarity stay visible and the filter chain is hidden — so an eight-way setup can show only the chain being tuned; folded blocks are remembered with the session and nothing about the channel changes while it is folded. Every channel in a project must share one sample rate — a measurement recorded at a different rate is refused; clear the existing sources first to switch the whole project to a new rate.

Each channel runs through:

• Gain (dB) — relative levels are only honest when the measurements share one playback chain; compensate any difference here
• Delay (ms) with a live mm read-out — the ruler check against the physical driver offset (343 m/s)
• Invert — the DSP polarity switch
• Crossover — Off, low-pass, high-pass, or band-pass; each edge picks Butterworth (6–48 dB/oct), Linkwitz-Riley (12/24/48 dB/oct), Bessel (6–48 dB/oct, near-constant group delay), or Chebyshev (6–48 dB/oct, with a selectable passband ripple for a steeper knee) with its own corner frequency
• All-pass — 1st order (180° of phase swing, −90° at the corner) or 2nd order (360°, −180° at the corner, with a Q setting how abruptly it turns). Magnitude is untouched; only phase moves, which makes it the tool for lining drivers up through a crossover where a delay (constant everywhere) and a polarity flip (180° everywhere) are both too blunt — a sub-to-midbass hand-off at 60–100 Hz is the classic case. It runs as its own stage, so it applies with the crossover off, and a live read-out shows the group delay it adds at its corner (≈ 4Q/ω₀) — the reason it works, and on a low corner its main risk
• PEQ — load a parametric EQ profile (any format the EQ Wizard imports) into the chain
• Mute — temporarily remove a channel from the plots, sum, loss metric, overlay capture, and Auto delay without clearing its source or settings
• Bypass — feed the channel's raw measured signal into the sum with the whole chain skipped (no gain, delay, polarity, crossover, all-pass, or PEQ), for an A/B against the processed result; unlike Mute, the channel stays in the sum. Auto delay refuses to run while any participating channel is bypassed — the proposed delay and polarity could not act on the raw signal, yet the channel would still steer every other channel's alignment
• IR polarity — a measured Normal / Inverted / Unknown indicator read from the transfer IR, independent of the virtual polarity switch

Because every stage is linear and the measurements are loopback-referenced transfer IRs, multiplying each measurement by its chain and summing the results as complex responses predicts the linear response the microphone would capture after dialing those settings into the hardware — the same math used by crossover design tools, applied to your own in-room measurements.

The filters are evaluated as the digital biquad cascades a real DSP runs (bilinear transform at the measurement sample rate), so the prediction matches miniDSP-class hardware up to Nyquist, not just an analog textbook curve.

The acoustic plot shows raw and processed curves per channel for the active side, the complex Sum, the opposite side's Sum as a dashed translucent curve (so the two sides' tunes compare at a glance without flipping back and forth), and the Sum loss curve (blanked where every channel is filtered more than 30 dB below the loudest level within an octave — out there the "loss" would be the phase arithmetic of noise floors, not audible summation; measuring against the local level rather than one global peak keeps a tilted in-room curve readable up top instead of blanking the treble), with a Phase view toggle to check that the channels track each other through the crossover region. Its Gate... dialog exposes Fixed / FDW, 4 / 6 / 8 cycles, and Off / Auto / Manual detrend alongside the IR preview, Tukey left / plateau / right controls, and gate offset. Where the gate SITS — its offset and the detrend τ — belongs to the side you are viewing: the left and right drivers sit at different distances, so their arrivals and their first reflections do not land together, and fitting the gate on one side no longer disturbs the other's traces. Everything that decides how the phase is READ stays project-wide — the Tukey lengths (they set the frequency resolution), the window mode, the detrend mode and the FDW cycle count — because two sides read through different windows could not be compared, which is what the view is for. Virtual DSP deliberately keeps phase wrapped to -180..+180 degrees. Auto chooses one common reference channel, displays the resolved τ, and applies that exact value to every driver and the complex sum; it never independently flattens the channels, so their relative phase and timing remain intact. A second plot shows each DSP chain's own magnitude and phase (without the driver) — or, on its Corr mode, the junction-correlation view of one adjacent channel pair (picked in the selector beside the mode switch): the pair's band-limited cross-correlation, raw and GCC-PHAT-whitened, drawn over ±1.5 crossover periods of extra delay on the upper channel — positive lobes are normal-polarity alignments, negative lobes the same alignments with the channel inverted — together with the junction's prior-free acoustic score for both polarities — the dip-penalized summation loss, honestly re-gated at every point (the same figure the engine's own lobe-hop and promotion gates compare). This is deliberately the acoustics ALONE: the searches additionally weigh the arrival prior and the lobe/onset/scene locks, so Auto delay's pick may sit off this curve's deepest lobe on purpose — the gap between the solid and dashed markers is exactly that envelope-versus-summation trade, drawn. The channels enter fully processed, so 0 ms is the alignment as it currently stands (the solid marker), the dashed marker is the band-limited envelope-arrival estimate the searches anchor on, and every near-tied comb lobe the log's [corr]/[phat] lines enumerate is visible as such. A Sum loss read-out (avg / dip per junction plus a total) turns tuning into numbers you can minimize.

A Junction phase block below it reads each adjacent pair's steady-state cross-phase — the regime sustained program material actually sums in. (Deliberately NOT the direct-sound / FDW phase: on field measurements the room adds several milliseconds of apparent group delay at subwoofer frequencies, so a direct-sound read would recommend a confidently wrong delay.) The read-out is analyzed in a time-sized window (~0.68 s of the processed IR), so the physical horizon — and the fix it recommends — does not change when the same measurement is captured at a different sample rate. Three figures per junction: φfc — the phase of the lower channel minus the upper AT the crossover: a weighted circular mean over a narrow (±1/6-octave) window around fc, deliberately a local measurement — a straight-line fit's intercept extrapolates through whatever interference notches and spectral gaps bend the band's phase, and on a real mid/tweeter junction read +158° where the handover itself stood near −15°. Its consistency R (how much the window's bins agree, shown in the tooltip) gates the figure: a low R dashes the column instead of presenting mush (≈0° means the handover is phase-aligned). A φ near ±180° does not by itself call for a polarity flip — an inverted channel and a half-period delay are identical at fc — so that decision comes from a whole-band score comparison, not from the angle; fix ms — the extra delay on the pair's LOWER channel that would maximize the overlap-band phase score, relative to the current settings (positive: delay it further; a negative fix advances the lower channel, so apply it as a +delay on the upper one when the lower is already at 0). A trailing mark carries the polarity call: i recommends flipping the lower channel (a whole-band score comparison, not the angle — an inverted channel and a half-period delay are identical at fc), ~ keeps the current polarity but warns that a flip nearly ties (an inversion and a half-period delay sum alike, common at a sub, so summation cannot settle the polarity). And lobe — how decisively that best delay beats the nearest same-polarity whole-period rival; below 0.10 it is flagged ! (the band is too narrow to rule the period hop out, so don't trust the fix). Read the columns right to left — lobe says whether to trust the delay, φfc says how the junction stands, fix says where to move — and treat fixes under ~0.1–0.2 ms as noise. The score is a phase-alignment measure (Σw·cos Δφ / Σw, −1…+1), not the magnitude coherence γ². The block is purely informative (nothing feeds back into Auto delay), and its tooltip — pinned while the mouse stays on the read-out — carries the full fit per junction: the phase score now and at the optimum, how close a flip scores, the rival and margin, the residual slope Δτ and the fit rms over the band.

A Δ L−R block below reports each stereo pair's final inter-side state: the two sides' band-limited envelope arrivals in the pair's shared band (fully processed chains included) with their difference — positive means the right side leads, the same sign convention as the scene offset, so after a stereo Auto delay every row should read the offset — and, below the timing, a Level Δ L−R row per pair: the gated band-level asymmetry of the two sides (positive: left louder). Timing (ITD) and level (ILD) steer the image together, so this is the read-out for the by-ear gain trim that finishes the centering; note a single microphone underestimates the binaural difference (no head shadow), so expect to trim a little more than it shows. A side whose arrival cannot be measured reliably (a silent band, or a near-noise record) shows an honest dash instead of a precise-looking number, and a side whose full-band envelope timed the room's modal build-up rather than the direct rise (its upper-half read lands much earlier — the same detection the alignment engine's cross-side links run) is marked with ~: the number is what the envelope measured, but the sides then compare different features and the Δ overstates the true skew, so trust the engine's log over that row.

Editing a chain recomputes the prediction on a background task, so dragging a gain, delay, or crossover value stays responsive even with several channels loaded; a burst of rapid edits is coalesced into a single trailing redraw that always lands on the latest settings, and the previous curves stay on screen until the new frame is ready.

A calibration selector (Off / 0° / 90°) applies your microphone correction to the magnitude curves, drawing on the same profiles configured in Record Settings; it defaults to Off because the measurements are loopback-referenced.

• Auto crossover... estimates each channel's usable band and driver type (subwoofer, woofer, midbass, midrange, or tweeter) — the band read is the most prominent contiguous segment of the response, and when the measurement carries coherence (γ²), a frequency the measurement did not trust cannot anchor a band edge, so a noisy or non-linear resonance can't stretch the band. It then asks which filter families to allow (Butterworth / Linkwitz-Riley / Bessel), the crossover-frequency window, and whether the two sides of a junction may take independent slopes (on by default: a driver's high-pass and low-pass may differ, so a woofer can low-pass steeply to the midrange while high-passing gently from the sub; turn it off to tie each driver's two shoulders to one slope). Different drivers always stay free to take different slopes. It searches the crossover frequency, family, and slope to flatten the summed magnitude (a plain amplitude sum — the assumption that Auto delay will bring each junction to its best alignment), penalizing wide band overlap and keeping a practical minimum slope, so it lands on a tight, engineer-sensible split rather than shallow filters that only look flat by overlapping widely. The gains, though, follow a car target curve rather than a flat sum: the midrange and tweeter are levelled to each other (the louder attenuated), and the subwoofer anchors the bass at a chosen elevation over that reference — the Sub level over mid/treble field defaults to (and is capped at) the measured elevation, so out of the box the sub keeps its own level and you only trim the field down if you want less bass. The remaining drivers are fit onto the resulting slope cut-only (a driver already below the target keeps its level — no measured dip is boosted), and every gain is a cut, so the result is headroom-safe. Handovers stay within the sensible range for the two driver types (so a woofer is not crossed up in its roll-off), land on human-friendly frequencies (5 Hz steps below 100 Hz, 10 Hz below 1 kHz, 50 Hz above), and a slope is allowed only while the filter's peak group delay stays within 10 ms — a bound on the delay itself, not the frequency, so the same steep slope is fine at a 250 Hz woofer/mid handover (~5 ms) but excluded at a 75 Hz sub/woofer one (~17 ms), and a low-group-delay family (Bessel) can go steeper down low than a Linkwitz-Riley can. The budget caps how much steeper than the practical floor the search may go; the floor itself — 12 dB/oct, the gentlest slope every family offers — is always kept, since anything gentler would break the overlap rule, so at a very low junction the floor's larger group delay is inherent to crossing there. Group delay is identical for the low-pass and high-pass sides, so with matched slopes a driver is held to the gentler of its two junctions; a steep woofer low-pass paired with a gentle high-pass needs independent slopes on. A shallow filter that lets a driver bleed into a non-adjacent driver's band is still heavily penalized (a woofer should not be audible up at the tweeter). Placement heuristics steer the handovers further: a junction landing in the ear's most sensitive band (2–4 kHz) is penalized, and two drivers that share a wide band are crossed low — letting the upper (smaller) driver take over as early as it cleanly can, for better dispersion and less excursion on the lower driver — except the subwoofer, which is nudged UP toward the ~80 Hz localization limit rather than pulled low. So a capable tweeter is crossed down toward its 1.7 kHz sensible floor (out of the ear band) when its measured band supports it, and a low tweeter handover has to earn its slope against the driver's resonance: the high-pass must give at least 22 dB of attenuation at the tweeter's estimated Fs, so the closer the crossover sits to that resonance, the steeper the filter the wizard demands. The same low-handover logic applies below: a midrange with headroom down to its 200 Hz sensible floor lets the woofer/midbass hand over early, before its cone-breakup region, so the wide woofer/mid overlap does not linger up where it interferes badly. In a stereo system both sides of a driver get the same crossover — a crossover is one electrical filter, so only delay and level differ per side. Narrowing the window past an outer driver adds a subsonic / brickwall band-limit on that channel. Apply does more than take the flattest magnitude candidate: the wizard expands ~50 near-optimal variants (always including the conventional all-LR24 setup) and re-ranks them by the junction loss actually achievable after the best per-junction delay, measured on your impulse responses with the same alignment search Auto delay runs — so a candidate whose slopes cannot phase-align at the handover loses to one that can, before you ever run Auto delay. Ties go to the conventional 24 dB/oct proposal.
• Auto delay aligns in two stages: band-limited first arrivals — refined by a GCC-PHAT cross-correlation where its dominant extremum, of either polarity, is a reliable, unambiguous read: a genuinely inverted junction (a subwoofer against its midbass is the classic) seeds from the trough's position exactly as a normal one seeds from the peak's, with the polarity decision left to the sum search. A junction whose corners leave a spectral gap degenerates the correlation into near-equal lobes, and a near-tie — peak against trough, or against the same-polarity lobe one period over — sends the seed back to the arrival estimate. The coarse offsets set, a fractional-delay search minimizes the sum-loss metric at each junction, reading the same direct-sound gate as the displayed metric so late room reflections the alignment cannot change do not steer it. At mid/tweeter-class junctions with sharp impulse fronts, the search is additionally locked to the drivers' broadband IR onsets — the wavefronts you would align by eye on the impulse plot — so the summation comb (whose near-equal lobes repeat every crossover period) can fine-tune only within the physically correct lobe and can never walk a driver a whole cycle off its neighbor. Each candidate is scored by its in-band average loss and how far its deepest smoothed notch falls below that average, so a solution that only looks good on average while hiding a sharp cancellation at the crossover loses to a slightly lossier but flat one. It weighs every near-optimal candidate against an arrival-based prior and a physical tie-break — a score near-tie goes to the candidate nearest the measured arrival regardless of polarity, because fractions of a dB can never choose between comb lobes — so it does not add delay or flip polarity without a real improvement, sidestepping the flip-plus-half-period impostor a steep crossover can otherwise hide. Where an untrusted coarse seed widens a low junction's search window toward a half period, the window itself spans foreign comb lobes, and fractions of a dB slip past both the prior and the tie-break — so a pick beyond the trusted window's own reach is held to the promotion standard: the hop must be plainly better on the prior-free score, or the best arrival-adjacent candidate stands. Each polarity seeds its own candidates, so the non-inverted optimum is always on the table for that preference even where the inverted curve edges it everywhere. The search runs on a background task with a busy indicator, so the window stays responsive during the few seconds it takes. If the resulting delays span more than ~10 ms — usually a sign that one channel's crossover has excessive group delay (a narrow or steep low-frequency band-pass) — a banner flags the lagging driver so you can soften its filter instead of dialing in an absurd bulk delay. With stereo pairs, Auto delay tunes both sides in one run, and an LHD / RHD toggle says which seat you are tuning for. The driver's side is the reference: it aligns first, the top pair is bridged to it by band-limited envelope arrivals, its polarity is matched, and the far side then descends junction by junction from that bridge. On LHD that reference is the left side and the right leads; on RHD it is mirrored. The scene offset is entered as a non-negative magnitude — how far the far side leads — so switching LHD/RHD never means re-entering a sign, and the level tilt is entered the same way, as a cut on the near side. Pairs whose shared band reaches the localization region are pinned to the scene: the far channel lands exactly the offset ahead of its near counterpart (±0.05 ms of junction fine-tuning), because the stereo image outranks the handover there. Pairs living entirely below that region are pinned too, but only loosely — to the arrival's lobe — and fall back to a free junction search when no reliable cross-side arrival exists. A final scene-preserving pass may then shift BOTH sides of a pair by one shared delta — which cannot touch the image — to recover junction summation the pin cost. A Mono channel (the shared subwoofer) is walked by the left pass; its junction against the right side is measured and reported. A final mono co-move then gives that right junction its vote: one shared channel's delay and polarity cannot touch the L/R scene, so the pass sweeps both across half a junction period and keeps the best mean of the two handovers — the compromise that previously had to be dialed in by hand when the sides disagreed about the sub.
• Capture to overlay saves the predicted sum as a Captured overlay in Frequency Response — compare it against real measurements and target curves, or feed it onward to the EQ Wizard.
• Audition track… renders a music file (wav/mp3/flac/m4a and friends) through the tune and writes a stereo WAV: each program channel is convolved with the summed processed response of its side — left sum to channel 1, right sum to channel 2 — with the microphone calibration optionally baked in as a linear-phase FIR, so what you hear matches the calibrated curves on screen. The track is converted to the project's sample rate when needed (the measured responses are never resampled), both channels share one normalization gain so the L/R balance survives, and re-rendering with a different calibration or after a settings tweak is one click — a quick A/B between tune variants. Subtract cabin optionally removes a typical body-style cabin transfer function (average sedan, hatchback, SUV, and friends — the pressure-zone bass rise that reaches +15…+27 dB at 20 Hz) as an inverse linear-phase FIR in both kernels (folded into the calibration FIR, so both corrections cost one filter): the raw render reproduces the full in-car bass rise as headphone boom the in-car listener never perceives, while the subtracted render leaves this car's deviation from the typical curve audible. The subtracted render is level-matched to the same tune without the subtraction, so an A/B between cabin choices differs in tone, not loudness — the removed bass reads as quieter bass, not a track the normalizer turned back up. Listen through headphones only: each ear gets the measured acoustic path of its side at the microphone position (drivers, cabin, and capsule included) — a stereo auralization of the two sides, not a binaural head simulation — and playing it back through the car would convolve the cabin twice.
• Export… writes the whole setup as a tuning sheet (printable PDF or plain text): for every side of every pair (a mono pair prints once) the gain, delay in ms and mm, polarity, crossover filters, the all-pass stage, and PEQ bands — exactly the list you type into the DSP, both sides in one sheet.
• Save session... / Load session... exports or imports the complete session JSON (sources, chains, gate, and view state) for sharing or archiving.

The tool's autosaved state (sources, chains, gate, and view flags) persists in tools/virtual-crossover.json and survives restarts. Accuracy holds within the usual physics: one microphone position, the same playback chain for every measurement, and the linear (non-clipping) regime.

Calibration

Resonalyze applies a microphone (or measurement-chain) frequency-response correction during logarithmic resampling. Configure up to two calibration profiles in Record Settings — one for 0° (on-axis) and one for 90° (grazing) measurements — by browsing to a correction file for each.

Correction files are read leniently in the common plain-text formats (.txt, .cal, .frd, .csv): frequency level pairs, with comments, headers, a decimal comma, comma / semicolon / tab delimiters, and extra columns all handled.

In the Frequency Response and Live Spectrum settings a calibration selector picks which profile to apply — Off, 0 degrees, or 90 degrees. A profile is offered once a file exists for it, with one deliberate exception: 90 degrees is also offered when only a 0° file is configured, in which case the 90° curve is approximated from the 0° one rather than measured. The selector does not currently distinguish the two cases, so if you never supplied a 90° file, treat that entry as an estimate. The selected mode is saved with the measurement settings, and every plot is routed through the matching file.

For a source checkout, a legacy source/calibration.txt beside the executable is still honored as the 0° profile when no 0° file is configured; the project copies it to the build and publish output automatically. Replace its example data with the correction curve for your microphone, or point the 0° / 90° profiles at your own files.

Sound Pressure Level (dB SPL)

The microphone calibration above corrects the response shape; an SPL calibration anchors its absolute level, so the Frequency Response and the Live Spectrum RTA can be read directly in dB SPL instead of relative dB.

In Record Settings, next to the microphone-calibration files, a Calibrate button listens to an external acoustic calibrator (a 94 / 104 / 114 dB tone at 1 kHz) and records the microphone's digital level at that known pressure. The listen is capture-only — it plays nothing — and the tone level is read from a flat-top power spectrum, whose peak bin holds the true amplitude within hundredths of a dB wherever the tone lands between bins (a Hann or rectangular window would read a few dB low). The capture rejects anything that is not a clean, dominant, on-frequency tone, and fails on clipping or an unsteady level rather than storing a wrong number.

What is stored is the anchor's ingredients — the reference and measured levels, the tone frequency, and the digital capture identity — not a baked "shift by N dB" value. The Frequency Response is a loopback-referenced transfer function, so turning the anchor into an SPL shift also uses each measurement's own loopback level; the Live Spectrum RTA is the plain microphone spectrum, so it needs only SPL = mic level + anchor offset, with no loopback term.

The anchor is valid only at the gain it was captured at. Resonalyze records the digital input identity so a changed input is flagged — the Calibrate button turns gold — and the dialog warns that the analog preamp gain, which software cannot see, must not move after calibrating. Where an SPL reading cannot be honored (no anchor, or one captured on a different input), the plot silently falls back to relative dB and says so, instead of showing an unbacked absolute scale.

In the Frequency Response and Live Spectrum settings a Scale control then switches between relative dB and dB SPL; the dB SPL option is offered only while a matching anchor applies to the current measurement or live input. The anchor is saved with the measurement settings and stamped onto every captured impulse response, so a saved measurement keeps its absolute reference.

Architecture

Resonalyze/
|-- source/                 WinForms application: composition root, measurement
|   |                       lifecycle, and plot presentation
|   |-- Diagnostics/        Profiling hooks (Tracy build configuration)
|   |-- History/            Measurement history snapshots and persistence
|   |-- LiveSpectrum/       Live analyzer orchestration
|   |-- Measurements/       Sweep/noise orchestration, signal generation, IR files
|   |-- ModeSwitching/      The analysis-mode catalogue and tab controller
|   |-- Options/            Measurement and visualization settings panels
|   |-- Overlays/           Persistent overlay slots and calculated overlays
|   |-- Plotting/           OxyPlot model creation, annotations, and adapters
|   |-- Settings/           Settings file, schema migrations, update checking
|   |-- Shell/              Main form, title bar, commands, and docked settings
|   |-- Storage/            Application data paths and safe file writing
|   |-- TimeAlignment/      Loopback delay measurement UI and orchestration
|   |-- Tools/              EQ Wizard, Signal Generator, Virtual DSP, PEQ import/export, tuning sheets
|   |-- Ui/                 Reusable WinForms controls and dialogs
|   `-- Resonalyze.csproj
|-- dsp/                    Reusable signal-processing library (no UI, no audio)
|   `-- Resonalyze.Dsp.csproj
|-- audio/                  Audio drivers and device access (NAudio lives here)
|   `-- Resonalyze.Audio.csproj
|-- tests/
|   |-- Resonalyze.App.Tests/    File-format and application tests
|   |-- Resonalyze.Audio.Tests/  Audio internals: PCM decoding, sessions, WASAPI
|   `-- Resonalyze.Dsp.Tests/    Synthetic DSP tests
|-- installer/              Inno Setup script for the Windows installer
|-- assets/                 Images used by the README and the application
|-- .github/workflows/      CI builds and automated tagged releases
|-- global.json             Pinned .NET SDK version
`-- README.md

The three projects have deliberate boundaries. Resonalyze.Audio owns every audio driver — MME, ASIO and both WASAPI modes — along with device enumeration, format negotiation and capture lifecycle; NAudio is confined to it and is not even referenceable from the application at compile time. Resonalyze.Dsp is pure signal processing with no UI and no audio dependency: FFT analysis, windowing, calibration, smoothing, logarithmic resampling, impulse processing, phase analysis, group delay, crossover and EQ design. The application project wires the two together and owns the measurement lifecycle and plot presentation.

Technology

• .NET 10
• Windows Forms
• NAudio
• NAudio.Asio
• Math.NET Numerics
• OxyPlot
• NetSparkle — in-app updates
• YamlDotNet — CamillaDSP profile import/export
• PDFsharp / MigraDoc — tuning-sheet PDF export

Third-party package licenses are listed in THIRD-PARTY-NOTICES.md.

Contributing

Bug reports, reproducible measurement cases, DSP corrections, and focused pull requests are welcome. Known technical debt and improvement ideas are collected in TODO.md — a good place to look for a first contribution.

When reporting a measurement issue, include:

• audio interface and driver
• sample rate and bit depth
• measurement mode
• relevant analysis settings
• expected and actual behavior
• a screenshot or exception stack trace — unexpected errors are appended to crash.log in the application data directory (%LocalAppData%\Resonalyze, or beside the executable in a portable build), so check there for the full stack trace

License

Resonalyze is available under the MIT License.