Plugin Analysis

Choose Help → Plugin Analysis… to open an independent analysis window. You do not need a project. Add effects using the Mixer plug-in picker and slots, then open their native or parameter editors. The silent experiment chains do not change your project.

Plug-in editors open as independent windows. Activating one does not bring the main Editor forward, and you can keep using Analysis with the Editor minimized or closed. Removing the effect or closing Analysis also closes its plug-in editor.

The left side uses the same Audio FX slots as the Mixer. Click the first empty slot to select a vendor, effect and audio mode. Rescan refreshes both bundled and external effects. Unavailable effects show their load/probe reason; they are not silently omitted from the directory.

Set the measurement conditions

Open Measurement settings at the top right to set linear excitation (Sweep, Delta or Random white noise), FFT size (16384, 32768 or 65536), processing speed (Realtime, 2×, 4× or Ultra), sample rate, block size, frequency range, sweep duration and settle/tail duration. Defaults are a 1 s sweep and 0.25 s settle/tail. The Sweep input level slider accepts −60 to +12 dBFS peak. At 0 dBFS, the peak amplitude is 1.0; +6 dBFS is approximately 2.0. Heron preserves these floating point values to test limiters above full scale, without input clipping or normalization.

Enable Compare chains and select Chain 1 or Chain 2 above the slots to edit each independently. Each chain accepts up to sixteen effects. 1 | 2 overlays both reports; 1 − 2 measures the actual audio output difference, including phase and latency differences. It does not align the chains' latency. The individual-chain controls also select which 2D spectrum to inspect. Reports are published together after the complete comparison batch finishes.

Choose Analyze. Auto analyze starts after an effect is added, removed, reordered or bypassed. Parameter, preset and measurement changes are analyzed after editing stops; the input-level slider commits when released. Continue editing while it runs; superseded reports are marked as describing the previous state. Repeat analysis reruns the complete measurement until cancelled. Cancel stops repetition and pending work and requests cancellation between processing blocks. A plug-in call in progress must return before cancellation completes.

Read the report

Every axis and tick shows its unit; phase is in degrees (°). Linear amplitude uses FS, where 1 FS is amplitude 1.0 (0 dBFS peak). The spectrum color scale uses dBFS. Automatic scaling and zoom keep minimum grid steps of 1 Hz, 1 dB/dBc/dBFS, 1°, 0.01%, 0.01 FS, 0.01 ms, 0.001 s, 0.1 μs or 1 sample, depending on the axis. Small measured values are preserved without magnifying numerical noise into a full-scale response.

Hover over a curve to see the series name and measured coordinates in the displayed units. In the 2D sweep spectrum, hover over a cell to see its measurement time, frequency band and peak level in dBFS. Color-range limits change the colors; the tooltip keeps the original measured level.

Continuous response curves use light display smoothing; measured coordinates and unavailable bins are preserved. Phase, impulse and captured waveform views retain straight sample connections to preserve discontinuities.

• Linear: frequency response, phase and IR. Select L→L, L→R, R→L or R→R. Use the controls below the main graph to switch Magnitude, Phase or IR; L + R overlays the two direct paths. The L/R–M/S toggle in the report header reroutes the excitation and folds the stereo output to mid and side, so every channel selector reads M/S instead of L/R. Select a rectangle on any graph (top-left to bottom-right) to zoom in, drag bottom-right to top-left to zoom out, drag bottom-left to top-right or double-click to reset, and use the wheel to pan or Ctrl+wheel to zoom an axis. Store retains a reference curve until Clear or a response-view change. Phase uses at least a ±3° range so small numerical errors do not dominate the display. Compensate measured delay to compare phase.
• Harmonics: H2–H8 in dBc and summed THD, measured with stable sine signals. 2D sweep spectrum plots the actual output against time and frequency, with color representing peak spectral dBFS. Select L/R, linear/log sweep, and color-range limits. Each sweep is measured independently. Harmonic and reflected aliasing lines can be seen across the whole output spectrum. Harmonics 1D overlays H2–H8 versus input frequency; frequency supports linear/log axes and magnitude supports percent of fundamental/dBc. THD shows the summed distortion. Display changes use the retained report. Orders beyond Nyquist are omitted rather than counted as zero.
• Hammerstein: each input power has its own independent 512-tap FIR filter, with a fitted DC offset. The static view projects each filter to its DC gain. Select the polynomial order from 3 to 7, and compare model predictions against an independent measured validation signal. Poor fits highlight the error readout. Per-order response overlays each order's contribution to the filter response. Compression, modulation and long tails often need more complex models.
• Distortion: single-tone THD and THD+N at a configurable test frequency, plus two-tone intermodulation distortion at fixed 60 Hz and 7 kHz carriers. Switch THD/IMD to inspect the measured output spectrum, including noise, harmonics, reflected aliases and modulation sidebands.
• Oscilloscope: sine, square, saw and triangle captures of the chain output. Time plots output against time with the input overlaid; Waveshaping plots output against input with a display delay that defaults to the reported latency.
• Dynamics: the Ramp view sweeps the input from −100 to 0 dBFS and plots the peak output transfer curve against a 1:1 reference; Attack / Release feeds three level segments and plots the input and output envelopes. Measurement settings configure ramp start/end, step and time per level (0.4–1.5 s), plus each segment's level and duration (0.01–5 s). Consecutive ramp levels retain signal history. Choose times longer than the effect's attack.
• Performance: average/P95/P99/maximum processing time, deadline budget use, reported latency and memory estimates. Measurements use the selected block size and the plug-in's real-time processing mode. The block-size graph compares average and P99 callback time at 64, 128, 256, 512 and 1024 samples. Processing speed controls elapsed time independently; pacing is excluded from callback timings.

Reports record actual input level and conditions. Nonlinear responses depend on level and signal history. Increase settle/tail duration for long delays and reverbs. Noise or large repeat differences reduce frequency/phase confidence.

Memory changes belong to the Electron main process, including shared libraries and concurrent work. They are estimates rather than exact plug-in allocations. Concurrent playback and system load also affect processing time.

The chain and report last until the window closes. Closing releases editors and instances. Unconfirmed resource release quarantines the session; restart Heron to reset it.