M1: DSP completion (compressor, listen mode, 4-voice doubler) + docs + tests#8
Merged
Conversation
Adds two new DSP stages requested for M1: a sibilance-listen/solo mode on the de-esser (DeEssListen) for tuning DeEssFreq by ear, and a hand-rolled, zero-latency, bit-exact-bypassable broadband downward compressor (GentleCompressor, Comp 0-100%) inserted between Air and the Doubler for glue before the signal is duplicated into doubled voices. Wires both through SeraphEngine, PluginProcessor's APVTS plumbing, and the v0.1 slider/toggle editor.
Grows Doubler from a symmetric two-voice L/R pair to four independently modulated voices, each at its own fixed pan position (a small-choir spread reached at DoubleWidth == 100%) rather than a single hard-panned pair. Gain-staging is compensated via a 2/numVoices factor so the added level exactly matches the original two-voice design when numVoices == 2. The outer pair keeps its original base delay/LFO-rate constants; two new inner voices were added at distinct delays and rates so all four voices decorrelate over time. True formant-preserving detune (the other part of the M1 doubler ask) is deferred - see docs/architecture.md and the M1 issue comment for the reasoning.
Extends coverage for the new DSP (GentleCompressor gain reduction and bit-exact bypass, De-Ess Listen output, four-voice doubler stereo image) and adds the M1 "broaden test coverage" scope in a new CoverageTests.cpp: a null-test sweep and a hot-settings finite-output sweep across the full documented sample-rate range (44.1-192 kHz), prepareToPlay sample-rate-change robustness, mono/stereo/rejected bus layouts, and a ~10.7 s simulated long-run stability sweep with a slowly modulated frequency checking for NaN/Inf and unbounded growth. Also extends the existing extreme-parameter, rapid-automation, and state-round-trip tests to cover the two new parameters. All 28 tests run in ~5 s locally.
Adds docs/manual.md: what Seraph is, where it sits in a symphonic-metal vocal chain, signal flow, a full parameter reference (name/range/default/unit/musical purpose), and mixing tips. Updates docs/architecture.md's diagram and per-stage sections for the Gentle Compressor and De-Ess Listen additions and the four-voice doubler (including the formant-preserving-detune deferral note), README's feature list/parameter table, and CLAUDE.md's Status/DSP sections to match the current signal path.
Documents the M1 DSP additions (Gentle Compressor, De-Ess Listen, four-voice doubler), the broadened test suite, and the new user manual, plus the formant-preserving-detune deferral, under a new [0.1.0] - 2026-07-14 section per Keep a Changelog. CMakeLists.txt's project/plugin version was already 0.1.0.
src/dsp/GentleCompressor.{h,cpp} (the compressor implementation added
in f886590) were left untracked by mistake; the build only worked
locally because they were already on disk. Adds them now.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Implements the M1 "DSP completion & test coverage" milestone and prepares Seraph for a v0.1.0 tag.
DSP (
Complete and refine the DSP, #1)Comp, 0-100%): a new, hand-rolled, zero-latency, feed-forward broadband downward compressor inserted between Air and the Doubler, so all four doubled voices track a consistent main signal.Comp == 0%is a bit-exact bypass (same convention asDeEss/Double). No auto makeup gain -Outputcompensates.DeEssListen, bool): solos the de-esser's detected sibilance band soDeEssFreqcan be tuned by ear, independent of the currentDeEssamount.DoubleWidth == 100%), replacing the single symmetric L/R pair. Gain-staging is compensated (2/numVoices) so it reduces to the exact original 2-voice behaviour atnumVoices == 2.Deferred (issue #1 left open): true formant-preserving detune (LPC/cepstral spectral-envelope correction). This is a substantially larger DSP feature that doesn't fit safely into the doubler's modulated-delay architecture without risking the plugin's zero-latency and bit-exact-bypass invariants. Full reasoning in
docs/architecture.md's Doubler section and in the issue comment.Tests (
Broaden test coverage, #2)Catch2 suite grown from 18 to 28 test cases:
prepareToPlaysample-rate-change robustness.Docs
docs/manual.md: full user manual (what Seraph is, where it sits in a symphonic-metal vocal chain, signal flow, complete parameter reference, mixing tips).docs/architecture.md,README.md,CLAUDE.mdupdated to match the new signal path and parameter set.CHANGELOG.md:Unreleasedcontent moved into a new[0.1.0] - 2026-07-14section documenting everything above, including the deferral.Version
CMakeLists.txt'sproject(Seraph VERSION 0.1.0 ...)was already 0.1.0 - no change needed.Test plan
cmake --build build --target Tests Seraph_Standalone --parallel 3- clean build, no warnings on touched files.ctest --test-dir build --output-on-failure- 28/28 tests pass locally (~5 s).Closes #2