scripts/fatness.py measures momentary bass fatness of rendered files with
the final model from Schmedeke's dissertation (section 5.4.3, adjusted R²
.773): two logistic terms on the C-weighted level and the MPEG-7 harmonic
ratio below 200 Hz, over 50 ms windows, after normalising to -23 LUFS
(BS.1770 integrated, gated).
The thesis does not state the dB(C) of its reference, so --offset maps
dBFS to dB(C) (default 100). On the seven Devath references, drops read
4-6, breakdowns near 0, and kick gaps show as 50 ms dips; SPL<200 medians
of 67-69 dB(C) suggest the offset may be 2-5 dB low.
Also corrects docs/ideas.md, which quoted the thesis's preliminary linear
model instead of the final one.
Adds a Research findings section with sourced numbers (Schmedeke fatness
model, virtual bass weighting, group-delay audibility, timbral tension,
coupled controls, layer coherence, heard timing of layered hits, loopback
FM, beat rates of detuned layers) and where each goes in the suite.
New problems 19-21, new ideas Braze, Arc and the fatness lane, additions
to Anvil, Weld, Crucible, Forge and Loupe, and a research track in the
build order. Also updates stale release notes for Patina and Planish.
Upward and downward compression in four bands whose edges are harmonic
numbers of the tracked note (default 1, 2-4, 5-16, 17 and up), so the
same setting squashes the same harmonics on every note. A fixed-band
OTT puts harmonic 2 of F1 and of C2 in different bands.
- Three Linkwitz-Riley splits at sqrt(k(k+1)) x f0 with allpass
compensation; at Depth 0 the bands sum within 0.05 dB.
- Band level is the RMS over one period of the note, so heavy settings
do not ripple at f0: a 40 Hz sine at Drive +12 dB, Time 10 % gets no
harmonics above -110 dB.
- At Depth 100 % on a saw, harmonics 1-12 change by the same amount on
F1, C2 and G2 (within 0.2 dB).
- Depth, Drive (on the detector only), Upward, Downward, Time, a gain
per band; edges Low 1-8, Mid 2-16, High 3-48 with four splits.
- Pitch from audio with about 67 ms lookahead, or from MIDI.
- Editor: harmonic ruler with draggable edges, level and gain per band.
- Catalog, workflows, editor-check, docs, signal chain and site copy.
docs/chain.md gives the plugin order for a bass track and the bus, the
latency each plugin adds, and the measured octave error of downstream
note trackers after Knurl: from C2 up the tracker reads an octave low
about 85 % of the time, and timing jitter adds 5-25 false note starts
per second. Knurl and Anvil guides point there and recommend MIDI as
the pitch source after Knurl.
docs/ideas.md gains the production-flow audit, next steps for every
released plugin, new ideas (Forge, Planish, Quench, Scribe, Bevel,
Peen) and a build order.
Knurl follows the played note and treats every cycle of the wave on its
own: a repeating level pattern over 1-16 cycles makes subharmonics in tune
with every note, random level and timing per cycle add rattle, and a noise
burst per cycle, highpassed at a multiple of f0, adds grit. Amounts are
counted in cycles, so the texture is the same on F1 and on C2.
- Cycle clock from the shared note tracker, restarted at each note; the
audio is delayed by the lookahead plus 4.2 ms of timing-jitter headroom.
- Per-cycle randomness from a hash of the cycle number: Same every note
or Free.
- Patterns Off, Double, Triple, Fry (default), Gallop, Stutter or drawn;
grit switchable per step. Pattern persisted as knurl-pattern.
- Editor: last eight cycles overlaid, pattern bars with the playing step.
- Tests match the prototype: Double gives f0/2 at -8.19 dB, Triple f0/3
at -10.68 dB, equal on F1 and C2, no added clicks, bit-exact when off,
no allocation on the audio thread.
Weld hears the other bass layer through the sidechain and fits polarity
plus up to four RBJ allpass sections to the measured cross-spectrum, so
the layers sum instead of cancelling. No latency; the layer on its own
keeps its sound.
- Worker thread: Welch cross-spectrum and coherence, Nelder-Mead fit with
1/f weighting, a per-band loss penalty and model selection by share of
the achievable gain; GCC-PHAT lag detection flags a late layer.
- Audio thread: allocation-free chain with a 40 ms warm-up and 40 ms
crossfade per new fit; mono blocks go to the worker through a lock-free
queue.
- Follow / Learn / Hold; the fit is persisted and restored without a
crossfade.
- Editor: per-band loss and recovery against a perfect alignment, phase
before and after, recovered dB, time offset and status.
- Tests for flat magnitude, crossfade, fit targets, Follow hysteresis,
Learn, silence, realtime allocation and the persisted fit in the host.
Set the CLAP vendor of all seven plugins to "Bass Foundry" and name the
combined preview ZIP bass-foundry-windows-x86_64.zip. CLAP IDs and saved
state stay unchanged.
Bring development.md and releases.md up to date with all seven products,
the shared crates, the test types, the steps to add a product and the
published releases. Add texture problems and seven texture plugin ideas
to ideas.md.
Lathe places EQ bands on harmonics 1-32 of the played note, so a cut on
harmonic 3 stays on harmonic 3 on every note. Temper normalises each
note's level before an oversampled shaper and tunes its floor, push and
bright filters to the note, so every note distorts the same.
The note tracker moves from Anvil into common-dsp, together with a
state-variable filter, 4x oversampler and harmonic meter. Anvil's
behaviour and editor render are unchanged.
docs/ideas.md collects all plugin ideas and their status.
nvrl