feat(weld): phase alignment for layered bass, release 0.1.0
Check and Windows preview / check (anvil) (push) Successful in 2m0s
Check and Windows preview / check (crucible) (push) Successful in 1m29s
Check and Windows preview / check (lathe) (push) Successful in 1m27s
Check and Windows preview / check (patina) (push) Successful in 1m53s
Check and Windows preview / check (rivet) (push) Successful in 1m29s
Check and Windows preview / check (temper) (push) Successful in 1m29s
Check and Windows preview / check (weld) (push) Successful in 2m38s
Check and Windows preview / previews (push) Successful in 1m28s
Independent plugin release / plan (push) Successful in 8s
Check and Windows preview / windows (push) Successful in 3m14s
Independent plugin release / build (push) Successful in 3m50s
Independent plugin release / publish (push) Successful in 9s

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.
This commit is contained in:
nvrl committed 2026-11-06 08:31:41 +01:00
1 parent 0a0917ac56
commit 9b3ddeaf56
26 files changed
+2898 -144

No files matched your search

+1 -1
View File
@@ -32,7 +32,7 @@ NIH-plug Git dependency stays out of the root workspace. Host identities and
saved state are separate for every product.
The engines run allocation-free on the audio thread. Analysis that is too heavy
for the audio thread (Rivet's generation jobs, and Weld's planned phase fit) runs on a worker
for the audio thread (Rivet's generation jobs, Weld's cross-spectrum and fit) runs on a worker
thread. Data goes to the editor through lock-free queues.
## Checks