With Blackman-Harris window and proper cutoff (~0.24), 32 taps still
provides 60+ dB stopband rejection. Halves the per-sample MAC count
from 192 to 96 across the three resampler channels.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Signed-off-by: Stan Grams <sjg@haxx.space>
Increase polyphase resampler phases from 32 to 128 for finer fractional
sample positioning. Replace Hamming window with Blackman-Harris for ~92 dB
stopband rejection. Add pilot notch on composite signal before diff demod
to prevent 19 kHz × 38 kHz intermod products in the stereo difference
path.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Signed-off-by: Stan Grams <sjg@haxx.space>
Increase polyphase resampler taps from 16 to 64 for sharper anti-alias
stopband rejection. Upgrade sum/diff lowpass filters from 4th-order to
6th-order Butterworth (three biquad stages) for ~36 dB/octave rolloff,
improving stereo separation by better rejecting the 38 kHz subcarrier
residuals.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Signed-off-by: Stan Grams <sjg@haxx.space>
The fixed WFM_RESAMP_CUTOFF of 0.94 passed frequencies up to 94 kHz at
200 kHz composite rate, while the output Nyquist is only ~24 kHz. The
38 kHz demod products in the stereo diff path were only ~31 dB attenuated
by the Butterworth and aliased back into 10-20 kHz audio, causing treble
corruption in stereo mode. Now the cutoff is computed as
audio_rate / composite_rate, properly anti-aliasing the polyphase
resampler output.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Signed-off-by: Stan Grams <sjg@haxx.space>
32 taps caused audio silence on real signals. Revert to 16 taps
which works correctly.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Signed-off-by: Stan Grams <sjg@haxx.space>
Replace the low-accuracy 0.273 linear atan approximation with a
7th-order minimax polynomial (max error ~2.4e-7 rad vs ~0.004 rad).
Use branchless |y|>|x| argument reduction instead of y/x division
with quadrant fixup, avoiding division-by-zero and NaN branches.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Signed-off-by: Stan Grams <sjg@haxx.space>
Replace fast_atan2 polynomial approximation with f32::atan2 in the WFM
stereo decoder's FM discriminator and pilot PLL. The approximation
introduced harmonic distortion (~0.22 deg error) that manifested as
treble artifacts on strong/overdeviated broadcast signals.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Signed-off-by: Stan Grams <sjg@haxx.space>
Normalize IQ samples exceeding unit magnitude before the FM
discriminator. The discriminator only uses phase, so clamping
amplitude prevents overdeviated signals from producing clipped
composite baseband without losing frequency information.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Signed-off-by: Stan Grams <sjg@haxx.space>
Doubles the polyphase FIR length for the composite-to-audio rate
converter, improving stopband rejection from ~25 dB to ~50 dB.
This reduces treble distortion from imaging artifacts.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Signed-off-by: Stan Grams <sjg@haxx.space>
- Set STEREO_DIFF_BW_HZ = AUDIO_BW_HZ so both filter paths have
identical group delay (improves multitone separation by ~10 dB).
- Zero out STEREO_SEPARATION_PHASE_TRIM (unnecessary with matched filters).
- Replace gradual blend ramp with binary blend: full stereo at pilot
lock, mono when unlocked. The hysteresis thresholds already handle
noisy signals.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Signed-off-by: Stan Grams <sjg@haxx.space>
Test L-only and R-only signals with tones at 400, 2000, 8000 and
14000 Hz to catch frequency-dependent group delay and phase trim
issues that the single 1 kHz test misses.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Signed-off-by: Stan Grams <sjg@haxx.space>
Restore AUDIO_BW_HZ to 15.8 kHz for cleaner mono path, widen
STEREO_DIFF_BW_HZ to 18 kHz for better high-frequency stereo detail,
and raise STEREO_MATRIX_GAIN from 0.30 to 0.50 (mathematically correct
unity gain for the L=(S+D)/2 stereo matrix).
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Signed-off-by: Stan Grams <sjg@haxx.space>
Raise both AUDIO_BW_HZ and STEREO_DIFF_BW_HZ to 18 kHz so the L+R and
L-R filter paths have identical group delay across the full audio band.
The previous mismatch (15.8 vs 14.5 kHz) caused frequency-dependent
phase errors in the stereo matrix that degraded real-world separation.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Signed-off-by: Stan Grams <sjg@haxx.space>
Raise the stereo blend floor from 0.55 to 0.75 at pilot lock and lower
the full-blend ceiling from stereo_detect_level 0.92 to 0.70. This
gives real-world signals with moderate pilot strength much better L/R
separation (~17 dB immediately at lock vs ~5 dB before) and reaches
full unity blend sooner.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Signed-off-by: Stan Grams <sjg@haxx.space>
Fix pre-existing compilation failures in four test call sites that were
missing the wfm_denoise: bool argument added to ChannelDsp::new() and
SdrPipeline::start().
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Signed-off-by: Stan Grams <sjg@haxx.space>
- Raise audio LPF cutoff from 15 kHz to 17 kHz to pass full FM stereo
audio bandwidth without excessive HF rolloff
- Replace 2-point linear interpolation resampler with 4-point Hermite
cubic spline for a much flatter passband up to 17 kHz
- Add FM discriminator gain normalization (fm_gain = fs / 150000) so
±75 kHz deviation maps to ±1.0 regardless of composite sample rate,
stabilizing stereo carrier amplitude reconstruction
- Double pilot PLL proportional (0.0015→0.003) and integral
(0.00002→0.00005) gains for faster lock and better tracking
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Signed-off-by: Stan Grams <sjg@haxx.space>
The 19 kHz pilot notch was applied only to the L+R sum path, introducing
~22° of phase shift at 15 kHz relative to the L-R diff path. This phase
mismatch caused interchannel crosstalk (≈ −14 dB separation at 15 kHz).
Fix: remove the notch from the sum processing chain so both sum and diff
pass through identical 4th-order Butterworth LPFs, giving phase-coherent
demodulation across the full audio band. The notch is relocated to the
mono output branch where phase alignment with the diff channel is not
required. Pilot rejection on the stereo L/R outputs is still adequate
(~28 dB) from the combined LPF + deemphasis response at 19 kHz.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Signed-off-by: Stan Grams <sjg@haxx.space>
Add a server-side toggle for the multiband stereo denoiser so it can be
enabled or disabled at runtime without restarting the server.
Backend (trx-backend-soapysdr):
- Add `denoise_enabled: bool` to `WfmStereoDecoder`; gate multiband
blend behind it (falls back to uniform single-band blend when off)
- Add `set_denoise_enabled()` method on `WfmStereoDecoder`
- Propagate `wfm_denoise: bool` through `ChannelDsp`, `SdrPipeline`,
and `SoapySdrRig`; add `set_wfm_denoise()` at each layer
- Include `wfm_denoise` in `filter_state()` so it flows into snapshots
Protocol / core (trx-core, trx-protocol, trx-server):
- Add `SetWfmDenoise(bool)` to `RigCommand` and `ClientCommand`
- Add default `set_wfm_denoise()` trait method to `RigCat`
- Handle `SetWfmDenoise` in `rig_task.rs` and update `RigFilterState`
- Add `wfm_denoise: bool` (default `true`) to `RigFilterState`
Frontend (trx-frontend-http):
- Add `POST /toggle_wfm_denoise` endpoint
- Add "Denoise On/Off" button next to the stereo/mono audio picker
- Sync button state from SSE filter snapshot (`update.filter.wfm_denoise`)
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Signed-off-by: Stan Grams <sjg@haxx.space>
Split the L-R diff channel into three frequency bands at audio rate and
apply SNR-weighted blending per band driven by pilot magnitude:
- 0–2 kHz: blend¹ (most stereo — low frequencies have best SNR)
- 2–8 kHz: blend² (moderate noise reduction)
- 8–15 kHz: blend⁴ (aggressive noise reduction — hiss-prone range)
Move blend from composite rate to audio rate so the crossover filters
(2nd-order Butterworth at 2 kHz and 8 kHz) operate at 48 kHz and the
pilot blend is linearly interpolated per audio sample for smooth
transitions. Unblended diff is now stored in prev_diff; prev_blend
tracks the blend value for the same interpolation.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Signed-off-by: Stan Grams <sjg@haxx.space>
Three bugs made the AM path sound wrong:
1. AGC attack too fast (5 ms). The slowest audio a broadcast AM station
can transmit is ~50 Hz (20 ms period). A 5 ms attack lets the AGC track
individual audio cycles, which causes severe pumping and amplitude
distortion. Change to 500 ms attack / 5 s release so the AGC only
responds to slow carrier-amplitude fading, not the audio modulation itself.
2. Bandwidth too narrow. The IQ filter cutoff is audio_bandwidth_hz / 2,
so the previous 6 000 Hz setting gave only 3 kHz audio bandwidth.
AM broadcast sidebands extend to ±4.5–5 kHz; raise the default to
12 000 (cutoff 6 kHz) to cover the full audio band.
3. DC blocker rate inconsistent. For AM the demodulated magnitude is
always ≥ 0 and the DC component equals the carrier amplitude; only true
DC needs removing. Unify all non-WFM modes to r = 0.9999 (corner
≈ 0.76 Hz @ 48 kHz), which strips carrier DC without touching any
audible bass content.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Signed-off-by: Stan Grams <sjg@haxx.space>
CW signals in SDR are centred at an audio offset (e.g. 700 Hz) by the
upstream FIR filter, so demodulating as USB (taking the real part) produces
the correct side-tone. The previous magnitude/envelope approach produced a
DC pulse per key press with no audible tone.
Re-enable the DC blocker for CW/CWR (r = 0.9999): the output is now audio
that can carry a DC offset from BFO frequency error, identical to USB.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Signed-off-by: Stan Grams <sjg@haxx.space>
Add SoftAgc — a fast-attack/slow-release envelope AGC with a max-gain cap
— to all demodulated audio paths so that switching between modes (WFM, AM,
SSB, CW, FM) no longer produces large volume jumps. AGC is applied after
every demodulator, including WFM, with a shared target level of 0.5.
Add per-mode DC blocking (DcBlocker) for USB/LSB/AM/FM/DIG paths to remove
carrier frequency-offset DC from the FM discriminator and LO bleedthrough in
SSB. CW is excluded because high-passing a non-negative envelope creates
negative-going artifacts on each key release; WFM already has internal DC
blockers on each output channel.
AGC time constants are tuned per mode:
CW/CWR – 1 ms attack / 50 ms release (follows individual dots/dashes)
AM – 5 ms attack / 200 ms release (tracks fading carriers)
all else– 5 ms attack / 500 ms release (suits voice and data)
Simplify demod_am and demod_cw: remove per-block peak normalisation and DC
removal that caused block-boundary level discontinuities ("pumping"). Both
now return raw magnitudes and rely on the downstream DC blocker and AGC for
normalisation.
DIG is already wired as Passthrough (identical to USB); no change needed.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Signed-off-by: Stan Grams <sjg@haxx.space>
Replace one-pole sum/diff filters with a proper 4th-order Butterworth
cascade (Q = 0.5412 / 1.3066) at 15 kHz. This reduces pilot tone
leakage from −4 dB to −12 dB at 19 kHz and suppresses the 38 kHz DSB
carrier from −9 dB to −32 dB, significantly improving stereo crosstalk.
Add a biquad notch at 19 kHz on the L+R channel to eliminate the residual
pilot tone that would otherwise be audible after downsampling to 48 kHz.
Replace nearest-neighbor (sample-hold) resampling with linear interpolation
inside WfmStereoDecoder. The output sample is now placed at the exact
fractional position between the two adjacent composite samples using the
phase accumulator state, removing timing jitter and harmonic distortion on
sustained tones.
Add DC blockers (pole at 0.9999, corner ≈ 0.75 Hz at 48 kHz) to all audio
outputs to remove carrier frequency-offset DC from the FM discriminator
without any audible bass roll-off.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Signed-off-by: Stan Grams <sjg@haxx.space>
sjgandClaude Opus 4.6