gbozo/no-sdr v2.3.0

on GitHub

What's New

Server-side RDS Decoding

• RDS decoder fully wired into the Opus pipeline (was silently emitting null since day 1)
• Complete rewrite of the Go RDS decoder to match the IEC 62106 standard — same algorithm as the working TypeScript client-side decoder
• PTY name lookup table, proper biphase clock recovery, differential Manchester decoding, correct CRC syndrome calculation
• PS, RT, PI, PTY fields streamed to client as JSON on group completion (~11 Hz max update rate)

Wire-Driven Audio Protocol

• Every frame is now self-describing — client is a dumb terminal
• MSG_IQ (0x02): 6-byte header with sample rate, channels, reserved
• MSG_IQ_ADPCM (0x09): 10-byte header with sample count, sample rate, channels, reserved
• MSG_AUDIO_OPUS (0x0C): unchanged (was already self-describing)
• Client demodulator sample rate driven entirely from wire headers — no proactive resets needed

Opus Chipmunk Fix

• Root cause: UpdateSampleRate(240000) was called before SetMode('wfm') on mode switch, so decimFactor = 240000/48000 = 5. FmStereoDemod already decimates 240k→48k internally, so the pipeline applied 5× decimation again → audio at 5× speed (chipmunk). Fixed: SetMode() now recalculates decimFactor with the correct new mode.
• Opus decoder lifecycle is now owned exclusively by MSG_AUDIO_OPUS handler — setMode(), tune(), setIqCodec() no longer touch the decoder
• opusDecoderChannels updated eagerly to prevent double-init races

Admin FFT Hot-Apply Fix

• Saving fftSize / fftFps in the admin panel now immediately applies to the running FFT processor
• Root cause: pointer aliasing — d.profile and the updated config were the same pointer, so the old/new comparison was always equal. Fixed by unconditionally rebuilding FftProcessor on SwitchProfile.

Other Fixes

• Fixed 50 Hz ticking / rumbling on ADPCM and raw IQ codecs — updateResampleRatio() was called on every frame, resetting resamplePhase at 50 Hz. Now only called when sample rate actually changes.
• stereoEnabled propagated to OpusPipeline at construction so first packets use correct channel count
• ws.Client defaults StereoEnabled = true so mono pipeline is never built before client sends preference
• All mode-change filter state leaks fixed: rumbleFilter, autoNotch, hiBlend IIR delay lines cleared; softMuteGain reset; stereoDetected store flag cleared
• 8 pre-existing TypeScript type errors resolved (clean build)