a-kgeorge

2 maanden geledendiscuss

This is fantastic. I love how it makes Earth’s rotation feel immediate and visceral with zero equipment. The controls and overlays are really well thought out. It would be cool to have a search box to jump to a specific star or coordinate, even if most things in the ribbon are obscure.

a-kgeorge

2 maanden geledendiscuss

[dead]

a-kgeorge

8 maanden geledendiscuss

Who It’s For • Node backends running streaming DSP • Systems needing stateful pipelines + fault tolerance • Teams avoiding separate Python/C++ services • IoT gateways w/ intermittent workloads

Not ideal: • Browser-only DSP (use WebAudio) • Hard realtime (<1ms) embedded audio

Help Wanted • Portable NEON optimization (FFT + convolution) • Kafka integration (batching / backpressure / checkpoints) • Production war stories

Status STFT / Mel / MFCC Convolution + FFT (native C/C++) Redis persistence stable ARM functional (not optimized) 891 tests passing Kafka integration (seeking input)

Repos: https://github.com/a-kgeorge/dspx https://github.com/a-kgeorge/dspx-benchmark

a-kgeorge

8 maanden geledendiscuss

Code Example:

const pipeline = await createDspPipeline({ redisHost: "localhost", redisPort: 6379, stateKey: "dsp:user:ch1", });

pipeline .filter({type:"butterworth",mode:"bandpass",lowCutoffFrequency:20,highCutoffFrequency:450}) .stft({windowSize:512,hopSize:160}) .melSpectrogram({numMelBands:26}) .mfcc({numCoefficients:13});

await pipeline.process(audioData, {sampleRate:16000}); const state = await pipeline.saveState(); // worker dies → new worker await pipeline.loadState(state);

Platform • x64 + ARM • Tested on Pixel 9 Pro XL (stable; NEON tuning pending) • Node 18/20/22 • N-API v8 • Prebuilds for x64 (Linux/macOS/Win) • ARM compiles once → works like normal

Why not WASM? • Native threading • AVX2/SSE2/NEON • Zero-copy buffers to/from Redis • Clean integration w/ Node streams

a-kgeorge

8 maanden geledendiscuss

DSP Primitives: • STFT, Mel-spectrogram, MFCC • FIR/IIR (Butterworth, Chebyshev, Parks-McClellan) • Time + FFT convolution • Decimation, resampling, rectification • Hilbert transform, wavelet decomposition

Infrastructure • Multi-channel streaming (audio, biosignals, IoT) • Redis-backed persistence • TypeScript-first API w/ full type safety • SIMD: AVX2 / SSE2 / NEON

Performance (i5-12600T • Node 22.21.1 • 16GB RAM): • 1 pipeline → ~34M samples/s • 32 pipelines → ~101M samples/s • Small inputs → sub-ms p50 • With Redis → ~31ms p50 / ~39ms p99 • Zero leaks over 10k+ save/restore cycles

Benchmark repo: https://github.com/a-kgeorge/dspx-benchmark

a-kgeorge

8 maanden geledendiscuss

I built dspx after hitting a wall with real-time EMG processing in Node.js.

Problem: 2 kHz × 8 channels → filters → decimation → FFT. JS is too slow; the Python sidecar is complex; WASM lacks native threading.

Solution: Native C++ DSP with Redis-persisted state. Workers process a chunk, save state in ~1–2ms, die, and resume elsewhere. Serverless DSP without losing filter history.

Repo: https://github.com/a-kgeorge/dspx Benchmarks: https://github.com/a-kgeorge/dspx-benchmark npm: https://www.npmjs.com/package/dspx

Submissions

[untitled]

[untitled]

Show HN: Dspx – Native C++ DSP Library for Node.js Built for AWS Lambda (1.3MB)

Show HN: dspx — Serverless-friendly DSP for Node.js (native C++ + Redis state)

comments