doppler-dsp 0.3.6

Dead-simple, ultra-fast digital signal processing.

doppler

Dead-simple, ultra-fast digital signal processing.

doppler is a lean C99 signal processing library built for one goal: maximum throughput with minimum friction — from any language. The full DSP stack lives in one portable core with paper-thin Python bindings and a Rust FFI. No runtime surprises, no framework lock-in.

What's inside

• NCO / LO — 32-bit phase accumulator, 2¹⁶-entry LUT, AVX-512 batch generation, FM ctrl port
• FIR filter — AVX-512 complex taps, CI8/CI16/CI32/CF32 input types
• FFT — 1D and 2D, selectable backend (FFTW or pocketfft)
• Polyphase resampler — continuously-variable rate, built-in 4096-phase × 19-tap Kaiser bank (60 dB)
• Halfband decimator — dedicated 2:1 decimator exploiting halfband symmetry; 375 MSa/s at 60 dB
• DDC / DDCR — digital down-converter for complex and real ADC input; Architecture D2 for ~2× savings on real input
• Accumulator — F32 and CF64 running accumulators with configurable window
• Delay — CF64 sample delay line
• Signal streaming — low-latency ZMQ transport (PUB/SUB, PUSH/PULL, REQ/REP)
• Circular buffers — double-mapped ring buffers for zero-copy, lock-free IPC (F32/F64/I16)
• Multi-language — clean C ABI; Python bindings for all modules and Rust FFI

Benchmarks

Throughput depends heavily on hardware, compiler, and SIMD availability. On a Ryzen 7 AI 350 (16 GB, -O2), typical figures range from hundreds of MSa/s (FFT, FIR complex, resampler) to tens of GSa/s (raw NCO phase accumulator).

To measure on your machine:

make bench              # C + Python; saves JSON to benchmarks/history/
make build              # then run C binaries directly, e.g.:
./build/native/src/fir/bench_fir_core
./build/native/src/hbdecim/bench_hbdecim_core
./build/native/src/resamp/bench_resamp_core

Historical Python benchmark snapshots are in benchmarks/history/.

Quick example

C:

#include "fft/fft_core.h"
#include <complex.h>

dp_fft_t *fft = dp_fft_create(1024, -1, 1);
dp_cf32_t in[1024], out[1024];
/* ... fill in[] ... */
dp_fft_execute_cf32(fft, in, 1024, out);
dp_fft_destroy(fft);

Python:

import numpy as np
from doppler.spectral import FFT

x = np.random.randn(1024).astype(np.complex64)
spectrum = FFT(1024).execute(x)

DDC (tune a carrier to baseband):

from doppler.ddc import DDC
import numpy as np

ddc = DDC(norm_freq=-0.1, num_in=4096, rate=0.25)
x = np.random.randn(4096).astype(np.complex64)
y = ddc.execute(x)   # CF32, len ≈ 1024

Documentation

Full docs at doppler-dsp.github.io/doppler

Document	Contents
Quick Start	Build, install, run the examples (Docker quickstart included)
Build Guide	CMake options, platform notes, Python setup, Docker details
Architecture	Design overview and layer diagram
Examples: C	C code examples — FFT, FIR, NCO, streaming
Examples: Python	Python code examples
Examples: Streaming	PUB/SUB and PUSH/PULL examples

| CLAUDE.md | Development notes and project context (for contributors) |

Build

make          # build (Linux/macOS; MSYS2 on Windows)
make test     # run CTest suite

Or with Python bindings:

make pyext && uv sync

See Build Guide for platform-specific instructions and all CMake options.

Licensing

The doppler source code is MIT-licensed.

If built with FFTW support (default), the resulting binary links against FFTW, which is licensed under the GNU General Public License (GPL). In this case, the distributed binary is covered by the GPL.

If built with -DUSE_FFTW=OFF, the pocketfft backend is used instead. pocketfft is BSD-3-Clause-licensed (see POCKETFFT_LICENSE) which is compatible with the MIT-license and so the resulting binary remains MIT-licensed with BSD-3-Clause licensed FFT features.

See Build Guide for details and the installed LICENSE files.