algorithm-evidence-matrix.md

DSP Algorithm Evidence Matrix

This matrix shows how each DSP algorithm in the repository is backed by implementation, tests, benchmarks, plots and possible FPGA mapping notes.

The goal is to make the repository reviewable as an engineering portfolio artifact, not only as a collection of C++ source files.

Evidence matrix

Algorithm / area	Implementation	Unit tests	Benchmark	Plot / report	FPGA mapping	Next improvement
FIR low-pass design	active	active	active	FIR response figure	partial	add coefficient quantization example
Direct convolution	active	active	active	benchmark report	possible	compare against FFT overlap-save by signal size
FFT overlap-save convolution	active	active	active	benchmark dashboard	future	add block-size tuning notes
Goertzel detector	active	active	active	tone-detection plot	possible	add fixed-point detector notes
Batched Goertzel	active	active	active	benchmark report	possible	add SIMD comparison table
GCC-PHAT delay estimation	active	active	partial	delay-estimation plot	future	add noisy/multipath robustness sweep
Rational L/M resampler	active	active	active	generated response / demo	possible	add alias rejection metric
IQ file analysis demo	active	smoke/demo	planned	demo output	not direct	add small public CI16 vector
SIMD / AVX2 path	optional	active where supported	active	performance report	not direct	add scalar-vs-AVX2 summary table
CMake package export	active	install smoke possible	not applicable	README example	not applicable	add downstream consumer CI smoke

Reviewer checklist

A reviewer should be able to answer:

• Which algorithms are implemented and tested?
• Which algorithms have performance evidence?
• Which algorithms have generated visual artifacts?
• Which algorithms are candidates for fixed-point or FPGA mapping?
• Which claims are protected by CI and which are only documented?

Recommended artifact standard

Each major algorithm should ideally have:

source implementation
unit test
benchmark case
small generated plot
short engineering note
known limitations

Suggested next additions

1. Add a scalar-vs-AVX2 benchmark summary table.
2. Add a small deterministic IQ vector under data/ or tests/data/.
3. Add fixed-point notes for FIR and Goertzel.
4. Add a downstream CMake consumer smoke test.
5. Add a bridge page that maps this repository to zynq-sdr-course labs.