- Add 15 Tier 5 pipeline co-sim tests (cross-layer): decimator, MTI,
Doppler, DC notch, CFAR, and full-chain golden match
- Add 4 SoftwareFPGA tests (test_v7): golden fixture integrity,
process_chirps wiring validation against manual chain
- Wire compare_doppler.py into run_regression.sh with 3 scenarios
(stationary, moving, two_targets) replacing single unvalidated run
- Harden _load_npy to raise on missing files; expand sentinel check
to all 14 golden artifacts
- DC notch test verifies both zero and pass-through bins
Test counts: 172 python + 55 cross-layer + 21 MCU + 29 FPGA = 277 total
- Correct carrier from 10.525/10 GHz to 10.5 GHz (verified ADF4382 config)
- Correct range-per-bin from 4.8/5.6/781.25 m to 24.0 m (matched-filter)
- Correct velocity resolution from 1.484 to 2.67 m/s/bin (PRI-based)
- Correct processing rate from 4 MSPS to 100 MSPS (post-DDC)
- Correct max range from 307/5000/50000 m to 1536 m (64 bins x 24 m)
- Add WaveformConfig.pri_s field (167 us PRI for velocity calculation)
- Fix short chirp chirp_complete deadlock (Bug A)
- Remove dead short_chirp ports, rename long_chirp to ref_chirp (Bug B)
- Fix stale latency comment 2159 -> 3187 cycles (Bug C)
- Create radar_params.vh as single source of truth for FPGA parameters
- Lower RadarSettings.cpp map_size validation bound from 1000 to 100
- Add PLFM hardware constants to golden_reference.py
- Update all GUI versions, tests, and cross-layer contracts
All 244 tests passing (167 Python + 21 MCU + 29 cross-layer + 27 FPGA)
Add SoftwareFPGA class that imports golden_reference functions to
replicate the FPGA pipeline in software, enabling bit-accurate replay
of raw IQ, FPGA co-sim, and HDF5 recordings through the same
dashboard path as live data.
New modules: software_fpga.py, replay.py (ReplayEngine + 3 loaders)
Enhanced: WaveformConfig model, extract_targets_from_frame() in
processing, ReplayWorker with thread-safe playback controls,
dashboard replay UI with transport controls and dual-dispatch
FPGA parameter routing.
Removed: ReplayConnection (from radar_protocol, hardware, dashboard,
tests) — replaced by the unified replay architecture.
150/150 tests pass, ruff clean.
Bug #1 — Range calibration for Raw IQ Replay:
- Add WaveformConfig dataclass (models.py) with FMCW waveform params
(fs, BW, T_chirp, fc) and methods to compute range/velocity resolution
- Add waveform parameter spinboxes to playback controls (dashboard.py)
- Auto-parse waveform params from ADI phaser filename convention
- Create replay-specific RadarSettings with correct calibration instead
of using FPGA defaults (781.25 m/bin → 0.334 m/bin for ADI phaser)
- Add 4 unit tests validating WaveformConfig math
Bug #2 — Demo + radar mutual exclusion:
- _start_demo() now refuses if radar is running (_running=True)
- _start_radar() stops demo first if _demo_mode is active
- Demo buttons disabled while radar/replay is running, re-enabled on stop
Bug #3 — Refactor adi_agc_analysis.py:
- Remove 60+ lines of duplicated AGC functions (signed_to_encoding,
encoding_to_signed, clamp_gain, apply_gain_shift)
- Import from v7.agc_sim canonical implementation
- Rewrite simulate_agc() to use process_agc_frame() in a loop
- Rewrite process_frame_rd() to use quantize_iq() from agc_sim
v7/__init__.py: wrap workers/map_widget/dashboard imports in try/except
so CI runners without PyQt6 can still test models, processing, hardware.
test_v7.py: skip TestPolarToGeographic when PyQt6 unavailable, split
TestV7Init.test_key_exports into core vs PyQt6-dependent assertions.
Jason