v1.0 completion gates¶
Evidence snapshot: 2026-10-29, based on the committed reports linked below. This
updates the completion plan originally reviewed at c7d5542 on 2026-09-19.
It does not establish the current GitHub issue state or announce a release.
The existing v0.1.0 release instructions remain separate.
| Workstream | Evidence already recorded | Remaining acceptance or follow-up |
|---|---|---|
| #48 OFDM | Fixed-point models, RTL and implementation reports; DMA fabric loopback on a board: 0 errors in 3,072,000 bits. The radio image has been built. | Run the radio image through a protected, attenuated RF loopback, then two boards; report BER/counts, detection failures, EVM/CFO, clipping and spectrum. Fabric loopback does not qualify the RF receiver. |
| #29 gain/overload | Two-board overload grid and gain steps at 433/915/2400 MHz. Levels are relative. | Measure the complete passive path S21 with its reference plane and uncertainty before converting settings to absolute input power or claiming connector-to-connector limits. |
| #37 NanoVNA | RU/EN Labs 10.5/10.6, parsers and templates. The conducted-tone report records a subsequent 30 dB attenuator check from 50 MHz to 1 GHz. | Archive the calibration, photos and Touchstone/CSV exports; measure cable and splitter loss as well. An attenuator check alone does not calibrate the complete path. |
| #58 AGC | Two-board gain transitions and frame-aligned level steps compare manual, slow and fast gain modes. | The slow-attack phase step still needs a common-reference measurement. Keep RFIC gain behavior separate from the modem's level-drop failure and its DC-blocker follow-up. |
| #55 mailbox | The message modem report records the mapped mailbox, board images, one-board use and two-board mailbox reception: 498/500 messages at the stated operating point, no wrong payload accepted. Chat/ARQ and fragmentation extend the measured path. | The original message-delivery target has recorded board evidence. Verify deployed-image RX_STOP/turnaround and chat/fragment latency for the later #71 change; the older mailbox measurements do not validate that change. |
The course status summarizes block readiness. This table tracks the acceptance workstreams; the linked reports retain the configurations, counts and limitations of individual experiments. Historical reports are not rewritten as though their measurements used a later image.
Repeatable offline preflight¶
For environment setup on Windows and Linux, use the reproducibility guide. Choose the checks that match the change:
python tools/run_local_ci.py --software-only
python tools/tasks.py docs
python tools/run_local_ci.py --hdl-only
The default python tools/run_local_ci.py also rebuilds representative Python
labs. HDL checks cover the canonical Block 5 suite, CSS and default OFDM benches
with radio parameter variants. Separate Actions workflows cover additional
OFDM configurations. These commands run software or simulations, not a board.
Closing rule¶
For every hardware step keep the commit/bitstream hash, configuration, raw-data checksums, number of all attempts and reasons for losses. Compare evidence with the exact acceptance criteria before closing a workstream. A synthetic self-test, a constellation without counters or a series from another image is insufficient. Mark MATLAB checks separately from Python/RTL and state missing licences or artifacts explicitly. Complete these acceptance steps before expanding the PHY.