Extends the existing FM/DAB ANTCAP EEPROM persistence pattern
(Eeprom24aa::writeFmAntCap/writeDabAntCap) to the crystal trim found by
POST /api/tuner/xtal-calibrate, which previously only applied live and
was lost on every reboot.
- Eeprom24aa gains readXtalCalibration()/writeXtalCalibration() at word
addresses 0x02 (ibias), 0x03 (ctun), 0x04-0x07 (xtalFreqHz,
big-endian), right after the existing FM/DAB ANTCAP bytes.
- HardwareBootstrap::boot() now boots ADAU1701 before Si4684 (needed so
the EEPROM read, which borrows ADAU1701's I2C bus, can happen before
Si4684's boot() call, which takes the crystal trim as an argument),
loads the saved trim if present, and falls back to the compiled-in
defaults (ibias=72, ctun=0, xtalFreqHz=19199750) otherwise.
- POST /api/tuner/xtal-calibrate now persists every successful live
recalibration automatically ("persisted":true/false in the response)
via a new saveXtalCalibration()/net::AntennaCalibration::saveXtal
bridge, mirroring the ANTCAP save pattern.
Verified live: boot log confirms "Xtal not calibrated" before the first
save, "Xtal calibration loaded: ibias=72 ctun=0 xtal_freq_hz=19199750"
after, surviving a reboot; DAB/FM tuning unaffected (DAB CNR 17-19dB,
locked).
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
281 lines
15 KiB
Markdown
281 lines
15 KiB
Markdown
# TODO — DigiRadio firmware
|
||
|
||
Agent task list and hardware-in-the-loop backlog. Working directory for all
|
||
commands is `Software/`.
|
||
|
||
**Current firmware:** `0.9.0` — everything in 0.8.5, plus: Si4684 RF
|
||
blackout root-caused and fixed (real FM/DAB lock and audio on real
|
||
hardware), DAB service list fixed (two rounds), FM ANTCAP antenna
|
||
calibration persisted to EEPROM, generic ADAU1701 parameter API, phone PCM
|
||
streaming, BLE Wi-Fi provisioning, full FM band scan, BT1035 boot retry.
|
||
See "Post-0.8.5 hardware-in-the-loop findings" below and
|
||
`docs/si4684-rf-investigation-report.md` for the full story.
|
||
|
||
**Before writing code, read `AGENTS.md`, `.cursor/rules/`, and
|
||
`instructions.md`.** Definition of Done: Apache header, doc blocks,
|
||
`doxygen Doxyfile` exits 0, host tests pass, `check-manual-sync.py` and
|
||
`check_si4684_blobs.py` pass, no plaintext secrets.
|
||
|
||
---
|
||
|
||
## Completed agent tasks (T1–T12, fw 0.7.1–0.8.5)
|
||
|
||
| Task | Version | Summary |
|
||
|------|---------|---------|
|
||
| **T1** | 0.7.1 | Doxygen warnings cleared |
|
||
| **T2** | 0.7.1 | CI workflow (host tests, Doxygen, manual sync) |
|
||
| **T3** | 0.7.2 | Preset reorder API/UI, DAB playing ids in status |
|
||
| **T4** | 0.8.0 | RDS/DLS broadcast metadata |
|
||
| **T5** | 0.8.1 | `IntegrationService` — startup, preset recall, last-preset NVS |
|
||
| **T6** | 0.8.2 | Tabbed configuration Web UI (REST coverage) |
|
||
| **T7** | 0.8.2 | Si4684 blob policy — gitignore, docs, `check_si4684_blobs.py` |
|
||
| **T8** | 0.8.3 | NVS + flash encryption — `initEncryptedStorage`, security docs |
|
||
| **T9** | 0.8.4 | Dual-OTA partition table + `dsp` blob slot, rollback Kconfig |
|
||
| **T10** | 0.8.4 | EEPROM EUI-48 identity — SoftAP/BT/hostname/serial |
|
||
| **T11** | 0.8.4 | Updatable ADAU1701 program — `POST /api/dsp/program`, DRAD blob |
|
||
| **T12** | 0.8.4 | ESP32 OTA — `POST /api/system/ota`, rollback confirm on boot |
|
||
|
||
**0.8.5** (hardware/doc alignment): BT1035 boot uses `AT+AUXCFG=3` +
|
||
`AT+I2SCFG=67` (I2S from ADAU1701, not Line-In); I2C pull-ups R1/R16
|
||
confirmed 2\,kΩ; `Hardware/DATASHEET/` bundle + manual cross-refs.
|
||
|
||
Also landed (not numbered): BT1035 pairing (`BluetoothService`), station presets
|
||
(fw 0.7.0), companion-chip boot (Slice 3), ADAU1701 runtime (Slice 5).
|
||
|
||
---
|
||
|
||
## P4 — Hardware-in-the-loop (when PCB arrives)
|
||
|
||
Manual validation only — does not block host CI.
|
||
|
||
### H1. Encrypted NVS boot path
|
||
Follow [`docs/security-flash-nvs.md`](security-flash-nvs.md): first flash with
|
||
`idf.py erase-flash flash`, verify boot logs, Wi-Fi provisioning survives
|
||
reboot, presets and `last_preset` survive power cycle.
|
||
|
||
### H2. End-to-end listening
|
||
Si4684 DAB/FM tune, ADAU1701 profile apply, BT1035 A2DP to headphones,
|
||
now-playing metadata in UI and `/api/tuner/status`.
|
||
|
||
### H3. OTA and DSP program update (on hardware)
|
||
Push a known-good `.bin` via `POST /api/system/ota`, confirm rollback after a
|
||
deliberately bad image. Upload a DRAD blob via `POST /api/dsp/program` and
|
||
verify ADAU replay after reboot.
|
||
|
||
### H4. Production flash encryption (optional)
|
||
After H1 passes, trial build with `sdkconfig.defaults.production` overlay on
|
||
a sacrificial unit; confirm RELEASE mode policy before shipping.
|
||
|
||
### H5. Si4684 FM/DAB no-lock — RESOLVED, was firmware after all
|
||
**Superseded verdict (2026-08-13): blob OK → suspected U6 RF ground, PCBWay
|
||
dispute opened.** That verdict was wrong. The actual cause was
|
||
`writeCommand()`'s ARG1 byte being mis-offset across FM/DAB tune, seek, and
|
||
several status/ack commands — the chip always answered correctly, so every
|
||
signal pointed at hardware, but it never actually tuned. Fixed; real FM
|
||
lock, real DAB ensemble lock, real audio confirmed live on the same board.
|
||
No PCB rework was needed. Full investigation, the wrong initial verdict,
|
||
and the eventual root cause: [`docs/si4684-rf-investigation-report.md`](si4684-rf-investigation-report.md).
|
||
|
||
---
|
||
|
||
## Post-0.8.5 hardware-in-the-loop findings
|
||
|
||
The board arrived and testing against it (not just host tests) found real
|
||
bugs the host-testable core couldn't catch, since they live in
|
||
ESP-IDF-only drivers. Full detail and evidence in
|
||
[`docs/si4684-rf-investigation-report.md`](si4684-rf-investigation-report.md).
|
||
Short version:
|
||
|
||
- Si4684 total RF blackout (H5 above) — firmware bug, fixed.
|
||
- Si4684→ADAU1701 digital audio silence — `PIN_CONFIG_ENABLE` mutual
|
||
exclusion + `SerialInputRegister` polarity, fixed.
|
||
- DAB service list empty/garbled — response-parsing offset bugs (two
|
||
rounds) plus `DAB_EVENT_INTERRUPT_SOURCE` (0xB300) never configured,
|
||
fixed.
|
||
- FM front-end auto-tune measurably suboptimal on this board's actual
|
||
matching network — ANTCAP calibration swept and persisted to EEPROM,
|
||
`POST /api/tuner/calibrate-antenna`.
|
||
- BT1035 total boot silence — root cause found and fixed (2026-08-20):
|
||
the module's spontaneous boot banner (`+VER=...`, `+DEVSTAT=1`) doesn't
|
||
appear until ~18-24s after RESET# releases (full BT stack init, not
|
||
just the internal regulator), but the boot code only waited 3.5s before
|
||
cutting power and restarting — so every attempt, in every prior session,
|
||
cut power before the module could ever finish booting even once. Power
|
||
rails (VBAT_IN/SYS_CTRL/VDD_IO/1.8V_OUT) and TX/RX wiring were all
|
||
independently verified correct with a multimeter first — the module and
|
||
PCB were never at fault. Fixed by waiting up to 25s for the banner
|
||
(`kBootBannerWaitMs`); boot now succeeds on the first attempt.
|
||
- **BT1035 — a second, harder failure mode confirmed intermittent, not
|
||
hardware (2026-08-21).** Distinct from the banner-timing bug above: even
|
||
with the 25s wait already in place, boot sometimes still gets zero UART
|
||
bytes at all — no banner, no AT response, silent across all 8 probed
|
||
baud rates (9600-921600). Root-cause evidence this session: VBAT_IN
|
||
(3.3V), 1.8V_OUT (1.8V), SYS_CTRL/RESET (~3.27V, matching the firmware's
|
||
own GPIO readback), and BT1035 TX (idle-HIGH ~3.29V, no short/float) all
|
||
measured normal with a multimeter. The BT1035's 32 MHz crystal is
|
||
integrated inside the sealed Feasycom module (confirmed via the module's
|
||
own datasheet block diagram — no external crystal on our schematic), so
|
||
it can't be inspected or reworked from our side; a marginal
|
||
oscillator-startup margin inside the module is the leading suspect.
|
||
**Decisive evidence it's intermittent, not a dead unit**: the exact same
|
||
physical module booted cleanly (banner + all AT commands `OK`) on one
|
||
attempt and went totally silent on the very next attempt, no physical
|
||
changes in between. A replacement module is therefore not a guaranteed
|
||
fix — the same defect class could recur on a different unit. Mitigated
|
||
(not fixed) by an indefinite background retry task
|
||
(`hardware::bt1035RetryTask` in `main/hardware_bootstrap.cpp`): if the
|
||
initial `Bt1035Driver::boot()` fails, a background FreeRTOS task keeps
|
||
calling `boot()` again with no artificial delay between attempts (each
|
||
attempt already takes ~25-60s on its own) until it succeeds, while the
|
||
rest of the system (tuner, Wi-Fi, web UI) stays fully usable in the
|
||
meantime. Turns a permanent-until-manual-power-cycle failure into a
|
||
bounded, self-recovering delay. See
|
||
`docs/si4684-rf-investigation-report.md` (2026-08-21 entry) for the full
|
||
session narrative, including a UART TX/RX loopback test attempt that was
|
||
inconclusive (bridging the ESP32's own TX/RX pins from cold boot caused
|
||
an unrelated, reproducible, harmless early-boot hang, not yet explained).
|
||
- **BT1035 — git archaeology + minimal patch, follow-up (2026-08-21).**
|
||
Traced the full commit history of `Bt1035Driver.cpp` from the last
|
||
documented-good boot (`6ca40f1`) through the regression (`6f7b6dd`, a
|
||
redundant `AT+RESET`) and its fix (`fd9d4ae`, 5/5 clean boots — removed
|
||
the `AT+RESET` and introduced the boot-banner listen at 3500ms in the
|
||
same commit). Comparing `fd9d4ae` to this session's working tree found
|
||
one real structural difference beyond the justified 25s banner window:
|
||
today's earlier commit (`3a58d33`) had added an intra-`boot()` retry
|
||
loop (2 attempts, only 300ms between hardware reset pulses) that never
|
||
existed in the validated baseline — shorter than the BT1035 datasheet's
|
||
own "Reset Protection timeout (typically >1.8s)", so the second pulse
|
||
may not have reached a clean power-off state. **Fixed**: removed the
|
||
intra-`boot()` retry loop entirely (`kBootAttempts`/`kBootRetryDelayMs`
|
||
deleted); `boot()` now makes exactly one attempt per call, matching
|
||
`fd9d4ae`. Retries remain exclusively at the `bt1035RetryTask` level
|
||
(whole clean `boot()` calls, never re-pulsing pins faster than one full
|
||
cycle apart — confirmed live, ~31.8s between attempts). Host tests
|
||
(20/20) and firmware build green; flashed and observed live. **Result
|
||
inconclusive on hit rate**: a 20-minute post-flash window captured 31
|
||
consecutive silent retry attempts, zero successes — worse than earlier
|
||
the same day. The patch is kept because it's structurally correct (only
|
||
known deviation from the historically validated design removed), not
|
||
because this sample proved a better success rate. Root cause of the
|
||
underlying intermittent silence is still open (see entry above).
|
||
- **BT1035 — RESET#/SYS_CTRL redesign, CTS/RTS diagnostics, Feasycom
|
||
escalation (2026-08-22).** A sibling project's PinScope netlist report
|
||
(RESET# pulled to GND, SYS_CTRL pulled HIGH by stray resistors) was
|
||
checked against our own schematic and does **not** apply to us — our
|
||
RESET#/SYS_CTRL wiring is correct, verified via netlist. Its RF_OUT/pin
|
||
51 floating finding **does** also apply to us, but is a separate,
|
||
RF-range-only concern (datasheet documents both internal- and
|
||
external-antenna variants; can't tell which we have), not the cause of
|
||
the digital/UART boot silence. As a diagnostic test, RESET# is now
|
||
never driven at all (floating input, relying on the module's own
|
||
internal pull-up per §4.8) and SYS_CTRL now does a genuine LOW(2.5s)→
|
||
HIGH power-cycle on every retry attempt (previously asserted once ever
|
||
and left alone, meaning retries never actually power-cycled the
|
||
module). Also added read-only diagnostics on the previously-unused
|
||
CTS/RTS pins (physically wired, named in `board_pins.hpp`, never
|
||
configured by any driver code, host flow control disabled) — live
|
||
readings were perfectly stable (CTS=HIGH, RTS=LOW) across ~12 samples
|
||
over 30+ minutes, arguing against pure floating-noise. **Across all of
|
||
today's changes combined, zero successful boots were observed in
|
||
cumulative 45+ minutes of live testing** — inconclusive-to-negative,
|
||
not proof any change helped or hurt. Escalated to Feasycom support with
|
||
a detailed email (drafted, kept outside the repo) covering the
|
||
symptom, everything ruled out, the CTS/RTS open question, and the
|
||
antenna-variant question; paused further live experimentation pending
|
||
their reply rather than keep permuting timing parameters blind.
|
||
- **Still open**: intermittent multi-second HTTP unresponsiveness under
|
||
load; DAB signal quality still antenna-limited; 24 KB `nvs` partition
|
||
may be undersized (`saveProfile()` `store_failed` seen intermittently,
|
||
error code never captured); BT1035 intermittent total-silence boot
|
||
failures (mitigated via background retry, not root-caused — see above).
|
||
|
||
---
|
||
|
||
## Open firmware polish (non-blocking)
|
||
|
||
Done in fw 0.8.5 unless noted:
|
||
|
||
- BT1035 boot — I2S slave init (`AT+AUXCFG=3`, `AT+I2SCFG=67`) per PCB routing (0.8.5).
|
||
- FM seek down — `POST /api/tuner/seek` with `{"direction":"down"}` (0.8.4).
|
||
- BT1035 — query/set name, paired list (`AT+PLIST`), auto-reconnect
|
||
(`AT+AUTOCONN`) per Feasycom BT1035 manual (0.8.4).
|
||
- Si4684 — `STOP_DIGITAL_SERVICE` (0x82) before FM band switch when DAB
|
||
audio is active; ensemble metrics remain via `DAB_DIGRAD_STATUS` in status (0.8.4).
|
||
|
||
---
|
||
|
||
## DONE (2026-08-24) — calibration functions are now permanent, in-firmware, on-demand tools
|
||
|
||
Both ANTCAP calibration (`tools/si4684_antenna_calibration.py`) and Si4684
|
||
crystal calibration (`tools/si4684_xtal_calibration.py`,
|
||
`POST /api/tuner/xtal-calibrate`) currently exist as **host-side Python
|
||
scripts driving live-but-unpersisted HTTP endpoints** — they compute a
|
||
result but the operator has to hand-edit firmware source (constants in
|
||
`Si4684Driver.cpp` / the `gSi4684.boot(...)` call in
|
||
`hardware_bootstrap.cpp`) and reflash to make a result permanent.
|
||
|
||
**Wanted instead**: both calibration procedures should be triggerable
|
||
on-demand *from the device itself* (an HTTP endpoint is enough — no UI
|
||
required yet) and, once a result converges, **write the result to the
|
||
24AA025E48 EEPROM** (same chip/pattern already used for ANTCAP
|
||
persistence, see `Eeprom24aa::writeFmAntCap`/`writeDabAntCap`) so it
|
||
survives a reboot without a firmware reflash. `recalibrateXtal()`
|
||
(`Si4684Driver.cpp`) already does the live re-boot-with-new-params part;
|
||
what's missing is EEPROM persistence for **all three** crystal
|
||
calibration parameters -- `ibias`, `ctun`, AND `xtalFreqHz` (not just
|
||
XTAL_FREQ; confirmed explicitly 2026-08-24 that all three need to
|
||
persist, not only the one this session happened to tune) -- plus loading
|
||
them at boot the same way `main.cpp` already loads the saved FM/DAB
|
||
ANTCAP into `TunerService` before the first tune. ANTCAP already
|
||
persists this way (2 bytes/band, word addresses 0x00/0x01) — the xtal
|
||
calibration needs its own new EEPROM word address(es) alongside those
|
||
(ibias fits in 1 byte, ctun in 1 byte, xtalFreqHz needs 4 bytes -- 6
|
||
bytes total, or pack more compactly if EEPROM space is tight). Also
|
||
still needed: deciding whether the FREQOFF-averaging/damping/
|
||
convergence-loop logic (currently in `tools/si4684_xtal_calibration.py`)
|
||
moves into firmware, or stays host-side with just an EEPROM-persist step
|
||
added at the end of the existing HTTP flow.
|
||
|
||
**Resolved 2026-08-24.** `Eeprom24aa` gained `readXtalCalibration()` /
|
||
`writeXtalCalibration()` (word addresses 0x02 ibias, 0x03 ctun, 0x04-0x07
|
||
xtalFreqHz big-endian, right after the existing FM/DAB ANTCAP bytes at
|
||
0x00/0x01). `HardwareBootstrap::boot()` now reads the ADAU1701 boot
|
||
earlier (moved before Si4684's, since the EEPROM read needs ADAU1701's
|
||
I2C bus) and loads the saved crystal trim before calling
|
||
`gSi4684.boot(...)`, falling back to the compiled-in defaults
|
||
(ibias=72, ctun=0, xtalFreqHz=19199750) when the EEPROM has never been
|
||
calibrated. `POST /api/tuner/xtal-calibrate` now persists every
|
||
successful live recalibration automatically (`"persisted":true/false` in
|
||
the response) via a new `saveXtalCalibration()` /
|
||
`net::AntennaCalibration::saveXtal` bridge, mirroring the existing
|
||
ANTCAP save pattern. The convergence-loop logic
|
||
(`tools/si4684_xtal_calibration.py`) stays host-side, unchanged — only
|
||
the persistence gap was closed.
|
||
|
||
See `docs/si4684-rf-investigation-report.md`'s 2026-08-23 entry for full
|
||
context on why this calibration was needed and how it currently works.
|
||
|
||
---
|
||
|
||
## Quality gates (run from `Software/` before merge)
|
||
|
||
```bash
|
||
cmake -S components/core/test -B build-host && cmake --build build-host
|
||
ctest --test-dir build-host --output-on-failure
|
||
doxygen Doxyfile
|
||
python3 tools/check-manual-sync.py
|
||
python3 tools/check_si4684_blobs.py
|
||
```
|
||
|
||
After editing the web UI: `tools/gzip-www.sh`.
|
||
|
||
---
|
||
|
||
## Notes for the agent
|
||
|
||
- Extend existing patterns (`AudioProfile` / `IAudioProfileStore` shape).
|
||
- Never invent Si4684 opcodes — cite AN649.
|
||
- Never invent BT1035 AT strings — cite Feasycom BT1035 programming guide.
|
||
- One logical change per commit; 50/72 messages.
|
||
- Update `ch-classes.tex` / `ch-api.tex` when public API or HTTP changes.
|