check / check-without-hardware (push) Successful in 6m45s
- README, CHANGELOG, RELEASE_NOTES, THIRD_PARTY_NOTICES and docs/*.md translated to English; cross-markdown anchors updated - C/C++/Python/shell comments and user-facing messages translated - identifiers, protocol names, commands and literals unchanged - native tests re-verified: 24/24 PASS; tpms_frame.py self-test PASS
130 lines
7.1 KiB
Markdown
130 lines
7.1 KiB
Markdown
# Changelog
|
||
|
||
## Work in progress after 2.0.0-rc.1
|
||
|
||
- Micro-naps instead of one long sleep (`config.h`: `NAP_PERIOD_MS`):
|
||
the LED blinks approximately every 2 s also between transmissions, and after
|
||
every timer wake a blink is visible (`NAP_AWAKE_MS`). Sleep is the remainder
|
||
of the period after boot time + window; TX of frames remains on the interval
|
||
(sleep counter in RTC memory, `interval` takes effect on the next cycle).
|
||
`power_deep_sleep` takes milliseconds. Previously, waking from the timer
|
||
turned the LED off in the same `loop()` — the blink lasted a few ms and the
|
||
wake-up was invisible.
|
||
- Button: the board's BOOT button by default (`PIN_BUTTON = 0`, GPIO0) —
|
||
EXT1 behaviour confirmed on hardware. An external button on GPIO3 is
|
||
selected with `-DPIN_BUTTON=3`. Note: GPIO0 is a strapping pin (held at
|
||
power-up/reset it starts the bootloader). Every button gesture opens a 10 s
|
||
console window (per `POWER.md`) — the device does not fall asleep
|
||
immediately after, e.g., leaving LEARN/SINGLE. LEARN/SINGLE still do not
|
||
sleep (unchanged).
|
||
- Calibration of the pressure field `P` against the car display: the rtl_433
|
||
scale (2.5 kPa/bit) is not confirmed (the captures have `P=0`), and the car
|
||
showed 284 kPa for 260 kPa. The scaling correction
|
||
`TPMS_PRESSURE_KPA100_PER_RAW` (273 = 2.73 kPa/bit; `P=95` for 260 kPa) was
|
||
applied in the firmware, the Python model, and the `verify_rtl433.py`
|
||
expectations. Confirmed on hardware: 260 kPa → 260 kPa on the car display.
|
||
- Hardware–firmware integration audit (2026-10-28), consistency of
|
||
ESP32-S2-MINI + CC1101 SMA + LM317T:
|
||
- EXT1 wake: the RTC pulls for the button and LF are configured before deep
|
||
sleep. A digital pull-up (`pinMode`) does not work with the RTC_PERIPH
|
||
domain disabled (ESP-IDF documentation), so the wake pins floated and
|
||
EXT1 could wake randomly. An external 10 kΩ is also recommended.
|
||
- Console: number parsing rejects NaN and infinity (previously
|
||
`pressure nan` / `rf freq nan` passed validation; NaN ended up in NVS as
|
||
pressure, and `rf freq nan` blocked later writes). The parsing logic was
|
||
extracted into the portable `parse` module with native tests.
|
||
- Power: ADC divider R2 270 kΩ → 180 kΩ. The previous divider saturated the
|
||
ADC (2.5 V full scale) already above about 11.8 V and did not measure the
|
||
12 V installation under a 12.6–14.4 V operating range.
|
||
- `tools/package_release.sh`: the full `merge_bin` image uses `--flash-freq
|
||
80m` consistent with the `lolin_s2_mini` manifest; the previous `40m`
|
||
overwrote the bootloader header (80 MHz) with a parameter that differed
|
||
from the build.
|
||
- `tools/tpms_frame.py`: temperature saturation to 0–255 consistent with
|
||
the firmware (temperatures ≤ −19 °C corrupted the frame in the Python
|
||
model).
|
||
- Tests: the negative checksum case now inverts the whole Manchester cell
|
||
(previously it exercised a Manchester violation instead of the sum); a new
|
||
`test_parse` suite (ranges, NaN/inf); corrected the bit diagram in
|
||
`PROTOCOL.md`.
|
||
- Target hardware changed to ESP32-S2-MINI (ESP32-S2FN4R2, e.g.
|
||
ARD-ESP32-S2-MINI, board `lolin_s2_mini`); updated the PlatformIO target,
|
||
native USB CDC, SPI/button/ADC pinout, and built-in LED polarity
|
||
(GPIO15, active HIGH). Supply ADC on GPIO1, optional LF on GPIO14.
|
||
`btStop()` is compiled only on chips with Bluetooth — the ESP32-S2 has none.
|
||
- CC1101 433 MHz SPI radio module with an SMA connector and antenna
|
||
(ARD-CC1101-ANT).
|
||
- External power through an adjustable LM317T (TO-220 metal, 1.25–37 V,
|
||
1.5 A): the correct ADJ–VOUT–IN pinout, TAB = VOUT, a 240/390 Ω divider
|
||
giving about 3.28 V, capacitors, and protection and thermal requirements.
|
||
- Target power from the car cigarette-lighter socket: a USB charger (variant
|
||
A) or the LM317T (variant B). With an ignition-switched socket the emulator
|
||
draws no current while parked; the lack of parking transmission is
|
||
acceptable because the car ECU does not read TPMS with the ignition off.
|
||
- ESP32-S2 release images use `esptool --chip esp32s2` and a bootloader at
|
||
address `0x1000`; older images for ESP32-C3 and ESP32-WROOM are incompatible.
|
||
|
||
## 2.0.0-rc.1 — 2026-10-27
|
||
|
||
Release candidate without a physical CC1101 test and a car test.
|
||
|
||
### Added
|
||
|
||
- Persistent NVS configuration with versioning, CRC32, range validation,
|
||
unique IDs, factory reset, and safe defaults.
|
||
- Console: Clone IDs, a new-ID generator, pressures/temperature, intervals,
|
||
limited RF, LEARN/SINGLE, and diagnostics without recompilation.
|
||
- Portable state machine with fake clock/radio, timeouts, and the
|
||
FL→FR→RR→RL order. Restart, ACC, radio-failure, and `millis()` wrap tests.
|
||
- Independent validation of four reference recordings and synthetic CU8 with
|
||
the pinned rtl_433; negative checksum/Manchester/truncation tests.
|
||
- `tools/check_all`, Dockerfile, Gitea Actions, image scripts, and SHA-256.
|
||
- User, hardware, power, safety, and HIL documentation.
|
||
|
||
### Changed
|
||
|
||
- Defaults of 260 kPa and −10 dBm, OOK only. The 30-byte Manchester encoder
|
||
is preserved and passes independent decoding.
|
||
- The CC1101 waits for `MARCSTATE=IDLE`, checks underflow and leftover bytes,
|
||
instead of relying only on timing and `TXBYTES=0`.
|
||
- A plain RTC wake does not open the console window; transmission and the
|
||
return to sleep limit the estimated consumption. The console window remains
|
||
open after user interaction.
|
||
- Full build with `-Wall -Wextra -Werror`; cppcheck with zero defects.
|
||
|
||
### Removed / fixed
|
||
|
||
- Removed the guessed `TPMS_FLAGS_LOWBAT_MASK` and the coupling between the
|
||
emulator's low supply and the simulated sensor. The supply is only a local
|
||
warning.
|
||
- Removed FSK from the main flow: no evidence for the analysed profile.
|
||
- Removed `PARKING_SHUTDOWN`; every deep sleep has an RTC alarm, so an ACC
|
||
restart cannot leave the device permanently asleep.
|
||
- The default, unvalidated LF input is disabled. Basic LEARN does not require
|
||
the EL-50448.
|
||
- Removed the recommendation of MCP1703 + 24 V TVS as automotive protection
|
||
and the promises of multi-year battery operation. The 12 V supply has
|
||
design requirements, not HIL.
|
||
|
||
### Audit iterations
|
||
|
||
1. **Initial state:** 7/7 native and build PASS, but configuration through
|
||
recompilation, a monolithic state machine, lowbat, and power needed
|
||
repair. Created `archive/pre-opencode`.
|
||
2. **Data model:** added NVS/CRC/validation and portable modes. The first
|
||
test caught wrong test expectations for a pending tick and immediate TX;
|
||
after correction, 20/20 native and the ESP32 build PASS.
|
||
3. **RF validation:** an audit of four recordings confirmed 16 repetitions
|
||
and variable gaps. A single vector was replaced with the full script;
|
||
64/64 references, 3/3 synthetic, and 3/3 negative PASS.
|
||
4. **Energy:** a console window after every RTC wake was detected; it was
|
||
limited to interaction. Consumption is still clearly marked as estimated.
|
||
5. **Static analysis:** cppcheck flagged memset-initialised float structs;
|
||
replaced with value initialisation and reached 0 defects.
|
||
|
||
### Not physically verified
|
||
|
||
CC1101 on an antenna/analyser, RTL-SDR with the prototype, consumption,
|
||
brownout, ACC, a flash update, Clone IDs and LEARN with the specific Opel. No
|
||
"validated" label.
|