mikkeli
fc9f93336e
/ crosscheck (push) Successful in 0s
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01DDwfd6Ki8UfG4hK6W5YpJz
59 lines
3.3 KiB
Markdown
59 lines
3.3 KiB
Markdown
---
|
||
type: plan
|
||
status: active
|
||
updated: 2026-10-14
|
||
---
|
||
# Car Power: Ride Through Cranking, Shut Down at Ignition Off
|
||
|
||
## Power chain
|
||
|
||
```
|
||
cigarette socket (ignition-switched, 65 W class, crank-sensitive)
|
||
→ USB-C PD charger → 15 V PD trigger (15 V / 3 A) → DC jack
|
||
→ Waveshare UPS Power Module (C), 3S 21700, 15–19 V in, pass-through
|
||
→ Orin Nano carrier
|
||
```
|
||
|
||
Socket power is ignition-switched, so **UPS input loss = ignition off or cranking**. The UPS rides through cranking
|
||
(bench: ≥ 20 s holdover across a supply swap). See [BOM](../../BOM.md#waveshare-ups-power-module-c).
|
||
|
||
## Detection
|
||
|
||
INA219 on I²C bus 7, `0x41`, monitor-only (the UPS has no control lines). Battery current from the shunt-voltage register (0x01, 10 µV/LSB):
|
||
about 0 or positive (charging) with input present, **clearly negative** (discharging) when input is lost. Battery voltage (0x02) is the backstop.
|
||
To confirm on the bench: the sign convention and the discharge magnitude at the Orin's real load (pull the UPS input and read).
|
||
|
||
## Policy (to build)
|
||
|
||
A small host service polling the INA219 about every 1 s:
|
||
|
||
| State | Condition | Action |
|
||
|---|---|---|
|
||
| on input | current ≥ about 0 | nothing |
|
||
| on battery | discharging, < grace | publish `system/power` "on battery"; **a short episode that recovers is a crank**: stamp it into the recording |
|
||
| ignition off | discharging continuously for grace (start at 60 s; tune after the car crank test) | publish "shutting down" → stop capture (close the session cleanly) → optional parked-at-home step → `systemctl poweroff` |
|
||
| battery low | **built 2026-10-14** (svc-core `orin-status --poweroff-below 11.0 --poweroff-after 30`): below 11.0 V *while discharging*, 30 readings (1/s) in a row; a missing reading restarts the count | state `stopping` on `yukart/system/orin-status`, then `systemctl poweroff` through svc-gw's polkit grant (no sudo) |
|
||
|
||
Optional parked-at-home step: if home WiFi/LAN is reachable at ignition off, stay up long enough to copy recordings off over the LAN
|
||
(same LAN-only rule as map sync, project-yukart decision 10), within a hard time cap, then shut down.
|
||
|
||
The battery-low row came first: on 2026-10-14 the pack ran flat overnight with the Orin up and nothing to stop it,
|
||
and the boot's journal ends mid-stream at 06:29:22 (ext4 replayed its journal on the next boot; nothing lost).
|
||
11.0 V while discharging at the Orin's ~0.45 A is about 10 % charge, a margin for a clean shutdown; 30 s rides out a
|
||
crank or a load spike. The ignition-off rows are still to build.
|
||
|
||
## Open: getting the Orin back on at ignition on
|
||
|
||
After `poweroff` the UPS keeps supplying the board from its battery:
|
||
|
||
1. **Standby drain:** the halted Orin still draws some power, so the pack drains while parked, until the UPS's over-discharge protection cuts out.
|
||
2. **No auto-start:** the dev kit powers on when power *appears*. If the UPS output never dropped, ignition on doesn't boot the Orin.
|
||
|
||
Candidate fixes (hardware, to evaluate):
|
||
|
||
- Cut the UPS output after shutdown (a latching relay or a timer on the output), so the next ignition-on is a real power-up.
|
||
- Pulse the carrier's power-button header from a small circuit when socket power returns.
|
||
- Suspend (SC7) instead of poweroff, with wake on a GPIO tied to socket power. Untested on this board, and still drains.
|
||
|
||
Until this is solved, ignition-off shutdown leaves the car unit off until someone power-cycles it.
|