Skip to content

Ring clock sometimes stops: could the app detect it and warn? #142

Open

Description

@Albabit

My ring's clock has stopped 4 times since August (in August, on Oct 12, on Oct 15, and on Nov 9 around 09:00). When it happens the ring still connects and syncs, but its clock no longer moves. So the ring takes no automatic HR and records no sleep. In the app it just looks like I didn't wear the ring: this morning there was no night and no HR since yesterday morning.

How I confirmed it this time:

  • The last automatic HR in the ring's memory is stamped 09:00:00 on Nov 9 (they normally land between :20 and :59 past the hour). Nothing after that, and no sleep session for the night.
  • The app's scheduled SpO2 kept running, but each reading got the time PulseLoop writes on connect instead of the time the measurement ended. Example: connection 00:38:21 to 00:39:48, reading stamped 00:38:25. Before the freeze: connection 08:43:21 to 08:44:48, reading stamped 08:44:44.
  • On Oct 16, two spot readings taken 4 minutes apart came back as one single record, stamped with the time PulseLoop had last set.

What fixes it:

Setting the time on connect (01 00) doesn't. What worked the last three times: turning the ring off with 03 16 01 (what SmartHealth sends from its power off menu), putting it on the charger so it turns back on, then letting PulseLoop connect and set the time. Nothing is lost from the ring's memory. Done again this morning: the next automatic HR came at 11:00:41, and the app's 10:30 SpO2 is stamped 10:31:26, right when the measurement ended (connection 10:30:03 to 10:31:30).

Suggestion:

I couldn't find a command to read the ring's clock, but the app could spot the pattern:

  • a scheduled SpO2 reading stamped before the measurement actually ended (right at the last time set), which looks like a clear sign;
  • or, as a weaker hint, no automatic HR for more than twice the HR interval while the ring keeps syncing.

Then show something like "Your ring's clock seems to have stopped" with a "Restart ring" button that walks through the fix:

  1. the app sends 03 16 01 and the ring turns off;
  2. it tells the user what to do next: "Put your ring on its charger to turn it back on";
  3. it reconnects and sets the time by itself, then checks that the next reading gets a proper timestamp.

I'd go with "Restart" rather than "Turn off", since the ring has to come back on for the fix to work. The charger step is the one people would miss: without it the ring just stays off.

Activity

  1. foureight84 commented on Nov 13, 2026

    @foureight84
    Owner

    @Albabit this one's a great find, thanks for working out both the pattern and the fix 🙌

    v2.11.3+85-rc1 (versionCode 85, PR #144) has a first version of what you suggested:

    • Detection: went with your strong signal. When the ring finishes a spot reading (04 0e), PulseLoop checks the copy the ring logs into its history. A working clock stamps it when the measurement ended. A stopped clock stamps it at the last time we set, or it doesn't land anywhere near our reading. Two of those within 24 h show a card on Today: "Your ring's clock seems to have stopped". A normal reading afterwards clears it. Each reading is only judged after a sync that actually pulled HR/SpO₂ history, against the closest copy in that sync, so opening the Sleep screen or an all-day sample sitting next to it won't trigger a false alarm. I left out the weaker "no automatic HR" hint for now, since a ring sitting on the desk looks exactly the same.
    • Restart ring: the card's button walks through your three steps. It sends 03 16 01 (matches SmartHealth's appShutDown(1)), then keeps "Put your ring on its charger to turn it back on" on screen, since that's the step people would miss, as you said. Then it reconnects and sets the time on its own. If the next readings still come back frozen, the warning returns.

    With scheduled SpO₂ on, the warning should show up within a couple of runs the next time the clock stops. When it does, could you try the Restart button and let me know whether it shows up, whether the flow makes sense, and whether the next SpO₂ comes back with the right timestamp? Also tell me if it ever warns on a ring that's actually fine. Thanks again!

  2. Albabit commented on Nov 13, 2026

    @Albabit
    Author

    Thanks a lot for this! Both the detection and the guided restart look exactly like what I had in mind.

    I installed v2.11.3+85-rc1 this morning, with scheduled SpO₂ on (every 2 hours). The ring's clock is running fine at the moment, so there's nothing to test yet.

    As soon as it stops again, I'll report back here: whether the card shows up, how the Restart flow goes, and whether the next SpO₂ comes back with the right timestamp. I'll also let you know if it ever warns while the clock is actually fine.

  3. foureight84 commented on Nov 14, 2026

    @foureight84
    Owner

    @Albabit thanks, sounds good! FYI v2.11.3+86-rc2 replaces RC1. The clock-stall detection and Restart flow are unchanged; RC2 only adds the Bluetooth-off status fix from #136, so it's safe to update and keep waiting for the next stop.

  4. Albabit commented on Nov 14, 2026

    @Albabit
    Author

    @foureight84, A side note on scheduled SpO₂, since the clock check relies on it.

    Last night (RC1, HR every 60 min, scheduled SpO₂ every 2 h), none of the 4 night slots ran. The last reading was at 22:47 (22:30 slot), the next one at 08:31. During the day all 7 slots ran, just a bit late.

    What I see: with the phone in Doze, PulseLoop only got to run at 00:57, 02:58 and 06:59 (the background syncs connected then, and nothing ran between 02:58 and 06:59). Each time that falls inside the 5 min + 75 s margin before the ring's own hourly reading. So if the SpO₂ run was released at the same time, isClear would skip it, and the next slot is 2 h later. That matches losing 00:30, 02:30, 04:30 and 06:30. I couldn't see the "Skipped" line itself, logcat didn't go back that far.

    Two ideas, if it's worth it:

    • The ring's reading starts around hh:00:20 to hh:00:45, and a scheduled run takes about 85 s here (connect to 04 0e). A run starting at 00:57 or 02:58 would have finished before it, so the margin before the ring's reading could maybe be closer to the run's real length.
    • Or, after a skip, retry shortly after the ring's reading instead of waiting for the next slot.

    Not urgent at all, the daytime runs are fine.

  5. foureight84 commented on Nov 15, 2026

    @foureight84
    Owner

    @Albabit great catch on the overnight Doze behavior — losing all 4 night slots would also starve the clock-stall check overnight.

    Both of your ideas (plus a wakeup race I spotted alongside them) are in v2.11.3+87-rc3 (versionCode 87, PR #144):

    • Tighter pre-reading clearance (105 s): ScheduledSpo2Slots.isClear now reserves 30 s for connect + sync before the 75 s measurement ceiling instead of 5 m + 75 s. A Doze maintenance wakeup at :57 or :58 now starts instead of skipping, and a second 75 s check (isClearToMeasure) runs right after startup sync drains before taking the sensor.
    • Retry at :05 instead of dropping the 2 h slot: when a run does get skipped for landing too close to the ring's reading (like 06:59, either at start or after sync), it now schedules a retry right after the ring's reading clears (nextClear, e.g. 07:05), and the run after that snaps back onto the normal :30 grid.
    • Avoid losing the Doze window to background sync: ScheduledSpo2Worker now raises isMeasuring as soon as it starts (before its Room queries), so a RingSyncWorker released in the same Doze maintenance window sees it and yields GATT immediately instead of racing for the link.

    Still keep an eye out for the clock freeze + Restart button whenever it next happens, and let me know if overnight scheduled SpO₂ runs reliably through Doze on RC3!

  6. Albabit commented on Nov 17, 2026

    @Albabit
    Author

    @foureight84, Two nights on 87 so far (app SpO₂ every 2 hours, ring HR every 60 min).

    The retry after a skip seems to work:
    On Nov 16, Doze released the worker at 03:02:51, less than 5 min after the ring's own 03:00 reading. The app connected for 24 s only (sync, no measurement), then connected again at 03:05:01 and got a 98 at 03:06:21. That's exactly the new retry time. I had no logcat for that night, so I only have the connection times, not the "Skipped" line itself.

    Night readings (00:30 to 06:30 slots):

    • Nov 15→16: 1 (03:06, the retry above). There was also a connection at 00:57 with no reading saved. I can't tell why, the app log in my export only starts the next morning.
    • Nov 16→17: 2 (00:56 for the 00:30 slot, 04:08 for the 02:30 slot). At 08:08 the worker tried twice but the ring didn't answer (status 147, I was probably out of range), and the next reading came at 08:36 when Bedtime ended.

    For comparison: 0 of 4 on 85, 2 of 4 on 86. The ring's hourly HR was there every hour on both nights, so no clash with the ring.

    What's left is Doze:
    Last night the phone only woke PulseLoop at 00:13, 00:54, 02:05, 04:07 and then 08:08, so the gaps grow to 2 and then 4 hours, and the readings just land in those windows. That's fine for me, I don't think it's worth fighting Doze over.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions