[BUG] RK2023 and RGB30v1 battery drain while off

[sydarn]

Describe the bug

RK2023 and RGB30v1 drains battery while being turned off.

How to reproduce

Steps to reproduce the behavior:
Leave the devices off, preferably with low charge for quicker feedback wait a few days and it's dead.

Information

• ROCKNIX Version: [e.g. 20240517]
• Hardware Platform: [e.g. Powkiddy RK2023, Powkiddy RGB30v1]

Log file

Not available

Context

Theory: RK2023 and RGB30v1 uses a different CPU voltage regulator than all other rk3566 devices.

[macromorgan]

Just throwing my two cents in... I wonder if the device is really shutting off when you do a shutdown? My hypothesis if this turns out to be the case is a problem in the Arm Trusted Firmware. How long do you have to hold down the power button to turn it back on after a shutdown? If it's less than 2 seconds it was truly shut down, if it's more than 4 seconds the device was never really off to begin with.

[sydarn]

I do not notice any difference in time I need to hold power to boot between the rgb30v1 and RG-Arc. A fast tap to the power button doesn't boot either, slightly longer 1-2 sec maybe will boot both.

[macromorgan]

Damn, okay. Might have to probe the PMIC outputs one by one when it's off to see if we can figure out the power drain. Do you know if the X55 has a similar issue?

Thank you.

[sydarn]

X55 is the only PK rk3566 device I don't own, though we have several contributors that own it. And we have no reports of draining while off. The rgb10max3 and rgb20sx (with exception of wifi identical to a rgb30v2) does not have this issue.
So it seems to be isolated to these 2 revisions which are differing on the regulator.

When looking at the regulator in the dts, I saw that it specifies min voltage to 0.7125 volts (I know this is from BSP), while I think it's specced to 0.76V. BSP also have something called init voltage which doesn't exist on mainline. I will test raising the min voltage an do a test, but it takes a day or two.

[sydarn]

I am somewhat optimistic, but the battery reading was wrong when I started. So it had more juice than expected. This was the rgb30v1.

I am starting a test on the rk2023 which was dead now, I updated and charged it up to 7%. I will check every 24h from now on.

[sydarn]

No the min voltage on the regulator had no effect, today I tried starting both of them after ~48h. They were both drained. :(

[Bapf]

Can confirm this on my RGB30v1.

[Soundtoxin]

I believe I'm running into this on my RG353M as well. I don't recall having issues on JELOS, which I was using up until moving to ROCKNIX's latest release two days ago. Now I've had one day of the device being too dead to turn on without being plugged in, and another day where after 8-10 hours or so I'm at 36% when I had powered off around 80 or 90%.
Also, I seem to have a harder time turning my device on since moving to ROCKNIX, unsure if that's related. Pressing and holding the power button seems to do nothing a lot of the time and I have to really fight with it for a minute or two before I see any signs of life. I swear just holding the button a second or two worked consistently before.
edit: seeing the comment above about how if it doesn't turn right on it wasn't really off makes me think it really isn't turning off properly when I choose to shutdown in the menu. I've held the power button 8 seconds or so with it "off" now and will check back in some hours on the charge.
edit2: 8 or 9 hours later and I've managed to maintain most of the charge, within 3% of what I left it at. It also turned on after only a second or two holding the button, and on the first try. So I think the issue is indeed that it's not shutting down normally via software.
edit3: I thought I did the workaround the same way but after letting it sit a day and a half or so after powering it off on full charge it was bone-dry dead again and wouldn't turn on until I plugged it in.

[Bapf]

Updated to 20240612 yesterday.
Shut the device off this morning after charging, still at 100% 10 hours later. 🤞

[sydarn]

The battery indicator isn't the most reliable, I usually left them on low charge 10-30% (hard to know what). Left it 2-3 days, and when I tried to start them they were dead.

[sydarn]

There several user report of this also happening on RGB10 Max 3, though I never experienced it.
Additionally there is one user report on RK2023-WiFi model NOT having this issue.

[drmindflip]

Just wanted to chime in and clarify - futher to my comments on the other thread - this is definitely happening with my RGB10 Max 3. Turned off with a decent charge remaining two days ago, dead and unresponsive today, had to plug in for a bit before powering on.

(on 20240612)

[Bapf]

Ok, maybe a wild guess, but maybe someone could also confirm my findings:

I had my RGB30 lying on my table for the last few days -> almost no battery drain when turned off.
Then I put it back in my bag -> empty after 24 hours of being turned off.

Since the device has no mechanical on/off-switch: Is it possible that the new kernel monitors all buttons (not only the on/off button) when turned off and every time a button gets pressed (which happens quite often in my bag) the device wakes up for a short time which drains the battery?

[Soundtoxin]

I believe this is now fixed for me on the latest release with my RG353M. I updated within a half hour of 20240702 coming out. After sitting one day turned off, I didn't lose any charge. Two more days being off again it still seems fine.

[sydarn]

I never had any issues with other than rk2023 and rgb30v1, but I just started another test run on linux 6.10 (newer than the release)

[drmindflip]

Can confirm that this is still happening on Powkiddy RGB10 max3 - running 20240702

Turned off device with 60-70% power left, left in one place without moving or using it at all, totally dead after a couple of days - won't turn on.

[sydarn]

After updating linux 6.10, u-boot 2024.07 and latest ATF from rockchip the issue still persists.

[Zeudan]

Hello, I have the same issue on the latest 20240702 on my RK3566 Arc-D. Went from 50% to 0% in one night

[sydarn]

Unless it's dead, it's most likely just a battery reading bug.

[Bapf]

Unless it's dead, it's most likely just a battery reading bug.

Is is dead!
To be clear: It is dead / the battery is empty in a way that even the RTC no longer works.
(To check: After charging the device again and booting it up at a location where there is no wifi available when I play a game and check the "last played" menu afterwards the game gets sorted to the middle/end of the list instead of to the top where it's supposed to end up. When wifi is available after powering the former drained device up this does not happen since it instantly syncs the time again after booting.)

[Zeudan]

Unless it's dead, it's most likely just a battery reading bug.

Is the workaround to this issue to reconnect the battery? I've done that, and it showed the correct battery level for a few days. However, today the issue has reappeared.