Bug 11980 - Controller in suspend mode drains battery longer than it should; needs to shut down at certain level
: Controller in suspend mode drains battery longer than it should; needs to shu...
Status: CLOSED FIXED
Product: SB Controller
Classification: Unclassified
Component: Power Management
: unspecified
: PC Windows XP
: -- normal (vote)
: 7.3.3
Assigned To: Richard Titmuss
:
Depends on:
Blocks:
  Show dependency treegraph
 
Reported: 2009-05-05 17:58 UTC by Ross Levine
Modified: 2009-09-08 09:27 UTC (History)
6 users (show)

See Also:
Category: ---


Attachments
the chart I failed to attach (199.28 KB, image/png)
2009-05-06 11:08 UTC, Ross Levine
Details
18 hour chart of 3993 vs 5730 (145.50 KB, image/png)
2009-05-14 12:10 UTC, Ross Levine
Details
serial capture (85.00 KB, text/plain)
2009-05-14 12:50 UTC, James Richardson
Details
6038 v 3993 graph after 8 days (94.22 KB, image/png)
2009-06-18 17:59 UTC, Ross Levine
Details

Note You need to log in before you can comment on or make changes to this bug.
Description Ross Levine 2009-05-05 17:58:21 UTC
A suspended controller should wake itself check its voltage and shut down if below a certain level. Based on this chart we believe controllers in suspend mode are discharging the battery too long. 

I'll repeat this test more aggressively waking Controllers and take notes as to what Controllers wake, or don't wake, at each interval.
Comment 1 Blackketter Dean 2009-05-06 07:51:47 UTC
Richard: the theory here is that the rounded curves that drain the battery over 6 hours until it's dead are as the CPU is suspended, can't wake but also won't turn off.  This is bad for the battery.  

The proposed solution is that the CPU wake every couple of hours long enough to measure its battery level and do a graceful shutdown if it's below a certain level.

Caleb: What voltage do you recommend?

Oh, and Ross, please attach the chart.  :)
Comment 2 Ross Levine 2009-05-06 11:08:09 UTC
Created attachment 5199 [details]
the chart I failed to attach
Comment 3 Ross Levine 2009-05-06 11:40:50 UTC
(In reply to comment #1)
> Caleb: What voltage do you recommend?

Spoke with Caleb, he says 3.2 volts based on the battery spec. 

I've got some more testing to do.
Comment 4 Ross Levine 2009-05-07 13:18:22 UTC
Using a variable power supply and a volt meter I measured what I believe to be the ranges of battery level indicators on the Controller UI. 

Full battery icon 3.7V+
3/4 battery icon 3.620V - 3.700V
1/2 battery icon 3.535V - 3.620V
1/4 battery icon 3.480V - 3.535V
Red battery icon 3.420V - 3.480
Battery low please charge me 3.420V or less.
Comment 5 Ross Levine 2009-05-07 13:23:18 UTC
Also this is the wiki entry I can never seem to find when Dean is over my should asking for this information. 

http://wiki.slimdevices.com/index.php/SqueezeOS_Internals
Comment 6 Ross Levine 2009-05-08 12:01:35 UTC
Standing by to test new changes.
Comment 7 Richard Titmuss 2009-05-11 14:24:07 UTC
Fixed in r5716-r5718. The controller now wakes up every three hours, checks the battery and turns off if it is too low. Ross is making some measurements that will allow us to fine tune when we turn off.

I am going to be testing this tomorrow myself, before releasing to QA.
Comment 8 Ross Levine 2009-05-11 16:34:27 UTC
Fluke measured voltage, jivectl 17 output (decimal base 10):
4.2     949
4.15    938
4.1	928
4.05	916
4	905
3.95	894
3.9	883
3.85	865
3.8	858
3.75	847
3.7	835
3.65	826
3.6	815
3.55	800

At this point system goes down and I can't jivectl 17 anymore.
Comment 9 Ross Levine 2009-05-14 12:10:03 UTC
Created attachment 5216 [details]
18 hour chart of 3993 vs 5730

Something is very wrong with 5730. After about an hour I wasn't able to wake 4/4 controllers with 5730. I connected serial to one of them and pulled the batteries out of all 4 to force a reboot. The serial one (G) then didn't reproduce the problem! 

This morning the 3/4 controllers running 5730 without a serial connection are all below 3v and won't turn on.
Comment 10 James Richardson 2009-05-14 12:50:45 UTC
Created attachment 5217 [details]
serial capture

Richard: Here is a serial log of the error.  This is captured right from the Jive.  Notice at the end, it just dies during the capture.  At this point, the jive is non responsive

Holding power for +10 seconds does nothing
Placing the controller in the Shuck does nothing
Plugging power into the Serial Dongle does nothing

Removing cycle battery is the only way to bring the controller to life again.  Battery is at 4.062 v

BTW: Jive 7.4r5580 does not have this issue // Jive 7.3r5577 does not have this issue
Comment 11 Ross Levine 2009-05-19 16:23:18 UTC
Richard I have questions about the implementation of the fix for this bug. From what I understand currently you've set Controller to wake (sort of) every hour on the hour, check its battery status, and returns to suspend or shuts down based on the voltage reading. I was under the impression the screen and backlit keys should only be on for 1/4 of a second or a very brief time. 

What I'm seeing is every hour on the hour my Controller wakes and stays awake until it sleeps, and then goes into suspend based on the normal timers in factory test settings. I'm using r5772 and I have made changes to the factory test timers.
Comment 12 Richard Titmuss 2009-05-28 06:44:05 UTC
*** Bug 12094 has been marked as a duplicate of this bug. ***
Comment 13 Richard Titmuss 2009-05-28 10:34:37 UTC
Hopefully! fixed in 7.3 r5921.
Comment 14 Spies Steven 2009-05-28 11:31:34 UTC
Richard, just to clarify should that be 7.3 r5921 or 7.3 r5922?
Comment 15 Ross Levine 2009-05-29 17:33:50 UTC
I'm seeing some problems with r5922 that are hindering verifying this as fixed. After almost 4 hours of testing 100% of the r5922 controllers I'm testing won't wake from suspend. Lets talk next week Richard.
Comment 16 James Richardson 2009-06-17 09:37:29 UTC
This bug has been fixed in the 7.3.3 release version of SqueezeCenter!

If you haven't already. please download the new version from http://www.logitechsqueezebox.com/support/download-squeezecenter.html 

If you are still experiencing this problem, feel free to reopen the bug with your new comments and we'll have another look.
Comment 17 Ross Levine 2009-06-18 17:59:52 UTC
Created attachment 5349 [details]
6038 v 3993 graph after 8 days
Comment 18 Blackketter Dean 2009-06-18 20:41:54 UTC
So, is that chart good?
Comment 19 Ross Levine 2009-06-19 11:25:23 UTC
(In reply to comment #18)
> So, is that chart good?

I think this is bad, but need Caleb to confirm, he said he'll have time next week. The 6038 batteries were drained to the 2.265v shutdown after 6-7 days.
Comment 20 Ross Levine 2009-06-24 16:54:50 UTC
Measured the voltage at which each of these 8 Controllers under test indicate please charge me. Also measured the current when powered off and in suspend mode, hopefully this will help explain the two oddities in my graph. Thanks Caleb!

controller A
please charge me 3.56v
powered off .68mA
suspend 8.72mA

controller B
please charge me 3.57v
powered off .73
suspend 6.41mA

controller C
please charge me 3.6v
powered off .72mA
suspend 6.56mA

controller D
please charge me 3.56v
powered off .77
suspend 7.84mA

controller E
please charge me 3.56v
powered off .76mA
suspend 7.32mA

controller F
please charge me 3.6v
powered off .75mA
suspend 6.96mA

controller G
please charge me 3.58v
powered off .71mA
suspend 6.96mA

controller H
please charge me 3.58v
powered off .77 mA
suspend 8.81mA

Should I look into the current draw in suspend/off modes of more Controllers? Some of these Controllers are MP (RMAs from James) and MPQ, some look to be older.