Bug 11619 - Controller doesn't suspend consistently after r3993.
: Controller doesn't suspend consistently after r3993.
Status: CLOSED FIXED
Product: SB Controller
Classification: Unclassified
Component: Product Quality
: unspecified
: PC Windows XP
: P2 major (vote)
: 7.3.3
Assigned To: Richard Titmuss
http://forums.slimdevices.com/showthr...
:
Depends on: 11824
Blocks:
  Show dependency treegraph
 
Reported: 2009-04-02 12:48 UTC by Ross Levine
Modified: 2009-09-08 09:11 UTC (History)
7 users (show)

See Also:
Category: ---


Attachments
Caleb's script log file (1.66 MB, application/octet-stream)
2009-04-08 18:18 UTC, Ross Levine
Details
2nd test, new batteries (156.13 KB, text/plain)
2009-04-13 11:45 UTC, Ross Levine
Details
Ross's 2nd test data made all pretty with graphs and stuff (12.72 MB, application/vnd.ms-excel)
2009-04-16 09:02 UTC, Chris Owens
Details
full battery test (514.10 KB, image/tiff)
2009-04-21 18:05 UTC, Ross Levine
Details
first 3 hours (569.17 KB, image/tiff)
2009-04-21 18:05 UTC, Ross Levine
Details
last 30 hours (539.70 KB, image/tiff)
2009-04-21 18:05 UTC, Ross Levine
Details
first 50 hours of 3993 vs 5497 (3.57 MB, image/png)
2009-05-05 17:01 UTC, Ross Levine
Details
resized (266.34 KB, image/png)
2009-05-05 17:02 UTC, Ross Levine
Details

Note You need to log in before you can comment on or make changes to this bug.
Description Ross Levine 2009-04-02 12:48:13 UTC
I put two Controllers side by side and Caleb's script is measuring their voltage and recording every two seconds using the DLPIO8. I will upload the results when the 3993 Controller finally dies; right now the 4824 Controller is dead, the 3993 one looks to have about half its life left according to on screen meter. The plan is recharge and switch batteries then repeat the same test, once this one is complete.
Comment 1 Ross Levine 2009-04-08 18:18:53 UTC
Created attachment 5084 [details]
Caleb's script log file

Switched batteries between Controllers, now the 4824 lasts longer than 3993. Note that fully charged the batteries differ 4.06 to 3.91. 

Richard / Dean, what voltage should I expect to see from a fully charged brand new battery? Also, at what voltage should I expect the Controller to shut off? From what I see it looks like around 3.6V they shut down. Sadly this test may have not been complete, we experienced a power outage when the 3993 Controller was totally dead but the 4824 still had life left.
Comment 2 Blackketter Dean 2009-04-08 22:14:17 UTC
Sorry, I'm confused.  Did the battery life move with the battery?
Comment 3 Ross Levine 2009-04-09 11:10:00 UTC
Yes Dean, I'll get a hold of some brand new batteries and try this again today.
Comment 4 Ross Levine 2009-04-13 11:45:22 UTC
Created attachment 5107 [details]
2nd test, new batteries

2nd run of this test. I got some new batteries from Diane and they both reported similar voltage with the meter at full charge. r3993 Controller started at 4.06v, r4824 started at 4.09. Unfortunately my PC suspended sometime during the test. Now the r3993 Controller is dead and reporting 3.48v (3.42 from DLPIO8), the r4824 Controller reports 3.71 (3.66 from DLPIO8) and shows about 1/4 left on the screen. 

I'll start this test once more tonight.
Comment 5 Chris Owens 2009-04-16 09:02:56 UTC
Created attachment 5132 [details]
Ross's 2nd test data made all pretty with graphs and stuff
Comment 6 Chris Owens 2009-04-20 09:20:10 UTC
Ross, what is the goal with this bug now?  What were we really trying to learn with this experiment?  What action is left?
Comment 7 Blackketter Dean 2009-04-20 09:30:54 UTC
I can chime in here:

- First we want to see the impact the fixes that richard put in to improve responsiveness (at the expense of some battery life)

- Second we want to take this opportunity to understand the battery discharge profile over time and user activity and make sure we're making the right trade offs.

Based on the last set of measurements, it looks like the first part is that the changes don't appear to have a significant impact, though we should probably measure the device with more use.  I suggest a motion table to keep the device awake.

And looking at the same charts, it seems that we can probably improve our overall battery life significantly by reducing the time-to-suspend in some cases.

For this, I propose that we suspend after 5 minutes if woken from suspend without the user interacting with the device.
Comment 8 Chris Owens 2009-04-20 10:12:11 UTC
See also bug 11824
Comment 9 Ross Levine 2009-04-21 18:05:13 UTC
Created attachment 5150 [details]
full battery test
Comment 10 Ross Levine 2009-04-21 18:05:34 UTC
Created attachment 5151 [details]
first 3 hours
Comment 11 Ross Levine 2009-04-21 18:05:50 UTC
Created attachment 5152 [details]
last 30 hours
Comment 12 Ross Levine 2009-04-21 18:53:35 UTC
Excel file is 40mb. A little more information. I started this test at 6pm on 4/17, I measured each battery (and labeled them) with the fluke meter:

controller A battery 3 (4.093 fluke) 
controller B battery 5 (4.190 fluke)
controller C battery 6 (4.098 fluke)
controller D battery 7 (4.165 fluke)
controller E battery 8 (4.129 fluke)
controller F battery 9 (4.126 fluke)
controller G battery 10 (4.174 fluke)
controller H battery 11 (4.034 fluke)

1pm today here is the status of each Controller:

A turns on, displays charge me!
B turns on, 50%
C turns on, 50%
D turns on, 25%
E won't turn on
F won't turn on
G turns on, 10-25%
H won't turn on

Controllers A through D are r3993, Controllers E through H are r4824.
Comment 13 Mickey Gee 2009-04-23 11:15:09 UTC
Have confirmed findings on these versions of firmware. Suggestions for improving overall battery life now in bug 11824. Closing this one to point to 11824.

*** This bug has been marked as a duplicate of bug 11824 ***
Comment 14 Ross Levine 2009-04-23 11:27:59 UTC
Turns out this is actually not a duplicate. There are two open battery issues currently. First this bug, 11619. Battery life is worse in r4824 than it is in r3993. It appears from the charts, particularly comment #10, that 4824 Controllers don't suspend after 60 minutes. The other bug 11824 is a more general bug about Deans changes we should make to default timers to improve battery life.
Comment 15 Chris Owens 2009-04-24 11:20:21 UTC
Would a log file of a unit that is not suspending on time be helpful, Richard?
Comment 16 Chris Owens 2009-05-04 09:12:59 UTC
Ross should we change the summary of this bug to 'Controller doesn't go into sleep on time' or something more specific?
Comment 17 Ross Levine 2009-05-05 17:01:09 UTC
Created attachment 5195 [details]
first 50 hours of 3993 vs 5497

At 3.65-3.7 volts the Controller will no longer turn on. During this test I would wake the Controllers every 2 hours or so. 3993 Controllers wait 60 minutes to suspend, 5497 Controllers wait 20 minutes (while not playing). This chart shows the 3993 Controllers reaching non-waking voltage levels somewhere around 28 hours while the 5497 Controllers continue to wake up at the 50 hour point.
Comment 18 Ross Levine 2009-05-05 17:02:49 UTC
Created attachment 5196 [details]
resized

Sorry this is a more reasonable sized image.
Comment 19 Ross Levine 2009-05-06 11:58:00 UTC
Fixed in r5497, Controller now consistently suspends when it should with new timers. Bug 11824 to remain open for other general power consumption changes, namely how much and which direction movements to wake on. Also bug 11980 for newly discovered potential battery life shortening bug.
Comment 20 James Richardson 2009-06-17 09:37:19 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.