[Gta04-owner] GTA04 suspend power consumption

[Andreas Kemnade]

On Thu, 12 Nov 2015 22:26:10 +0100
"H. Nikolaus Schaller" <hns at goldelico.com> wrote:

> Hi,
> 
> Am 02.11.2015 um 08:05 schrieb H. Nikolaus Schaller <hns at goldelico.com>:
> 
> > 
> > Am 30.10.2015 um 19:18 schrieb H. Nikolaus Schaller <hns at goldelico.com>:
> > 
> >> Hi,
> >> I have aquired a thermographic camera and made some pictures of a running GTA04
> >> (without case and battery connected through external cables so that it does not cover
> >> the chips on the PCB). 
> >> 
> >> Here are some first findings and conclusions (which may be wrong):
> >> 
> >> running from battery:
> >> 	<http://download.goldelico.com/gta04/images/FLIR00003.bmp>
> >> 
> >> USB charging:
> >> 	<http://download.goldelico.com/gta04/images/FLIR00004.bmp>
> >> 
> >> suspended (power LED still on):
> >> 	<http://download.goldelico.com/gta04/images/FLIR00007.bmp>
> >> 
> >> tps65950 during suspend:
> >> 	<http://download.goldelico.com/gta04/images/FLIR00012.bmp>
> >> 
> >> resumed:
> >> 	<http://download.goldelico.com/gta04/images/FLIR00013.bmp>
> >> 
> >> chip placement to compare:
> >> 	<http://download.goldelico.com/gta04/images/DSC01393.jpg>
> >> 
> >> During operation it is visible that the TPS65950, the charging control FETs, the
> >> CPU+PoP and the USB3322 are "hot spots".
> >> 
> >> During suspend it was really unexpected. The TPS65950 stays at around 30°C
> >> (23°C surrounding temperature) which means quite some energy is "consumed".
> >> 
> >> And, it is visible that the chip has 4 "quadrants". Looking up the quadrants deeply
> >> in the board layout and TI documentation (BGA ball assignment) makes me think
> >> it is the VDD2 regulator which drives VDD_CORE of the OMAP3.
> >> 
> >> VDD_CORE drives all the peripherals, GPU and some other components of the 
> >> OMAP3 SoC while VDD_MPU drives the Cortex A8 and 1V8 drives the DRAM.
> >> 
> >> What I don't exactly know is which power rail is driving the built-in DSP and camera
> >> image signal processor.
> >> 
> >> So my first conclusion would be that the 25 mA beyond 3.7 optimization we still
> >> see comes from a GPU or DSP that is not properly suspended or disabled.
> >> 
> >> Well, we have no drivers for that...
> >> 
> >> We don't have them on a 3.7 kernel either, but 3.7 might handle some minor
> >> clock/enables/resets differently to get a better situation.
> >> 
> >> So this are first results. More experiments may be able to prove or disprove this
> >> idea. But kernel developers may be able to build on this and better focus their
> >> search for the missing bit.
> >> 
> >> If I find time, I can try to prepare (restore) an old µSD to boot 3.7 and do a
> >> comparable experiment.
> > 
> > 
> > Ok, yesterday I found some time. Didn't need as much as expected.
> > 
> > I just had to recompile and install the old 3.7-neil-plus kernel [1] and remove the
> > device tree binary so that u-boot starts the kernel with platform data.
> > 
> > Results are again very interesting and confirm what I already was assuming:
> > 
> > 3.7-neil-plus running from battery:
> > 	<http://download.goldelico.com/gta04/images/FLIR00015.bmp>
> > 
> > 3.7-neil-plus suspended (power LED still on):
> > 	<http://download.goldelico.com/gta04/images/FLIR00017.bmp>
> > 
> > 	Note: the rs232 driver is also not suspended as you can easily spot.
> > 	The reason is that the ./measure-suspend script can not find the controlling
> > 	GPIO on the 3.7 kernel at the same location.
> > 
> > 3.7-neil-plus resumed:
> > 	<http://download.goldelico.com/gta04/images/FLIR00018.bmp>
> > 
> > There is no hot spot in the tps65950 during suspend!
> > 
> > So it might also be an issue with the twl4030/tps65950 suspend control.
> > Or the sound subsystem (e.g. speaker amplifiers not turned off?).
> > Or the OTG port remaining active?
> > 
> > Now it is up to the kernel specialists (and everyone who wants to become one)
> > to draw the right conclusions and propose fixes. If they compile, I can test them.
> 
I am wondering how to systematically do that. Maybe avoid enabling stuff in uboot
like no lcm init, and then trying to use feature-reduced kernels, like
without audio or without video. 
Or kexec'ing such a kernel from a 3.7 kernel with most stuff powered down by that?
Does kexec work with device tree stuff?

Regards,
Andreas Kemnade
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 490 bytes
Desc: not available
URL: <http://lists.goldelico.com/pipermail/gta04-owner/attachments/20151118/638d31cc/attachment.asc>