Suspended mode

[Andy Green]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Somebody in the thread at some point said:

>> IO_3V3 is ON and 0.97 mA  

Actually, this leaves quite a lot of current unaccounted for.  I somehow
thought 1.8V rail was coming from IO_3V3 but it isn't - DOWN2IN is
VB_SYS which is Vb effectively if no USB is hooked up.  So there is
still a mystery here about:

   970uA - 200uA - 140uA = 630uA

on IO_3V3 (2mW or > 30% of suspend power accounting for regulation loss)
we didn't explain yet.

>>   S3C2442 VDDOP   ==>0.2mA             <== OK

Maybe this is a little bit suspicious now we get down to caring about
very small currents.  CPU VDDOP == "Vdd for IO Block", it pulls 200uA,
dunno if that is just normal or we can suspect at least one pin from the
CPU is routing some static current.

>>   Gensor (U7801,U7802) ==>0.07mA *2    <== Hm OK

Not OK -- datasheet says 5uA -- and we do put it into Power Down mode in
suspend.  Looking at it, I think R7809 and R7810 are unnecessary and
should be made NC, it may or may not affect the 70uA figure on IO_3V3.

After removing them, we should best drive GPGD12 and GPGD13 as output 0V
(not "KEEP") during suspend.

Also I wonder if we are meant to keep GSENSOR_3V3 powered to get the 5uA
figure, and 65uA flows inside the chip between the core and IO voltage
domains.  But we will lose many uA in the regulator keeping it powered,
so it isn't great.

We could do better on the motion sensors with circuit change -- by

 - tying VDD_IO on the chips to the same GSENSOR_3V3

 - disconnect INT2 on each chip, change INT1 to active high push-pull in
software

 - changing R7807 and R7808 to pull DOWN

 - ensure we drive GPG5 - GPG7 to 0V in suspend

with this method we give no "high" level anywhere to that chip when
GSENSOR_3V3 is not powered and absolutely 0uA current can flow.  But I
think we missed that opportunity since it certainly isn't important
enough to force a PCB change now.  Still that's 140uA at 3.3V == 462mW (12%
of our suspend consumption now!) right there burning for nothing.  Let's
try to remove R7809 and R7810 and see if it helps.

>> IO_1V8 is ON and 0.5mA                 <== SDRAM OK

Maybe this isn't crazy... we expect 250uA for selfrefresh on SDRAM.

But it only comes to 0.9mW out of our entire measured power consumption
of >6mW.  So 85% of our suspend power usage is blown on something other
than keeping the SDRAM happy, which is the only job we even wanted!  It
strongly suggests that we can still cut the 6mW in half or more.

>> I think these other currents maybe is pull high, pull down resistor, and some digital logics...etc
> 
> Agree, I'll take a last look around with this really excellent list in mind.

Yes this list of the power tree currents is great for this job.

- -Andy
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.7 (GNU/Linux)
Comment: Using GnuPG with Fedora - http://enigmail.mozdev.org

iD8DBQFHss0rOjLpvpq7dMoRAqy5AJ4ujyNVMKx/oBbxsqKQ/T+M53nPggCfeO9Y
hcOorNOuKpLn6OQaeOmBD0g=
=u2eU
-----END PGP SIGNATURE-----