From: (wrong string) én <simel346_at_student.liu.se>
Date: Sat, 25 Jan 2003 15:25:29 +0100

Andrew Jamieson wrote:
> OK. For this build (in idle state):
>
> No power save mode: 110mA
> CPU Sleep: 98mA
> CPU Stop: 95mA
> CPU Standby: 81mA

Huh?

Now, that doesn't seem right... the CPU is supposed to draw avg 0.01, max
300 *uA* in standby mode! That should make a saving of about 60 mA, not
30... Anyone have any ideas what this might be caused by?

> So, it seems having a more efficient scheduler to increase the time
> spent sleeping does not seem worthwhile, as the best we can get it
> only 3mA more than what we have at the moment

Agreed.

> It should be noted that changing the states of
> various port pins during sleep may have an effect on this current
> consumption, causing it to deviate from documented values. However,
> any further savings would apply equally to the readings of 'sleep'
> and 'stop' modes.

Wouldn't they apply also when we are not sleeping at all, or am I missing
something? It still seems to me that the *difference* between "No sleep" and
"Sleep all the time" should be 20 mA, not 15, no matter what other power
saving measures are taken. (At least if the numbers from the data sheet are
to be trusted.)

> Clearly, standby would be awesome if we could implement it, saving up
> to a whopping 20% on the best we can currently achieve.However, the
> reality of the HW is something of an obstacle. I will do some more
> hunting on my Player to see if I can find any ways to reduce the
> sleep current, and look for hope as to a standby mode. Updated
> schematics will follow.

6k_schematic.pdf still looks wrong when I'm awake. The 12 MHz crystal is
connected to NMI and WDTOVF, not to EXTAL and XTAL which are three pins
above NMI and WDTOVF. Assuming NMI is actually unconnected on the player,
there seems to be no hope of awaking from standby without a hardware mod. On
the recorder, however, it seems we could wake up by a RESET signal from the
watchdog timer in the RTC. We must sleep for at least 1/16 second, but that
should be enough if we are just waiting for keypresses. I'll look into this
some more.

--
Simon Elén

Received on 2003-01-25