AT%SLEEP=n power consumption

[Alastair Johnson]

Joerg Reisenweber wrote:
> Am Mo  20. Oktober 2008 schrieb Alastair Johnson:
>> Werner Almesberger wrote:
>>> I order to evaluate the impact of the AT%SLEEP=2 work-around, Joerg
>>> asked me to measure how the sleep mode affects the current drawn by
>>> the system.
>>>
>>> Test setup:
>>>
>>> - device is idle but (for convenience) not suspended
>>> - GSM is active and has registered with the network (AT+CFUN=1,
>>>   AT+COPS=0)
>>> - the test setup does not suffer from #1024
>>> - power is supplied to the battery terminals from a lab power supply
>>>   set to 4.0V
>>> - USB is not connected
>>>
>>> Battery current of the whole system in mA. Values are from measurements
>>> averaged over about 15 minutes each:
>>>
>>> at%sleep=       avg     min     max     samples
>>> 4                96.0    91.6   198.0   20006
>>> 3                98.4    91.6   214.6   21432
>>> 2               102.2    97.6   209.3   21428
>>> 1               104.0   101.9   197.2   21352
>>> 0               120.3   118.8   213.0   21452
>>>
>>> Integration period is one power line cycle (20ms). There are transients
>>> much faster than 50Hz. E.g., sampling at 2.5kHz finds more noise but
>>> does not change the difference in the averaged value:
>>>
>>> at%sleep=       avg     min     max     samples
>>> 4                95.1    63.1   242.5   210207
>>> 3                97.1    62.8   242.9   215460
>>> 2               101.4    68.7   243.4   208230
>>>
>>> Note that this is an idle but not suspended system, hence the large
>>> minimum current.
>>>
>>> So the difference between sleep=2 and sleep=4 is about 6mA. Joerg
>>> mentioned that the idle current of the GSM subsystem should be 4mA,
>>> so this would be an increase to 250%.
>>>
>>> Andy, you once measured the power consumption when the overall system
>>> is perfectly suspended. Do you still remember how much it was ?
>>>
>>> - Werner
>> Can you repeat the test on a handset that is suffering from #1024? I 
>> suspect that the frequent reregisters increase power consumption, but I 
>> don't have the equipment available to make the measurements. It 
>> certainly produces more instances of GSM interference on my PC speakers 
>>   which suggests more transmit operations. If this is the case then for 
>> those wth #1024 SLEEP=2 may be no worse, or possibly better, than SLEEP=4.
> 
> that's the idea behind Mickey's recent bandaid, switching to SLEEP=2 on 
> detecting #1024 behaviour
> /j

I thought the idea was to make GSM usable for people suffering from 
#1024 :-) Mickey suggests the bandaid may make power consumption worse 
in comment 45.

I may have missed something, but I haven't seen any hard evidence on the 
effect of #1024 on power consumption. Werner's results show the 
difference between SLEEP=4 and SLEEP=2 for a phone without #1024. I 
would like to see the same test done with #1024 to provide evidence on 
the effect of reregistration on power consumption. I have intermittent 
problems with #1024, and the patches won't make any difference if the 
phone is suspended when the problem starts. I also see big variations in 
battery life while suspended. If #1024 has no significant effect on 
power consumption then I can rule it out as a possible cause and look 
elsewhere.