[Gta04-owner] The last bug - or is there always one more?
Dr. H. Nikolaus Schaller
hns at goldelico.com
Fri Jan 20 10:04:34 CET 2012
While testing the GTA04A4 boards waiting to be shipped,
we observed some strange behaviour of the new circuit
that controls the power of the UMTS module.
Severals days of analysis revealed that we had expected
that the module goes to a low-power mode (in uA range)
if the reset pin is active. But it turned out that this is not the
case. If we reset the module by the hardware control line, it
draws ca. 16 mA.
This means that if the CPU is completely shut off, the battery
is drained within 75 hours.
So we had to develop another fix that can be easily retrofitted
into the boards that have already produced (unfortunately
not for the GTA04A3 boards). It will be a small NPN transistor
and a small wire replacing 3 resistors.
With this addition it is possible to shut down to (measured!)
250 uA. This is the remaining power demand of the oscillators
(RTC) and some other circuits in standby mode.
Adding this NPN needs some time, but So we have to fix all
the boards waiting to be shipped and this will be done until
end of next week (because we have to order the NPN in
a small package that can be easily installed). Sorry for the
delay to all those who have already ordered the "from stock"
devices and did expect to receive much earlier.
So we will have boards and new users to meet at FOSDEM!
And I am now confident that this is the last important hardware
bug we have. So let's see if we have another one...
For the software side I think I also have an explanation of
the 70 mA reported for the suspend state of the 3.x Kernel.
During the analysis mentioned above, I found that the module
basically has these current levels:
some uA if powered off
3-9 mA if registered and nothing to do
30-40 mA in some intermediate state
100-120 mA when trying to register
more during UMTS data call
And the most important observation was that the module
may be in an alternating state where it draws 3-9 mA over
the first 35 seconds and 100-120 mA for the second 25 seconds.
Because the board was operated without antenna in a
room with weak reception, I think the following happened:
The device receives one base station (receive only with 3-9 mA)
and then tries to register with it. But since there is no antenna
the base station will not receive the request.
So the module tries to register forever, alternating between
listening and transmission mode.
This gives an average current of 50 mA.
If we assume that the CPU, DRAM etc. still draws 20 mA
during suspend, it could well be that the module is in the
same state as mentioned above.
This also reminds me about reports about other cellular
devices, where the battery drains much faster if the device
is stowed away in a position with weak reception.
More information about the Gta04-owner