[Gta04-owner] GTA04: switching off the VAUX2 supply for the Gyro and Compass may not save power as you would expect...
Dr. H. Nikolaus Schaller
hns at goldelico.com
Sun Apr 22 17:16:02 CEST 2012
VAUX2 drives the power of the ITG3200 gyroscope
and the HMC5883L compass. Both chips are specified
for ca. 2.2 - 3.6 V. VAUX2 is typically 2.8V.
All this is ok - until we think we can reduce power by
switching off VAUX2 completely. This is done by default
in all current kernels (hw-validation and 3.3).
If VAUX2 is switched off, both chips still get the 1.8V logic
supply and violate the specification of VIO <= VDD.
This means that there may flow some current from the
1.8V rail through parasitic diodes into the chips to the
VDD rail. Although it is a quite low voltage (I did measure
0.4 - 1 V), this appears to be sufficient to drive the
charge pump of the ITG3200 so that I could see approx.
3.6V at C1313 (it is 25 V in normal operation). This means
the chip is drawing some energy in this state.
So, we must never shut off the VAUX2 regulator (except
Therefore, I have set the regulator code to always_on
in the hw-validation kernel (1):
With this patch, VAUX2 is always enabled and the chips
draw some current (6.5 mA typ. for the ITG3200 and ca. 100 uA
for the HMC).
But how can we get the chips reduce power? Both must
have the drivers installed and initializing. And a suspend operation
must set the correct sleep bits (e.g. register 62 for the ITG / idle
mode for the HMC5883L). I.e. without loading a driver they
may not go to sleep.
This also means that the description at
is no longer valid. Sleep/Power-Down mode is set by
setting the sample rate to 0.
I don't know if the suspend scheme of the Kernel-3.x already
includes the drivers and sends the chips to sleep mode during
initialization and standby. So we may be able to save some more
(1): we could even remove it completely since VAUX2
comes up with 2.8 V by default (PMU and U-Boot)
so we do not need any active code in the kernel.
More information about the Gta04-owner