tps65070 touch pressure with battery

MIke, 

If I understand correctly, you are reading the touch screen through the TPS65070 with your OMAP and have looked at the pins TSX# and TSY# with a scope.

Is it possible you have changed the contents of ADCONFIG [3:0]  to 4b1111 instead of 4b1110?

Along with verifying the contents of ADCONFIG, I would verify that the I2C data stream matches what you are reading with your processor. 

Just verified ADCONFIG, it read 0xae.  I am using the touch driver that came with the linux psp, so it shouldn't change behavior based on my power source.

Also verified the i2c data stream using a logic analyzer.  It matches what the process is reading.

Also of note, TSC standby does properly detect touch, even when using only the battery.  The pressure mode reading though is still 1023, so I cannot detect a release.

Mike, 

Per my message, will you please provide the schematics and layout along with any supporting scope shots.

Thanks,

Daniel

Daniel,

I just tested using the TPS65070EVM-430.  I see the same problem.  I attached a 4V power supply to vbat and held S1 to turn on.  Wrote 0x02 (pressure) to TSCMODE (0x08) and 0xEE to ADCONFIG (0x07).  Read ADRESULT back as 0x3FF.  Repeating the same exact steps, changing to 5V on the AC pin, changes the ADRESULT register to 0x000.

Mike

"Interestingly enough, this does make a difference, at least on the EVM.

If I write 0xEE (vref not explicity enabled), I get 0x3ff.  I then change to 0xFE.  The next couple readings may still be 0x3FF, but it does settle down to 0x000 after a few tries (human speed, write 0xFE, read ADC, write 0xFE, read ADC, etc – slow enough when switching back and forth between tabs on the GUI).  The reverse is true, disabling VREF gives a gradual rise in the ADC result back to 0x3FF.

I will try this on our L138 SOM and see if it works…"

"Indeed, I modified the linux driver to force VREF on, and it appears to work.  Thanks Daniel!"

I see the same behaviour with the AM1808 reference platform with the Windows CE BSP.  Forcing the reference to on by |= 0x10 seems to fix the problem.