The AGC is off according to the registers read back through debugfs. Still, is it reasonable to expect the output of the PGA to be dead silent (I mean digital zeros) with no inputs connected and 40dB of gain?
Also in an earlier post I suggested that the noise floor was down below -100dBFS. According to Figure 6 of the datasheet we only get 15dB of SNR improvement from the PGA at 40dB of gain (vs 0dB of gain). Is the original noise floor more like -80dBFS?
In regards to ground noise, one of our inputs comes out to test points which are very close to the codec. I soldered the caps from those test points to a solid analog ground also very close to the codec. The noise was unchanged compared to the noise on the actual used input (ac-coupled to ground).
I've also tried to power the analog section of the codec from a battery. This didn't improve the noise either. In this case, I only powered AVdd and connected battery ground to analog ground. We have a simple 1206 ferrite tying analog to digital ground so I could try to disconnect AVss from digital ground and repeat the battery test.
Otherwise, I'm open to suggestions as to locating the source of this noise. Can anyone at TI review the schematics/layouts?