I use differential inputs LINP/LINN and RINP/RINN. Full scale is 2.83Vpp (AVDD=3.3V, 0dB – pg.5 of datasheet ADAU1361). But on page 31 of datasheet in division “ADC Full-Scale Level” said about -1.38dB.

What codes I get from ADC, if I apply on LINP=2.5+2.83/2 and LINN=2.5-2.83/2?

Registers setting:

R8=R9=0100011 (Differential Input Volume 0dB),

R11=0 (ALC off),

R20=R21=0 (Input Digital Volume 0dB)

It will be 0x7ffff or 0x6d3262?

If I apply a little above 2.83V: LINP=2.5+2.84/2 and LINN=2.5-2.84/2, will be it clipping? What codes I get from ADC in this case: the same (0x7ffff or 0x6d3262) or 0x6d9529?

If I set R8=R9=00111111 (Differential Input Volume -0.75dB), what input full-scale level I get? It will be 3.085Vpp and the same ADC codes (0x7ffff or 0x6d3262)?

There is parameter for differential PGA input on pg.5 “Gain Error -14%”. What does it mean (-14%=-1.38dB)? But for single-end input “Gain Error” is -12%...

Alexander

If you input the datasheet's defined full-scale input, which is 2.83 Vp-p centered around CM which is 3.3/2, then you should see a peak level of 1.0 (in 5.23 format) in the DSP. This 1.0 value can be read back as 0x00800000. Remember that the DSP core has 28 bits of resolution, not 24.

If you exceed the full-scale value of the ADCs, the ADC output to the DSP will saturate (clip) at 0x00800000.

I believe the gain error listed in the datasheet represents the deviation between the desired PGA gain and the actual gain achieved, but this error can vary depending on the properties of the silicon from chip to chip. If you have a chip that has a gain error of -10%, for example, and you set the gain of the PGA to be +6 dB (a linear multiplication of 2), then the actual gain achieved by the PGA may be 2 * (90%), or 1.8.

The gain error is listed as a typical spec in the datasheet because it represents the average deviation from ideal PGA gain measured over a large number of devices.

So, it's possible that if you have a gain error of -10% and you input the full scale signal to the ADC via the PGAs, then the value you read in the DSP will be 1.0 * 90% = 0.9, which is represented as 0x00733333 in hex.

I think the explanation on page 31 needs a revision. The full scale level of the ADC should not be described using Vrms units. It simply doesn't make sense, unless you clearly define the waveform as being a sine tone. I need to check with the design team about this, because the apps engineer who wrote the datasheet is no longer in our group. For now, please use the spec table as a guide.