i have problem with AD7606. On my board I have two AD7606 with sampling 16-channels in same time. I use internal reference voltage. When I'm reading data from ADC (SPI at 21MHz, oversampling off) I have DC histogram like this (in 5 sec time window):
-3LSB = 16
-2LSB = 3'670
-1LSB = 80'610
0LSB = 267'492
+1LSB = 147'294
+2LSB = 13'050
+3LSB = 164
In data sheet on page 30, AD claims that with no oversampling maximal spread is -2 LSB to +3 LSB. If I scale my readings to total occurences from datasheet I have:
-3LSB = 0.064 (DataSheet = 0)
-2LSB = 14.67 (DataSheet = 3)
-1LSB = 322.25 (DataSheet = 131)
0LSB = 1'069.35 (DataSheet = 928)
+1LSB = 588.83 (DataSheet = 887)
+2LSB = 52.17 (DataSheet = 97)
+3LSB = 0.655 (DataSheet = 2)
I wonder if obtained accuracy is maximum from this ADC with internal VREF and with no oversampling or my board isn't good and I must modify it.
Can anyone help me with this matter?
I have tried so far:
1) change sampling ratio between 12.8kHz, 25.6kHz and 51.2kHz - same results
2) increase analog line filtering capacitor - same results
3) significantly reduced common mode noise - same results
4) added snubber on dc/dc converter - removed peaks but same results
5) added LC filter on dc/dc output - less noise but same results
6) removed L filter from ADC analog voltage - less analog voltage ripple but same results
so - I have no more idea what is wrong. It can be clock jitter ? but I measure shorted to ground inputs, so this is DC.
I have attached schematic of ADC with inputs and power source of ADC.