I'm using a ADA4392 to convert a single ended input to a bidirectional output suitable for a AD7641.
You will find a circuit diagram attached here. (Which is only a part of the whole, but sufficient I hope).
The problem is, that the difference of the positive and the negative output is not Null but about 40mV instead. With a Vocm of 1,25V I would expect, that both ouputs are on 1,25V. But one is slightly (a bit more than 20mV below, and the other above (about 16mV).
As I use this Amp for an 18 Bit converter a difference of that magnitude will have an remarkable impact (assumed a 2,5V reference).
I use 0,1% tolerated resistors. From AN-1026 I concluded that even considering resistor tolerance and the impact on gain mismatch would give me at most a difference voltage of about 2,5mV.
I saw that differential output when both inputs are connected to ground and either Vocm set to 1,25V (as shown in the diagram) or set to Gnd level.
The outputs are slightly unbalanced too, but this amounts to about 2mV.
The same effect occured on another prototype board, but with about 20mV difference.
I made the test, mentioned here, with either no load or about 1k Ohms.
I had a look at the supply. Although there is some noise, it does amount only to about 100mV or so. No oscillations or so.
I further reduced the resistor values to about 550/560 Ohm. This reduced the difference to about +5/-2,5mV from Vocm. Funny! I didn't realize that resistor noise would come up to such an amount.
Measuring was done with a Keithley 2400 (just calibrated) and a LeCroy WaveSurfer 44Xs.
Any hint or idea?
BTW: What is the correct term for naming an unwanted/unexpected difference on a differential output?