Hi,
You do not need to use the SPI port to program internal registers to get basic operation of the ADS5400. Our ADS5400 EVM will operate with the TSW1400 or TSW1405 capture cards without having to use the SPI GUI to program the ADC on the EVM in any way.
I notice on your schematics that the SPI pins are floating - it might be safer to pull SCLK low or high. I doubt that it would cause a problem as is - usually single ended inputs do not float to a level where the input might 'chatter' but it would be safer to tie off the input to be sure.
The sample clock to the ADC in your schematics is not differential. It is not advisable to drive a single ended swing into one side of the differential input. The inputs are biased to a common mode voltage, but with a signal on one side and not the other, the common mode voltage at the inputs is no longer a constant. This would be particularly bad on the analog input, but also the clock input wants to see a balanced differenital signal. I don't know if doing this to the clock input would have an effect on the devices VCM used to bias the analog input, but i will talk to the characterization team about it.
Now - regarding your description of what you see on the sample data - I cannot tell from the descripiton what it is you are seeing. There *will* be some DC offset up to the max specified in the datasheet so you will not see an exact mid-scale code. And there will be some idle-channel noise so the lsb's will have some random activity. And - if the DC offset happens to be smaller than the idle channel noise, then as the idle channel noise may bounce around the midscale code which mkes *all* the sample bits switch as the code goes from below 0111 1111 1111 to above 1000 0000 0000.
Beyond that - i would need to see an example of what you are seeing on the sample data to comment further.
Regards,
RIchard P.