AD8397 is "rail-to-rail" output but not rail-to-rail input. The amplifier's input voltage range is ~1V from the supply rails. Means that the amplifier can only accept voltage levels within this range. Beyond this, the amplifier output would be limited by the IVR (Vout = Gain * IVR) until of course, the Output Swing limitation takes over.
There are actually two solutions that you could implement to have the amplifier operate linearly:
1. Use dual supplies (Vs+=11, Vs-=-1) instead of single. Vs = + 6V (or Vs+=11and Vs-=-1) instead of Vs = 12V. or
2. Provide sufficient bias voltage on the inputs to make sure that the signal of interest is within the IVR and OVR of the amplifier.
I hope this helps.
Thank you very much.
Current soltution,I used the dual supply(V+=12V, V-=-3V) to implement the function.
and backup solution is that TI LM7322 took the place of AD8397.Because LM7322 is cheaper than AD8397.
if used LM7322, use single supply(V+=12V, V-=GND) to work normally. That can save cost of power IC(-3V). The total cost is lower than current solution.
i stumbled over almost the same problem which is described by the thread starter. Obviously the IVR is still not included in the AD8397 datasheet...
One thing is very strange: we wanted to use the AD8397 as an output buffer for a resolver excitation circuit. This is described in your circuit note CN-0276 "High Performance, 10-Bit to 16-Bit Resolver-to-Digital Converter". The schematic we used could be found in figure 2. We did a spice simulation of this circuit (with the AD8397 spice model) and found this thing with the limited input voltage range of the AD8397: input voltages approx. lower than 1.4 V will result in a clipped output signal (as you described in a previous post). For me it seems quite clear, that the proposed circuit in CN-0276 could not work as suggested. But if you have a look at figure 5, 6 and 7 in CN-0276, there are measurements from the input and output of the non-inverting configuration of the AD8397. It is clearly stated that the input signal is below 1 V and the output signal is not a bit distorted or clipped. Unfortunatelly i cant get this together. The simulation validates your statement but CN-0276 says something completely different.
Maybe someone could bring light into the dark concerning this issue..?