I'm using an AD8330 to amplify small signals (pk-pk of some tens of mV) and I need a very clean step response. The analog chain is built with a 5 V/V preamplifier with a differential output, an AD8330, and a 125 MSPS ADC (an ADS6125 if that matters). Please see the following graph:
The orange trace is the signal applied to the input of the chain, the blue trace is the acquired signal scaled to overlap the input. The AD8330 gain is set to 1 V/V (VMAG=0.5V and VDBS=1.5V).
The input signal actually has also a small offset (about 20 mV) that has been removed from the plotted data in order to "zoom" the end of the pulse with a log scale.
You can see how the two signals are initially identical (when the signal is "big" and therefore small distortions don't matter much) but when the signal approaces zero, there is an overshoot.
The overshoot is quite small in size (about 200 uV referred to the input signal, i.e. about 1 mV referred to the AD8330 output) but it is quite long (7 us).
I'm sure that the cause of the overshoot is the AD8330 for two reasons:
1) Changing the AD8330 gain to 10 V/V, and scaling again the acquired signal to the input signal, the overshoot decreases in relative magnitude. This should mean that the phenomenon is not generated at the input by the preamplifier:
2) Bypassing the AD8330 completely (connecting INHI to OPHI, INLO to OPLO, and biasing the preamplifier with the ADC common voltage) the overshoot almost disappears:
So, after all of this description, my question is simple: is there a way to compensate this phenomenon or is it inherent to AD8330?
I've tried filtering the input and/or output signal, but since this overshoot is very slow (a 7 us transient need a <1 MHz low-pass filter) removing it will destroy my real signal.
Thank you for the patience. Any suggestion is very welcome.