I'm working on a design using the AD8310, attempting to detect very low power RF pulses, on the order of -75dBm with pulse widths ranging from 400 to 500-nsec. My log detector output is buffered with a differential op-amp (in a single-ended to differential configuration) and then digitized with a 12-bit ADC.
When I disconnect the log detector from the buffer, the ADC reads out +/-2 counts. With the log detector connected to the buffer input (log detector input 50-ohm terminated) I am seeing a noisy waveform with peaks around +/-150 ADC counts. I disconnected all other circuits in the signal chain up to the log detector from the power supplies, so I'm convinced that what what I am seeing is coming from interactions among the log detector, buffer, and ADC.
I probed the log detector input and output with a spectrum analyzer, and did not see any frequency content above the noise floor of the analyzer. When I probe the buffer output, I see many harmonics of the 25-MSPS sampling clock.
When I probe the log detector output with a digital scope, I am seeing ripple with fast (~20nsec) rising edges and slower (100s of nsec) RC-looking falling edges and amplitudes on the order of 10's up to a couple hundred mV. I acquired an AD8310 evaluation board, and see the same fast-rise/slower-fall waveform when probing the evaluation board output with the scope, leading me to think the choppy waveform may be caused by my power supply or the scope.
Has anyone else seen this behavior and found a good way to mitigate it?