AD9286 for capturing 2ns pulses

Document created by analog-archivist Employee on Feb 23, 2016
Version 1Show Document
  • View in full screen mode

It was suggested to use analog filter for connecting signal to the AD (AD9286)
converter board. My input signal has high bandwidth (500 MHz) and it is
Gaussian pulses with 2 ns duration. I sometimes can not capture the pulses
completely with the capture board and some is missed. It would be grateful if
you make comment on it. What sort of filter is recommended there? Currently I
am not using any filter, the signal is directly connected.
I am not looking at FFT. I am just capturing 500 MHz ADC output and programming
it with my own code. The code delivered by ADI is fine and with it I was able
to see the pulses (2 ns) with scope. We sample incoherently to synchronise the
input signal to the ADC clock.
Sometimes we are having perfect capturing sometimes not. Do you have an idea
why filtering is recommended for the input signal before applying it to the
ADC? I am wondering if we would need to do that.


Filtering is recommended because input sources have harmonic content that, when
unfiltered, will show up in the FFT and degrade the converter performance. For
this application, which is time domain, I don't think a filter is necessary. If
you are dealing with short pulses with fast rising edges, you want a lot of
bandwidth so the pulse edges do not get distorted (a filter would cut out high
frequency content which would round the rising and falling edges).
I am not an FPGA expert, but I suspect it may have something to do with the
data capture as to why you aren't seeing all the pulses. From my basic
understanding, there is a FIFO that gets filled in the FPGA and is polled to
release the data. Perhaps there is something in the timing there that is
causing some pulses to be missed. Also, I don't know what/how you did
programming with your own code. There is a protocol that is needed from Visual
Analog to read the data back from the FPGA (proprietary and we normally don't
give that information out). It could be that your own programming code is not
'talking to' Visual Analog correctly.