Other Parts Discussed in Thread: OPA860, OPA695,
Hello experts,
I am working in a tough design where we are trying to measure the peak value of very fast pulses. The pulses are shaped as a double exponential, with a rise time of around 1.2 ns and a fall time around 8-10 ns. The pulses arrive randomly, I mean, with a random delay/interval between consecutive pulses. The application is nuclear spectroscopy, that is, we are not interested in the time of arrival of the pulses however, we want to measure as precisely as possible the height (peak) of each incoming pulse.
Additional info:
- The expected input pulse peaks can be 50 to 200 mV. I can (voltage) amplify the pulses with much problem prior to feeding the pulse peak measurement circuit.
- We also have a pulse trigger/detector, so that I can sample and hold the output of the peak measurement circuit a few nanoseconds after the pulse arrival, when the output of the peak detector has settled.
- After the sample & hold, we have a 100 MHz digitizer to capture the pulses peak information, for further processing.
I am yet reading about three possible topologies,
A) Open loop peak detection, with a high bandwidth high slew rate OPAMP/TIA and a very fast RF diode, for instance, OPA695 or OPA860. I assume this may be the fastest solution, however, we will have to characterize each circuit, measure and/or compensate the offset and diode voltage drop, which sounds like not very precise and labor intensive (!)
B) Closed loop peak detection with a pair of, again, OPA695 or OPA860. I am not sure if there is a more performant IC amplifier for the task. This approach has the great advantage of removing the diode drop of topology A, however, I am afraid I will not be able to get the required measurement speed.
C) Using the OPA615 as integrator (figure 47 of the datasheet) or peak detector (figure 48 of the datasheet). This is the topology I am less familiar with, however, it looks like the engineers who wrote the datasheet were targeting a class of problems similar to the one I am trying to tackle.
Any kind of feedback is highly appreciated, specially which approach A/B/C is more performant in terms of speed & accuracy.
Kind regards,
Jose Jimenez
