Comparator excites op amps in sensitive radiation detector

[MarcinWachowiak]

Hello,
I have encountered a problem while trying to assemble semiconductor radiation detector that I described here:
https://hackaday.io/project/159909-gamma-pin-semiconductor-radiation-detector
There were some problems at the beginning with the 3- stage op amp, because the gain was set too high - 1000x an there probably occurred a crosstalk that excited it all.
I reduced it to 100x and now it behaves fine.

While trying to assemble the comparator segment I found out that it gets excites and affects the op amps even after a single pulse.
The first three op amps serve to amplify the signal then the last one is designed as a subtractor to lower the upper noise level just below 0V/ground so that the pulses from radiation are easily distinguishable above the "floor". The comparator has little hysteresis and is supposed to detect positive pulses, but instead after a single detection it goes on a rampage and generates a square wave which causes the signal to go full sine from rail to rail.

I will increase the caps in the three op amps to 100pF, and the hysteresis loop to 10mV, but I have no further idea what else may be causing the oscillations.
I have a different design already in mind with a layout consideration, but I would like to make this one work too...
Can you advise me, please?

[iMo]

I did it with 3 transistors in past getting 10-300mV pulses from a single PIN diode. The 10 diodes in parallel - the total capacitance must be great.. You can hardly get a pulse out of it.. Over-engineered, imho.

I would advice you to make a simulation (ie LTSpice) with your string of opamps and comparator. Increasing the capacitors in the feedbacks may affect the amplitude and shape of the pulses you want to measure.

I would not use the TL3016 "ultrafast 7.6ns comparator" with 47k/1ohm feedback. The pulses are generally "slow" (when talking their frequency and edges). An old LM311 would be ok there, imho.

[iMo]

Here is an app note from Maxim re pin diode gamma detector:
https://www.maximintegrated.com/en/app-notes/index.mvp/id/2236

[ConKbot]

Try adding more decoupling. A single 100n probably isn't enough to properly supply 6 different ICs, which include relatively fast opamps,  fast comparators and the like.