ADC/OPAMP Selection Help

[sgt_bear]

Hello

I want to do a project with an STM32, a ADC and a Opamp. Basically i have a given pulse from a SIPM (similar to an Avalanche Photodiode). The pulse is negative, so an inverting opamp is necessary.

Find attached a image of a small pulse. To have security in the circuit, the ADC should have the capability to measue a signal that is one third of the amplitude and time of the displayed signal
Also attached a persistant screen where you can see the range of pulses. Since the pulses already make it to 1.3V sometimes, the inverting opamp has to have a -gain since most ADC's are around 2V max input.


For calculations i used the displayed 15us long singal, one third would be a signal lenght of 5us which equals a frequency of around 200kHz, therefore according to my knowledge (not that much) the ADC should be at least 10Msps. Is that correct?  How many bits would be needed? Sadly high speed, high bit adc's become fairly expensive.

For my application, the uc (i would take an STM32F7) would only have to calculate the height of the signal. So i want just to know the pulse height, after finding the pulse height, all other informations about the pulse can be wasted.

Any tips which op-amp would be good?  Do you reccomend to "deform" the pulse before measuring?

[Kleinstein]

It is possible to do some pulse forming. That is a kind of filtering on the pulse, usually to make the pulse a little longer, by removing the very high frequency part.

The pulse has a length of something like 5 us for most of the energy. If the demands are not that high, one could get away with a ADC speed of something like 5 MHz-10 MHz.  Some of the STM32F4... (e.g. STM32F407 up to 7.2 MHz) have a reasonable fast internal ADCs, that might be sufficient.

The OP for amplification should be at least as "fast" as the ADC. So it would be more like an GBW in the 10-20 MHz range that is suitable.

[David Hess]

What exactly are you trying to capture?  Timing, height, energy, distribution of these?

What is the minimum time between pulses?

I would be inclined to use an operational amplifier integrator instead of amplifier so it captures the area of the pulse instead of the height relaxing the timing requirements.  A fast peak detector is also feasible.  If there is a minimum time between pulses, then the integrator can be reset after every measurement.

[sgt_bear]

I'm trying to measure the height of the pulse.  The SIPM is places behind a scintillation material. Which puts out light. The output pulse height is inherit to the amount of light that hit the SIPM. Only the height is relevant, so the pulse can be modified/formed in any way, because the original signal is not relevant.

The minium between pulses is hard to say.

Technically from the SIPM, the microcell refresh time is 35nS, but since the scintillation material has a longer decay, it doesnt matter.
According to the situation, pulses can occour 10times per second, to 10k per second.


I'm currently trying to capture more pulses and then import them to LT-Spice to make som experiments

[sgt_bear]

I made experiments with LT Spice, i just picked an Opamp that looked good  and tried.

I made it so far that i can invert the pulse, and then make it more looking like a positive sine wave (i heard you can only read sines with an adc, since you can apply math functions to calculate the exact height without measuring the exact peak -> lower ADC bits).

But as you can see the green trace in the screenshot the part of the pulse becomes negative.... but if my ADC is voltage span 0-2V, it should read 0000 for negative pulses, or?

[Kleinstein]

So far the pulse shaping part is not doing very much - the pulse still looks rather similar. Usually the aim is to make the pulse a little longer if needed, so that the ADC can be slower. Even the original pulse shape does not look that bad, if the ADC is reasonably fast. The main purpose of the pulse shaping would be reducing frequency components above the ADCs limits.  This could be as little as an additional capacitor in the FB path of the inverter. The high pass filter function can be done in software as well, by looking at the base line before and after the pulse. To avoid negative values to the ADC one might want to add a little offset.

[f5r5e5d]

lots of peak detector circuit options at that speed - choosing requires more specs - what resolution/accuracy is needed for what range of pulse heights/widths, min reset times

the peak dector would allow much slower on-chip ADC to be used, with a fast uC in the picture various calibrations and delayed reset options are possible too