Evaluation of a SIPM vs GaAsP PMT

[Michael Giacomelli-2]

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
Post images on http://www.imgur.com and include the link in your posting.
*****

Last year there was a fascinating discussion of silicon photomultipliers (multi pixel photon counters) and their theoretical advantages over PMTs:

https://lists.umn.edu/cgi-bin/wa?A2=CONFOCALMICROSCOPY;931e01af.1805

It stuck in my mind, so over the summer, I tested a current generation SIPM against a brand new 'P' model GaAsP PMT for two photon imaging to see how theory met practice:

http://dx.doi.org/10.1117/1.JBO.24.10.106503

As mentioned by James Pawley, the excess noise on PMTs is quite large when not photon counting.  I measured an effective sensitivity of about 60% of what would be expected from the photocathode QE alone.  Due to this effect, surprisingly the SIPM actually had a higher shot/excess noise limited sensitivity than GaAsP.  I didn't test against GaAs or bialkali, but the sensitivity advantage should be much larger due to the lower photocathode QE.

The SIPM has other advantages, including very low cost, higher dynamic range, lower operating voltage and the ability to saturate without damage when exposed to bright lights. They are also very nearly without excess noise, so the number of photons in a pixel can be easily calculated from the shot noise variance which is very convenient.

There are significant disadvantages however.  At low to medium speeds (less than about 1.5 MP/s for a 3 mm, red-enhanced model, about half that for shorter wavelength models), dark counts on the SIPM hurt dynamic range without sub-ambient cooling.  The detector capacitance is very large compared to a PMT (10-100x), which means that most transimpedance amplifiers will work poorly or not at all unless tuned for the specific model of detector (which took me some effort). Finally, they have a multiexponential single photon response, so an equalizing (PZC) filter is required above ~ 5 MHz bandwidth, at least for the model I evaluated.   If you can use descanned detection and a smaller detector, these will be less of a concern.

I don't think the moderate sensitivity advantage is a good reason to throw out a GaAsP detector, but for a new system operating at higher speed, with descanned detection or longer wavelengths, it may be time to think about silicon detectors instead of vacuum tubes.

  --
Michael Giacomelli, Ph.D.
Assistant Professor
Department of Biomedical Engineering
University of Rochester

[Craig Brideau]

*****
To join, leave or search the confocal microscopy listserv, go to:
http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
Post images on http://www.imgur.com and include the link in your posting.
*****

Very interesting Michael! Thanks for sharing this: I've been very curious
about silicon vs. PMTs for a while now and your report on the current
'state-of-the-art' was very helpful!

Craig

On Mon, Oct 21, 2019 at 7:38 PM Michael Giacomelli <
[hidden email]> wrote:

> *****
> To join, leave or search the confocal microscopy listserv, go to:
> http://lists.umn.edu/cgi-bin/wa?A0=confocalmicroscopy
> Post images on http://www.imgur.com and include the link in your posting.
> *****
>
> Last year there was a fascinating discussion of silicon photomultipliers
> (multi pixel photon counters) and their theoretical advantages over PMTs:
>
> https://lists.umn.edu/cgi-bin/wa?A2=CONFOCALMICROSCOPY;931e01af.1805
>
> It stuck in my mind, so over the summer, I tested a current generation
> SIPM against a brand new 'P' model GaAsP PMT for two photon imaging to see
> how theory met practice:
>
> http://dx.doi.org/10.1117/1.JBO.24.10.106503
>
> As mentioned by James Pawley, the excess noise on PMTs is quite large when
> not photon counting.  I measured an effective sensitivity of about 60% of
> what would be expected from the photocathode QE alone.  Due to this effect,
> surprisingly the SIPM actually had a higher shot/excess noise limited
> sensitivity than GaAsP.  I didn't test against GaAs or bialkali, but the
> sensitivity advantage should be much larger due to the lower photocathode
> QE.
>
> The SIPM has other advantages, including very low cost, higher dynamic
> range, lower operating voltage and the ability to saturate without damage
> when exposed to bright lights. They are also very nearly without excess
> noise, so the number of photons in a pixel can be easily calculated from
> the shot noise variance which is very convenient.
>
> There are significant disadvantages however.  At low to medium speeds
> (less than about 1.5 MP/s for a 3 mm, red-enhanced model, about half that
> for shorter wavelength models), dark counts on the SIPM hurt dynamic range
> without sub-ambient cooling.  The detector capacitance is very large
> compared to a PMT (10-100x), which means that most transimpedance
> amplifiers will work poorly or not at all unless tuned for the specific
> model of detector (which took me some effort). Finally, they have a
> multiexponential single photon response, so an equalizing (PZC) filter is
> required above ~ 5 MHz bandwidth, at least for the model I evaluated.   If
> you can use descanned detection and a smaller detector, these will be less
> of a concern.
>
> I don't think the moderate sensitivity advantage is a good reason to throw
> out a GaAsP detector, but for a new system operating at higher speed, with
> descanned detection or longer wavelengths, it may be time to think about
> silicon detectors instead of vacuum tubes.
>
>   --
> Michael Giacomelli, Ph.D.
> Assistant Professor
> Department of Biomedical Engineering
> University of Rochester
>