SPAD array as an image sensor...

March 16, 2016

 

So, while the SPAD array I developed during my PhD was intended for optical communications, I purposly included circuitry to allow the low-level SPAD counts to be read out in a manner in which the spatial relationship is still preserved. This circumvents the parallel summation tree used in the design. 

 

 

In this image a low-light level laser was focused onto the SPAD array. We can see that the sensor does operate as a low-resolution imaging sensor, which while done many times in the SPAD literature, was more for my own interest than for anything else. The shape shows the typical star diffraction pattern due to the cover glass in front of the sensor, shows a central hot spot and an -ish ary-disk-like profile.

 

The actual reason to implement the circuitry used here was to allow assessment of SPAD dark count rate and the non-uniformity of those SPADs in terms of detection rates. In the image below, 1/4 of the SPADs within the silicon are readout at a time due to the structure of the readout chain. The illumination was forced to be highly uniform allowing this image to represent the detection non-uniformity. 

The read out 16 by 16 image, can be combined with the other four readouts from the sensor, (spatial readout required selection of SPADs, then readout, then re-selection etc), to get overall non-uniformity. As can be seen some SPADs detect more, some detect less than the global average. 

 

It was interesting to assess how spatial non-uniformity impacted the error rates in the eventual optical communications application, however it seemed that significantly worse non-uniformity would be needed to impact the system. It was found that other non-idealities of the SPAD array caused more significant issues and blockaged to communications performance. 

 

 

In the third image, the signals from a particular SPAD was averaged over multiple acquisitions to obtain a good estimate of its detectivity. The signals from each SPAD from the 1024 in the SPAD array were then plotted as a sorted distribution and normalised to the maximum. Most SPADs have a detectivity around 0.6 on this normalised range, however there are a small number of SPADs that do not detect photons at the same rate as the others. Likewise there are a proportion of SPADs that have higher photon detection efficiency.

 

Thats all for now...

Please reload

Senior Electronics Design Engineer, (Digital)

Coda Octopus Products Ltd.

South Gyle, Edinburgh, Scotland, UK

  • Rg
  • orcidlogo_350x158
  • Twitter Social Icon
  • LinkedIn Social Icon

© 2019 by Edward Fisher. Proudly created with Wix.com

Dr Edward M.D. Fisher Ph.D MEng MIEEE MIET