Albert Theuwissen continues is review of IISW 2015 presentations. The latest part talks about Tohoku University, Dartmouth School of Engineering, CEA, Caeleste and ON Semi work on increasing the CG all the way up to 400uV/e- (Caeleste in 0.18um process).
I think Albert missed part of the Dartmouth talk. Our 2015 IISW presentation briefly included the summary results of measuring two different jot devices, each measured two ways. Using PTC, the CGs were 242 and 403 uV/e- respectively. Using a different method that we think is more accurate called PCH, we obtained 256 and 426 uV/e- respectively. Both devices exhibited quantized electron peaks in their histograms. One example (the lower performing jot) was reproduced in this blog a few weeks ago. To see peaks, you need less than 0.45 e- rms of read noise. Those brief slides will be posted to the IISW library with all the other papers and presentations in August sometime.
ReplyDeleteEric, I have based my review of what is available in the written publications + the histogram you published in this blog. The numbers you mentioned here in your comments are not listed in proceedings, or am I wrong ?
ReplyDeleteTwo years ago I wrote a conference review based on the Q&A after a presentation, and this was not appreciated by the authors. So for that reason I do base my reviews on the published data.
Thanks Albert, I understand and you are correct. The high CG and photon-counting jot results were briefly presented as breaking news and as such, did not appear in print.
Delete"With a noise floor down to 0.25 electrons, a “standard” CIS could be applied in single electron detection"
ReplyDeleteI think somehow even this definition is not correct. Just as QE is a quantum mechnical phenomenon, so too is electron "detection". Perhaps CIS pixel with single electron full-well capacity might be better?
I think about 0.15e- rms read noise is required to get reliable single electron detection. But at 0.25e- rms The detection error rate (BER) was calculated and shown in this paper:
DeleteE.R. Fossum, Modeling the performance of single-bit and multi-bit quanta image sensors, IEEE J. Electron Devices Society, vol.1(9) pp. 166-174 September 2013
At 0.15e- rms the BER is about 1/2500, and at 0.25e- rms it grows to 1/50 and at 0.3e- rms it is 1/20. So at some read noise level, electron counting becomes good enough to be useful for many applications.
Single electron full well capacity is easy. Just have a 1-bit ADC where the threshold between 0 and 1 is half an electron. But then this would be the single-bit QIS!