I do find this amusing. BTW, the comment about CMOS and TDI is incorrect. Photon-counting image sensors in CMOS (esp. the ones without avalanche) with deep sub-electron read noise (say 0.25e- rms and lower) will allow TDI in any track direction -- so better than CCD most likely.
Not true ! The important thing in TDI is the dynamic MTF which is optimal when the signal charge moves in smooth way. This can be approximatively true in CCD but not in CMOS. The signal deplacement in CMOS is a saccaded, so the MTF will be considerably degraded even the readout noise is not an issue.
@Yang, I think the "smoother" charge motion is a minor feature of TDI. The main thing is time-delay and integration. QIS, with subdiffraction limit pixels (jots) and oversampled in time probably will look smoother than CCD anyway.
@Albert, the chart says "change to CMOS unlikely" relative to TDI. Sounds quite definitive, and wrong.
"Time to use embedded CCD in CMOS ;-)" Not so easy and economical. The issue is that higher voltages (and heavy process optimizations) would still be required for fully-depleted operation/ excellent charge transfer. So after optimization, you would either end-up with a superior CCD channel with incompatibility with CMOS; or a CMOS compatible CCD structure that has mediocre detection performance.
I do find this amusing. BTW, the comment about CMOS and TDI is incorrect. Photon-counting image sensors in CMOS (esp. the ones without avalanche) with deep sub-electron read noise (say 0.25e- rms and lower) will allow TDI in any track direction -- so better than CCD most likely.
ReplyDeleteNot true ! The important thing in TDI is the dynamic MTF which is optimal when the signal charge moves in smooth way. This can be approximatively true in CCD but not in CMOS. The signal deplacement in CMOS is a saccaded, so the MTF will be considerably degraded even the readout noise is not an issue.
DeleteEric, can I buy such a device ? Not to my knowledge ... For sure in the future more and better devices will show up, but as of today ???
Delete@Yang, I think the "smoother" charge motion is a minor feature of TDI. The main thing is time-delay and integration. QIS, with subdiffraction limit pixels (jots) and oversampled in time probably will look smoother than CCD anyway.
Delete@Albert, the chart says "change to CMOS unlikely" relative to TDI. Sounds quite definitive, and wrong.
I am happy to change my mind once we get in products. BTW, the word "wrong" is also very definitive ;-)
DeleteTime to use embedded CCD in CMOS ;-)
ReplyDeletePierre Boulenc
"Time to use embedded CCD in CMOS ;-)"
DeleteNot so easy and economical. The issue is that higher voltages (and heavy process optimizations) would still be required for fully-depleted operation/ excellent charge transfer. So after optimization, you would either end-up with a superior CCD channel with incompatibility with CMOS; or a CMOS compatible CCD structure that has mediocre detection performance.
This might be of interest, CCD-CMOS TDI by Imec: http://image-sensors-world.blogspot.ca/2016/05/imec-publishes-combined-ccd-cmos-tdi.html?m=0
ReplyDelete