Vision Systems Design: Proceeedings of the IEEE publishes an open-access invited paper "
Bioinspired Polarization Imaging Sensors: From Circuits and Optics to Signal Processing Algorithms and Biomedical Applications" by Timothy York, Samuel B. Powell, Shengkui Gao, Lindsey Kahan, Tauseef Charanya, Debajit Saha, Nicholas W. Roberts, Thomas W. Cronin, Justin Marshall, Samuel Achilefu, Spencer P. Lake, Baranidharan Raman, and Viktor Gruev.
The paper proposes to use polarization imaging to detect cancer cells. For unknown reasons, the pixel is current-mode based:
![](http://1.bp.blogspot.com/-vC4X3_q-V-I/VC_AhHizBFI/AAAAAAAAKYg/XPlvyUQjG_A/s1600/Current-mode%2Bpixel.GIF) |
Current-mode pixel schematic and peripheral readout circuitry of the imaging sensor. The pixel’s readout transistor operates in the linear mode, allowing for high linearity between incident photons on the photodiode and output current from the pixel. |
![](http://2.bp.blogspot.com/-O_7_poJSjvo/VC_Bt86UyZI/AAAAAAAAKYs/CCGIVsFPjFI/s1600/Current-mode%2Bpixel-2.GIF) |
Cross section of the pinned photodiode together with the reset,
transfer, readout, and select transistors. The diode is an n-type diode
on a p-substrate with an insulating barrier between. The readout
transistor operates as a transconductor, providing a linear
relationship between accumulated photo charges and an
output current. |
The 7.4um-sized pixel in 180nm process shows excellent PRNU of less than 1% rms, in spite of Gm variations between the pixels:
![](http://2.bp.blogspot.com/-MdcJOd_5Tv8/VC_DrwNG7yI/AAAAAAAAKY4/nEdH43zhlT0/s1600/Current-mode%2Bpixel%2BPRNU.GIF) |
Histogram of all responses of pixels in the imaging array to a uniform
illumination at room light intensity. The fixed pattern noise of the
current-mode imaging sensor |
The current mode readou by using linear mode MOS should be firstly used by CalTech Mead team. I've experienced this also in an early circuit. In order to get good linearity, the drain voltage should be kept very small. In column parallel mode, the error on the clamping voltage will be transformed into gain offset. What is the fundamental advantage compared to classic voltage mode readout then V-I conversion please ? Tobi, you are expert on this, could you please say more ?
ReplyDeleteFor current mode, Lisa McIlrath reported on this in 1997: McIlrath, L.G.; Clark, V.S.; Duane, P.K.; McGrath, R.D.; Waskurak, W.D., "Design and analysis of a 512×768 current-mediated active pixel array image sensor," Electron Devices, IEEE Transactions on , vol.44, no.10, pp.1706,1715, Oct 1997 as did Junichi Nakamura in the same special issue on image sensors: Nakamura, J.; Pain, B.; Nomoto, T.; Nakamura, T.; Fossum, E.R., "On-focal-plane signal processing for current-mode active pixel sensors," Electron Devices, IEEE Transactions on , vol.44, no.10, pp.1747,1758, Oct 1997. While quite good results were obtained, esp. with the more complex circuits of Nakamura et al., current-mode is really not used often. I think the polarization sensors that Gruev's team have been developing are quite interesting and have come a long way in performance.
DeleteCalTech work was around 1985...1990
ReplyDeleteyes, but I think the Caltech retina work was quite different, focused on biomemetic circuits esp. logarithmic, adaptive and motion detection, usually with a BJT as the phototransducer. I think the work I cited above is much closer to what Gruev's group is using for readout. I have to say I am not 100% knowledgeable about all the things Mead's group looked at.
DeleteYes. In Mead team's circuits, all the pixels share a single transimpedance amplifier made by an OP-AMP outside (TL074). I've used such circuit too. The key advantage is that the bus voltage is clamped and you can get higher speed. It works pretty well till several Mhz.
DeleteIn this design also, the key advantage for me is that in current-mode, we can vehicle 2 signals at the same time: voltage and current, since an ideal current source value is not influenced by the voltage on the bus.
All these are the good things so long time ago !
-yang ni
One amplifier for all pixels ??
ReplyDeleteIn this case, it is possible.
PRNU is slightly less than 1% actually, ex. 93/9504*100%=0.98%.
ReplyDelete