Wednesday, January 12, 2011

WIN, FENA and UCLA Present Self-Quenching Avalanche PD Imager

Another interesting presentation at Taiwan Tsing Hua University Workshop on CMOS Applications in Astronomy and Space Sciences is done by Western Institute of Nanoelectronics (WIN), Center on Functional Engineered NanoArchitectonics (FENA) and UCLA:

Towards Ultimate Single Photon Counting Imaging
Kang L. Wang, Xinyu Zheng

A self-quenching avalanche photodiode is presented:


The silicon measurement proves that the self-quenching works:

12 comments:

  1. I like this UCLA work. Good for a jot sensor I think.

    ReplyDelete
  2. No one else refers to jots. This term is meaningless.

    ReplyDelete
  3. Actually, I'm using this term too. I think Eric's jot idea is quite widely known and many (most?) people understand to what he's referring.

    ReplyDelete
  4. Is it widely known because every time we see some small pixel work, Fossum refers to a jot? The people doing the work don't refer to it this way and many (most?) are not taking anything away from the half-baked paper where that term was defined.

    ReplyDelete
  5. OK, Mr. Anonymous Troll. You must be a grouchy person. But just in case you don't know, here is a definition for you.

    A jot is a binary pixel that is logically in a 0 or 1 state, depending on the presence or absence of a received photoelectron. Its size is well below the diffraction limit and probably becomes most useful at 10^9 jots/sensor or more.

    A jot sensor can be treated as a DRAM in many ways as far as readout. Integration, spatial and temporal, is done elsewhere. Thus pixels size and frame rate is fully flexible, dynamically changeable, and can be reconfigured for the even the same image.

    In a jot sensor, there is no full well issue except in readout speed. There is no analog circuitry per se, although bit sense amplifiers are analog circuits.

    I hope to get back to this research in 2011. Mesnwhile you may now understand why I like the UCLA work and why I think it might be good for a jot-based image sensor.

    ReplyDelete
  6. All of this can be done whether or you call it a jot or a pixel. Good luck with your research in 2011.

    ReplyDelete
  7. I hope that this will not be a joke sensor :)

    ReplyDelete
  8. Actually, we already have this sensor today. Take the latest 8MP Omni BSI sensor. Stop down the lens to F4 and only look at the MSB or set a threshold for high gain. Now you have a binary pixel with size well below the diffraction limit. Check out how useful that is. Then increase the pixel count to 10^9.

    ReplyDelete
  9. so many (25?) year ago, there was an interesting paper in the magazine Lambda (VLSI design). The author used a DRAM as image sensor by writing "1" into the cells and then read the data out, the incident light makes lost some of the data and gives a binary image...

    ReplyDelete
  10. There is a lot of prior art on binary image sensors and photon counting sensors.

    In the case of the jot the goal is to detect single photoelectrons and amplify them to a logical "1" from logical "0". This is quite different from thresholding a normal pixel signal where the threshold might be hundreds of photoelectrons. Such thresholding sensors are indeed easy to make and even defective DRAMS were proposed by Micron a long time ago as a sort of cheap image sensor. The use of gigabit DRAMS was reproposed recently by Vetterli and Charbon, for example, for "gigavision". Their emphasis is on image processing though, not on sensors in that work. There is also much current interest in photon counting, such as SPADs and the device posted here (all cool stuff!). Unfortunately most SPAD pixels are too large at this time to integrate a billion or so per sensor. Even a million is a long term goal last I heard.

    By the way, the word "jot" is a variant of the Greek "iota", which means "smallest thing".

    I started thinking about jot sensors when Junichi Nakamura asked me to write a chapter about future image sensors for his now-famous Digital Camera book. The reason I started working with Samsung in 2008 was to further explore this concept. Unfortunately that project got derailed in favor of 3D image sensing. I hope to get back to it at Dartmouth before long. I'd like to know if a jot sensor is a good thing for practical digital imaging, or just an interesting academic idea.

    ReplyDelete
  11. Thanks for posting your research interest. It sounds like a good thing for you to do. But there is no need to make up a new word for pixel even if it has Greek roots. No matter how small or insensitive you make it, you still have a pixel. In order to explore the value of this concept, you can start with today's sensors that integrate charge and configure/process them as if there were the SNR equivalent of one photoelectron. There is no sharp discontinuity in what you are interested in.

    ReplyDelete
  12. There are much research work in France (IEF & ENSTA) on binary smart image sensor. But it seems the potentiality is limited to some particuler applications. There is a similar case in fluid dynamic simulation, it's easy to produce conceptually the difficult phenomena by using cellular automaton concept but this method is hard to be harnished to a particular fluid. So the "analog" Navier-Stocks equation is still widely used in the real applications.

    -Yang Ni

    ReplyDelete

All comments are moderated to avoid spam and personal attacks.