Saturday, September 08, 2012

Delft University HDR and GS Sensor Theses

Delft University published MS thesis by Yang Liu "The Design of a High Dynamic Range
CMOS Image Sensor in 110nm Technology
", apparently based on CMOSIS work done in cooperation with CMOSIS. The sensor is made in 0.11um process with pixel size of 4.8um. The 6T pixel has 3.1e- noise and 84.5dB DR:


Two floating diffusions FD1 and FD2 are needed to provide high and low conversion gain paths for low light and strong light signals. The maximum charge is limited by either FD2 capacity or PD well. The reported pixel full well is 51750e-.

In order to reduce the image lag, the two transfer gates are placed in the corners on the same side:


Another recently published thesis based on CMOSIS global shutter work is "The Design of a Global Shutter CMOS Image Sensor in 110nm Technology" by Xiaoliang Ge. The thesis has many interesting details on shrinking CMOSIS 8T global shutter pixel to 4.8um in 0.11um process:


Thanks to TH for the link!

10 comments:

  1. Vladimir, the work is not only based on CMOSIS technology, the work was done in cooperation with CMOSIS. The 2 students worked for almost 1 year at CMOSIS in Antwerp.

    ReplyDelete
  2. I thought the 1st one was both GS and HDR, but it is not.
    It can be driven by GS mode, but then reset noise appears.
    3.1e- noise and 84.5dB DR are for RS mode.

    1st one is HDR, 2nd one is GS, so the subject is correct.

    ReplyDelete
  3. Will there be silicon results published in the near future?

    ReplyDelete
    Replies
    1. The students finished their projects (being design and lay-out), successfully defended their thesis and got their MSc degree. What is going to happen with the designs is out of my/our control.

      Delete
  4. The noise calculation of 3.1e- of the HDR thesis seems not correct. Correct me if I'm wrong: the source follower thermal noise, given the double sampling (reset noise + signal), has to be multiplied by square root of 2...The final noise should be ~ 4.2e-

    ReplyDelete
  5. The dominate source follower noise is 1/f instead of thermal noise. The effect of double sampling on 1/f noise is a bit complex, can not simply expressed by multiply by square root of 2. Actually in the 2nd global shutter thesis, there's a very good theory analysis about this.

    ReplyDelete
    Replies
    1. 3e 1/f noise with that fast readout? Sure the calculations are correct? Anyway, many thanks for the previous answer :)!

      Delete
  6. What is the relation between CMOSIS & Delft Univ?

    ReplyDelete
    Replies
    1. None. Students from TU Delft can do their master thesis work at any company or institute. By coincidence, last year two students went to CMOSIS for their thesis work, the years before I had students at other locations, and also this year, I have a student doing his thesis work in another company. So what is the relation between CMOSIS and TU Delft ? We are colelagues in the field of solid-state imaging, nothing more, nothing less.

      Delete

All comments are moderated to avoid spam.