Tuesday, July 14, 2015

Sub-0.3e- Noise in Quanta Sensor Reported

A preliminary open-access copy on IEEE EDL paper "Quanta Image Sensor Jot with Sub 0.3e- r.m.s. Read Noise and Photon Counting Capability" by Jiaju Ma and Eric Fossum is available on IEEE site. The paper has the latest results of pumped gate Jot sensor with measured noise down to 0.28e- in some pixel versions:

17 comments:

  1. These are typical jots in the arrays. The best jot we found so far is 0.22e- rms noise at 423uV/e- CG. We call it the "golden jot." The CG and voltage noise both have some scatter which will be published in a more detailed characterization paper when we are done. There is a lot of testing left to do. Dark current seems to be about 10pA/cm^2 at 60C, which is pretty low, I am happy to say, but not unexpected.
    A paper on the PCH and VPM measurement technique is also in the process of being written.

    ReplyDelete
  2. Eric, the FWC mentioned in the characteristics table, is it determined by the floating diffusion charge handling capacity or the charge storage capacity of the PPD please ??

    Thanks !

    -yang ni

    ReplyDelete
    Replies
    1. FWC was measured from the "end" of the PTC, corresponding to about 90 mV (or 74 mV) of swing after the SF. The FWC is limited by the SW capacity. It could probably be increased but for QIS application we felt it was more than sufficient.

      Delete
    2. Thanks Eric !

      -yang ni

      Delete
  3. The CG (uV/e) is probably different between PCH-VPM and PTC due to uncertainty in the SF/ADC gain. Thoughts? PCH-VPM should be more accurate.

    ReplyDelete
    Replies
    1. I don't think the accuracy is related SF/ADC gain since calibration has to be done for both cases, although I note that our CG is after the in-pixel SF.
      But, CG is easy to read by reading the peak locations in PCH. And noise from VPM is independent of CG. PTC requires the best fit line through literally a lot of noise. There is always noise in the noise unless you take a lot of data, and PRNU and FD non-linearity can both reduce accuracy, esp. when you only care about the very low light portion of the PTC curve.

      Delete
  4. For PTC, by the way, I am looking for the earliest reference to that method. The best I can find is a paper by Alan Fowler in 1981 with theory and measurement. But, at the same SPIE meeting, there is a paper by Janesick, Hynecek and Blouke which also uses the method. I asked Jerry and he thinks it might have originated at TI but does not really remember. Fowler is/was an astronomer at Arizona and does not cite any prior work so he probably thought he was first.
    JPL used PTC extensively and gave it its moniker a few years later. Anyway, anyone know of earlier publications?

    ReplyDelete
    Replies
    1. Eric,

      You might want to try to track down Bob Gazerro through Gamma Scientific. Bob introduced us to the PTC method in the early 80's during standards meetings at the Automated Imaging Association. I think he was working for Kodak Megaplus in San Diego at the time where they were using the method.

      Bob may remember where he first learned of PTC.

      Delete
    2. Thanks Dave. The multi-prong approach in tracking Bob down was successful, but unfortunately he does not know of earlier references. However, he claims the recalls articles from the 1930's an 1940's that used "PTC" on single detector devices, and says he used this technique of "PTC" to get CG himself in the 1970's. I also know it from a long time ago but no idea from where.

      Delete
    3. Eric, ask Jim Janesick !

      Delete
    4. I know you are half joking but sure, I will give that a try.

      Delete
  5. Hi Eric, 403uV/e- is an amazing value.
    Would you please separate the capacitance into the diffusion, the wiring and the SF gate?
    Thanks.

    ReplyDelete
    Replies
    1. Actually it is 426uV/e-. Please refer to the March JEDS paper for the rough breakdown.

      Delete
  6. And theoretically, as CG increases, noise also increases by the feedback from the signal. Especially 1/f and RTS tend to remain. Do you have an experimental data?
    Thanks!

    ReplyDelete
    Replies
    1. You are right that there are diminishing returns, but they don't just cancel out.
      We are preparing a longer paper on characterization that will provide additional experimental data. It is almost done and will be submitted to IEEE JEDS.

      Delete
  7. PTC history update:
    1. Morley Blouke says he first learned about the technique from a visit to JPL in the 1974-1976 time frame, in a lecture by Fred Lindauer and Jim Janesick. He thinks Fred said it came from vidicon characterization.
    2. Jim Janesick says he also thinks it came from vidicon characterization, but also referenced a paper by George Root from JPL from 1976:
    http://www.imagesensors.org/Past%20Workshops/Marvin%20White%20Collection/1976%20Papers/1976%208%20Root.pdf
    3. I looked for vidicon papers but only found reference so far to a "Light Transfer Curve" which seems to mean signal v. light.
    4. I found some papers from the 50's on characterizing the human visual system and talks about measuring variance and signal and getting the human eye response value but no data or plots so far so hard to say if it is the same, plus the language they use is so far removed from the language we use it is hard to know really what they are talking about.

    So, for now, the root paper for PTC seems to be...the Root paper form 1976.

    ReplyDelete
  8. More than interesting research after the origin of the PTC ....

    ReplyDelete

All comments are moderated to avoid spam.