Friday, December 31, 2021

2021 in Review

Here is my list of the most significant achievements in image sensor industry this year:

19 comments:

  1. Gigajot quanta probably important as spad progress

    ReplyDelete
  2. Thanks for sharing your views, and Happy New Year! What do you think of important industry achievements in HDR sensors in the past year?

    ReplyDelete
    Replies
    1. HDR has been evolving in the same directions as in the previous years. I have not seen any new revolutionary approach popping up in 2021.

      Delete
    2. Thanks for the reply!

      Delete
  3. Regarding RTS/RTN, I thought the dominant RTN observed in CIS dark readout noises originated from the source follower, not the DC-RTS.

    ReplyDelete
    Replies
    1. Right. Hope it will be expanded to the source follower RTN too.

      Delete
    2. I agree with anonymous -- RTN is a behavior, not a mechanism so the dark current RTN can have the metastable mechanism that you mentions, but that does not mean that SF RTN needs to be explained in the same way (and is likely not).

      Delete
    3. Yes, I fully agree that RTN is probably the dominant RTS phenomenon in mass market CIS. However, DC-RTS (or more generally Leakage Current RTS) coming from the bulk or the oxide interface can be the dominant RTS mechanism in some imaging applications (e.g. space applications, IR detectors, SPADs, some large pitch pixels, some global shutter designs...) and can be an issue in other nodes than the photodiode if significant storage times are used on any floating node in the IC (or if a significant electric field is present in the junction). It is for example a significant issue in DRAMs.
      Sense node junction leakage RTS can also be an issue in CIS sometimes.

      Regarding RTN, I also agree that the simple ideal interface state that traps and emits a channel carrier is sufficient to explain most of RTN behavior. Nevertheless there are also more exotic RTN traces with non stationary behavior or with more than 2 levels (before CDS) that probably include a structural reconfiguration of the trapping center. Such type of RTN could benefit from the atomic scale simulation analysis presented in this work to clarify the reason behind these exotic RTN that disagree with the single point interface defect model.

      Delete
  4. What about a "prediction contest" for 2022? What are your guesses for what will be "the highlights of 2022"? Then on 1.1.2023 we meet again and check what became reality ;-)

    ReplyDelete
    Replies
    1. Yes, it might be interesting. My predictions for 2022:

      1. Few companies would come up with EUV sensors. As EUV lithography becomes a mainstream, all kinds of inspection equipment needs EUV or UV sensors. Sony has already announced UV in 2021. I'd expect that EUV versions will follow from Sony and other companies.

      2. Smaller pixels and more megapixels for mobile sensors. Probably, 0.5um pixels appear this year. Samsung 0.56um presentation is already in ISSCC agenda

      3. China:
      - Smartphone companies come up with more experiments substituting RGB filter for RG-W or similar.
      - More iToF and dToF sensors for various applications.

      4. Each LiDAR company reinvents the system performance benchmarks, so that its specific technology shines brighter than others in its own benchmark.

      5. "Me too" progress:
      - Sony adopts TSMC 16nm after Samsung announced 17nm stacked process last year
      - Omnivision presents its SPAD products rivaling Sony in PDP

      6. Some potentially new ideas that might appear in academic publications:
      - Event-driven SWIR sensors that might alleviate the problems of high dark current
      - Microbolometric event-driven sensors
      - Various attempts to reduce power consumption in SPAD array sensors

      Delete
    2. Thanks Vladimir. One guess I would like to add is: we might see a high(er) volume SWIR sensor somewhere by end of 22. There were a lot of announcements and papers about quantum dot based sensors recently. I'm still curious what type of application will get adressed by this sensors.
      I ask myself the same question for the event sensors. What will be the high volume application that gets adressed by event sensors?
      Not sure about EUV. The litho machines are miracles, it would be very interesting to learn what sensors are used there and the problems they have to solve to reach this incredible accuracy (low single digit nm overlay accuracy, right?). I guess the currently available UV image sensors that work down to about 200nm are still for 'lens based optics' (the change is a different cover glass and microlens material transparent in UV) while EUV could also work with a Xray camera approach? Would a scintillation layer on top of a classic CIS work at 13.5nm? A interesting thought that there might be an extension of image sensors into this spectral range.
      And thanks once again for the work you put into your blog, it is very helpful that so much high quality information gets collected in one place.

      Delete
    3. "Probably, 0.5um pixels appear this year."
      Well, in your other post on 01/04/22, VPS just rolled out 0.5um pixels ...

      Delete
  5. @ I ask myself the same question for the event sensors. What will be the high volume application that gets adressed by event sensors?

    In theory, event pixels could reduce the price of microbolometric imagers. Currently, microbolemetric sensors need per-pixel calibration over the temperature range that makes them quite expensive in production. Having an event-driven pixel might eliminate this costly stage and dramatically reduce the price of thermal cameras.

    @ while EUV could also work with a Xray camera approach? Would a scintillation layer on top of a classic CIS work at 13.5nm?

    I think that EUV sensors for mask and process inspection should have very high resolution and small pixel size, just because of huge amount of the details and high throughput requirements. I'm not sure that scintillation layer can work in case of small pixels.

    ReplyDelete
  6. How do you expect event sensors to solve bolometric calibration?

    ReplyDelete
    Replies
    1. In the event-driven version, the bolometric sensors would only react on the temperature change, rather than the absolute temperature. It might be good enough for a lot of applications, but not for all, of course.

      Delete
    2. If one only needs to consider relative changes, I wonder why one couldn't already do that with bolometers. You refer to calibration of "normal" bolometers. If you think of offset/gain/linearity - the offset wouldn't matter in either case and gain/linearity would still be present in both. So I still don't see how EVS would help.

      Delete
    3. Staying with frame-based paradigm to trace the changes is possible. But how many frames back do you trace? Is this history depend on local change rate? Or just dull frame-to-frame subtraction?

      Delete
  7. My list would include canidates from ToF, event-driven and SWIR imaging. I note that there are a growing number of papers presented in each and that SONY now lists products in each. Therefore 2021 seems to be a time for each becoming mainstream.

    ReplyDelete
  8. one area we missed in our 'prediction contest' was mergers, takeovers etc. Intel buying Tower (http://image-sensors-world.blogspot.com/2022/02/intel-gets-into-cis-foundry-business.html) is definitely a big surprise. Anyone has a guess what else could happen in this area in the next year?

    ReplyDelete

All comments are moderated to avoid spam and personal attacks.