Friday, December 27, 2013

Experiment: Pinned PD Paper Peer Review

Eric Fossum and Donald Hondongwa submitted Pinned Photodiode review paper for publication in open-access IEEE Journal of Electron Devices Society, J-EDS.

The experiment is to publish a draft version of the review paper and solicit input – especially on things that ought to be referenced or discussed – in order to create the best possible review paper for our community. A list of references is quite extensive and currently counts 102 items. Here is the link to the draft version:

"A Review of the Pinned Photodiode for CCD and CMOS Image Sensors"

Abstract: The pinned photodiode is the primary photodetector structure used in most CCD and CMOS image sensors. This paper reviews the development, physics, and technology of the pinned photodiode.

30 comments:

  1. Just in case it wasn't clear, please post any comments here, or email me privately. While I hope I did not miss anything too important in the paper, I would really welcome your suggestions. I received some great input from the people already listed in the acknowledgments.

    By the way, Donald did the sims and tracked down a few references and I wrote the paper.

    Philosophically, I like the idea of having a paper open for comments for a while before cementing a final version into the archival literature. It was something suggested by Johannes Solhusvik when we were contemplating the open access journal on image sensor technology (JIST?) for IISS. It is pretty unusual so thanks in advance for participating in this experiment.

    ReplyDelete
  2. A good and interesting paper, thanks for sharing Eric.

    Some comments:

    In the historical development section I could find no reference or mention of early CMOS passive pixel sensors of the type made by VLSI Vision.

    In Fig.3 the p-substrate is not indicated.

    In Fig. 4 the X and Y axis dimensions are not labelled. I presume they are um. For those not familiar with TCAD, it may be helpful to add in the legend that green is oxide, and that hot colors are n-type and cold colors p-type doping. Maybe also consider overlaying text on the relevant PD areas such as TG, STI, SW, FD etc.

    Maybe Fig. 10 & 11 should indicate the sensor manufacturer illustrated in the image from chipworks.

    In section III.F you state that "Thus, blooming in the PPD in a CMOS image sensor is not as consequential as it is in a CC ILT pixel where the excess carriers bloom into the readout CCD and contaminate other pixels in the image."

    I agree that blooming in a CCD is very bad, but blooming in a CIS also contaminates other pixels in the image but just neighbouring ones rather than a whole row/column. Blooming is also bad in pinned PD CIS's and I don't think this sentence quite gets that across.

    ReplyDelete
    Replies
    1. thanks for the figure improvement suggestions and I agree about needing to modify the blooming comments.

      Delete
    2. Chapiter III Section A, "nm" should Angstrom, no ?

      Yang Ni

      Delete
    3. yes, there is an extra zero on the pinning layer thickness range. thx!

      Delete
  3. Eric, from my point of view, the Virtual Phase and OPP should be considered as the origin of today's modern CMOS charge transfer based pixel design. Because both of them contain the essential characteristics, knowing, 1) surfacing potential pinning and 2) complete charge transfer (fully depleted junction).

    -Yang Ni

    ReplyDelete
    Replies
    1. I generally agree which is why Hynecek's VP-CCD and Hagiwara's prior work are discussed.

      Delete
  4. This paper is a good read, but I do not understand what it presents beyond the state of the art. All of these are well known. This looks like a review paper and might attract some citations, but I would not have accepted if I were a reviewer..

    ReplyDelete
    Replies
    1. I suppose you are asking what the benefit is of a review paper. A review paper is not supposed to present anything new. It is for the convenience of the community, and in particular, for new members of the community. Perhaps such a paper would have been useful to you when you were a newbie.

      Delete
  5. In the last sentence of section IV, and the captions for Fig.9 and 10, should it be 1.375T (1+3/8) instead of 1.35T?

    ReplyDelete
  6. Eric,

    I apologize in advance if this comment is stupid but I have the different sort of suggestion. I'm an engineer but I am not a physicist or a sensor designer. I do like to understand the mechanisms that make sensors work well because I still sell sensors.

    I think the audience of people like me, including most camera designers I have ever met, would appreciate a little more of the how pinning works, why it works and why each of the structures discussed results in pinning. This means that when an invention is cited that fixes a problem, it would be interesting to know why the problem occurred in the first place and how the invention cured it.

    I know that sensor designers can visualize all of this from their experience and fundamental knowledge but I can't and I suspect that a lot of people interested in this history can't either. Maybe this just isn't the place to publish this expanded content and I should just take a course or reread a couple of books and merge the content but please consider adding some educational materials if you can.

    Regards, Dave

    ReplyDelete
    Replies
    1. Thanks for your comments Dave. You are right that a tutorial paper is different from a review paper, and the material that you are reasonably suggesting would not fit due to page count limits. Then, there is is the question of what level to start at with a tutorial. It is probably important to understand MOS device physics before understanding how valence band pinning works, but at least you can find that in normal semiconductor device textbooks (e.g., Streetman). The whole idea of how a pinned photodiode works is not actually easy to understand. I remember trying to get the team at Micron to make a pinned photodiode and it took a long time (>6 months) for them to understand what they were supposed to be making and what was important in the device. And that was a bunch of fairly smart people with me traveling to Boise weekly to re-point them in the right direction..
      So, I agree that a tutorial paper or monograph needs to be written. One just needs time, motivation, and a starting point - all difficult to figure out. I will mull it over but perhaps this discussion will inspire someone else to write something.

      Delete
    2. Dave,

      If you want to kow some fundaments about Pinned photodiode and CIS, there might be some useful books. For example, Image sensors ans signal process for digital still camera. I am sure you can find the the mechanisms that make sensors work well. Hope it is useful.

      Regards,
      Jun

      Delete
  7. Thank you for your comments. I have replaced the posted version with one containing revisions based on public and private suggestions. Note this paper is still unreviewed by IEEE and may or may not be published in IEEE JEDS.

    ReplyDelete
    Replies
    1. Dear Eric,

      I feel very confused that an important feature of the FSI PPD is that blue light is not blocked by polysilicon-gate layers above it, nor is it substantially absorbed in the pinning layer. For the FSI, the light enter from the front side, this means it must pass the gate to PD. Is my understanding right? Please let me konw if I have misunderstanding. Thank you very much

      Delete
    2. You need clearly to read carefully this review paper, "3 times", prescribed Dr. Fossum.

      Delete
    3. I agree with your opion,carefully reading is necessary. Would you just brief explain it how to understand it?

      Delete
    4. Dear Eric, I still very confused about that in Page 4-- "an important feature of the FSI PPD is that blue light is not blocked by polysilicon-gate layers above it, nor is it substantially absorbed in the pinning layer". One of the advantage of the BSI is the poly-silicon gate is absent and Qe is enhanced, especially the blue light.

      For the FSI, the light path should be Micro lens--coloe filter--gate material-p+layer.

      BSI, Micro lens--coloe filter-silicon surface. right.

      Is my understanding right?

      Delete
    5. Yes, you are correct but perhaps slightly confused. Prior to the PPD, the main way to get light into the image sensor was either through a polysilicon CCD gate, or using a pn junction. The CCD poly gate allows complete charge transfer, no lag, no noise, but suffers blue light absorption in the poly gate. The pn junction had some blue absorption in the heavily doped surface, and suffers lag and noise. So the advantage of the PPD (and VP CCD) is that the blue light is not blocked by poly gates and the thin heavily doped region does not absorb too much blue light, but complete charge transfer, no lag, and no noise becomes possible.

      So, the advantage is relative to CCD-type poly gate. Does that help?

      Delete
  8. Eric, do you plan on providing open access to your corrected/final version? If not, I suggest that you please consider doing so. Thanks

    ReplyDelete
    Replies
    1. IEEE JEDS is an open access journal so if accepted, then yes. If not accepted, at a minimum it will be on my website. So, also open access. I believe scientific papers should be open access, that is, freely available to all. It is why IISS papers are completely open access.

      Delete
    2. Thanks. I agree with you that scientific papers should all be open access. If only you could get APS, IEEE, Elseveier to buy this philosophy ...

      Delete
  9. Dear Eric, thank you very much for your reply!

    ReplyDelete
  10. Dear Mr.Fossum,

    You said that increasing the lenght of the TG would improve the barrier control (totally agree), but would not it also increase the dark current and slow the charge transfer (and also increase the possibility of charge trapping)?

    ReplyDelete
    Replies
    1. Yes. But usually this is a slight increase so the relative increase in dark current and potential interface trapping is small. Like everything in pixel and readout design, there are tradeoffs, starting with pixel size!

      Delete
  11. Thank you all for your comments. I have taken the draft off the web for now. If it is accepted for publication in IEEE JEDS and published, I will let you know. And if it isn't, I will also let you know and probably just publish it on my own website. Either way, in a few months it will be available. If you need a copy of the draft, email me at Dartmouth. I think the experimental "open review" was useful and will pass that information on to IEEE and IISS.

    ReplyDelete
  12. Dear Eric,

    Some pixel designs use partial PPD. If you could add a small section about the partial PPD, I think it would be more complete.

    ReplyDelete
    Replies
    1. A mention and reference to the partially pinned photodiode was already included in the latest revision thanks to a prior suggestion. This is just a form of the buried photodiode introduced by Noble in 1968 for 3T APS but improved and patented by Kodak.

      Delete
  13. The "Review of the Pinned Photodiode" paper has been reviewed by IEEE J. Electron Devices Society (JEDS), improved, and accepted for publication in a future open-access issue. A preprint (prior to IEEE typesetting) is available if needed at:
    http://ericfossum.com/Publications/Papers/2014%20IEEE%20JEDS%20Pinned%20Photodiode%20Preprint.pdf
    We hope this becomes a useful reference for newcomers and oldtimers alike in our community.

    ReplyDelete

All comments are moderated to avoid spam and personal attacks.