Tuesday, July 03, 2012

Will CCDs Survive?

DALSA blog post called "Are All CCDs Dinosaurs?" discusses the future of CCDs. The post author Nixon believes that one niche where CCDs are superior to CMOS is noiseless TDI: "so like the dinosaurs that survived the great meteor impact (which many now call “birds”), I am betting that CCD TDIs will stick around for quite a while."

14 comments:

  1. "Of course, CMOS active pixel image sensors were not easily nor well received in the CCD community. At the 1993 Workshop, Savvas Chamberlain and I had a public debate on stage on CCDs v. CMOS. I remember at the time
    just a few brave souls in the audience raised their hands when we asked who thought CMOS was an interesting avenue to take. I also remember as various CCD community leaders, one by one, conceded the advantages of using CMOS image sensors. I, on the other hand, imagined CCDs would be long gone by now but such extinction does not seem imminent, despite the volume dominance of CMOS active pixel image sensors."

    from "The 25th Anniversary of IISW: Reflections on Directions," 2011 IISW.
    http://ericfossum.com/Publications/Papers/2011%20IISW%2025th%20Anniversary.pdf

    CCD TDI is hard to beat but "noiseless" TDI will be possible with the QIS in an arbitrary track direction, at least conceptually.

    ReplyDelete
  2. CMOS APS today has nothing to do with the CMOS APS 10 years ago. All the CCD fondamental concepts have been adopted by CMOS APS now. The main challenge is the compatibility between the basic charge transfer structure with CMOS sub-micron transistors. As in all such evolutions, 1+1 is always > 2. Eric's work on the intra-pixel charge transfer is indeed visionary at that time.

    -yang ni

    ReplyDelete
  3. "CMOS APS today has nothing to do with the CMOS APS 10 years ago."
    Yang, I appreciate your kind comment but a quick look at the papers from 2001 or 2003 CCD/AIS Workshop (precursor to IISW) will show that is just a silly and completely incorrect statement right away. Not to say that there has not been great progress, but your statement is rather exaggerated I think.

    ReplyDelete
    Replies
    1. yes indeed, maybe more than 10 years ago. The time flys so fast!

      -yang ni

      Delete
  4. As long as one can make money from making CCDs than CCDs will be around even if only in niche markets. That being said there aren`t many companies left making CCDs.

    ReplyDelete
  5. CCD will be reborn as an analog memory. (^^)

    ReplyDelete
  6. In "amateur" astrphotograpy CCDs still rule. Even for planetary imaging where typical fast machine vision cameras are used Sony CCDs dominate. Some CMOS based (Aptina Micron, Sony IMX) cameras exists, but tests doesn't turn out that nicely (either rolling shutter and or high read noise or very small pixels...). IMHO at the moment only Cmosis or e2v CMOS sensors look good for fast planetary/lunar/solar imaging (+ Sony CCDs).

    ReplyDelete
    Replies
    1. Piotr, are you saying that a Canon camera attached to a telescope does not yield good images compared to a camera using a Sony CCD? I am having a little trouble understanding what the basic issue is since there is no rolling shutter issue (mechanical in case of Canon), low read noise, and large-ish pixels. I think any DSLR CMOS image sensor would work pretty well in astronomy, and even better if cooled. What is the problem?

      Delete
    2. DSLR are used - like Canon 350/450/1100 D - in the more "entry level" Deep Space imaging. The problem with those cameras is that they are not cooled and exposures time is for example 5-10 minutes. There is for example QHY8 DS camera with APS-C sized ICX413AQ or ICX453AQ. It's better than a DSLR because it's cooled (and it provides standard in astronomy 16 bit output image). I use Atik 314L+ with mono ICX285 and that when working at around -10C has minimal dark current and around 4e read noise which is very nice ;) And color cameras have some limitations in astrophotography as nebulas tend to glow in bands, not continuum.

      For planetary imaging (what I wrote above) the cameras are different - fast, 8-bit machine vision cameras. The exposures usually don't exceed 30 ms. Rolling or even more interlaced shutter aren't good there as unstable air gives unstable planetary image. Also if the camera/sensor is noisy it will give more grainy noise on the final image. So Sony CCD wins again. Some CMOS can or are used - like global shuttered MICRON in entry-level planetary cameras and guiders. e2v Sapphire CMOS turned out to be very good (but still not widely used) and Ruby version or CMOSIS big sensors await tests in a form of nice and easy to use cameras. :)


      Now if you want to make a CMOS based DS camera... like using big nice mono square CMV4000 then you put it in a cooled camera and compare with for example square KAI 4022 (Atik 4000). Will the dark frame will be relatively low in noise for a 5 - 10 min exposure? Atik 4000 prices start at 2700 EUR - and quite likely a lot of that price is the price for the KAI sensor.

      Delete
    3. Piotr, these seem to be camera issues, not sensor issues. Seems like an interesting boutique business opportunity to build amateur astronomy cameras with CMOS image sensors. Or, build sensors AND cameras.

      I still don't understand why you can't use an off-the-shelf Canon camera for planetary imaging but it looks like you are targeting the lower priced cameras. I suppose cost is a big factor for amateur astronomy.

      Again, best not to confuse image sensor issues with off-the-shelf camera issues.

      Delete
    4. DSLR aren't used for planets as planets are very small even with big scopes. 330kpix ROI video mode on such DSLR would catch the whole planet. And you need a lot of frames collected in a short time window (limited by planet rotation). In my case I use a mono camera recording uncompressed 2000 frames for every R,G,B channel and around 3000-4000 frames for one of luminance channels. Such AVIs and then stacked into images of much better quality than a single frame (+ bad frames rejected). For Saturn that's doable at 30 FPS and for brighter Jupiter - 60 FPS. It's very specific imaging on insane resolution.

      Antenna photographed with a DSLR + lens: http://rkblog.rk.edu.pl/site_media/resources/rkblog.rk.edu.pl/images/photography_day.jpg
      Antenna photographed with a DS imaging setup: http://rkblog.rk.edu.pl/site_media/resources/rkblog.rk.edu.pl/images/photography_ds.jpg
      Antenna photographed with a Planetary setup: http://rkblog.rk.edu.pl/site_media/resources/rkblog.rk.edu.pl/images/photography_planetary.jpg :)

      Delete
    5. OK, fun with optics.

      Meanwhile, if you are going to do planetary images, this is my preference. Still a CCD, so I guess you are right...

      http://photojournal.jpl.nasa.gov/jpeg/PIA15689.jpg

      Just like being there.

      Delete
  7. "IMHO" indeed!

    ReplyDelete
  8. Most of the comments appear on CMOS vs CCD but not where you need CCD TDI or CMOS TDI. Low light fast systems where number of accumulated electrons produced due to integrated light in single exposure time is too small and get buried in read noise, TDI sensor becomes a necessity. A good TDI sensor that can compete with CCD TDI is yet to be commercially available .

    ReplyDelete

All comments are moderated to avoid spam.