Tuesday, May 17, 2011

Sony Security Sensors to Shift from CCD to CMOS

Sony released a CCD to CMOS sensor comparison document targeted to its security and surveillance customers. Some comparison figures from the document:



Talking about the future trends Sony says that as acceptance of HD video in security continues and CMOS video quality improves, the market is shifting toward CMOS sensors. It's nice to see this statement coming from the world's security CCD king.

15 comments:

  1. So in what markets/applications are CCDs still preferred over CMOS? Are there any that CCD has a long-term advantage? I.e. won't be displaced at some foreseeable point by CMOS?

    ReplyDelete
  2. What about atronomy?

    ReplyDelete
  3. Sony has been selling the IMX035 to security customers for some time now. I believe I was aware of this 18 months or more ago, when I was at my last company and also selling CMOS sensors to security customers.

    ReplyDelete
  4. I know of no competitive CMOS to replace CCDs with 24um pixels for taking a 5 minute exposure of extremely faint bio-luminescent samples (e.g. in vivo imaging). Admittedly a very specialized ,niche market, I doubt too many CMOS manufacturers would be interested.

    ReplyDelete
  5. The fundamental advantage of CCD is that the CCD cell has only two discret states (charge storage or no storage). If the noise on the control signals doesn't exceed a limit, this noise will not go inside the video signal. So there is a big noise immunity on CCD sensor.

    But for CMOS, this is not the case. All the noise in the control signals, on the power supplies will all go inside the video signal.

    From this point of view, CCD is always better than CMOS in noise performance when the image frame rate is not too high.

    The fundamental advantage of CMOS is that the image line can be loaded in parallel into line buffer. The equivalent bandwidth can be limited which reduces the noise.

    For classic video surveillance resolution and TV frame rate, CCD is still much better than CMOS.

    -yang ni

    ReplyDelete
  6. PPD in CMOS has also only two discret states, charge storage or no storage. The noise of control signals does not affect the charge storage state, so that CMOS should be comparable with CCD in terms of noise. Do you agree?

    ReplyDelete
  7. There is only a source follower and floating diffusion for readout. The CMOS readout chain is much more complexe and has multiple sample&hold which will get the noise from control and power supplies. Inside a 4T pixel, yes, you are right.

    -yang ni

    ReplyDelete
  8. By the way, PPD has only two states only when it can be fully depleted.
    -yang ni

    ReplyDelete
  9. That is actually the idea of PPD to be fully depleted, isn't it?

    4T pixel is standard today. So, the noise performance approaches the one of CCD. The complex readout structures mainly eliminate non-uniformities over the whole pixel array, but, if well designed, they should not significantly contribute to the noise. Particularly they should not add more noise than the readout chain in a CCD imager.

    ReplyDelete
  10. The biggest issue with PPD is image lag because the pinning potential (as the name suggests) cannot be independently controlled, affecting charge transfer. This is true also for the presence of charge pockets in the PPD. In a CCD, gate control is possible even with a buried CCD (albeit with the use of higher voltages). The left over charges in the PPD contribute to "noise"..

    ReplyDelete
  11. rolling shutter is a problem for app domain (eg computer vision..)

    ReplyDelete
  12. Frankly for 'consumer' astronomy THE chip to go for is the KAF8300. 0.01e/s dark current @-20C 7e noise, good QE into the NIR lots of pixels and complete cameras for <£2k

    Sony will have to up their game if they are to compete.

    The only alternative to CCD really is SCMOS, but that's still expensive.

    ReplyDelete
  13. Is there an Exmor R version of the sensor?

    Theoretically, it should provide better sensitivity given that it is a BSI sensor.

    ReplyDelete

All comments are moderated to avoid spam.