Tuesday, July 11, 2017

HDR Pixel Thesis

Glasgow University, UK, publishes a PhD thesis "High dynamic range image sensor using tone mapping operation" by Waqas Mughal. Here is the HDR pixel principle:

"WDR capture can be performed by introducing a monotonically increasing reference signal Vref. It is possible to capture high intensity information by comparing the integrated voltage at node N to a reference voltage Vref.

The pixel output follows a known reference signal, which is sampled and held at a value when the photo-generated signal on the diode becomes lower than the reference voltage. The potential at which these two signals are equal is recorded and is used as the pixel’s response. In the pixel, M1, M2, M3 and M4 are reset device, reference voltage switch, source follower and row select switch.
"


The Vref sweep function can be used for the tone mapping, once the pixel FPN issues are solved:

7 comments:

  1. Interesting.
    I'd like to play with such a sensor.

    ReplyDelete
  2. You want to play with the huge FPN :)-

    ReplyDelete
  3. FPN is not the most difficult thing to compensate for. Remember pure log sensors.

    ReplyDelete
    Replies
    1. Albert Theuwissen - Harvest ImagingJuly 12, 2017 at 1:00 PM

      Arnaud you are right, but every compensation will cost you time, power, memory, processing and last-but-not-least signal-to-noise ratio ! To keep the FPN low, the best thing to do is to avoid it.

      Delete
    2. I think log pixels example is the perfect illustration of how FPN is not a palatable tradeoff. It is difficult and costly to compensate, especially if it is prone to temperature dependency. Not impossible, but apparently not a lot of users willing to deal with it.

      Delete
  4. Albert Theuwissen - Harvest ImagingJuly 12, 2017 at 11:05 AM

    Did David Stoppa something similar in his "Time-Stamping CMOS Imager" ?

    ReplyDelete
  5. We can reduce the FPN in our Log sensors to a level lower than RMS noise.

    -yang ni

    ReplyDelete

All comments are moderated to avoid spam and personal attacks.