Friday, February 25, 2011

ISSCC Review, Part 4

Part 4 of ISSCC review by Albert Theuwissen covers three presentations:

Prof. Etoh of Kinki University "A 16Mfps 165 kpixel Backside-Illuminated CCD"

Y. Yamashita (Canon) "A 300mm Wafer Size CMOS Image Sensor with In-Pixel Voltage Gain Amplifier and Column-Level Differential Read-out Circuitry"

R. Walker, J. Richardson (University of Edinburgh) "A 128 x 96 Pixel Event-Driven Phase-Domain Delta-Sigma-Based Fully Digital 3D Camera in 0.13 mm CMOS Imaging Technology"

10 comments:

  1. the full wafer sensor seemed rather pointless. They had 160 micron pixels and used a lens with 1600mm focal length and f/11 to take video in the dark.

    they could have used smaller pixels, had a smaller sensor and used a smaller lens with a faster focal ratio and still have the same SNR in the same exposure time.

    it seemed to me to be simply bragging rights to build a 1.6mpixel sensor using a full 300mm wafer

    ReplyDelete
  2. Maybe they don't want to see in the visible spectrum ...

    ReplyDelete
  3. the sensor was made of silicon. it was optimized for visible light....maybe this is what happens when you have sensor designers that know nothing about designing imaging systems... it seems that the systems engineering was non-existent.

    ReplyDelete
  4. You underestimate too much Canon's engineers and designers ...

    ReplyDelete
  5. well I heard the paper and it had 160 x 160 micron pixels and they showed video taken using 1600mm focal length f/11 lens

    The presenter was asked a question by Bryan Ackland if they could have accomplished the same sensitivity using a pixel of smaller size and a faster focal ratio. The presenter was simply stumped by the question and said he needed time to think about it and would prefer to discuss at the author interviews.

    with a pixel of half the size per linear side (one quarter of the area per pixel) and a lens of half the focal ratio (f/5.5 instead of f/11) they would capture the same number of photons per unit of time.....

    The optics would be cheaper and they could have build four such sensors in the same area as the one they built....

    I personally think Bryan asked the classic "gotcha" question, what ISSCC is well-know for...

    I think the presenter was terribly embarrassed by the question.

    I think that the fellow that suggested it was made purely for bragging rights got it correct.

    ReplyDelete
  6. sorry i meant that with a pixel of one quarter the area and a focal ratio twice as fast they would collect the same number of electrons provided the QEs were the same

    ReplyDelete
  7. Try to keep the ratio of pixel pitch to F# constant. This is the collection ratio, or QE*p/F. The larger the number, the more photons you collect. While it seems obvious to drop the F#, there are plenty of design issues to consider. The question is actually not that easy to answer. On paper, it is simple. You could take this all the way down to F1 and the sensor could be 100 times smaller for the same sensitivity. That camera has totally different properties from an imaging point of view. This camera is the one on your cell phone. It is not about bragging rights. It is a large format and the camera has different properties. It is very difficult and expensive to build a lens with lower F# at those long focal lengths. It is not obvious that F5 at half the focal length is cheaper. Look at the cost of lenses for lithography. But cost is not the point. The idea is to build a long focal length camera. That camera can be used for things that one with half the focal length cannot be used for.

    The question that you think was so brilliant is not brilliant at all. It should be completely obvious that the format size creates a different camera. Otherwise, why didn't he just ask the question: "Why are you bothering with a wafer sized sensor? Why don't you just build sensors for camera phones like everyone else?" We can all see that this is not a good question at all. I would say the question asked was a gotcha for the person asking. I would answer back: The question you should be asking is "What are the reasons to build a large format camera?". If the presenter cannot answer that question, then may be others can answer that for him.

    ReplyDelete
  8. the japanese presenter found that the question is so naive and didon't know how to answer. By japnese style politesse, he could not say that the question is non sense!

    ReplyDelete
  9. .... or, the imaging with a lens is not the real application. And I think this is the case. They only showed images of night vision and astronomy to avoid talking about the real application. BTW, one should also understand that unexperienced presenters at ISSCC are pretty nervous if they get questions of experienced ISSCC attendees. So it does not surprise me that a foreign presenter does or can not answer a question asked in a foreign language .... imagine we have to do a presentation in Japanese and get the questions in Japanese as well.

    ReplyDelete
  10. they did mention astronomy as a potential application. Those are big pixels though and if you need any sort of fine resolution for the image scale that's going to be a pretty large telescope.

    here are some numbers:

    for a 10 meter f/7.5 (75meter focal length) you would get 0.45arc-sec/pixel. That will give you reasonable detail of a galaxy that is fairly close-in.

    for example here is an example of 0.45"/pixel on a galaxy with seeing in the 1.5" range

    http://www.narrowbandimaging.com/ngc7331_mk1sn2_ml4022_rgb_page.htm

    if they wanted to image planets, and I think I remember them saying something about that too, then 0.45"/pixel is inadequate: you'll not Saturn or Jupiter spread over enough pixels to make it interesting.

    They probably need to be at f/20 or more for that: for example you get 0.17"/pixel at 200Meters focal length.

    The way people often take planetary images is to shoot video and select the best frames and then combine them together. I don't remember their frame rate but seeing that it is only 1.6mpixels, i'd expect it to be reasonably fast (but don't have the paper or digest with me where I am at the moment).

    so for a huge telescope this could be an interesting sensor. I certainly don't think that it makes no sense, but it is obviously very specialized.

    ReplyDelete

All comments are moderated to avoid spam.