Tuesday, February 09, 2010

Sony Presents Image Sensor Future at ISSCC 2010

EETimes reports from Tomoyuki Suzuki, Sony VP pleanary talk at ISSCC 2010 that future sensors are proposed to be bent into a spherical shape, enabling it to capture wide-angle images like an eyeball. The performance targets of such future device include 4K x 2K video capture, less than 0.1 Lux video capture and 1,000 frames per second.

Another private report from ISSCC that I received talks about very impressive spec of the new Sony 10.3MP BSI sensor featuring in a number of recently announced cameras:

Resolution: 10.3MP
Pixel size: 1.65u BSI
Sensitivity: 9890 e/Lux*s @ 3200K, 650nm IR cut-off filter, F2.8
Full Well Capacity: 9130e (linear)
Conversion Gain: 75uV/e
Random Noise: 1.7e RMS with 16x analog gain
Dynamic Range: 71dB
Dark Current: 3e/s @ 60C degrees

7 comments:

  1. Sensitivity is way too high to be true. Without any lense, only about 27,000 photons strike to each 1.65um pixel for 1 lux of D65 light during 1 second. With F2.8,IR filter and 3200K light, no way to get almost 10,000 e/lux*s unless QE>>100%. But 3e/sec DC at 60C is impressive if it is true.

    ReplyDelete
  2. I believe the sensitivity number is calculated with illumination at the sensor plane with no lens. Then all the numbers start to make sense. I'm not sure why F2.8 is mentioned here. May be the lens CRA was used when illuminating the sensor in this measurement.

    3200K light slightly improves the sensitivity number over D65 you used.

    ReplyDelete
  3. actually 3200K+IR has a photon flux of about 4800 photons/sec/lux/um^2. So a 1.65um will get about 13kphotons/sec/lux or an equiavlent qe=76%.

    ReplyDelete
    Replies
    1. Hi Peter:
      I believe peak QE may reach 76% at 550nm, but it's really amazing to have "average" QE=76%

      Delete
  4. the sensitivity is always given with faceplate illumination level ...

    ReplyDelete
  5. Peter, how did you get 4800 photons/sec/lux/um^2?

    ReplyDelete
  6. From Suzuki-san's presentation it looks like CMOS imaging power is around 0.83mW/Mpixel/fps assuming everything scales linearly from his 60fps graph.

    Is this accurate?

    If so a mobile phone or camera capturing 3D 180p at 60fps would dissipate around 400mW just for the sensors???

    Also his slide showing an 8MP sensor (4k x 2k) @ 240fps would dissipate 1.6W???

    ReplyDelete

All comments are moderated to avoid spam.