Wednesday, February 03, 2016

Panasonic APD-CMOS Sensor Sees Colors under 0.01 lux Illumination

BusinessWire: Panasonic developed APD CMOS sensor with avalanche photodiodes (APDs) in each pixel. The in-pixel APDs multiply the photoelectrons by 10,000 times. This is said to allow getting sharp color images even under starlight (illuminance of 0.01 lux) or in other similarly dark places. Variable sensitivity technology in which multiplication is controlled in a thousandth of a second by controlling the voltage applied to the APDs, realizing 30-fps video imaging that can follow the changes in illuminance.

The new sensor has been presented on the 2016 ISSCC on February 1, 2016.

This development brings the following advantages:
  • Highly sensitive color imaging:
    40 million/lux・sec・μm2 (10,000 times sensitive in comparison to conventional devices)
  • Wide dynamic range:
    100 dB (+ 40 dB in comparison to conventional CMOS image sensors)


  1. 0.01Lux starlight! This is ridiculous!!!

  2. More details at the Panasonic PR

  3. Wikipedia lists starlight at 0.0001 Lx, which while not the most reliable source suggests perhaps a typo? [ ]

    Lux units are normalized by the response of the human eye, but of course the sensor isn't the human eye but silicon. Wouldn't quoting light levels in radiometric units be preferable?

    1. people describing sensor should get rid of lux units asap as they only bring confusion and misunderstandings. A power density in W/m^2 would be much better.

    2. at which wavelength? ;-)

    3. Illumination with these kinds if image test setups tends to be white light. Often halogen video continuous lights etc.

      Lux is useful to describe to a human how much light we are talking about, starlight, direct midday sunshine etc, but being weighted towards green its not what the silicon is actually receiving.

      Having said all that, its good to see electron multiplication. Yes it has an excess noise factor but if it achieves gain we wouldn't be able to achieve with the NF of conventional analogue gain circuitry then free game.

      It also avoids the radically different readout structures needed for SPAD array readout.


All comments are moderated to avoid spam and personal attacks.