Monday, February 20, 2012

Samsung Presents BSI ToF Sensor

The updated Samsung Imaging web site lists BSI ToF sensor, the S5K32A. The previously shown ToF sensor had different part number S5K32D, and does not mention BSI technology. I'm not sure why one needs BSI for an IR sensor having 14um pixel size, but probably there is a reason. The QVGA 1/3-inch ToF sensor is said to have embedded depth output processing.


  1. I think that BSI can increase IR response and decrease visible response if the silicon thickness is adapted.

    -yang ni

  2. How can one decrease visible response in BSI? Very thick active silicon > 200 um? (average diffusion length = 100-200 um)

  3. BSI works better than FSI but I am not at liberty to explain why, except it is the same sort of reason as for visible image sensors. Within Samsung, BSI is a readily available process.

  4. Isn't it a mistake?
    BSI isn't suitable for global shutter.

  5. Customised anti-reflective coatings are a big advantage of bsi.

  6. BSI tends to have lower response in the blue spectrum because the "sweet" region of the photodiode is farther away from the back side surface, and silicon just sucks up blue light- so blue is absorbed at very shallow depths. Longer wavelengths do just fine. With BSI, you have much more room to play with AR coatings, and you don't have to worry about film variations due to structure on the wafer. To my knowledge, small pixel BSI is not (yet) compatible with global shutter, but if you have a 14u to play with, who knows. 14u is like having room for an aircraft carrier or two.

  7. Well, 14 um is not so big when you consider the scarcity of photons. With less than a one watt light bulb, TOF tries to see things that 5-7 m away. Try that at home with a 10 W night light as your flash light and try taking 30 fps video from objects 15-20 feet away.

    1. From that perspective, agreed! I was only thinking about what we have to work with. Interesting to think that one of the rages right now is 0.9um pitches in the visible...
      Side question- how fast does TOF need to capture (A nice Fermi-type question)? Commercial film is 24 fps. I was trying to come up with a good estimate of the minimum capture needed to do a good job, including system processing requirements to translate motion into an "action". The human body apparently has about 1/30 second tremor, just from being alive, but I don't think this is a good estimate figure. The fastest human punch was 78mph (~3.5e3 cm/s), and if we assume we have to capture 1cm of motion between frames, I get ~3e-4 seconds to capture this. Certainly, this is off... -Technophile

  8. One more reasen I suspect BSI is a mistake.
    Infrared light will diffusely reflect through several um substrate for BSI.


All comments are moderated to avoid spam.