Sunday, February 14, 2010

ISSCC 2010: Sony 1.65um BSI Pixel Cross-Section

Sony seems to be targeting Chipworks' bread and butter when it published its newest 1.65um BSI pixel corss-section at its ISSCC 2010 presentation:

A 1/2.3-inch 10.3Mpixel 50frame/s Back-Illuminated CMOS Image Sensor
Hayato Wakabayashi, Keiji Yamaguchi, Masafumi Okano, Souichiro Kuramochi, Oichi Kumagai, Seijiro Sakane, Masamichi Ito1, Masahiro Hatano1, Masaru Kikuchi, Yuuki Yamagata, Takeshi Shikanai, Ken Koseki, Keiji Mabuchi, Yasushi Maruyama, Kentaro Akiyama, Eiji Miyata, Tomoyuki Honda, Masanori Ohashi, Tetsuo Nomoto

The thickness of silicon layer appears to be about 3um, which should give Sony excellent QE. Indeed, red QE approaches the green one, according to the relative graphs in the presentation. No absolute QE numbers are given. Metal aperture does not seem to have an electrical contact to the backside. The backside planarization looks is next to perfect. It seems that Sony has created the ultimate Bayer pixel - at least I do not see what else can be improved here.

Some corrections/additions to the Sony spec I posted earlier:

Process: 0.14um 1P4M
Full well: 9130e at 60C (Saturation)
Dynamic Range: 71dB at 1x gain
Sensitivity: 9890e/Lux*s, 3200K light with IR-cut filter
Supply: 2.7V/1.8V
Power consumption: 375mW, HD video at dark

The low-light luma-SNR after AWB and CCM is 14.5dB is 10 Lux, as measured at 18% gray patch from a 5500K light source through F2.8 lens.


  1. why the metal shield is needed on the silicon back side ???

  2. Metal shields between pixels reduce crosstalk. Metal shield on the periphery reduces light-generated leakage current in the sensitive analog circuits.

  3. Metal shields also reduces sensitivity, and the advantage of BSI structure.

  4. Unless you can control charge-spread, its a trade-off between QE and cross-talk

  5. Even if you can control the charge spread, QE-XTLK trade-off still does not go away.

    There appears to be a spacer between microlens and color filter layers - a clear sign that Sony tries to move light rays away from rough borders between RGB colors. These borders, other then being undefined color, scatter the light all other the array. The metal shield between pixels serves the same purpose - blocks light from the RGB borders. I think Sony engineers made a right choice to sacrifice a little bit of QE here.

  6. If you look really very carefully, you can see that there is quite a thickness variation between the two color filters shown. The layer between the color filters and the micro-lenses is a planarization layer to get rid of the surface topography introduced by the color filters. A.T.

  7. Yes, it can have a planarization function too. It looks too thick to me, if its only purpose is planarization. Look how thin is the planarization over the backside metal - a quarter micron or so. The layer above CF is about 0.5um thick. Normally the planarization is not that thick.

  8. BTW, Canon is all set to release its EOS550D. Coincidentally, they have "gapless" micro lenses in this version as compared to the older EOS500D which had gaps.
    Waiting for the full technical review of this much awaited camera!

  9. What is the exposure time used in SNR10 test ??

  10. Great.. this is what I expected. OVT BSI looks a fake when it compare with Sony..

  11. If metal shields between pixels reduce crosstalk,
    how to reduce light-generated leakage of MOS device in the pixel region


All comments are moderated to avoid spam and personal attacks.