Monday, July 23, 2018

Sony Announces 48MP 0.8um Pixel Sensor

Sony announces 1/2-inch IMX586 stacked CMOS sensor for smartphones featuring 48MP, the industry’s highest pixel count. The new product uses the world-first pixel size of 0.8 μm.

The new sensor uses the Quad Bayer color filter array, where adjacent 2x2 pixels come in the same color. During low light shooting, the signals from the four adjacent pixels are added, raising the sensitivity to a level equivalent to that of 1.6 μm pixels (12MP).

12MP sensor (left) vs IMX586 (right)

20 comments:

  1. Looks like nice work and a significant jump ahead, again.

    ReplyDelete
  2. also this 4 macro-pixel for each RGB color should improve significantly the color accuracy since each pixel block from the same color "sees" twice less crosstalk from other different colors: can anyone comment on the improvement related to color accuracy of such 2x2 macro block of RGB pixel pattern?

    ReplyDelete
    Replies
    1. Sure, it improves the color fidelity and sensitivity with respect to a 0.8um pixel with the classical per-pixel Bayer pattern. However, if we compare it with 1.6um pixel made with the same process complexity (lightguides, DTI, same separation materials, etc.), the 1.6um pixel would have better color and sensitivity.

      Delete
    2. And that's because the virtual small green is an interpolation of 2 big greens and DTI of 0.8um pixel which makes up a quad 1.6um takes up twice as much space as DTI of a single 1.6um pixel thus reducing the PD node and sensitivity.

      Delete
    3. with good ulenses and binning sensitivity should be nearly the same as 1.6u pixel, shouldn't it?

      Delete
  3. The release indicates that the device can adjust the binning locally for enhanced intrascene dynamic range. This will enable a significant improvement in HDR imaging.

    ReplyDelete
    Replies
    1. I can't find it in the PR. Could you give a quote?

      Delete
    2. Four times greater dynamic range than previous products and real time output

      Original Sony exposure control technology and signal processing functionality are built into the image sensor, enabling real-time output and a superior dynamic range that is four times greater than conventional products. Even scenes with both bright and dark areas can be captured with minimal highlight blowout or loss of detail in shadows while viewing the image on the smartphone display.

      Delete
    3. Does it say something about adjusting the binning locally? I meant that I do not see it in the PR.

      Delete
    4. I agree with Vladimir, Sony's words just sound like they are comparing 1.6um quad to a 0.8um pixel.

      Delete
  4. That is a lot of data that needs to be read out in 33 ms and surprised that the output configuration can support it.

    ReplyDelete
    Replies
    1. "MIPI C-PHY accomplishes this by departing from a conventional differential signaling technique on two-wire lanes and introducing three-phase symbol encoding of about 2.28 bits/symbol to transmit data symbols on three-wire lanes, or “trios”, where each trio includes an embedded clock. Three trios operating at 3.5 Gsym/s achieve a peak data rate of about 24 Gbps over a nine-wire interface."
      Not clear if this is v1.0 or v1.2 from MIPI alliance website.

      IMX586 is running on C-PHY 3 lanes, 23.94Gbps so plenty fast.

      Delete
    2. Thank you for the information.

      Delete
  5. I am wondering if this sensor support 4x sensitivity at low light conditions by binning 4 pixels and it support the option to increase full well capacity(FWC) by 4 times at high light conditions. if then, it works as if 1.6umx1.6um big pixel do.

    ReplyDelete
  6. As for HDR, maybe they can control exposure time for each of the four pixels separately.

    ReplyDelete
  7. How good does this "array conversion" work (from 2x2 Bayer to classic Bayer)? I could imagine that this is quite tricky, especially in the edges of the image where there might be some color aberration from the lens. But I think there are other products with similar arrays already on the market, so there should be some experience?

    ReplyDelete
    Replies
    1. I think they use demosaicking for this conversion.

      Delete
  8. I bet they will be using some sort of spatial exposure multiplexing with neighboring same color pixels in order to bump up the dynamic range. There are a lot of patents out on that and I think Sony owns quite a few of them (both natively theirs and purchased from others).

    ReplyDelete
  9. I am not sure why my comment is not visible. I am guessing it's more of a site (blog) glitch, than anything else.

    ReplyDelete
    Replies
    1. Please re-post it. Can be some sort of glitch, indeed.

      Delete

All comments are moderated to avoid spam and personal attacks.