Saturday, June 29, 2013

Rambus Blog Post on Binary Pixels

Rambus' blog post talks about the personal aspects of Binary Pixel team. Few quotes:

"The imaging team set out to solve this problem by applying our knowledge of chip design, signal processing & managed dataflow learned from our 100+ years of combined experience designing sensors and camera systems."

"Binary Pixel technology builds upon the visionary works of imaging and signal processing experts including: The Gigavision Camera by Professor Martin Vetterli at École Polytechnique Fédérale de Lausanne (EPFL) and Professor Edoardo Charbon at Delft University of Technology & EPFL; The Digital Film Sensor by Dr. Eric Fossum, pioneer in the modern CMOS active pixel image sensor."

Rambus: "Our test-chip demonstrates a 14 times improvement
in dynamic range, enabling pixel performance in mobile phone
cameras that exceeds most DSLRs"

12 comments:

  1. Well Rambus is revoltionning all the imaging industry in 3 months by using 3 papers. I suggest that they stay humble that so many talent people have passed all of their life in more than 40 years to arrive the stage of today. Your tiny test chip is like a sand on the California beach.

    ReplyDelete
  2. And don't forget that of all the talented people, Mr. Anonymous is the most talented :) and since he couldn't do this, how could Rambus presume that they could out-innovate here...

    ReplyDelete
  3. Of course these talented engineers can not produce the faked photos that you have shown in your blog. Yes you are right, we don't know how to make such kind of "innovations" according to your definition. We have simply produced the solid-state image sensors which surpass the silver film by our hard working.

    ReplyDelete
  4. Nothing happens until you sell something.

    ReplyDelete
    Replies
    1. Amen brotha! One can always claim to have the made the "best" sensor in the world, but selling it and making money is a whole different ball game. Based on Rambus' financial performance over the course of the past 16 years, they are have not succeeded on making money yet.

      Perhaps the "14 times improvement in DR" may offer some value to the end user. Although I would think in the cell phone industry, partnerships would matter more than "I can do it better" statements.

      Delete
  5. I don't think Rambus has announced any sort of product or product plan, and have just released results of their binary pixel investigation. Personally, I find it interesting how things are progressing. I think it is far more likely than not that oversampling in time and space will lead to higher image quality. Few that really understand the whole imaging process would disagree since the evidence is just overwhelming. the Nokia Pureview camera is one approach. The Rambus binary sensor is another approach. The QIS and multibit QIS is yet another approach. Both latter approaches address the question of what to do as full well capacity shrinks (due to pixel shrink) well below the level required to achieve acceptable SNR. In this sense, let me predict that conventional approach to image sensors is another future dinosaur and all you naysayers ought to wake up and smell the coffee. I don't know which approach to oversampling will prevail down the road but I am sure some oversampled approach will become the norm in the next decade.

    ReplyDelete
    Replies
    1. I liked the point made by Paul Danini at the London conference: differentiation between players is low, and that creates a competitive opportunity. If someone can figure out how to break away from the APS crowd in a meaningful way that brings tangible benefits, it may be a lucrative commercial opportunity. I doubt we will simply coast on APS down the shrinking pixel curve. The only question is who will come out with something new and commercial first. Eric, I'd add burst fusion to your list of oversampled approaches. This is likely to come to the fore first, and some products are already taking advantage of it.

      Delete
  6. My 50 cents or 50 photons: One day there will be an ultimate imager that can count every photon from the scene - no matter how low or how high the photon rate is across any portion of the scene.
    Today’s technology is asymptotically approaching that goal and the rate of improvement is becoming slower.
    As with the previous two digital imaging breakthroughs, the inventions of CCD (Nobel Prize) and 4T (Nobel Prize2?), the third one may be coming again from the unexpected place and technology. While the light to charge to voltage conversion may remain the same, the challenge for the ultimate imager that counts every photon seems to be in the digital domain. While its sensitivity needs to be high enough to count every photon, and thus its well capacity need not to be more than one or a few electrons deep, it will need to have enough sampling (counting) speed at enough pixels (Gpix?). The challenge is how to process in real time the vast amounts of photon events in real time (3D IC?).

    ReplyDelete
    Replies
    1. uh, I presume you're just joking around with me. Maybe you ought to call it, like, a Quanta Image Sensor or QIS (quiz) for short...

      Delete
    2. 3D-stacked IC equals lower yields.And considerable amount of time developing it.

      Delete
  7. Sure, I have used the wrong acronym, QIS instead of Gpix.

    ReplyDelete
  8. The point was that the ultimate imager solution may not be an analog sensing problem - it may be a digital acquisition problem.

    ReplyDelete

All comments are moderated to avoid spam.