Saturday, January 28, 2006

Eric Fossum on Image Sensor History

Eric Fossum posted on dpreview.com public forums a few interesting bits on image sensors history:

1st post about CCD vs. CMOS:
"I invented the CMOS technology so of course I feel pretty strongly that CMOS is better. We have had a lot of discussion about this in the imaging community but I think I have prevailed. I sold my company, Photobit, to Micron. Micron is now the biggest sensor manufacturer in the world, at least the past few quarters, and of course uses my CMOS technology. Sony just invested the better part of a billion dollars in a plant to make CMOS sensors. Truthfully, I am surprised by how fast CMOS sensor technology has entered into the dSLR market, but that is mostly due to some excellent CMOS sensors built by Canon (and now Sony). Most camera phones also use my technology and the sensor quality there is excellent. Too bad the rest of the camera does not measure up but that is just a matter of economics and not fundamental.
I started in CCDs and love that technology. My job at JPL was to come up with a better technology and I think we did, with the CMOS active pixel sensor. It has just taken 10 yrs or so for it to catch on. My only disappointment is that NASA is so so so slow to adopt new technology, even technology it spawned."

2nd post, CCD vs CMOS:
"The CMOS active pixel sensor was designed to take advantage of all the things we knew worked well in a CCD (e.g. pinned photodiode) and yet avoid the problems of a CCD which center around charge transfer. By putting the amplifier in each pixel ("active pixel") the signal could be read out of the sensor across wires instead of by using charge transfer. This saves power (important in cell phones and interplanetary spacecraft) and allows for fast readout (most high speed cameras use CMOS active pixel sensors now), reduced sensitivity to radiation (spacecraft again), and most importantly, manufacturability on process fabrication lines that are like mainstream microelectronics vs. the specialized line for CCDs. This makes them affordable. It also brought image sensor technology back to the USA, for those of us that care about that."

3rd post, Who invented pinned photodiode in CMOS sensor:
"I was there. Tom [Lee] and I were at the whiteboard and he and I invented the CMOS APS with pinned photodiode. I am not sure Paul [Lee] was even in the room at the moment, although he was part of the project team. Don't get me into a discussion of why Caltech (JPL) was not included on that particular patent despite a co-authored NASA New Technology Report that was filed.....politics..."

4th post, conitinuation of the above:
"In my opinion, Tom was the key CCD technologist at Kodak until he retired about 10 years ago. Paul Lee and Russell Gee (from JPL) did most of the layout work - Paul on the pixel and Russell on the surrounding circuits. We reported the first CMOS APS using a pinned photodiode at the 1995 IEEE Workshop on CCDs and Image Sensors in a joint JPL/Kodak paper. Kodak did almost nothing with the technology but did some tech transfer to Motorola in a short lived 3-way collaboration with Photobit. Photobit did sensor design, Motorola did wafer fab. It is a shame Kodak did not move more quickly or efficiently. They could have been way ahead of the rest of the world. Sigh. Natural selection I guess. But, a nice bunch of guys at Kodak and a pleasure to work with."

Post on Micron's UDTV project:
"At the Japanese Aichi World Expo this past summer, NHK demonstrated their 32 megapixel Ultra-Definition Television (UDTV) system that uses 4 8Mpixel CMOS active pixel sensors with a dichroic prism color splitting optical system. They also produced a big screen display system to demonstrate the (breathtaking) video footage to large audiences. The line to get in was one of the longest at the expo. We built these sensors around CY2000 for the NHK project.
-Eric (on behalf of the Photobit/Micron engineering team) "

Post on DSLR sensor opportunities:
"Canon makes all its own DSLR sensors. Samsung can make its own sensors (they have a huge CMOS image sensor group). But I agree with your point. Nice opportunity for a pure play sensor company like Micron.
btw, don't forget IBM makes sensors for Kodak, and DALSA (formerly Philips) makes high performance CCDs. There is still places for the DSLR manufacturers to shop for sensors."

Update Jan. 30: Post on Motorola problems:
"Moto fab was problematic in my opinion for 3 main reasons. (1) Very slow turnaround times, (2) technical problems with dark current and lag and (3) Very expensive wafer pricing (not sure this ever really emerged as a practical problem since eclipsed by 1 & 2. You may remember that Fisher came to Kodak as CEO from Moto and I think that was a key reason Kodak went this route despite our best advice."

Update Feb. 6: Post on trusted people network in CMOS imaging:
"When we started transferring the CMOS active pixel sensor technology out of JPL to US industry, one of the transferees was National Semiconductor. The principals there were Kevin Brehmer and Dick Merrill. Kevin went on to start his own CMOS sensor company (I forget the name) and soon sold it to Zoran (I think). Dick went and co-founded Foveon with his multi-layer photodiode idea, using the same readout architecture we transferred to National. Getting color information from a stacked multiphotodiode structure is an older idea, but Dick was the first to incorporate it into the CMOS APS architecture. Meanwhile, Carver Mead was on one of the JPL advisory boards and became a good friend and supporter of our CMOS APS work at JPL. Furthermore, their recent VP for Biz Dev was someone I hired and worked with at Photobit. That guy lasted longer than I could at Micron, but eventually left Micron for Foveon. (and now he works for me again). So, you are not going to find me saying too many negative things about Foveon. Good group of guys and I feel I had some influence on its creation. I just hope they can overcome the sensor limitations I refer to so they can be financially successful."
The company whose name Eric forgot is Pixelcam led by Kevin Brehmer. In more recent history, Zoran closed its image sensor department as failing to win any significant marketplace.

3 comments:

  1. Kevin was Founder and CEO of PixelCam, which was sold to Zoran in 2000.

    Carver Mead was working with National at the time Gil Amelio was CEO. Gil connected Carver to Kevin to continue imaging technology development at National. Kevin was on the executive team that developed the business plan at National Semiconductor that spun out Foveon from National. Some patents invented by Kevin and Dick at National became some the early IP for Foveon. Some of Kevin's team from National were part of the original founding team at Foveon.

    ReplyDelete
    Replies
    1. It's nice to hear from you, Kevin.

      What's your explanation of why Zoran was not successful in image sensors? I've heard Zoran's explanation, but it would be nice to hear your version.

      Delete
    2. The PixelCam image sensor advantage was good image quality without column FPN, a scalable column-wise architecture with a 12bit ADC and a larger image format. We lost the combined Zoran COACH image processor plus PixelCam sensor advantage and could not keep pace with smaller pixel size sensor development.

      The sensor market preferred price over image quality which required deep pockets, access to advanced processing technology and a long-term investment. It took Sony approximately 15 years to regain their image sensor dominance with CMOS image sensors.

      The future image sensor market looks exciting and will require new partnerships, new systems, new processors and new sensors for the automotive, machine vision, robotics, drone and 3D-Video markets.

      Delete

All comments are moderated to avoid spam and personal attacks.