Monday, March 15, 2021

Omnivision's Early Years

SinaFinance publishes a biography of Datong Chen, Omnivision's Chairman, including the early ears on the company:

"In 1995, a coincidental opportunity came. A friend of Chen Datong wanted to start a company to develop image sensors using CMOS technology. He wanted to invite Chen Datong to be the co-founder. Chen Datong has never learned this technology. The other party even asked him to be the co-founder and take care of the most critical technology. This seems very incredible.

Later, Chen Datong thought that it should be because of two things:

One is that this friend had asked him for technical consultation before, and he got through and solved the problem in two or three weeks for the technology he hadn't touched, which proved his ability. The second is that he helped the other party a lot, and he didn't ask for it when they were paid.

In this way, in 1995, Omnivision Technology was established in Silicon Valley, USA, and Chen Datong became the co-founder, passively starting his first business venture in his life.

"At that time there were probably 20 or 30 companies like this in the world doing the same thing, but since they are all on this boat, do it hard."

In Omnivision's team, 80% of the engineers are Chinese, 80% of the Chinese are foreign students, and 80% of the foreign students are Tsinghua [University] graduates. The team is extremely creative and works very hard. Finally, they became the first company in the world to turn CMOS technology into a product.

This company has triggered an industrial revolution: through the commercialization of CMOS technology, the power consumption of image sensors has been reduced by several hundred times, making mobile phones usable as cameras.

In 2000, Omnivision Technology was listed on NASDAQ and successfully listed."

13 comments:

  1. The number of companies that claim to have been the first to "turn CMOS technology into a product" is quite amusing. I would argue that it was VVL that had the first commercial CMOS sensor product. Albeit a passive (3T) array one.

    ReplyDelete
    Replies
    1. VVL, or maybe IVP (Sweden), had the first commercial products, although IVP was focused on machine vision applications. Both were PPS technology. I remember that Peter Denyer, VVL founder who became a friend, felt that Omnivision just copied VVL's PPS technology. Omnivision and Photobit were founded within a few weeks of each other in the Spring of 1995. Omnivision eventually followed Photobit's CMOS APS technology and some legal action followed, resulting in a settlement in favor of Photobit. Omnivision's signal chain was analog color, like VVL, but Photobit's was digital. VVL also eventually followed the Photobit pixel approach but VVL was a very good competitor to have, based on mutual respect (company, and personal) and both spurring each other to improvements. Photobit never initiated legal action against VVL. All in all, the quotations from above are sort of a fantasy posture and rewrite of history. To suggest that Omnivision alone triggered an industrial revolution is silly. But, in all fairness, I do credit Omnivision/Howard Rhodes/TSMC with being at the forefront of BSI mass production, even if HR took a lot of Micron know-how with him to Omnivision when he was pushed out.

      Delete
    2. Was Howard really pushed out? From my point of view, Howard wanted to push the bleeding edge of BSI (and other ideas), and Micron just wouldn't support him. I saw a lot of frustration from Howard at the end of his tenure there when he wanted new equipment and Micron just insisted on giving him the DRAM fab hand-me-downs... I thought it was mostly Howard that left in frustration. But maybe I didn't see the part you saw...

      Delete
    3. I only know what I heard from people that should know, and it makes sense to me. I am also 100% sure that HR had his own narrative to spin. Let's let him rest in peace.

      Delete
  2. In Japan, Hitachi was selling passive pixel MOS senser 1960s or 1970s. Matsunaga converted Toshiba's CCD group to CIS in1994. Their first product was 1997 or 1998 with active pixel with a transfer gate and with 2 share.
    Sony started CIS 1998. First product was around 2001, with pinned PD. But it was current-output pixel, not column parallel. Their first 4tr pixel product was around 2004. They were using negative bias TX from their second product around 2002 earliest similar time with CANON, which made their image quality reputation quite high. They applied BSI patents for CIS around 2001. The first product was 2008.

    ReplyDelete
    Replies
    1. I think your dates are a little off. It would be amazing if Hitachi was selling its MOS sensors in the 60's. Toshiba's product came out around 2000 as I recall. It was a momentous occasion when they joined "the active pixel revolution" and added a lot of legitimacy to what Photobit/VVL/OVT had started. No idea about Sony's timeline. Photobit applied for a BSI patent in Jan 2000. (US 6,429,036) but the conceptual idea of BSI for CMOS was published much earlier, maybe 1994, just as sort of an obvious thing carried over from CCDs. I think Sony's biggest strength has been making other people's ideas manufacturable and put into solid products along with commensurate long-game investment.

      Delete
    2. Eric,

      I think that Toshiba's CIS was earlier than 2000. I remember that they presented 1T 5.6um active pixel sensor at ISSCC and you were the chairman of this session and you said that you just started Photobit. That should be 1996 and Toshiba showed their product at this conference.
      There was also the presentation of capacitive FP sensor which became eventually TouchID of Apple.

      Delete
    3. hi Yang Ni, we may be talking about somewhat different things, or at least my memory gets a bit fuzzy between different things. At 1997 ISSCC Toshiba presented essentially a 3T CMOS APS. I think the 4T PPD device came out a few years later. I remember HR at Micron (so that would be 2002) working hard to reverse engineer that device, and I think it was also the first product on the market with shared readout (even if invented by Kodak, I think). Looking back, it is amazing how fast the progress was for CIS in those days. https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=585324

      Delete
    4. This comment has been removed by the author.

      Delete
    5. Hi Eric,
      No, 4T came later. The pixel from Toshiba uses 1 single NMOS, the photodiode is coupled to SEL line via a capacitor. So there are 3 voltage levels: readout, integration and reset. It was the first time that 5.6um pixel was presented ... and we have got a long time silience in the conference room.
      This pixel has been presented also in IISW also:
      http://www.imagesensors.org/Past%20Workshops/1997%20Workshop/1997%20Papers/05%20Sasaki%20et%20al.pdf

      Delete
    6. hi Yang, the '97 ISSCC paper introduce 3 types of pixels and then states: "The sensor uses a transistor address cell" which is the 3T type unless I have missed something. Anyway, 3T devices were well known and demonstrated well before 1997. E.g. my plenary talk at the 1995 IEDM or the AT&T plenary talk at the 1996 ISSCC (not to mention Noble's work). I think we got to this discussion because someone claimed Toshiba did a 4T transfer gate shared readout device in 1997 which I disagree with. I think that did not happen until 2000+. Toshiba did do a 3T in 97 and I guess also did a camera product in 97 (which I assume was 3T, too). Thanks for the walk down memory lane.

      Delete
    7. The fact is: Toshiba's first CIS product was 1997 (or earlier) with active pixel with a transfer gate and with 2 share.

      Delete
  3. Toshiba's digital still camera "Allegretto" PDR-2 was sold in 1997. So their CIS production was 1997 (or earlier).

    ReplyDelete

All comments are moderated to avoid spam and personal attacks.