Sunday, July 26, 2009

Pixart Presentation

Pixart is one of the very few profitable mass market image sensor businesses these days. A relatively recent presentation gives some insight about the company structure and expenses. The last quarter update is here.


  1. Some observation I have about Pixart in the last 6-7 years.

    Pixart started off making mouse sensors, nothing fancy, or high end, but enough to sustain the company. Pixart's first serious image product was a image sensor (array only)and image signal processor IC combo. It was a very simple Image signal processor IC companion chip that mated with Mediatek's cellphone baseband which at the time did not have ISP functionality. Unfortunately for Pixart, no one wanted to mate Pixart's ISP with Pixart's CMOS as the sensor part was just shockingly bad. As a side note, Ommivision's OV7660 was chosen by most makers in China and thus cemented Omnivision's dominance in the earlier half of this decade in the China market. Once Mediatek's cellphone basebands boomed in China, Mediatek added ISP funcationality into their next generation baseband and Pixart suddenly found itself left hanging with a rapidly dwindling revenue stream from their ISP business, and no significant design wins for their underperforming sensor business.

    Pixart's luck changed about 3-4 years ago by landing a multi-year deal with guaranteed prices with Nintendo in supplying the sensor for the Wii controller. I guess this was fine for Nintendo as Pixart's sensor was not used as a photographic tool, just a rough IR detection sensor.

    Pixart were very luck in getting the Nintendo deal and god knows what deeds Pixart had to pull in order to win Nintendo's business.

    Once the Nintendo Wii deal moves on, I'm not sure what Pixart can do in order to keep their revenues growing. The only major business they have left is in mouse sensors. My guess is Nintendo's next generation might also has similar technology, so Pixart might live to fight another day if they can win anymore contracts

    I personally think UMC's foundry capability in fabricating Pixart's CMOS sensors is to blame for Pixart's crappy performance. But Pixart is stuck with UMC as they are part of the UMC group.

  2. Pixart mainly produced 3T pixel sensors, that's why they were unable to capture any significant market share on the camera phone market. Up until recently UMC had no competitive 4T pixels, so Pixart was handicapped in that respect. However, the recent rumor is that UMC was finally able to make progress and create lag-free and high QE 4T pixels. Now Pixart is preparing to re-enter camera phone market with much better imagers. Let's see what happens then.

  3. UMC has unwittingly killed many sensor design houses in the past. Back in 2001/2002, SETi (formerly CI Sensor) also fabbed their VGA sensors at UMC. The process node used was 0.25um and gave 1/4" VGA sensors. After the fact, CI Sensor went to Dongbu but never fully recovered from the UMC fiasco and was eventually bought out by SETI

  4. Photobit was fabbing at UMC before Pixart was formed. We were highly dismayed when UMC executives helped form Pixart soon after Photobit had taught UMC how to make reasonably good image sensors. Early Pixart designs used a pixel design that was remarkably similar to Photobit's. We were assured that there was a complete firewall between Photobit's designs and Pixart's activities and that any similarities were a coincidence and that UMC was fully committed to working with Photobit on new advanced designs.... Later, Photobit worked with TSMC and found much improved and consistent yields, a stronger team at TSMC, higher wafer costs, but not as much interest in our business. I am not sure when OVT started to work with TSMC, but it was a year or two later.

    This is the sort of thing a small fabless company with a new technology faces. Business ethics and practices vary around the world. It is challenge to find the right partner.

    Meanwhile, if Pixart is making money one must wonder if it is related to the UMC association and perhaps some very good wafer pricing.

  5. IC-Media was another successful UMC customer in 3T era. At some point IC-Media had half the market share size of Omnivision. However, it was unable to switch to 4T and eventually lost much of its sales because of that.

    These days UMC seems to get new customers. The rumor is that Pixelplus uses UMC as its foundry for 4T pixels.

    Getting back to Pixart, it's nice to see that despite all the limitations of its foundry partner, the company managed to find the creative solutions and remain profitable over many years. Many others fail to be so, even though their fabs are much better than Pixart's.


All comments are moderated to avoid spam and personal attacks.