Friday, July 02, 2021

CIS Stacking Patent Invalidated due to PCB-Based Prior Art

BloombergUS Patent Trial and Appeal Board (PTAB) dismisses Cellect stacked patent lawsuit against Samsung on the grounds of existing PCB-style prior art from 1995. Cellect's patent US 9,198,565 "Reduced area imaging device incorporated within endoscopic devices" proposes image sensor stacking as a way to reduce endoscope footprint:

PTAB concludes that 1995 camera arrangement proposing to stack drivers on PCB underneath the imager, such as described here, is a known prior art:

As an unrelated note, the PTAB document quotes Kodak market review from 1995, where the company rightfully forecasts that CMOS sensors would conquer the whole mass market, while CCDs would survive only in low-volume niche applications. As the history shows, having the right forecast at the right time has not changed the fate of the company:


  1. I think hybridized infra-red focal plane arrays (IRFPAs) using indium bump bonds to connect a detector layer to a readout circuit layer would be even stronger and earlier prior art for the concept of stacking. I worked on those devices in the early 80's, for example, in the aerospace industry. The method and density and materials for the detector were different, and stacking technology vastly improved since then, but seems a better example than PCB stacking. I understand having patent claims invalidated takes a lot of care when it comes to the words that are used, but I am just talking about the concept. BTW, anyone besides me remember Z-plane technology? A different sort of stacking concept, but perhaps still useful when it comes to CNN integration with image sensors, esp. when combined with modern stacking technology to achieve true 3-D interconnection within a volume of silicon.

    1. Irvine Sensor ... it was too early for that time. but I think that thermal dissipation in such configuration could be a serious problem.

    2. @yang ni - you win the prize! I agree heat dissipation (power dissipation) is important but maybe there are some innovations involving etched microchannels and active cooling that could be used. Plus power dissipation will continue to drop for advanced nodes on per-functionality basis.

    3. Take a look at industrial inkjet printheads, like the Samba from Fujifilm or W1-series from Xerox. Those are practically solid blocks of stacked silicon with microchannels. Just add the missing photodiodes.


All comments are moderated to avoid spam and personal attacks.