Friday, November 02, 2012

Altasens Talks about Stacked Sensor Advantages

Altasens' patent application US20120267511 "Image sensor with hybrid heterostructure" by Lester Kozlowski has few interesting claims:

First, a stacked sensor structure is proposed:

The pixel array uses "optimized" pmos transistors. An example pixel capable of GS, S&H and CDS is shown on Fig. 6:

The main advantages of the stacked structure are (1) very good global shutter efficiency and (2) lower noise due to a special pmos transistor "optimization":

"FIGS. 7 and 8 [not shown here] compare the read noise achievable with optimized PMOS technology vs. standard CMOS technology wherein the source follower is formed in NMOS having flicker noise lower than readily achieved in common foundry processes, i.e., very good compared to what is readily available. Even so, the PMOS global shutter can yield a read noise of 1 e− (or holes) at a sense capacitance of 5 fF. The NMOS global shutter circuit instead has read noise of 3.5 e− at 5.5 fF. More importantly for long term development, the PMOS solution goes well below 1 e−0 as the sense capacitance is reduced while the NMOS solution plateaus well above 2 e−."


  1. What storage capacitors do they need to use at the second layer to achieve < 1e- noise at that low conversion gain? 10pF per pixel? Please correct me if I'm wrong..

    1. Not necessarily that big cap. The pixel uses common source pmos instead of source follower. So the pmos part of the pixel might have a significant voltage gain, depending of ratio of FD capacitance over the dotted line one.

      If anything looks strange in this schematics, it's that bulks of all the transistors are tied to their sources. I'd guess it's just a quick and dirty sketch for simulation.

    2. If gain is high you have loss of dynamic range...

  2. what people try to apply for these days...

    hybrid imagers: done many times before

    PMOS sample&hold circuits: done before but not in a pixel because the nwell would take electrons from the PD

    combining the two is obvious: how can anyone claim this as an invention

    btw: some used PMOS SF in large pixels before, even in a non-hybrid structure

    Indeed to get to very low noise, you would need some gain before sampling

  3. Les Kozlowski was an expert in hybrid IR FPAs long before he co-founded Altasens. This application is really no surprise at all. That hardest thing he will have to fight against at USPTO is probably his own prior art.


All comments are moderated to avoid spam.