Lists

▼

Thursday, December 30, 2010

Samsung Proposes SEFET Sensor, Pixart Flips Color Filter and Microlens

Samsung patent application US20100320515 proposes a single-electron JFET-like structure (SEFET) to make small and sensitive pixel:

The photoelectrons are collected in N-doped area and drift close to the p-channel area due to the doping gradient. Then the collected electrons modulate the channel conduction by changing the N-gate potential:


Few different readout schemes are proposed, for example, this one (Vin is N-doped area voltage):

The SEFET pixel is said to be able to achieve a conversion gain of 1-9mV/e, as measured at the N-doped gate. Less than 1um pixel size is claimed to be possible.

One thing I was unable to find in the Samsung application is how the N-doped gate is reset before the integration starts.

Pixart application US20100320552 proposes to place color filter on top of microlens to improve the light passage through the metal stake:

"Compared with the prior art, such arrangement is advantageous in that the micro lens 18 is located much closer to the photo diode 12. As a result, the micro lens 18 has a wider chief ray angle, and the photo diode 12 has a better light collection efficiency."

at

7 comments:

  1. AnonymousDecember 30, 2010 at 11:47 PM

    SEFET = Charge Modulation Device.
    Flip ML/CFA = the color cross talk will be increased and depend on the ray angle.

    ReplyDelete
  2. AnonymousDecember 31, 2010 at 8:08 PM

    Are you sure 1-9mV/e? I think that m = micron here!!!

    ReplyDelete
  3. Vladimir KoifmanDecember 31, 2010 at 9:07 PM

    In the application it's written meV/e, as the gate is virtual and it talks about the conduction band barrier. The trick here is to have very small area transistor, so the effective gate capacitance is low.

    ReplyDelete
  4. Eric R FossumDecember 31, 2010 at 9:39 PM

    Millivolts. This is a step towards the jot-based image sensor. We need to be able to detect a single photogenerated carrier.

    ReplyDelete
  5. AnonymousDecember 31, 2010 at 10:22 PM

    1mV/e => 0.16fF on the detection node. Is this possible? Even in a 4T pixel, the floating diffusion has still 1-2fF.

    ReplyDelete
  6. AnonymousDecember 31, 2010 at 10:23 PM

    have they talked about the thermal noise in this structure please?

    ReplyDelete
  7. AnonymousJanuary 2, 2011 at 12:44 PM

    Very similar to DEPFET I guess

    ReplyDelete

All comments are moderated to avoid spam and personal attacks.