The presentation also tells a little bit about image stabilization feature: "A series of short exposed images are captured, aligned and summed up in the same pixel array. Pixel array is used as memory during the stabilization process"
Thursday, September 18, 2008
Advasense Explains its Technology
Advasense made quite a lot of noise with its impressive 1.4um pixel announcement a few days ago. The company has few presentations on its web site. One of them, "The Small Pixel Race" from over a year ago, compares Advasense feedback pixel with 4T pixel:
The presentation also tells a little bit about image stabilization feature: "A series of short exposed images are captured, aligned and summed up in the same pixel array. Pixel array is used as memory during the stabilization process"
The presentation also tells a little bit about image stabilization feature: "A series of short exposed images are captured, aligned and summed up in the same pixel array. Pixel array is used as memory during the stabilization process"
Is there real data that shows this does not have kTC noise or suffer trap and deep photocarrier lag? I am just wondering...because it looks like it should but maybe there is something else going on.
ReplyDeleteThey should submit a paper to the forthcoming IEEE TED special issue, or the 2009 IISW.
-EF
Is there real data that shows this does not have kTC noise or suffer trap and deep photocarrier lag? I am just wondering...because it looks like it should but maybe there is something else going on.
ReplyDeleteThey should submit a paper to the forthcoming IEEE TED special issue, or the 2009 IISW.
-EF
The presentation specifically denies the image lag, so I assume there is a data to support this claim. As for kTC noise, you are right, there is no mention of it in the presentation. On the other hand, what does the feedback do there, if not kTC suppression?
ReplyDeleteWell, to my eyes, this circuit diagram looks a lot like a passive pixel sensor with an amplifier at the bottom of the column to keep the column voltage constant during readout. It looks like their pixel is a pinned photodiode with transfer gate with all pixels in the same column tied together at the other side of the transfer gate.
ReplyDeleteSo, maybe they made some improvements and it works well or maybe I missed something. But, there is no data in the presentation and this is where the rubber meets the road.
Still, using a transfer gate and junction photogate aka pinned photodiode I agree they can improve full well and it is probably easier to optimize complete transfer and minimize lag. The problem comes from dumping a few electrons onto a large column bus capacitance and expecting not to get it lost in the noise. There is still a sort of kTC noise in this approach due to the feedback amplifier's gain-bandwidth product. So, let's see the data!
-EF
I guess I should add that if my description is correct, I don't see the advantage over a buried photodiode and normal transistor switch in the pixel. That is, I don't think there is any advantage in actually transferring the signal charge out of the junction photogate.
ReplyDeleteAgain, makes me think I am missing something.
-EF
Our pixel is not passive one. Layout- and schematic-wise it's quite similar to other 4T shared pixels. It has reset, select and amplifier transistor connected to few photodiodes with transfer gates.
ReplyDeleteThe kTC noise is reduced by feedback loop. Most other feedback schemes work only with 3T pixels. Ours works in a different way and it virtually eliminates kTC noise in 4T pixels.
Our PD works in charge sharing mode, rather than charge transfer. The PD is much deeper than most and relatively heavy doped. Most of the PD capacitance comes from its sidewalls. Deep and heavy doped PD is impossible design, if full charge transfer is needed.
The measured image lag is less than 0.5e under all conditions.
Vladimir Koifman, Advasense.
In continuation of the Advasense discussion I got the following mail from EF:
ReplyDeleteI keep trying to respond to the Advasense discussion but it wont let me.
"Thanks Vlad, I think I understand your approach now. It seems to be an interesting mix of active and passive pixel strategies. Keeping the cap. the feedback circuit has to drive small can lead to good kTC suppression.
It would still be good to see some data but I agree that competitive performance is possible. -EF"
Is it limited to 6 comments or something?
Thanks,
Eric
Image Sensor: I'm not aware of any limit on comment number. May be some glitch in Blogger server?