Yes, the current sensors have unconditional reset when their exposure time ends and the integrated signal is read out. Conditional reset is done not by time but when the integrated signal level crosses a threshold.
OK - so what is so special about this? This has been done by other randomly addressable sensors for extending linear dynamic range. RAMBUS is couples decades behind times on this concept....
OK, let's play this game. First, I really don't know what the Rambus IP is regarding the conditionally reset binary sensor (you know, the megapage patent app). Second, I have nothing to gain, win or lose, in this game. Third the opinions I express in this game will be my own and have nothing to do with those of Rambus, even if they happen to be the same.
So, you say "decades". Let's start then with your top 3 examples of prior art on conditionally reset binary sensors and let's see if we can distinguish the Rambus concept from those decades-old concepts.
Sounds to me that you are concerned and/or have something to personally loose/gain from this discussion. Yes, there are published articles about extending linear dynamic range by "selectively" resetting and and reading each pixel. Perhaps RAMBUS just twisted the words and called it conditional reset?
You mean link this one? http://ericfossum.com/Publications/Papers/Wide%20dynamic%20range%20APS%20star%20tracker.pdf Wide dynamic range APS star tracker Orly Yadid-Pecht, Christopher Clark, Bedabrata Pain, Craig Staller, and Eric Fossum
and the prior art referenced therein? Not binary output. Very different.
I was not familiar with this work so thanks for bringing it up. I looked through the entire thesis, which seems to be about predicting the optimum integration time for each pixel, and setting it to that value.
On pages 40-51: Part of this is review of prior art, and part describes his algorithm.
Aside from it being in the class of HDR image sensor approaches, and requiring different integration times for different pixels, I don't see this as being related to reading out binary data from each pixel and reconstructing images from the binary data.
The Rambus work is much closer to a sigma-delta per pixel on a scanned basis, except the delta is variable from pixel to pixel since all pixels get reset to the same voltage. In that case you could look at Mandl's work perhaps. But the Rambus work is significantly different from that CCD approach.
what is conditional reset? what is unconditional reset? Does the sensor in smartphone today use unconditional reset?
ReplyDeleteBesides, will this binary pixel work well with rolling shutter since it is nonlinear?
Yes, the current sensors have unconditional reset when their exposure time ends and the integrated signal is read out. Conditional reset is done not by time but when the integrated signal level crosses a threshold.
DeleteVery interesting!
ReplyDeleteOK - so what is so special about this? This has been done by other randomly addressable sensors for extending linear dynamic range. RAMBUS is couples decades behind times on this concept....
ReplyDeleteperhaps the fact that they took out a patent on the technique?
DeleteOK, let's play this game. First, I really don't know what the Rambus IP is regarding the conditionally reset binary sensor (you know, the megapage patent app). Second, I have nothing to gain, win or lose, in this game. Third the opinions I express in this game will be my own and have nothing to do with those of Rambus, even if they happen to be the same.
DeleteSo, you say "decades". Let's start then with your top 3 examples of prior art on conditionally reset binary sensors and let's see if we can distinguish the Rambus concept from those decades-old concepts.
Sounds to me that you are concerned and/or have something to personally loose/gain from this discussion. Yes, there are published articles about extending linear dynamic range by "selectively" resetting and and reading each pixel. Perhaps RAMBUS just twisted the words and called it conditional reset?
DeleteYou mean link this one?
Deletehttp://ericfossum.com/Publications/Papers/Wide%20dynamic%20range%20APS%20star%20tracker.pdf
Wide dynamic range APS star tracker Orly Yadid-Pecht, Christopher Clark, Bedabrata Pain, Craig Staller, and Eric Fossum
and the prior art referenced therein?
Not binary output. Very different.
I remember Prof Sodinis PhD student doing something like this a while back..
ReplyDeleteLet me see..page 40-51
http://dspace.mit.edu/bitstream/handle/1721.1/16612/55672778.pdf?sequence=1
I was not familiar with this work so thanks for bringing it up. I looked through the entire thesis, which seems to be about predicting the optimum integration time for each pixel, and setting it to that value.
DeleteOn pages 40-51: Part of this is review of prior art, and part describes his algorithm.
Aside from it being in the class of HDR image sensor approaches, and requiring different integration times for different pixels, I don't see this as being related to reading out binary data from each pixel and reconstructing images from the binary data.
The Rambus work is much closer to a sigma-delta per pixel on a scanned basis, except the delta is variable from pixel to pixel since all pixels get reset to the same voltage. In that case you could look at Mandl's work perhaps. But the Rambus work is significantly different from that CCD approach.