What set Aptina's announcement apart of some others is that the company explains how these improvements have been achieved, a rare example of openness (Youtube link):
“Designed by our Japan design center to provide the highest image quality possible, our new APS-C format sensor delivers the maximum signal-to-noise ratio across all scene lighting and ISO speed conditions,” said Junichi Nakamura, Aptina’s Japan Design Center Director. “The positive reception our sensor has received from DSLR camera makers underscores the SNR and inter-scene dynamic range performance we’ve been able to achieve.”
TSR estimates CMOS image sensor shipments for the DSLR market to be 12.5 million in 2013, up from 4 million in 2007, with much of this growth coming from the emerging compact mirrorless segment of the market. “Our new sensor is only the first step in our effort to secure a prominent position in this exciting market,” said Sandor Barna, GM of Aptina’s Camera Business Unit.
Aptina DR-Pix technology page describes the 2-gain approach:
Aptina’s innovative approach to improving low light performance without compromising image quality utilizes a dynamic response pixel, called Aptina DR-Pix technology. Aptina DR-Pix technology combines two modes of operation in one pixel design – low conversion gain (LCG) mode for large charge handling capacity in bright scenes and a high conversion gain (HCG) mode with increased sensitivity and low read noise for low-light scenes. The inter-scene DR equal to 82.9 dB due to the low read noise of the HCG mode. In comparison, without the HCG mode of operation, the sensor’s DR would be limited to 69 dB of intra-scene DR in the LCG mode.
The sensor's 4.79um-sized pixel schematics is given in Aptina's white paper:
Sampling now, Aptina’s new 16MP MT9H004 image sensor is available in a 116-pin CLCC package. Mass production is scheduled for Q1 2011.
The circuit looks similar to the lofic WDR developed by Fossum...But I guess in this case the DCG voltage is controlled manually right?
ReplyDelete@ the DCG voltage is controlled manually right?
ReplyDeleteYes, it looks so.
Is Aptina making profit ?
ReplyDeleteYes, looks similar. Would not be a surprise since Photobit -> Micron -> Aptina Japan office under Nakamura did the first prototyping of that circuit for me while I continued with Micron around 2002. I am sure they have improved it considerably. Also, Micron/Aptina is the assignee of the issued patents on my contributions. Hope this goes well for them.
ReplyDeleteSo the idea is that in high-gain mode, you shut off the DCG transistor and dump your photocharge through the transfer gate onto the relatively smaller capacitance of FD, while in low-gain mode you turn on the DCG transistor and dump your photocharge through the transfer gate to the relatively larger capacitance of FD in parallel with the unlabelled capacitor?
ReplyDeleteIt seems like this incurs a hefty chip area cost for the unlabelled capacitor, the DCG control line, and the DCG transistor. Plus you have to choose in advance whether you want low gain or high gain, with ratios that are fixed at time of manufacture.
What is the motivation for doing this rather than just doing a shorter exposure and a longer exposure with the high-gain portion of the circuit, where you can adjust the dynamic range allocation between bright performance and dark performance via the relative exposure times? (Aptina and Johannes Solhusvik got an award or award nomination for a chip with a triple-exposure approach like this earlier this year, I think.)
Well, I can't speak for the why's or even the how's for Aptina, although I would always give them the benefit of any doubt for the why's or how's.
ReplyDeleteIn my scheme, the pixel is first read with DCG off (high gain). DCG is then always turned on, and the pixel read again. If the signal is large, then incomplete 4T transfer results and the first read is sort of a saturated signal. Once DCG turns on, the rest of the signal flows out from the transfer gate and the signal is re-read. One can then externally fuse the two signal readouts. (minus the reset level readouts).
Thus the same signal is measured twice, once at high gain and once at low gain. There is no decision process. The signals also come from the same integration period so there is no motion skew. Lastly there is no need for external memory for storing the multi-exposure WDR signal.
Could be Aptina is doing something else. The white paper was a little vague.
If the PD has more charges than one can collect on the FD, adding a a switch with an additional cap to have a bigger bucket to collect PD-charges in a second read is obvious to any analog circuit designer...or at least, that's the first thing that came to my mind when I first got exposed to pixels a few years ago.
ReplyDeleteA small cap limits the FWC, adding a bigger cap limits the responsivity. So...you make the big cap selectable. duh...
Did the USPTO really accept a patent for that?
Preposterous, I'm glad I got out of this industry.
What's next, someone is going to patent the idea to put an A/D converter and an image array on the same die?
LOFIC does this without adding additional devices to the FD, that is a creative idea that deserves some merit. But adding a selectable capacitor...come on...
The interesting question is which of the DSLR makers will be buying these sensors from Aptina.
ReplyDeleteCanon, Sony, and Samsung make their own sensors and Nikon is (still?) buying from Sony.
That leaves Pentax as a potential buyer right now.
Maybe Nikon will be following suit ???
its look like Nikon, i think this may be result of their partnership with Nikon
ReplyDeleteAptina as such doesn't looks to be low on R&D stuff, only problem looks to be in their process technology. They seriously need to either acquire Micron fabs or build partnership with TSMC. I am still skeptical as to why they are not able to out BSI sensors
ReplyDeleteI guess Aptina is shrinking big time in mobile world and their new sensor is kind of an effort for them to launch into other markets. Interestingly, even this sensor seems to be at lower fps especially for future cameras. The market already has DSLRs with such specs. Unless they're already selling it to somebody, its kind of hard to believe that this product will be bought by anybody for DSLR cameras (definitely not beyond next year).
ReplyDeletethere's rumors that nikon is working renesas
ReplyDeleteThere is something I still dont understand with lofic: when you read the lofic cap you multiply the digitized voltage of the FD with Clofic/Cpd in order to have a linear light to code conversion. You need a good matching between Clofic and CPD for a correct conversion. How do you solve this?
ReplyDelete@ a little vague
ReplyDeleteMaybe they should be called "gray papers" for this very reason. :D
Any one knows the F no. of lens used in the examples of their whitepaper ???
ReplyDeletethanks a lot !
They mention low dark current, but no numbers. Does anyone know what it is at, say, 60C?
ReplyDeleteLooks like du*p sorry, High-ISO Capabilities showed at this video aren't very impressive, just normal standard CMOS stuff.
ReplyDelete