InVisage spent three years engineering the quantum dot material to produce highly-sensitive image sensors that integrate with standard CMOS manufacturing processes. "While other image sensor companies have been focused on increasing the number of pixels, InVisage is focused on bringing completely new thinking and technologies to CMOS image sensors -- like tackling rolling shutter," says Jess Lee, CEO, InVisage Technologies. "InVisage's QuantumShutter will bring better and more reliable picture quality to an industry that is desperate for innovation."
Update: Invisage web site states that the image above came from 1.4um pixel test chip. "This image is the first and only 1.4um pixel in world which can implement this [QuantumShutter] feature." QuantumShutter will be available as an option on InVisage's first product due to be sampling later this year.
Nice picture. Some details would be good.
ReplyDeleteI thought shutters were always quantized, like ON and OFF. But, I guess we needed a new trademarked name to keep the buzz going.
"InVisage is the first and only company to solve the rolling shutter effect in any device, from high-end cameras like digital SLRs to small everyday devices like cameraphones."
ReplyDeleteThis is an outright lie, starting with FT and FIT CCDs and moving right along to many CMOS image sensors.
For sure Aptina/Micron already have product with global shutter:
ReplyDeletehttp://www.engineeringtv.com/video/Aptina-TrueSNAP-Global-Shutter
What I want to know is whether the picture was actually taken with an InVisage sensor or is just a simulation for PR.
ReplyDeletewhere have they been last 5-10 years? at least on this planet many have tackled that problem before...
ReplyDeleteawesome!
ReplyDeleteIs the picture really taken with their quantum film sensor ?
website says its a 1.1um. isn't that kind of small for a global shutter pixel?
ReplyDeleteInvisage has a page dedicated to QuantumShutter on its web site:
ReplyDeletehttp://www.invisageinc.com/page.aspx?cont=QuantumShutter
It says regarding the image origins:
"The image below is from our test chip using our 1.4um pixel.This image is the first and only 1.4um pixel in world which can implement this feature. [global shutter, that is]"
I think that their image is a fake one. If you look at carefully these two images, they are exactly the same. How can you shot two image with exactly the same view point ? Please zoom in on these image, you can see that even the small details are coincided exactly !
ReplyDeleteThis comment has been removed by the author.
ReplyDeleteThis technology is being pitched as superior to in-silicon alternatives for fill factor reasons, but some or all of the performance gain could go away depending on the dark current characteristics of the QuantumFilm and whatever vias and other circuit elements connect it to the charge storage area.
ReplyDeleteIf the 1.4 um pixel with global shutter was implemented with sub 5 e- read noise and no lag then it is truly an important achievement.
ReplyDeleteOtherwise, practically any big player could make a 1.4 um pixel device with a global shutter and with kTC noise. Not much of an achievement in that case.
@ practically any big player could make a 1.4 um pixel device with a global shutter and with kTC noise.
ReplyDeleteAs far as I know, everybody uses 2- or 4-shared 4T pixels at 1.4um node. There is one floating diffusion per 2 or 4 photodiodes. How do you implement global shutter in that case?
Eric,
ReplyDeleteAs IISW said, it will be hard to do GS with shared FD.
Don't people typically use at least 5T (as opposed to 4T) pixels for snapshot?
The technology presented here may possibly have much better dark current at the storage node and superior shutter efficiency.
Does anyobody know what is the smallest true snapshot shutter pixel reported?
-wicky-
Dear Invisage people,
ReplyDeleteOn one hand the imaging community is constantly seeking to new ideas, technologies, breakthroughs in the field, but on the other hand, very often we were disappointed with announcements of these new breakthrough.
Help us to avoid a new disappointment this time with your technology. The imaging community would be very thankful if you can give us more technical data about the performance. We, engineers, love to see numbers. But there's also the saying : "Seeing is believing". Please provide us with a video to show the performance of your new technology. That will automatically stop all the negative publicity for your company and your technology. I am looking forward to it !
Albert.
Good try Albert. It is just not seeing, it is in the specs, and the proof is actually in selling to a (big) customer. Nevertheless, we should invite them to present something to their peers at IISW.
ReplyDeleteWicky, IISW <> ISW!
ISW, If you are going to allow kTC noise, why bother with a 4T pixel structure? Passive pixel, isolated sampling cap and 3T RO. It will work, just not so well. But, under favorable lighting conditions, it will look as good as 4T.
BSI can be used to ameliorate fill factor effects if need be, although the quantum film approach (like any overlayer approach) has a cost advantage over BSI. We would all be be making a-Si overlayer devices if there weren't those lingering pesky problems with a-Si material and overlayer readout porblems, such as noise and lag.
The test for Invisage will be noise and lag and eventually material stability. If all this looks good, it will be a real breakthrough. Right now the hype seems premature...but maybe they are running out of money or need to demo some next milestone.
So, a simple first question for Invisage - what is YNSR10 for your 1.4 device?
Eric,
ReplyDeleteIISW <> ISW: It was an honest typo...kind of like
YNSR10 <> YSNR10, but I would have to be out of arguments to react to such typo...
3T with switch+cap will always have worse conversion gain compared to a 4T, did anybody commercialize any snapshot products with such pixels?
In any case, the shutter efficiency of the proposed solution could be superior to conventional pixels.
What is the smallest report true snapshot pixel?
-wicky-
Wicky,
ReplyDeleteYNSR10 heh.
The conversion gain per optically generated electron will be worse of course. The node cap can be smaller.
The shutter efficiency will probably be better with a big metal plate blocking light (and QE in the quantum film higher) but the plate will still have light leaks. I don't think this is a big issue either way for consumer applications.
No clue about the reported smallest true snapshot pixel. Usually snapshot mode is most important for high speed imaging in which case pixels are large for sensitivity.
When using my cell phone camera for stills or video I have not yet seen the ERS effect so I am not sure how strong of a selling point this is.
Again, once we know "YNSR10" or something like that we can figure out the real impact of this technology.
Meanwhile I am impressed by the technical progress but their PR campaign is obnoxious. Foveon was also obnoxious but they didn't have outright lies in their press releases.
Frankly, I am kind of bored by CMOS image sensors in general and am looking forward to the emergence of a 3rd gen solid-state image sensor.
@ ISW, If you are going to allow kTC noise, why bother with a 4T pixel structure? Passive pixel, isolated sampling cap and 3T RO.
ReplyDeleteOK, I see. Still, it wan't be easy to fit everything in 1.4um pixel pitch and still keep a decent PD area.
@ BSI can be used to ameliorate fill factor effects if need be
The storage node light shielding would be difficult in BSI.
@ I am kind of bored by CMOS image sensors in general and am looking forward to the emergence of a 3rd gen solid-state image sensor.
Actually, CIS had a huge progress in the recent couple of years: 1.4um pixels matured to a degree where its QE is close to ideal. 1.1um pixels appeared and gathering good customer feedbacks. 0.9um pixels are in design everywhere. Resolution is approaching 200MP. 3D sensors are mature for mass market now. Consumer-grade BSI entered mass production. TSV and wafer-level optics are widely used. EDoF hit the mass market, finally. It's a log going on now, too much to be bored, in my opinion.
ISW - Well, I have never been happy doing the incremental work you describe. My Samsung project (disclosed in an earlier post) has its challenges and it is not your mother's CMOS APS device. Anyway, glad you and most others reading this are not bored.
ReplyDeleteI want to get back to jot-based imaging. Just need funding, an industry partner ot two, great students and a lab. It may not turn out to be the 3rd gen but I think it could be and the project is loaded with more questions than answers. Just the way I like it.
Eric, what do you mean "jot-based imaging" ?
ReplyDeletethanks !
@ what do you mean by "jot-based imaging"
ReplyDeleteSee the papers at Eric's links on the left side of the page.
Incidentally, maybe this QuantumFilm technology could be used to make a stacked structure resembling a jot. :D
the shade of a glass and a tin can looks darker
ReplyDeletewith their quantum shutter than that of conventional sensor. is that mean the sensitivity of their sensor is inferior to conventional one? I want to know more detail about their technology.
Would anyone be able to point me to a recent survey paper on amorphous silicon imagers if there is one, or else to some background papers where I could find out more about amorphous silicon noise and lag effects?
ReplyDeleteDo a search on e-Phocus. You can see some patents through 2009. This is part of Trex Enterprises from Hawaii. They showed their first a-Si imager (HDTV resolution) over 10 years ago.
ReplyDeletea-Si overlayer technology predates Trex by a long shot. a-Si photodectectors with CCD readout was explored in the 80's and I think NHK looked at a-Si with CMOS readout in the late 80's or early 90's. Xerox has been doing a-Si detector arrays with TFT readout (passive pixel) for x-ray apps for probably 20 years. Agilent did quite a bit of study on a-Si overlayers in the later 90's as well.
ReplyDeleteHow is it that with all the work on mos readout circuits before Fossum's efforts started in that area, that Fossum claims that he is the inventor. Is there a nice timeline of the actual meaningful inventions listed somewhere or is everything very incremental?
ReplyDeleteThanks for the pointers.
ReplyDelete