Photonis new fast 1MP xSCELL camera is powered by a proprietary CMOS-based technology named InXite. The 2e- low noise at 1000fps full resolution rate InXite sensor features 14 bits ADC.
The 1024 x 1024 pixels, 15 µm pitch array is said to deliver have than 70% (other source - 65%) peak QE. The sensor has 2x2 binning mode and also non-destructive read mode to reduce noise to less than 1e- at 10fps. The pixel full well is 10Ke-.
Awesome! Anyone know the heritage of this sensor and camera? You can always write me privately.
ReplyDeleteNice work!
I'm skeptical.
ReplyDelete"InXite", which the posted link claims as a trademark, does not appear to be an active trademark in the US, and the two patents in the US PTO database listing the company as assignee are both related to photomultiplier tubes. Otherwise there doesn't appear to be any background material - white papers, conference papers, etc.
Interestingly, trademark costs seem to be fairly arbitrary. "InXite", for instance, looks to be available in the US for $159, in the EU for $431.33, in Canada for $599, in Australia for $249, in China for $349, and in India for $199. I wonder if this reflects a long-standing US emphasis on branding combined with bilateral volume asymmetries (for Canada and Australia because of relative populations, for India and China because of relative newness to mass-market branding). Whatever the reasons, it's quite a revenue bonanza for a number of governments.
In the US, anyone can put the TM mark on anything. This is free. Only the registered trademarks wirh the circle R symbol have a registration fee. Most companies do not bother to register their trademarks because they are used narrowly and not for very long. Anyway, most trademerks are like most logos - just a diatraction from brand-building for the company name.
ReplyDelete15um pixel size with only 10Ke FWC????
ReplyDeleteUseful only under very low light condition?
I guess that the pixel is CTIA based.
ReplyDeleteWe need a post on the Olympus scandal!!
ReplyDeleteThe claim of 1000 fps under 2-e maybe the row rate for 1 line as a rolling shutter 4T pixel.
ReplyDelete@ "We need a post on the Olympus scandal!!"
ReplyDeleteNormally, I do not publish camera companies news, unless there is an interesting image sensor innovation involved.
what is Olympus scandal???
ReplyDelete@ "The claim of 1000 fps under 2-e maybe the row rate for 1 line as a rolling shutter 4T pixel"
ReplyDeleteOR... it could be a claim for the entire 1024x1024 array, which has been not been fully optimized yet, but has already clocked 1080fps at full resolution.
It's a shame that we've all become so jaded and suspicious of the camera market claims.
@ "15um pixel size with only 10Ke FWC????
ReplyDeleteUseful only under very low light condition? "
Indeed, with a 2e- read noise, it IS intended for low light levels only. Besides, at 1000fps, most light level become low light levels.
In any case, the sensor still provides >12 bits of DR.
is it possible to make non-destructive readout with a 4T pixel?
ReplyDelete@ "is it possible to make non-destructive readout with a 4T pixel? "
ReplyDeleteIt may be possible... but on the tough side...
somehow, charge would have to be spilled BACK over the TR gate into the pinned photodiode prior to a 2nd sample.
I don't know how to do this, but my vast ignorance has been pointed out to me in the past.
but this approach can be used only in CCD with a readout by floating gate. In a 4T pixel, I can not see how this is possible.
ReplyDeleteSept 1995
ReplyDeletehttp://ericfossum.com/Publications/Papers/CMOS%20active%20pixel%20image%20sensor%20with%20simple%20floating%20gate%20pixels.pdf
@ "OR... it could be a claim for the entire 1024x1024 array, which has been not been fully optimized yet, but has already clocked 1080fps at full resolution."
ReplyDeleteThis is an interesting maybe-from-the-inside comment. Has it been tested at 1080 fps? If so, what further optimization is possible considering that what's being advertised is an image sensor in a ready product? I would think everything is pretty much set in stone, er, silicon, at this point.
Is there any further information on the InXite technology? I conducted a minimal search, and found nothing about it. I would like to learn more.
@ "It's a shame that we've all become so jaded and suspicious of the camera market claims."
I'm sorry it comes off that way. For my part, I'm just interested in how stuff works and how it could be improved. I like picking through technical details. I don't especially like glossed-over marketing.
@ "is it possible to make non-destructive readout with a 4T pixel? "
ReplyDeletewhat made you think of non-destructive readout (NDRO)? is it mentioned in their press release?
do you know how to read the text written in English :) ?
ReplyDeleteEric, I've one question on your floating photogate pixel: what is the voltage after CMOS process ? Did you do some UV/Tunneling trimming for the floating gate voltage ?
ReplyDeleteThanks !
-yang ni
I mean the voltage on the floating gate ...
ReplyDeleteFirst of all, it was a long time ago so I don't know if I recall, or just imagine that I recall. (If you are 50 yrs+ you will understand what I mean.) But I was surprised at the more-or-less zero stored voltage (charge) and its uniformity. There was some work on floating gate synapses in Japan around the same time with similar results. Junichi might (imagine that he) recalls better than I do.
ReplyDeleteA couple of years ago my now-8-year-old punctuated a family conversation with the statement "That happened a really long time ago... Like, back in the 1990s!" He was so utterly earnest and sweet and didn't understand why the rest of us broke out laughing.
ReplyDeleteThe problem is, I don't remember what we were talking about, just his hilarious interjection.
Probably it was a CMOS image sensor paper. (So I don't get anonymous grief for making a non-CMOS-image-sensor comment.)
Thanks Eric ! You are not old at all :) !
ReplyDeleteBy the way, indeed there are some work on analog synapse implantation using floating gate as summing node. But in all the publications that I've read, there is no statement on this point.
If you can find some data, I'll be very interested to know them. Otherwise I can place a test cell on a shuttle MPW ...
-yang ni
The frame rate of 1000fps is for the full 1024 x 1024 image. It uses a unique multi-tap architecture to achieve such high speed at low noise.
ReplyDeleteYes, the well capacity is only 10ke-, but this sensor allows 2x2 ion chip binning (not very common in CMOS) to increase the well to 40ke- if required.
the readout is digital or analog please?? Why the FWC is so small with a so big pixel??
ReplyDeleteThe sensor readout is analog.
ReplyDeleteThe 15um pixel size provides much more light gathering capability than other low light CMOS sensors with much smaller pixels. The InXite technology also provides the ability to perform accumulations on chip using a nondestructive readout. So, for example, 10 accumulations would give you an effective FWC of 100ke- at 100 fps.