One of the features of the new sensor is a reduced dark current at low temperature: "By devising the driving method of the peripheral circuit, this product realizes reduction of dark current noise that still occurs even in low temperature condition. This makes it possible to use it for astronomical observation of a star with a slight brightness on the order of 10 stars and so on."
For the mentioned target applications like astronomy etc. a monochrome version should also be available!
ReplyDelete-dkf
There is one 35MMFHDXSM
DeleteI am a bit surprised about the full-well capacity. Only 61ke-, resulting in something like 350 mV saturation at gain 1. This is surprisingly low for such a large pixel, or do I miss something ?
ReplyDeleteI expected that these huge pixels should have a high QE, a high FWC, a high DR.
The low QE is probably caused by the lack of effective microlens at this pixel size.
DeleteAs for the small FWC and FD swing, it could be caused by the lateral field tricks that reduce image lag. They might be spending so much potential to create the lateral field that nothing is left for the FD swing:
https://image-sensors-world.blogspot.com/2017/09/canon-on-large-pixel-design-challenges.html
Why do microlens matter? They just collect photons. They don't store electrons. How can microlens have impact on the electrons full-well capacity?
DeleteMicrolenses explain only the QE. The low FWC is a separate paragraph and topic in my comment.
DeleteSo the output-referred dark noise is 12.32uV rms @ gain 16x ?
ReplyDeleteAnyone knows which fab Canon uses? Slap Nyxel technology on these and it'll benefit Astronomy by being more sensitive in the NIR when using NIR pass filters to get NIR exposures. Those pyramids also bump up the visible QE.
ReplyDeleteIs it FSI or BSI?
Hmmm, no ulenses available for 19um pitch? Really?
The 19um microlens are available, they are just not as effective as the smaller ones.
DeleteWhat pyramid size do you propose? 19um base or smaller?
I thought the pyramid size depends on the target wavelength? As such you'd have multiple peaks and valleys within each pixel, otherwise it would seem to me like the silicon thickness required to make the pyramid work would be prohibitively thick.
ReplyDeletePyramid size is optimized for pixel optics and light absorption path. I'm not sure what is the optimal size for 19um pixel case, but it's not likely that green or blue QE would be improved by pyramids. In Canon case, the QE is probably limited by the pixel fill factor.
DeleteAnother interesting question is why Canon does not use the small 3x3 or 4x4 groups of pixels instead of a single 19um one? These small groups could be hard-wired in binning mode and behave essentially like a large super-pixel. Then, a higher QE can be achieved and, possibly, larger FWC too.
That is an interesting point. Probably the small pixel 3x3 or 4x4 array is also easier to design from the process point of view.
DeleteCould it be the conversion gain gets worse because of the additional metal wiring and FD regions affecting the input referred noise ? ... but the low-light performance should be boosted by the improved QE because of better focused micro-lenses.
As of now, their conversion gain is 5.6uV/e. Even with 4x4 sharing, they easily get much better CG.
DeleteI wonder whether this sensor is able to show any better result in lowlight as compared with sony IMX235 from ILCE-7s (which, as far as I know has better QE and back illumination) (of course on condition that we previously bring sony's ⁴4K to 1080p either by viewing it on Full HD display or by a smart resize).
ReplyDelete