The 4MP camera sensor is based on 6.5um pixel with full well of 30,000e. The ORCA-Flash4.0 has QE of over 70% at 600 nm and 50% at 750 nm, and has only 1.3e of read noise at 100fps at full-resolution. This unique combination of high QE and low noise, in the absence of EM-CCD multiplicative noise, means that images are no longer limited by the camera.
The new camera datasheet has comparison with competitors:
The uniformity comparison:
A very nice whitepaper explains the comparisons and the new sensor features.
pls. check white paper link - may be broken.
ReplyDeleteSorry - it's fixed now.
ReplyDeleteIs the sensor itself from Hamamatsu, or is it the sCMOS publicized by Fairchild?
ReplyDeleteWhy do they don't show up any prices. It's fun to look at all those "0" in them :)
ReplyDelete@ "Is the sensor itself from Hamamatsu, or is it the sCMOS publicized by Fairchild?"
ReplyDeleteHard to say. The pixel size is the same, but Fairchild sCMOS is made in 5.5MP and 2.1MP flavors, while Hamamatsu one is 4MP. Hamamatsu QE is >70% vs Fairchild's 55%. Other than that the spec is quite similar:
http://image-sensors-world.blogspot.com/2011/08/bae-systems-imaging-solutions-offers.html
It doesn't make sense to me that Hamamatsu would use a Fairchild sCMOS sensor in a camera and then claim better performance than with a Fairchild sCMOS sensor. There has definitely been enough time for the sCMOS to have been taken apart and a competing chip designed working around or licensing any sCMOS patents. That's what I'd guess happened here.
ReplyDeleteThey used a SONY sensor in the first version ORCA camera. Then they have designed their own CMOS image sensor with better performance, better uniformity, etc.
ReplyDeleteThe Fairchild sCMOS approach required cooling to lower dark current. I tried to see if Hamatsu camera needed cooling, but it doesn't seem to mention this. Any ideas?
ReplyDeleteThe camera datasheet says that the sensor is air-cooled down to -10C.
ReplyDeletefor this kind of camera, the cooling is mandatory. Long exposure time will accumulate dark electrons. If you can 1-2e readout noise, then DC should be limited to the same order.
ReplyDeletethe "gen1 sCMOS" from Fairchild/BAE had a real MTF problem...
ReplyDeletedoes anyone know what is the MTF of the new "gen2 sCMOS"?
@ CDM
ReplyDeleteI agree that it does not make sense...
But... lo and behold!!! Fairchild/BAE recently comes out with "their own" 4MPix sensor and camera...
Funny coincidences are for both sensors:
6.5u pixels
identical QE curves
both pixels are now 4T and not 5T
identical noise numbers
Dark noise histograms look similar
I once heard that "if it walks like a duck, quacks like a duck..." etc.
Perhaps Hamamatsu is simply getting higher cosmetic grade devices?