# 10 Jan, 2018 06:35
I visited the Canon booth today at Consumer Electronics show and noticed they were showing off a 19micron 1080p CMOS camera used very dark video surveillance. Obviously the resolution is pretty low and the pixel size is extreme for astro-imaging.|
However, they were also showing off their new 122MP 2.2micron APS-H sensor and I was intrigued. 2.2microns is tiny but with 2x2 and 3x3 binning, we are getting into the sweet spot for pixel size without sacrificing resolution due to the 13280x9184 resolution of a 122MP sensor.
I talked to some of their reps and found out Canon would like to investigate the astrophotography market to see if they can partner with some companies to build astro cameras using their latest sensors. I would love to be able to purchase a CMOS with good pixel size, great resolution, low read noise and dark current, USB3.0 and mono!
I've only been imaging for about a year, so I wanted others with more knowledge to chime in. Would anyone else be interested in buying this? Of course, I have know idea what the cost would be.
Here's the sensor specs…
APS-H (29.22mm x 20.2mm)
Color or Mono
13280 x 9184 (122MP) effective resolution
10,000e lux/sec (color)
20,000e lux/sec (mono)
10,000e full well capacity
2.3e rms @gain x8, room temp dark noise
8.1e/sec @gain x8, 60C dark current
720Mbps in LVDS output 9.4fps @ 10bit
(Unfortunately the full well is low)
# 10 Jan, 2018 07:01
|As far as I know, real "CCD like" binning on CMOS sensor is not possible :/ only software level simulated binning, which is not the same as a pixel level real binning.|
# 10 Jan, 2018 07:06
|That's unfortunate. If we could wave a magic wand and design our own APS-H format CMOS, what would an ideal pixel size be while maintaining a good resolution?|
# 10 Jan, 2018 07:53
Here are a few points regarding the above discussion.
1. There already exist Canon EOS APS-C sized cameras with suitable pixel sizes for astrophotography. Why would one want to buy a camera with much smaller pixels sizes (supposedly more expansive) and then bin the pixels?
2. I was told that there are CMOS binning technologies being developed that can achieve CCD-like binning. However, questions #1 still remains. Plus, one has to wonder what it means by binning the pixels in a Bayer array? For example, each R pixel in a Bayer array is two pixels away from other R pixels. If you bin four (2x2) R pixels into one, you are no longer binning adjacent pixels like what you do on mono CCDs. So even if CCD-like CMOS binning technology exists, how it will be useful to astrophotography remains to be seen, unless we are talking about mono CMOS.
3. CMOS chips used in DSLRs are already being used in cooled astronomical cameras. But those are Sony CMOSs, not Canon ones. So if you want astro-cameras with low price, low noise etc, options already exist. You don't have to wait for Canon to offer their CMOS chips. Plus, Sony CMOS performs substantially better than Canon ones, which is pretty much now widely recognized.
The optical lab of my institute collaborates with the Japanese National Observatory to build a giant spectrograph for Japan's 8m telescope in Hawaii. In a camera subsystem we are building, we use Canon's 50 MP medium format mono CMOS. And we have used this chip for years, since the designing and testing phase of this project. So Canon is indeed making mono CMOS that's very suitable for amateur astrophotography. It's their business decision not to mass-produce these sensors, not only not for other camera companies, but also not for their own cameras. That chip exists for years, but we had never heard any Canon cameras using them, right?
I think waiting for good astro-cameras with Sony CMOS is much more practical. They already exist. And I heard that QHY is demonstrating some larger format CMOS cameras in some conference recently somewhere. We can have our fingers crossed.
# 10 Jan, 2018 08:09
Let me add one more point:|
CCD binning is useful largely because Kodak CCDs have huge readout noise (around 15 electrons). (This does not apply to recent Sony CCDs.) To overcome the large readout noise, one can bin the pixels to reduce the total amount of readouts for an image. For example, if you use a 4K*4K chip and bin it 2x2, the number of readouts required for an image reduces from 16M to 4M. This can greatly reduces the noise in an image, at a price of lower image resolution.
The same concerns do not quite exist on CMOS. Sony and Canon CMOS typically have readout noise of 2 to 4 electrons, depending on the ISO (gain) and camera model. This pretty much eliminates the need of on-chip binning for many (if not most) situations.
# 10 Jan, 2018 08:28
I now understand the difference between CCD hardware binning vs CMOS software binning. I mentioned binning in OP since 2.2 microns is so small. Is it too simplistic to say that 2x2 software binning would give you a 4.4 micron “superpixel” that may give you a better image scale to prevent oversampling?|
Sorry for the confusion, i am more concerned with image scale since read noise is basically irrelevant for modern CMOS sensors.
# 10 Jan, 2018 09:07
There can be many parameters in this game. If the two chips have equal quantum efficiencies, given that your subs are not under-exposed, software binning 2x2 2.2 um pixels should give you a results similar to a 4.4 um native pixel. However, we know that for chips of the same generation, there is a tendency that larger pixels have higher quantum efficiencies. For chips of the same size, those with more pixels can generate more heat and therefore have higher dark noise. So the actual situation is always more complicated.|
I think it safe to say that it is always more desirable to look for cameras with the right pixel size to begin with. Getting a camera with inadequately small pixels and then try to bin it (no matter hardware or software) can always lead to unwanted problems or less performance at least. There is really no point doing this.
# 10 Jan, 2018 09:19
|Ideal pixel size also depends on the focal length does it?|
# 10 Jan, 2018 10:06
|Right. And not only focal length, but also optical performance and atmospheric stability.|
# 11 Jan, 2018 07:50
|I'm surprised none of the vendors have picked up the CMV12000 (http://www.cmosis.com/products/product_detail/cmv12000) or CMV50000 (http://www.cmosis.com/products/product_detail/cmv50000) yet. In particular the NIR enhanced CMV12000-2E12M1PA. I reached out to some vendors and none of then considered it so far. Seems like an excellent chip and comes in mono. Even contemplated building my own with the https://apertus.org/axiom-beta kit.|
# 11 Jan, 2018 08:10
|Based on the data in the links you provided, the 12000 model has dark current about 1000 times worse than typical CMOS in Sony/Canon cameras. Its readout noise is about 5 times worse. The 50000 model has dark current on par with Sony/Canon CMOS, but its readout noise is still about 3 times worse. I think it is quite understandable why astro-camera companies do not prefer these chips, unless they are significantly cheaper.|
|You have no new notifications.|