CMOS versus CCD cameras for astro, what are the cons?

[gregbradley]

I have a little bit of experience now with CMOS cameras and I know a lot are using them now. They are cheap compared to their performance and we have all seen some great example images.

But are they really the end of CCD cameras?

Given Kodak and OnSemi plus Sony have ceased CCD production that would seem inevitable. But CMOS are not without their issues, much more so than CCDs.

So far it seems to me that CCDs evolved into a mature state. Firstly they are all 16 bit whereas only a small % of CMOS are 16 bit.

Secondly CMOS can have their own peculiar noise problems CCDs don't have like rain noise (also called walking noise) where slight shifts between images end up with ripple like bands in the image.

CMOS are more prone to poor star colours and oversaturation.

Full well depth tends to be a lot weaker in CMOS although QHY offers some readout modes that will increase well depth (at the expense of read noise).

CMOS cameras tend to be a lot lighter and smaller which is easier on focusers and mounts.

Narrowband QE can be very high in some CMOS sensors.

The 1600 sensors has issues with microlense flaring on bright stars.

Drivers can be weak to poor in performance and often still being developed on cameras that are over a year old.

Reliability seems poorer judging by the number of people complaining about camera failures in short period of use.

CCDs can develop vertical bright lines (full frame CCD type), CMOS don't seem to.

CMOS suffer amp glow on almost all models except a few of the higher models. This calibrates out pretty well but not 100% perfectly and is a problem to be watched.

CMOS 2x2 binning is not as effective as CCD.

CMOS cameras are depending on Sony sensors mostly. The trend from Sony is for more megapixels and this smaller pixels along with smaller wells although this can be compensated with QHY cameras allowing different read modes offering deeper wells at the cost of higher read noise).

So the future of CMOS would appear to be smaller pixels that are less suitable for an ever larger group of telescopes and could eventually be confined to small focal length imaging. Imagine trying to image with a camera with 1.3 micron pixels? You'd have to bin it 3x3.

Looking though quite a few images now from CMOS I see plenty of nice high res CMOS images but you can immediately pick the CCD images by their usually much nicer stars with better colour and better shapes.

These small pixel large full frame sensors are very much harder to adjust for tilt and that becomes a big problem to overcome especially for those not well experienced in handling tilt.

The smaller sensors don't really give you a hard time with tilt though.

Some CMOS sensors don't calibrate well and they have extra settings to match that CCDs don't so they are more complex and harder to calibrate.
More things to match. These extra settings are not well understood and most fumble about trial and error or ask on the internet of people who also probably don't know for advice. Each setting change requires a whole new set of darks, flats, flat darks. So the calibration library can get very large.

Sensitivity, low noise and low cost are really where CMOS cameras shine.
Amp glow, white stars, complicated settings, poor drivers, extra types of noise are where they are problematic compared to CCDs.

QHY is kind of funny. They post an "all in one firmware updater" that is in fact at least 2 in 1 as it does not do the job by itself and needs a 2nd update to work. Haha, they don't say that on their site so that's a little trap for you.

The other issue is horizontal random banding which most seem to suffer from, some worse than others. It mostly shows itself in dimmer narrowband images not brighter objects. It does not calibrate out. So these lower gain settings sound good because of low read noise but you have to use higher gain settings to reduce/overcome the horizontal banding.

Out of the current crop of CMOS the no amp glow 16 bit backside illuminated Sony sensored APSc and Full frame seem the pick of the lot.
But at 3.76 microns they are already too small for 2 metres or beyond focal lengths unless you bin 2x2 or have exceptional seeing. Still good though.

CMOS seem like a longer learning curve because of the extra settings and getting to know them (the difference between some settings is almost undetectable in a single download).

Despite all this they seem to be the way of the future but need more maturing but the long term outlook is not good with future smaller and smaller pixels.

But I won't be selling my CCDs just yet. All those extra settings and weak software means less bullet proof and moreover another thing that can go wrong along with cloud, tracking and gear issues, wind etc.

The above is my opinion only from handling 3 of these CMOS cameras so far - a ZWO183mm (easy to use), QHY294m (also easy, banding is something to watch and poor drivers from QHY need to be kept updated, a QHY600m not very problematic but lots of choices and very sensitive to tilt).

Greg.

[sunslayr (David)]

No I would not say CMOS is a scam, they a rapidly gaining on the features of CCD with all the advantages that CMOS brings. The high frame rates alone allow for lucky imaging that no CCD could ever hope to achieve. If the noise keeps going down and the full well depth keeps going up then it won't matter that the pixels are smaller thanks to binning. That being said the 294 is perhaps one of the more finicky models so I'm not surprised it might give a bad impression.

[glend (Glen)]

It comes down to the economics of manufacturing, and where the bulk of sensors are sold - into the everyday photographic market. The phase out of CCD production was a business decision.
You can rant all day long about it but it is not going to change anything. Sure there may always be a cult market segment that is willing to pay significantly more for limited numbers of CCD but they are no longer mainstream.
As far as picking easy targets with your complaints, the 294 does have cooling issues that result in banding when in long sub cooled used , this is the result of back lit architecture preventing full contact surface cooling by an adequately sized TEC. However, when used without cooling, in planetary, high frame rate mode, or for EAA use, it has little competition. It is in my view a niche camera, and I own the version without cooling, in fact the only astro camera I own.
Good luck with your crusade, but the market has spoken.

[Startrek (Martin)]

Greg,
Thanks for posting as this topic has come up at quite a few forums and Astro sites
My ZWOASI2600MC purchased last August is a beautiful camera and a quantum leap in CMOS technology from only a few years ago. Reports about the 2600MM are even better.
CMOS can only improve so I can’t image how good they will be in 3, 5 or even 10 years time
Never used a CCD but the images presented in various sites in the public domain from these ZWO and QHY next generation cameras is truly spectacular.
There’s no real reason to stop using your CCD’s unless they fail or are inadvertently damaged
Great post
Martin

[xthestreams (Paul)]

I’ve been watching and involved in a thread with Chris Anderson and Jon Rista amongst others picking their brains on just this topic. My concern post-OnSemi is that we’re forced to think seriously about what next?

FLI will tell you that CCDs degrade with age, eventually those column deflects won’t calibrate out so well and the chip is all but dead.

Verdict isn’t quite there yet one way or the other, but a critical perspective from the “experts” is that smaller pixels aren’t really an issue provided that their QE and full well depth per square micron is roughly the same or better than their CCD competitors, provided that’s the case there’s no real loss of dynamic range and quite possibly a lot to be gained from shorter subs.
The walking/rain noise is a concern for me, I see it on my 1600 a lot.
The 2600MC is amazing, I wish I had waited for the MM though!
Having said all of the above, a 16803 or 0900 would be a welcome edition any time.

[gregbradley]

Certainly with mirrorless cameras the high megapixel smaller pixelled sensors have performed beyond expectations. The traditional small pixel = noise has been bypassed to some degree by the electronics experts.
Specifically the BSI (backside illuminated) type sensors can have much larger wells than their pixel size would normally dictate.

Using a 183mm I found I could get away with the 2.3 micron pixels in reasonable to good seeing but if the seeing was poor then it was a waste with very soft images. Softer than the larger pixelled cameras would get.

Greg.

Originally Posted by xthestreams

I’ve been watching and involved in a thread with Chris Anderson and Jon Rista amongst others picking their brains on just this topic. My concern post-OnSemi is that we’re forced to think seriously about what next?

FLI will tell you that CCDs degrade with age, eventually those column deflects won’t calibrate out so well and the chip is all but dead.

Verdict isn’t quite there yet one way or the other, but a critical perspective from the “experts” is that smaller pixels aren’t really an issue provided that their QE and full well depth per square micron is roughly the same or better than their CCD competitors, provided that’s the case there’s no real loss of dynamic range and quite possibly a lot to be gained from shorter subs.
The walking/rain noise is a concern for me, I see it on my 1600 a lot.
The 2600MC is amazing, I wish I had waited for the MM though!
Having said all of the above, a 16803 or 0900 would be a welcome edition any time.

[gregbradley]

Certainly with mirrorless cameras the high megapixel smaller pixelled sensors have performed beyond expectations. The traditional small pixel = noise has been bypassed to some degree by the electronics experts.
Specifically the BSI (backside illuminated) type sensors can have much larger wells than their pixel size would normally dictate.

Using a 183mm I found I could get away with the 2.3 micron pixels in reasonable to good seeing but if the seeing was poor then it was a waste with very soft images. Softer than the larger pixelled cameras would get.

So yes you can get away with it up to a point but they are more sensitive to the seeing.

Greg.

Originally Posted by xthestreams

I’ve been watching and involved in a thread with Chris Anderson and Jon Rista amongst others picking their brains on just this topic. My concern post-OnSemi is that we’re forced to think seriously about what next?

FLI will tell you that CCDs degrade with age, eventually those column deflects won’t calibrate out so well and the chip is all but dead.

Verdict isn’t quite there yet one way or the other, but a critical perspective from the “experts” is that smaller pixels aren’t really an issue provided that their QE and full well depth per square micron is roughly the same or better than their CCD competitors, provided that’s the case there’s no real loss of dynamic range and quite possibly a lot to be gained from shorter subs.
The walking/rain noise is a concern for me, I see it on my 1600 a lot.
The 2600MC is amazing, I wish I had waited for the MM though!
Having said all of the above, a 16803 or 0900 would be a welcome edition any time.

[jwoody (Jeremy)]

Hello all
This is a great thread and quite timely for me. I am seriously considering the QHY268M to replace my QHY9M but I am just not sure.
The QHY9 is a very good camera, a proven track record, great cooling and large pixels especially as I mainly image at 2000mm.
How much better would the 268 be over the 9? If at all?
I can't decide yet.

Jeremy

[gregbradley]

Originally Posted by jwoody

Hello all
This is a great thread and quite timely for me. I am seriously considering the QHY268M to replace my QHY9M but I am just not sure.
The QHY9 is a very good camera, a proven track record, great cooling and large pixels especially as I mainly image at 2000mm.
How much better would the 268 be over the 9? If at all?
I can't decide yet.

Jeremy

Don't be put off by my frank assessment. The QHY268/ASI2600 is a late model,16 bit, no amp glow, high QE, low read noise and relatively large sensor at APSc which suits many scopes and is less fussy with tilt. I was more referring to the fact that these cameras are on a developmental path and not fully mature yet. This model though is one of the closest to maturity.

As to QHY versus ZWO best to read Cloudy Night assessments on this.
QHY has the readout modes, ZWO has a built in hub. Similar pricing.

I would go with the 268 over the QHY9 as it is higher res and low noise.

What settings to use is one area you'll have to study up on but there are plenty of users on the Cloudy Nights forums.

Sony has a large pixel sensitive sensor for the A7S111 camera. If QHY and ZWO get ahold of that one it may be a great camera for long focal length scopes.

Greg.

[benklerk]

I'll add a quick response to this.


I have a CCD camera which uses the KAF-16803 chip on my large 14" RC. I get really nice smooth images out of it.



I also have a ZWO 1600mm, its a great little camera for my wide field.


At the moment from what I have seen. CMOS isn't up for large long focal length telescopes yet. Unless you want to sell your kidney to get a Gsense 6060 CMOS which has large enough pixels and Full well to match a CCD.



So for short FL telescopes CMOS is better is most areas. Once you start to get into long focal lengths CCD is still better over CMOS. Using Bin1 on CCD instead of bin 3 for CMOS gives you much better resolution.