Is CMOS (rather than CCD) the future?

[gorann]

With great interest I read reviews of the new Sony A7S super sensitive DSLR, and followed the subsequent disussion on one of our fora and it made me thinking about the next step from my Canon EOS 60Da. I wonder if I should wait a year or two  for CMOS to finally take over from CCD?

I get the impression that the CCD era might soon be over. Kodak has given up producing CCD chips and much more effort and money appear to be put into CMOS development than into CCD development. CCD is the old technique and CMOS is the new, and initially CMOS was substandard but apparently no longer. For astrophotographers, I wonder if the problem at the moment is that no dedicated astrophotography company has got hold of the latest large low-noise CMOS chips (like the big pixel one in Sony A7S) to adapt them to astrophotography and even sell them in monochrome versions. From reading the reviews of the Sony A7S chip I get the feeling that a monochrome version of that chip would be the ultimate AP camera. The CMOS inherent low noise should put less demand on cooling to get a good S/N ratio (maybe no cooling at all) and I assume that prices would be quite a bit lower than for comparable large CCD AP cameras.

As a novice in AP I of course wonder if I am all wrong in my analysis of the situation?

[Tinker1947]

You going to waste 2 years wondering, buy second hand and sell later for more or less the cost of the postage.....

[ollypenrice]

One thing is for sure: there is nothing but disadvantage in having a fixed colour matrix in front of a chip for serious DS imaging. Luminance is 3x more efficient than RGB for global data collection and NB filtering is vastly better for finding structural emission contrasts. New chips, yes, great news - but please, monochrome. What are they like on well depth because this also matters, though playing with different sub lengths can mitigate this issue.

Olly

[perfrej]

The well depth is far more interesting than people think, and the reason is gradients. The smallest unit of light that can be detected is one electron, and most cameras deliver a 16 bit unsigned integer result, which in turn means light values of 0-65535. The best match for this number range is a well depth of 65535 electrons, obviously, because then the camera can paint a black to white gradient very well.

Now, the stuff we want to image is usually hidden deep down in the signal range. In essence, we are not using the dynamic range of the camera, so let's assume that the background ends up at 1000 ADUs and the nebulosity we want in the range of 1000-21000, which means 20000 different levels of gray in our target. The rest of the dynamic range available is used to not bloat the stars completely, so that's the reason for just 20000 for the target.

A camera with a well depth of 32768 (half of the integer range used to describe the pixel values) multiplies the values read from the chip by two in order to fill the range. At this point our 20000-range target is a 10000 range target, so we have introduced a step-gradient in our image.

This behaviour is why a large pixel camera (larger pixels -> larger well depth) produces a better image than a small pixel camera; it has a higher dynamic range.

Sony ICX814 has a well depth of 14000 electrons. Our target will have 4200 grays with this camera.

Kodak KAF-8300 has a well depth of 25500. Our target will have 8200 grays with this camera.

Kodak KAF-16803 has a well depth of 100000. Our target will have 20000 grays with this camera (note that the sensor is actually dynamically cropped by the 16 bit number used to describe the image pixels).

Sony A7s has a pixel size of 8.4u, which should put it at very high well depth, most likely at or above the integer limit of 65536 levels of gray.

So, is this really an issue? Well, it certainly is. At some point when pulling the faint signal out of the dark recesses of the background, gradient banding will occur. This is the same thing that happens when you edit your images in 8-bit color. Don't! Always edit with at least 16 bits per colour channel, or 16 bits of gray for monochrome.

Nothing like a morning rant

/per

[Horwig]

Well said Per, and good morning to you 

Huw

[perfrej]

Haha! Yes, thank you, and good morning!

/per

[gorann]

Thanks for interesting input on this topic!

If we are expecting a CMOS breakthrough sometime soon, it may be risky to buy a monochrome CCD now and expect to get a good price for it later. Who would buy my CCD camera if, for example, a monochrome version of a CMOS chip like the one in the Sony A7s becomes available, with lower noise, more well depth and lower price than the CCD I just bought? This is exactly why I hope someone has an idea about if and when such a camera will appear.

[Bubbles82]

I would love a big chip dedicated astro cam in the future. But the manufacturer passing on any saving by using cmos tech to the customer rather than putting it in there pockets is never going to happen especially with astro gear [emoji3]

Sent from my iPad using Tapatalk

[perfrej]

Well, it isn't quite that simple. 8.4u is a BIG pixel. With my TEC-140 I would get 1.77"/px which is a bit of a waste of resolution. Around 5u is better, so 29050 or 50100 sensors... Nothing like that in CMOS...

/p

[pete_l]

So, is this really an issue? Well, it certainly is. At some point when pulling the faint signal out of the dark recesses of the background, gradient banding will occur. This is the same thing that happens when you edit your images in 8-bit color. Don't! Always edit with at least 16 bits per colour channel, or 16 bits of gray for monochrome.

This is similar to a problem that audio engineers had when digital audio didn't have a very high dynamic range. When they were fading a digital recording, the sound level would jump from level to level. This was noticeable in quiet passages and was not the effect they wanted.

The solution was to add noise to the signal .

I don't recall the details as I read about this a long, long time ago. But I believe the added noise was of the order of 1 bit (makes sense!) and this made low-level fades sound smoother.

[dph1nm]

The solution was to add noise to the signal

Fortunately the astro world comes with noise ready built in!

NigelM

[Physopto]

Apparently Sony are going to stop producing CCD chips in 2017!

What then?

Derek

[Xplode]

I think there will be a lot happening to CMOS technology the next couple of years.

Sony has just announced the A7R II with the worlds first full frame back illuminated CMOS sensor, back illuminated sensors can have QE of over 90% compared to around 60% from the best front illuminated CMOS sensors.

The new Sony sensor is 42mpix and pixel size of around 4.5u.

Canon also seems to be working on back illuminated sensors and there's also rumours about a full frame Canon DSLR made for astrophotography, i hope they will go for a 6Da/6Da II instead of a overpriced camera like the Nikon D810a which is just a waste of features for astrophotography.

If companies producing astrocameras can get access to these newer back illuminated CMOS sensors there might finally be some  "affordable" large sensor cameras with new technology 

[Physopto]

I think there will be a lot happening to CMOS technology the next couple of years.

Sony has just announced the A7R II with the worlds first full frame back illuminated CMOS sensor, back illuminated sensors can have QE of over 90% compared to around 60% from the best front illuminated CMOS sensors.

The new Sony sensor is 42mpix and pixel size of around 4.5u.

Canon also seems to be working on back illuminated sensors and there's also rumours about a full frame Canon DSLR made for astrophotography, i hope they will go for a 6Da/6Da II instead of a overpriced camera like the Nikon D810a which is just a waste of features for astrophotography.

If companies producing astrocameras can get access to these newer back illuminated CMOS sensors there might finally be some  "affordable" large sensor cameras with new technology 

What is the well capacity of these sensors?

Derek

[Xplode]

Well capcacity is impossible to know before someone can get their hands on the camera and do the measurements.

Sony haven't even released any specs on the sensor so i had to calculate the aproximate pixel size

[Physopto]

I suppose I was getting a bit ahead. The future looks interesting as always. It doesn't matter what you do there is always something better in the wings. You can never keep up!

Derek

[Merlin66]

Yeah but........

The commercial camera requirements will always be for one shot colour chip solutions.....it's industry which MAY drive the needs towards a mono chip - certainly not the astronomy community.

[gorann]

When Sony and Canon produce these new CMOS chips, I would assume it would not be too much of a trouble for them to also make versions without the Bayer masks, but maybe it is more complicated than that?

[Zakalwe]

When Sony and Canon produce these new CMOS chips, I would assume it would not be too much of a trouble for them to also make versions without the Bayer masks, but maybe it is more complicated than that?

It'll be about economies of scale.

I'd love to see a mono version of this sensor with a Peltier cooler.

http://www.eoshd.com/2015/06/first-sony-a7r-ii-user-experiences-global-shutter-and-native-iso-800/

Back illuminated, 960fps (!!!) at 1136 x 384, native ISO800, full frame.

[davew]

While every one has been watching Sony and Canon the first Mono CMOS chip is already out ! It's in a Leica camera and is being produced by CMOSIS I believe. All 24 mb of it in  35 mm full frame format. I heard the pixel size is 6 um.

Back illuminated chips, as mentioned previously, have been out for six years to my knowledge. That means they've probably been around longer ! The last I actually read about in thinning was 65 nm by Samsung. Probably been beaten by now.

Lots more going on than just the big two

CMOS may well take over but I wouldn't put off buying a CCD camera, " Just in case "

Dave.

[gorann]

Thanks for the interesting info about the monochrome Leica. I see that this 24 Mpix CMOS camera (called Typ 246)  is replacing a monochrome Leica that came out in 2012 with a 18 Mpix CCD chip. So, also Leica is moving towards CMOS. What is quite sad is that they are asking 7500 USD for the camera, so it is very unlikely that it will be much used in astrophotography (at least not by me).

Here is a review of it:

http://www.imaging-resource.com/PRODS/leica-m-mono-typ-246/leica-m-mono-typ-246A.HTM

[davew]

I wouldn't suggest you rush out and buy one Goran

It's the technology that matters and this does at least show that Mono CMOS sensors are alive and working well in daylight at least. I don't imagine Leica will return to CCD in either mono or colour. They have a store of colour CCD sensors to replace the older ones in the M9. ( They break )

Just for reference, as they aren't yet available, Leica are going to offer colour filters. I will be sat down before I look up the price.

Dave.

[Big Jim Slade]

Red (who make digital cinema cameras) have been offering mono versions for a few years, both for their older Mysterium-X (13mp) and the newer Dragon (19mp). But both of these are Super35, which is nearer APS-C, rather than full frame. And this is an even smaller market than Leica's so I doubt there's any knock on effect to be had. They do take lovely pictures though: https://www.youtube.com/watch?v=IsUsVbTj2AY  (Apologies if you have an aversion to Justin Timberlake!)

This thread is interesting, but I've been hoping someone who has a real understanding of CMOS and it's properties would chip in and tell us how it compares with CCD in the areas we are interested in (well depth etc.) Maybe the question we should be asking is: what are the pro's doing? Are the big telescopes using CMOS detectors?

[riklaunim]

CMOS is easier to design and it's simpler than CCD. More components may be put on the sensor die and so on. CCD is fading away just like CD and DVD

[Macavity]

I sense the technologies and (happily) the "communities" are converging! 

Classical imagers acknowledging the capabilities of video imaging etc. 

One slight obstacle is the difficulty of comparing cameras because

of the diversity of UNITS used to measure the light sensitivity etc. 

If only I could quickly convert Quantum Efficiency to Volts / Lux-sec...

It seems easier to buy the cameras. And a great marketing strategy?