So, getting back to the original post. Yep, it is pretty obvious given the same flux, a dedicated medium QE CCD astrocam shows higher dynamic range, better QE, lower noise etc than on off the shelf Premium DSLR
Hello Peter
So what then was the purpose of the test other than comparing apples to oranges.
My reason for using the term subjective still holds as you obviously feel that comparing a CCD to a off the shelf DSLR is valid.
The test would be objective if you leveled the playing field.
For example you say you optimised the shot noise for both cameras by applying a software algorithm. Correct if i am wrong but shot noise or Poisson noise relates to the random variation in the incoming photon flux. That was on a par for both cameras to start with.
If you discount the IR cut off filter in the dlsr, the different QE's, the non linearities of QE/wavelength, the differences in gain or conversion factor, the differences in dark current, the differences in readout noise, the differences in bias current and then present 2 pictures one showing more red than the other then i fail to see objectivity. Your test was subject to vast differences in camera specification and then to illustrate something that should be obvious.
The high ISO or gain on the DLSR also adds extra noise and narrows the dynamic range of the DLSR to say nothing of the 16bit ADC vs the 12bit ADCf the DLSR.
So imo apples and oranges, hyundais and SLK,CCD and CMOS makes for an unfair and biased comparison which is not objective.
What the test illustrated for me is that the canon 5d produces a good image although a few tweaks would certainly narrow the gap substantially.
The current range of CCD's are certainly superior but this may not always be the case given that the CCD technology is in effect old technology and the CMOS technology is newer with a better future potential.
Of course eventually all image sensors will go the way of Moore's law in computers, they will get close to 100% QE and then thats it. They will never get any better.
Sadly canon doesnt make them, but Id love to see how one of their DSLR sensors would go if its bayer filter array was removed, making it a pure monochrome sensor.
I believe Astronomik will soon release a "multi emission line" filter, essentially a narrowband filter with all three emission lines, thus allowing a DSLR or CCD single shot colour camera to do true colour narrowband imaging where the red, green and blue pixels will get light.
A clever idia that never really took off was the "foveon" sensor where each pixel recorded all colours. Photons would penetrate to a certain depth according to colour and the sensor differentiated that. So far however the sensor doesnt seem any better, and isnt that widely used.
DSLRs have one advantage, economies of scale. Thousands of DSLR CMOS sensors come off the production line for each Kodak CCD sensor. DSLR companies can thus spend a heap more on R&D than Kodak can on its CCDs
Personally, if I was SBIG Id approach Canon as to making a run of monochrome 11 MP CMOS sensors, with 16 bit ADC, single stage peltier cooling, then proudly sell them for < $3000USD and still make a profit
As to the point, well there are several.
CCD's
Higher QE, lower noise, less noise structure, better colour saturation, higher dynamic range.
CMOS.
Low H-alpha sensitivity unless modified, Easier use, requires external guider.
I just read this page and most of the linked articles, and I am wondering how the astronomy market will fare down the track with regards to image sensors. Will it be like the current LCD V plasma battle with tv's?
Neither of the technologies are new, dating back to the late 60's / early 70's, but it wasn't until recently that CMOS sensors became viable because of reduced lithography sizes (the same reason why computer chips are smaller/more powerful/faster).
Some interesting facts:
* CCD will always have the signal to noise battle won simply by the fact that CMOS chips have more digital circuitry on the chip, which adds noise.
* For responsivity (the amount of signal the sensor delivers per unit of optical energy) CMOS chips have the advantage because it is easier to place the gain elements on the chip. I don't think this is the same as QE, and theoretically I don't see any difference in potential QE between the two technologies.
* For Dynamic range, CCD's win again. I guess integrating the circuitry to do a charge to voltage conversion for every single pixel takes up space where CCD's can utilise this elsewhere?
* Windowing - this is an interesting one, and I wonder of it is patented yet? For a CMOS chip it wouldn't be too much effort to designate different framerates and functions to different parts of the sensor. For example, I was thinking you could define your own guidecamera from any part of the chip, and you could do this every time you slew to an object to image it. Take an exposure, use a 25x25 pixel array centered on a star towards one of the edges, and that can operate at 2 fps for guiding while the other pixels operate to capture yout image. I don't think you can do this with CCD because you have to transfer a line at a time - with a CCD you need a seperate chip to guide with.
*Antiblooming - CMOS wins here as it can drain a pixel without affecting anything around it. It can be implemented onto CCD though, so I don't know if this is really an advantage.
The biggest impact in my mind, however, is market impact. Astronomy is a tiny market - the big market is from security cameras/phones/digital cameras etc. With the ability to integrate more functions onto the chip itself, I see CMOS winning more market space in the future, which could potentially isolate CCD to the specialist market, making it a more expensive technology.
Anyway, I don't know if this analysis is too off topic, but the research was interesting for me at least
Your test was subject to vast differences in camera specification and then to illustrate something that should be obvious.
Steve you raise many points, but the above is the gist of the matter, for me at least.
I did not build or design or modify either camera. They are just "as is" off the shelf models available to anyone. The images look different, because the devices are intrinsically different.
Nothing to discount or explain, just an illustration of how specifications make real world differences to an image.
I could also put a 50mm lens on the SBIG and try to take pictures at the Melbourne Cup..... something tells me in that application the result would not be pretty
Hi Peter and everyone else posting great replies on this thread, I know I am off topic a bit here regarding QE but I have very limited knowledge on these issue's you all are talking about but can you tell me how the mallincam colour video camera can capture M57 in the daytime and awesome views of objects such as M42 in 2 second exposures in colour. I am just curiuos as how this is possible. Is it a special CCD chip. I know it is a video system so it is not designed for proper imaging such as the fantastic results you display Peter but I ordered one to use with my young family to have fun with in our light polluted backyard in Brissy.
Keen to here back from your expert opinions on the way these things work, thanks again.
Interesting tests and discussion.
Lets hope Peter does more.
The SBIG with canon chips at 3000$ sounds like the ticket.
Getting to QE=100% may be close to if not impossible but then never say never.
So many cameras so many stars so little time
Regards to All
Steve
Hi Peter and everyone else posting great replies on this thread, I know I am off topic a bit here regarding QE but I have very limited knowledge on these issue's you all are talking about but can you tell me how the mallincam colour video camera can capture M57 in the daytime and awesome views of objects such as M42 in 2 second exposures in colour. I am just curiuos as how this is possible. Is it a special CCD chip. I know it is a video system so it is not designed for proper imaging such as the fantastic results you display Peter but I ordered one to use with my young family to have fun with in our light polluted backyard in Brissy.
Keen to here back from your expert opinions on the way these things work, thanks again.
Matt.
The CCD chip would most likely be a common CCD chip from one of the big name CCD chip manufacturers. What differentiates the mallnicam from other cameras (according to their website) is "Unique Custom Manufactured High Gain Circuitry Not Found on Any Competitor Video CCD Camera", which when combined with a high QE CCD and peltier cooling, allow alot of data to be acquired with relatively short exposure times.
In simple terms, I guess it is like taking the image and doing some image processing (similar to a stretch of the curves in photoshop to bring out detail) except it is doing it on a hardware level.
Hi Coldspace
From what i understand the mallincam uses a much higher gain than normal.
The exposure times are also very limited. I think about 12 seconds/frame max.
This means that you can collect lots of photons in a short time and then use software to reduce the noise.Operating at high gain means lots of noise as well as lots of signal.
Its almost like a Dolby noise reduction system for photons. Then you are taking lots of images and stacking them to improve the signal to noise ratio.
The quality of the image looks good but certainly not comparable to the quality you would obtain with dedicated CCD and CMOS still cameras.
Remeber that you view these images on a video monitor which does not have a very high resolution. The good old mark 1 eyeball and human brain does a lot of averaging and correctional computation for you.
Hope this helps.
Either way a good way to start of with a expensive and challenging hobby.
Once you are bitten you will always want more disposable income to upgrade your equipment.
Regards
Steve
Interesting arguments here, but I strongly suspect that in time they'll all be consigned to the dustbin of history as a new generation of smart sensor is developed, one that can mimic the remarkable responsiveness the human eye. For instance, I can walk outside tonight and see the nearly full moon and all its dark maria showing, I can see bright stars near it, I can see the moonlit sky and tall trees silhouetted against it. But I can't photograph it with a dumb sensor, CCD or CMOS.
Combine this with a smart telescope that knows where it is in the sky, rather than the present generation of dumb scopes that are told where they are in the sky (and may or may not be exactly there) and you'll have the brave new world of astrophotography.
Thanks for the replies Steve and Andrew, clears thing up a little bit for me. Just starting on the road of this great hobby. Anyway back to the thread that I interupted.
Hi Rob
Cameras have been around for about 150-200 years.
Even the latest doesnt come close to the elegance of the eye and the brain combined.
I think that by the time technology can reproduce that then we will have no need for telescopes or cameras as we would have evolved to something else or wiped ourselves out.
Nice thought though and in the meantime we can still be pragmatic and play with some pretty cool stuff.
Regards
Steve
Rob, there's some interesting comparisons of the Human eye to technology on the web, here is one for example. The human eye is equivalent to a 576 megapixel ISO 800 detector, but it's main advantage over film or digital sensors is it's dynamic range, which is why you can see so much more at night than you can capture with a camera
The U47 chip saturated at 30 seconds (!) , and the DeBloomer plug-in was used to remove the worst of it.
With the back illuminated device there was so much signal that it was hard to pick a black point and noise be came trivial. There is also evidence of scattering/flare in the U47 star images due too much signal, hence 2x 15 second exposures probably would have been better.
Retrospectively, either comparison, while amusing, does not say a lot about CMOS vs CCD but does give good insight to "off the shelf" implementations and how that effects the performance of these detectors.
Interesting comparison, one that would be expected with the U47 chip's high QE here page 4.
I still think if I was Apogee or SBIG I'd approach Canon with a view of ordering monochrome CMOS sensors.
This is interesting. Now change the unmodded to modded, the red peak would go up to about 40%, and the others would gain a few % as well. It would then be on par with the SBIG model, but at a fraction of the price.
I really do think that if SBIG, FLI, APOGEE and others arent prepared to investigate the use of the rapidly maturing CMOS technology, they risk loosing more and more of the astro imaging market to DSLR's. I will bet money the first manufacturer to start using large frame cooled 14 to 16 bit CMOS monochrome chips will cornet the astro imager market.
My Q guider has a QE or around 50% (mono CMOS chip 1280 x 1024), but is only 10 bit and not cooled. if it was 16 bit and cooled, it would make a lovely entry level dedicated astro imager. As it is it cost me $300. The currect chip cannot become 16 bit as it has an on chip 10 bit ADC, but if a 16 bit model of the same chip was available, all one would need to add is the peltier cooler and control circuitry. About a thousand dollars would pull it up I think. Add a 2nd 10 bit chip for autoguiding and a fully self guided 1280 x 1024 astro camera with 50% QE would be ours for the buying. Sadly none exists.
IMHO, it is like this:
DLSR sensor (having Bayer) has ~1/3 available area for each color. Hence, only 1/3 of photons detected, hence 1/3 of sensitivity.
Also, chip temperature is HUGE factor here: at -15°C the noise floor of CCD sensor is almost negligent.
If you take all these correction factors into account, plus additional reasons (H-alpha sensitivity, noise in amplifiers etc) the results are actually very close.
After all, there are only so many photons to capture...
Hello Peter
So what then was the purpose of the test other than comparing apples to oranges.
Yes i was also wondering why you are doing these comparisons between DSLRs and CCD Cameras ?
What are you trying to prove ?
Peter if you are only trying to highlight that a $10,000 purpose built astronomical CCD camera returns better images than an off the shelf $1,200 Digital SLR camera built for everyday photography, then I think most of us already know that the CCD camera is likely to produce "better" images.
The reality is (IMHO) that not everyone wants too or can afford to spend that much money and so as an alternative we can use something almost 10x cheaper - a DSLR.
Peter if I didn't know better I would think you are a CCD Camera Dealer and what you are doing here on IIS is sales work ?
regards
Paul
Last edited by PhotonCollector; 20-02-2008 at 03:04 PM.
Reason: spelling
If the Canon was exposed for 90 seconds, i.e. the time it took to get the RG&B channels out of a Mono camera, then it would have performed better, but that was not the point of the exercise.
I was curious to see how the two cameras would deal with the same amount of (limited) flux, and was admittedly surprised by the obvious differences in signal and noise between the two.
OK these are taken with my ed80 and a cooled modded canon 350D, please note the ambient air temp was 23 degrees I had the camera cooled to 5 degrees. these are unprocessed suburban images, so there are no darks, no bias's, no flats (and badly polar aligned)
Peter if I didn't know better I would think you are a CCD Camera Dealer and what you are doing here on IIS is sales work ?
regards
Paul
Nice try
Last time I checked my commercial associations ended with Apogee Instruments over 5 years ago. As to why... curiosity. Nothing more, plus a little bit of fun, until I can take some serious images.... that's if Sydney ever gets some reasonable weather...
Houghy, gee...for a change I can say they look a little too red to me
Quote:
Originally Posted by h0ughy
OK these are taken with my ed80 and a cooled modded canon 350D, please note the ambient air temp was 23 degrees I had the camera cooled to 5 degrees. these are unprocessed suburban images, so there are no darks, no bias's, no flats (and badly polar aligned)
But, O.K. I'm a convert....may have to get my 350D modified as well!
Houghy, gee...for a change I can say they look a little too red to me
But, O.K. I'm a convert....may have to get my 350D modified as well!
well Peter, unless i am mistaken i did say that they were unprocessed images, and the camera is modded to let in all the hydrogen light it can, hence the red tinge.
but i digress - you were comparing a vanilla flavoured dslr with a purpose built dealer approved st11 (which you sell). lets see my modded camera cost me all up 1200 for the twin lens kit then 1480 to get it doctored, so thats $2680. the st 11 is what 8800 plus filter wheel plus filters so can we say a tad over 10000. so thats a 70% cost difference, but then i suppose its bang for bucks.
Curiously the rgb the st11 took were 30 seconds each, so thats 90 seconds, yet you only took a 30second shot with the dslr? my maths was never my strong point
