Peltier Canon vs Dedicated AstroCCD

[MarkS]

I've refined my thinking on this and I now have a very controversial viewpoint on the DSLR vs dedicated CCD issue which I'm prepared to defend with some interesting maths (which I won't introduce here).

I reckon my forthcoming Peltier cooled Canon 350D (total cost £150) would give the QS583 a run for its money in many (if not most) imaging situations.  My reasoning is that the main limiting factors are:
1) Dark Current
2) Read Noise
3) Quantum Efficiency & Pixel size
4) Well depth
5) Sky background noise
6) Bayer matrix

Addressing these in turn:
1) The dark current of the Kodak chip is much higher than that of the Canon but once they are both cooled sufficiently this is no longer a limiting factor for either CCD.
2) The read noise of the Kodak (8e RMS) is higher than the read noise of the Canon (5e RMS) but as long as exposure times of the subs are adjusted for this it no longer becomes a limiting factor for either CCD.
3) The Kodak QE peaks at 60% whereas the Canon QE peaks at 30% (I think 30% - I hope to have some accurate figures on this quite soon).  This ought to give a big advantage to the Kodak.  But the Kodak pixel size is 5.4um against the Canon's 6.4um which reduces the light collecting ability by 30% - this eliminates a substantial part of the QE advantage.  It means that pixel for pixel, the Kodak won't collect sufficiently more photons to make a great deal of difference to signal to noise ratio for that pixel.
4) Well depth of the Kodak is 25000e but the Canon's ADU limits it to approx 5500e (at ISO 800).  All this does is to reduce the dynamic range of the Canon - so for a mix of bright and faint objects, exposure bracketing becomes necessary.
5) At the end of the day, sky background noise is the great leveller.  Unless you go to deepest Wales, deepest Scotland, some parts of France or some parts of Spain then light pollution will definitely be the main limit on the faintness of objects you can successfully image.  When using Luminance or R or G or B filters then either camera will receive the same ratio of signal photons to light pollution photons.
6) Bayer matrix - OK I admit it, this is the Achille's heel of the Canon.  This might make a big difference but only when using narrowband filters (because only 1 pixel in 4 is receiving any light).   In other situations I think its main effect is to reduce final image resolution slightly.  For example, in a 3 hour LRGB imaging session where the QS583 is using a combination of filters, the Canon gets a solid 3 hours of R, a solid 3 hours of G and a solid 3 hours of B.

My main conclusion is that where sky background noise is the main limiting factor (i.e. all the places I've ever imaged - I've never imaged in any really dark place) then there won't be a huge difference between the images produced by the two cameras for the same total imaging time.

Take a concrete example - my recent M31 image.  My calculations show that on that particular night, light pollution was by far the main limiting factor in resolving the faint outer regions.  I don't believe that a QS583 on the same ED80 scope would produce a much better image.  If we go to Riberac then dark current becomes the Canon's main problem (except on the coldest of nights).  But with Peltier cooling, light pollution again becomes the main limiting factor.  Under those conditions the Peltier Canon would produce a significantly improved image of M31 than the one I took last weekend (in terms of bringing out the faint areas against the background noise) but the QS853 on the same night on the same scope would not be significantly better.

But if we go to a truly dark site?  Then all bets are off!  Dark current would then become the main limiting factor and the far superior QS583 cooling would give it a huge advantage.

Let the arguments commence!

Mark

[Rocket Pooch]

noone is going to take my QSI out to parties and get it scratched unlike my canon...

[RobertM]

noone is going to take my QSI out to parties and get it scratched unlike my canon...

Well you can only be sure of that if the QSI either wears a chastity belt or is kept locked in a safe, the Canon on the other hand isn't fussy where you take it 

[Daniel]

I was wondering how astro camera's compare against dslr's in relation to ISO gain, I've been recently talking to Greg Parker on his site and he tells me he has quite bad light polution too but can do 10 minute subs with his hyperstar.

Even at ISO160 I can only go about 4 minutes maximum, does this mean that ISO 160 on a DSLR has more gain than an MX25C or could it be that the light polution in new forest is actually not that bad at all?

In any case, i'd love to peltier cool my 40D, but im not sure it would be possible to keep the camera size down enough to use on the hyperstar.

Oh, incidentally im VERY interested in your findings on this one as im looking to possibly buy a QSI 583 around January time.

Daniel
:O)

[MarkS]

Daniel,

Take a look at this map http://avex.org.free.fr/cartes-pl/france/visuel/pl/  (you can zoom in interactively) and you will see that you can't escape light pollution in the New Forest.  But your location will be an order of magnitude worse!

For example, here is a table of light pollution figures I have collected using the same camera (modified Canon EOS 350D) on the same scope.

They represent total flux (R+G+B) of the background light pollution measured in electrons for one pixel with a 5 minute exposure at F6.3 ISO800

Sidcup                5600
Sidcup (CLS filter)  650
Lydd                    300
Tuesnoad             280
Kelling Heath         150
Riberac                  80

But each pixel only receives red or green or blue.  So lets divide the above figures by 3 to get an average flux per pixel (ignoring the fact that red light predominates in light pollution)

Sidcup                1900
Sidcup (CLS filter)  220
Lydd                    100
Tuesnoad               95
Kelling Heath           50
Riberac                   30



Also of interest are equivalent figures for dark current at various ambient air temperatures - again measured in electrons for one pixel with the same 5 minute exposure at ISO 800:

22C  150
16C   80
10C   40
4C    20
-2C   10

So you can see that in a darkish place (Riberac) on a warmish night (16C) the dark current (80 electrons) is worse than the light pollution (30 electrons).

Now consider the following.  Look at the huge difference the CLS filter made in Sidcup.  I want to try using it at Lydd or Riberac to see if it has the same dramatic effect.  If it does, then the dark current will completely swamp the light pollution.  This is the key reason I want to try Peltier cooling.

Narrowband filtering reduces background light pollution by another order of magnitude - this is why dedicated astro-cameras put so much effort into cooling.

As you can see, I'm putting a lot of effort into understanding this.  I'm determined to push a DSLR as far as it can go - the maths tells me that the photos I'm producing so far are nowhere near that limit!

Mark

Later Edit:  In the above figures, ignore the fact that they were taken at ISO 800 - since I'm actually counting electrons the ISO makes no difference whatsoever.

[MarkS]

im looking to possibly buy a QSI 583

Let me put the other point of view.
For narrowband imaging, the QSI 583 will always beat the DSLR because:
1) The DSLR is hampered by the Bayer Matrix - just 1 in 4 pixels receive any light
2) The dark current now becomes the main limitation and the QSI has much better cooling

My previous arguments were really based on LRGB imaging. Given your location, narrowband is probably the only way to deal with light pollution so the QSI would be a good choice.

Mark

[RobertM]

Mark,

Just a few observations regarding your initial points:

1) Yes that is true but a cooled CCD can operate at a set temperature all year around which means only one set of darks.  This saves a potentially huge amount of time as temperatures vary enormously through the night - advantage CCD.

2) Agreed, exposure times can mitigate read noise.

3) It is true about the quantum efficiency and read noise, however that's also an advantage of a CCD as in camera binning can make the pixels 10.8, 16.2 microns or more in size - advantage CCD.

4) Well depth - means longer exposures are possible before saturation so much better for low contrast targets such a nebulae - advantage CCD.

5) Yes agreed - at the sensor surface.

6) Bayer Matrix - yes this makes things somewhat one sided.  If it weren't for LP then a cooled canon would indeed make more sense but we do have LP so NB imaging was born and CCD sensors make best use of that signal for reasons you have already explained.  In itself that is a massive bonus around where we live, at a dark site where we can do broadband imaging then the advantage is much less.  Then there is the resolution aspect.  With high quality low f/l optics a telescope can focus a single faint star onto a pixel.  So for the DSLR you have four to choose from and maybe a bit of each colour if you're lucky.  To have a chance of balancing colour the star has to cover four photosites which means dithering a lot of exposures, either that or defocus and or a bit of errant guiding.  As for getting a solid 3 hrs of each, that is technically correct though in fact you will get twice as much green signal as each of red or blue because of the nature of the colour matrix.

Don't get me wrong I think DSLRs are great for imaging, their portability for grab and go shots is unrivaled.  They also have a massive cost advantage, huge sensor size and of course you are getting all three colours in one hit.   To cool them removes a little of that flexibility but I'm still going to follow your experiments with keen interest as it dovetails into my plans nicely and is a whole lot cheaper than the nearest commercial alternative.

In three years time the DSLR may well be almost usable without cooling but on the other hand CMOS mono sensors may well have gone over the 90% efficiency mark so it's still going to be horse for courses.

btw. Very interesting figures you've quoted on background flux.  Though a lot depends on the conditions and time of year it's still gives a reasonable indication.

Daniel sorry whilst composing this you replied...

Robert

[Rocket Pooch]

don't forget the 12bit camera against the 16 or 32 bit astronomy camera and the contrast ratio and the -50C cooling and i think you have to divide the qe of the slr per colour to get a comparison per pixel as well

[MarkS]

don't forget ... the -50C cooling

The Kodak KAF 8300 chip needs 20 degrees C of cooling just to reduce its dark current down to Canon levels.

[RobertM]

You're not far off there Mark but I can operate my camera at -30C all year around as long as the temperature stays below +30C.  That means it's dark current is negligable and can for most practical purposes be ignored.  Of course in the winter I'll have the option of setting it to 60C below ambient but I don't anticipate that should ever be necessary.  The downside, as you're finding out, is that the cooling draws copious amounts of power....

Robert