Choosing my first dedicated astrocamera (upgrade from DSLR)

[StarMich]

I have a question about experiences with entry-level CCD / ColdMOS cameras and the selection process for upgrading to one.

I have been doing astrophotography for 9 years now, always worked with DSLR (uncooled, color, full spectrum mod). My main telescopes have a focal length of 812 mm (f/3.9) and 1000 mm (f/5).
Since about 5 years I started following the CCD world with the idea to eventually switch to a monochrome cooled camera for luminance, with the aim of more light sensitivity and higher resolution.
At that time my goal was to save for a camera with the well-known KAF8300 sensor, but in the meantime the CMOS sensors have started to compete more and more with the entry-level CCD cameras. Everywhere these days I see images popping up made with the well-known ASI1600M or QHY163M, the results usually speak for themselves and I had the idea to go for the ASI1600M Pro. I started some research into the specs of the sensor in these cameras (MN34230) and compaired these to the KAF8300, my current DSLR: a full spectrum Canon 1100D and other cameras in the price range ~ €2000 or lower. (< £1700)

After some calculations I started to have doubts about the choice for an ASI1600M, this was my reasoning:

Light output per pixel and consequently also the exposure time required to capture the signal from deepsky objects can be expressed (fast proxy) as the product of pixel surface and quantum efficiency. This is of course independent of thermal noise and read-out noise, but it seems to me that differences in this are not significant between different manufacturers that incorporate the same sensor into their cameras.

If there is anything wrong in my reasoning, feel free to correct me.  

Below the numbers I obtained:

• Canon 1100D: 9.1
• Canon 450D: 9.5 (is often "monofied" nowadays)
• ASI183M: 4.7
• ASI1600M / QHY163M: 8.5
• ASI174M / QHY174M: 26.1
• ASI294C: 16.0
• Atik 414EX Mono (ICX825 sensor): 22.0
• KAF8300: 16.8

 

The first thing that struck me here was the lower score of the ASI1600M compared to my current camera, of course I compare mono with color here but it makes me wonder if the ASI1600M is worth it in this price range (€300 1100D vs €1100 ASI1600M) uncooled)?

Second thing I noticed was that the ASI174M has a higher score than the Atik and it has more pixels.

And a third was that the KAF8300 is gradually being overtaken by a price range € 800 lower.

The high score of the ASI294C also makes this camera tempting, but I lose the resolution gain of the monochrome there.

The above reasoning makes me leaning towards a QHY174M (ASI is discontinued), resolution is then almost equal to the current one with the Canon 1100D and the image field is a lot smaller, but I do see a large potential light gain. The ASI1600M offers me the highest resolution possible for a decent sensitivity, but I still find the score quite low compared to the other cameras.

I would like to hear experiences about the above cameras / sensors and whether I would have forgotten another decisive factor in my numerical comparison.

Thanks in advance,
Michael

[Drakester]

Hi Michael,

Im in a similar boat to you, currently researching my first proper CCD/CMOS after predominantly using my Canon DSLR for the last few years. 

“Pixel scale” appears to be quite an important factor when considering matching CCD specs to specific scopes. 

There’s a useful calculator in the top section above under Resources> Astronomy tools> CCD suitabilty which may help. 

Hope this helps. 
 

 

 

[tooth_dr]

I’m afraid I can offer no theoretical reasoning, but I went from a canon 1000d to a KAF8300.  FOV is similar, pixel size similar, but image quality is very much superior.  I certainly won’t be going back to the Canon again. There was last nothing wrong with it, but set point cooling on a camera makes all the difference.

You mention the 1600 uncooled above, that would not a popular choice for deep sky imaging.

5 years is a long time to be deciding. I wouldn’t get too hung up on figures. You’ve seen the images online from these various cameras, that should help you decide, much more than comparing figures.  I have two 8300s and it is a noisy camera on paper, yet it calibrates well and produces excellent images.  With your scopes it would give a “/px of 1.11 and 1.37  which would be quite nice.

 

 

[vlaiv]

There is a lot more to consider when choosing suitable camera.

I will give you two examples where simple formula like pixel_surface*QE can go so wrong.

For ASI1600 you obtained figure of 8.5 - that presumably being 3.8 *  3.8 * 0.5= 7.22 (in my calculator, but maybe you used different QE figure there - I've found it to be around 50% in quotations on the net). Actual figure does not matter for what I'm about to say - why don't you repeat your calculation but this time take binned x2 pixel size? Instead of going for 3.8 * 3.8 * 0.5, go for 7.6 * 7.6 * 0.5 = 28.88!

Now we are talking, right? It's clearly the best of all on your list above.

But hold on, binning must be cheating right? No, it's not - it's legitimately used and indeed it has the same effect as using larger pixel size (for hardware binning, for software binning like with ASI1600 it acts as larger pixel size and larger read noise - by factor of 2 so it is 3.4e instead of 1.7e at unity gain - but like you said yourself - read noise is not as important because it can be sorted out with longer exposures up to a point).

Here you go - simple example that your metric of pixel_surface * QE does no hold much sense.

Let's go with another example.

Atik 414Ex - it has something like 9mm x 6.7mm or 11.22mm diagonal and "figure of merit" equal to 22

vs again

ASI1600 which has diagonal of 22.2mm (let's simplify to double that of atik - it's not quite but almost).

Let's now take some common type of telescope - like F/5 newtonian and compare how will two sensor fare if we decide to shoot same size FOV with each mounted on suitable newtonian. It is clear that scope paired with ASI1600 needs to have twice focal length of that paired with Atik to provide same FOV. Since both scopes are F/5 - one paired with ASI1600 will have twice aperture as well - or x4 light gathering surface.

Which sensor do you think is going to provide better image on same FOV in same time? I would guess - one collecting x4 the light.

[wimvb]

I wouldn't worry too much about those numbers, it's pure theory. What you should consider imo, are:

• Sensor size
• Pixel scale ("/pixel) 
• Read noise
• Dynamic range

I believe the 174mm was discontinued because ZWO had many models with small chips. They just cleaned their catalogue. 

[pete_l]

I agree with wimvb. It is easy to fall into the trap of decision paralysis based on the specifications of CCD/CMOS cameras. They tend to exaggerate the importance of small differences in the numbers. The underlying principles of all cameras is the same - with some advantages of CMOS (such as lower noise, the ability to "bin" in software and not lose much).

As for pixel size in general the "best" focal length for a given camera is the number of microns for a pixel multiplied by 200 (so a 5µ pixel is "best" with a 1000mm focal length). But the range over which it provides perfectly good and usually indistinguishable results is very, very, broad. I'd say from below 500mm focal length to over 2000mm - so the vast majority of imaging scopes.

I also would say that the technical difference between cameras becomes insignificant when compared to the results from manually processing the image and the skill and experience of the person doing that work. Unless you get a camera with a fault, almost any one will give you great results - once they are properly stacked and processed.

[wimvb]

You also have to consider software support. When I was scanning the market for an astro camera, I removed qhy from the list because there were too many reports from people who had problems with them under Linux. At the time, qhy's linux drivers were just not stable, and they didn't seem to have anybody working on it. This despite the fact that qhy cameras were better engineered than zwo. Qhy cameras had a window heater, amp glow elimination, and a memory buffer as standard, whereas zwo introduced heater and buffer only much later in their pro models. 

[vlaiv]

1 hour ago, pete_l said:

I agree with wimvb. It is easy to fall into the trap of decision paralysis based on the specifications of CCD/CMOS cameras. They tend to exaggerate the importance of small differences in the numbers. The underlying principles of all cameras is the same - with some advantages of CMOS (such as lower noise, the ability to "bin" in software and not lose much).

As for pixel size in general the "best" focal length for a given camera is the number of microns for a pixel multiplied by 200 (so a 5µ pixel is "best" with a 1000mm focal length). But the range over which it provides perfectly good and usually indistinguishable results is very, very, broad. I'd say from below 500mm focal length to over 2000mm - so the vast majority of imaging scopes.

I also would say that the technical difference between cameras becomes insignificant when compared to the results from manually processing the image and the skill and experience of the person doing that work. Unless you get a camera with a fault, almost any one will give you great results - once they are properly stacked and processed.

I think you are very much overestimating "best" focal length for a given pixel size. What is the criteria that you are using for this?

Sure you can use 5um pixel camera with 1000mm focal length for effective sampling resolution of 1"/px - but most of amateur setups are simply not capable of delivering that sort of resolution. One needs 1.6" FWHM stars in their subs to do that. Hence in most circumstances that will be oversampling.

Saying that you can equally use 2000mm of focal length with 5um pixel camera without mentioning how can you make it work is probably going to cause issues for most people that don't have deeper understanding of this topic.

I can imagine someone choosing C8 over 8" F/5 newtonian based on that reasoning and wondering why their images are so poor without realizing that at 0.5"/px they are spreading photons too much over pixels and that they need x4 more imaging time to match SNR of 8" F/5 newtonian with same camera.

[pete_l]

1 hour ago, vlaiv said:

I can imagine someone choosing C8 over 8" F/5 newtonian based on that reasoning and wondering why their images are so poor

Would they be poor? Most cameras have pixels in the 2.2-8µ range. When you look on astrobin you can see images at all ranges of pixel scale and it seems to me that pretty much any pixel scale can produce great images. A large amount of the success is in the skill of image processing.

Here's a link to NGC206 at .65 arcsec/px and here's an M31 at 4.8 arcsec/px (neither are my work). They are both very fine images.

[vlaiv]

17 minutes ago, pete_l said:

Would they be poor? Most cameras have pixels in the 2.2-8µ range. When you look on astrobin you can see images at all ranges of pixel scale and it seems to me that pretty much any pixel scale can produce great images. A large amount of the success is in the skill of image processing.

Here's a link to NGC206 at .65 arcsec/px and here's an M31 at 4.8 arcsec/px (neither are my work). They are both very fine images.

I'm not saying that one can't make a good image with long focal length scope if they know what they are doing. I'm just saying that if people are looking for advice on which camera will serve them best - saying that good rule is fl = pixel size x 200 - when that particular rule will lead to oversampling in 99% of amateur setups, or mentioning that it is ok to use even double focal length without explaining impact of such decision, or how it can be mitigated, will do disservice more often than not.

It is also the fact that explaining all the intricacies of choosing particular sampling rate and processing workflow (like binning, etc ...) can similarly confuse people and do them disservice and should be done carefully as not to confuse or deter from understanding.

 

[wimvb]

In my personal experience, moderately oversampled images have been easier to process than undersampled images. While there is a general rule relating pixel size to sampling "resolution" ("rate" seems inappropriate here) and star fwhm, that only takes into account the optical and mechanical part of data capture. Imo, one needs to optimize the entire process from data capture to final image. For me, the method of choice to sharpen images is deconvolution, and that works best together with (again, modest) oversampling. But as @vlaiv notes, oversampling will deteriorate snr, so there must always be a trade off between the parameters in a given configuration.

Furthermore, a configuration not only includes hardware and processing workflow, but also location (atmospheric conditions, light pollution) as well as intended targets. Nebulae generally don't need the same sampling resolution as galaxies and star clusters. All this has to be considered when configuring an imaging setup.

[StarMich]

14 hours ago, Drakester said:

There’s a useful calculator in the top section above under Resources> Astronomy tools> CCD suitabilty which may help. 

I'm using Astronomy Calculator 4.8, it outputs images at the actual resolution you can expect voor a given setup and object.

 

13 hours ago, tooth_dr said:

You mention the 1600 uncooled above, that would not a popular choice for deep sky imaging.

5 years is a long time to be deciding.

I mentioned uncooled to make it more comparable to my current DSLR, cooling is a must if I'm going to upgrade.  It's just recently I'm planning to actually buy a new camera, with those CMOS astrocameras popping up the last few years I intentionally waited longer.

 

13 hours ago, vlaiv said:

why don't you repeat your calculation but this time take binned x2 pixel size? Instead of going for 3.8 * 3.8 * 0.5, go for 7.6 * 7.6 * 0.5 = 28.88!

Now we are talking, right? It's clearly the best of all on your list above.

...

...

...

Let's now take some common type of telescope - like F/5 newtonian and compare how will two sensor fare if we decide to shoot same size FOV with each mounted on suitable newtonian. It is clear that scope paired with ASI1600 needs to have twice focal length of that paired with Atik to provide same FOV. Since both scopes are F/5 - one paired with ASI1600 will have twice aperture as well - or x4 light gathering surface.

Which sensor do you think is going to provide better image on same FOV in same time? I would guess - one collecting x4 the light.

First part: Didn't think about this, good point! Have to look more into that.  

Second part: I don't understand this entirely, do you mean if you would crop the ASI1600 image to the same FOV there would still be a benefit in light gathering power (unbinned)? When doing mosaics you indeed have a benefit, but is this also true for single pane shots?
 

5 hours ago, pete_l said:

I also would say that the technical difference between cameras becomes insignificant when compared to the results from manually processing the image and the skill and experience of the person doing that work. Unless you get a camera with a fault, almost any one will give you great results - once they are properly stacked and processed.

I know, it took me years to get the full potential out of my gear. But now I've reached the limit, it seems like it's time for an upgrade.  It's also hard to compare shots online made with different cameras as a lot of astrophotographers make use of some sort of noise reduction in their luminance. I personally prefer not to do any noise reduction as you see the traces almost everytime, I prefer a bit more noisy but sharper than smoother with noise reduction.

[vlaiv]

8 minutes ago, StarMich said:

Second part: I don't understand this entirely, do you mean if you would crop the ASI1600 image to the same FOV there would still be a benefit in light gathering power (unbinned)? When doing mosaics you indeed have a benefit, but is this also true for single pane shots?

I'm talking of comparing two sensors based on field of view and matching to particular scope.

Here is example:

Yellow FOV is ASI1600 + 10" Newtonian scope by Skywatcher, while red FOV is Atik 414 + 130PDS (again Skywatcher imaging newtonian) - which is 5.1" scope.

Now important thing to remember here is that each little bit of above FOVs (small surface making up picture - let's not call it pixel yet as it is not on sensor but rather on focal plane) will contain focused light from whole respective aperture.

With use of binning, and especially fractional binning - you can match sampling resolution of two cameras, so we can convert our "small surface at focal plane" to actual pixel that is "equal in size" (or rather resolution "/px). We still have difference in aperture and light gathering so more light will end up in pixel of the image with ASI1600 + 10" scope than with Atik 414 + 5.1" scope but both images will show same thing - same FOV sampled at same pixel rate.

How does that make ASI1600 inferior to Atik414 (as above numbers would suggest)?

[sharkmelley]

Your "per pixel" metric is misleading you.  A better metric is the total number of photons captured by the whole sensor.

My advice is to choose whether you want to go OSC or Mono and then choose a large sensor.  Otherwise you will miss the field-of-view of your Canon.

Mark