Moving from DSLR to pure astro cam- things to consider

[uhb1966]

OK, so i am considering a long term move from my trusty DSLR to a pure astro cam.

I sort of figured out the following things (just my opinion!):

• A DSLR like my Canon EOS 550D gives a large, decent sensor at an affordable price - zero (if you already own one).
• In order to really step things remarkably up, i'll probably have to go Mono and cooled - it does not make sense to shell out 4 figures for a small change.
• I would also like to go for the hubble palette- lets do it right the first try...
• This means that the investment for me will be at least camera + computer controlled filter wheel + 3 narrow band filters + who knows what else!
• If i go for the ASI 1600 cool, it looks like the total investment would be positively north of 2200 EUR/GBP- right?

I'd love to have feedback from the community!

PS.: lots of thanks to the SGL community- since i have started reading SGL, my astro malinvestments have gone down to zero (so far). Thanks again!

[Adriano]

Hi. I would go with the ASI1600 if it was a 16-bit ADC camera (it is a 12-bit ADC camera). Read this article about color depth: https://en.wikipedia.org/wiki/Color_depth

[iwols]

yep the asi 1600 looks a good camera not sure the downside of 12bit v 16 bit though

[uhb1966]

Well, it looks like the filter wheel will set me back around 500 euros and the filters another 700, so the total will be at least 2800 eur or 2330 gbp. This looks like automn 2018 :(((

[geordie85]

4 hours ago, uhb1966 said:

Well, it looks like the filter wheel will set me back around 500 euros and the filters another 700, so the total will be at least 2800 eur or 2330 gbp. This looks like automn 2018 :(((

When I was contemplating the leap from DSLR to CCD I kept eyeing up the atik 383l bundle from Ianking as it pretty much came with everything. CCD, lrgb and Ha, 0111 and s11 filters and filter wheel all for just over 2k.

Whilst saving for this I came across a mono 383L for £850 which isn't too far from half price so I had to buy it. 

Then a member on here put an ad in the classifieds for a starlight xpress filter wheel, 9 position carousel and full set of 36mm unmounted filters for just shy of £600. Needless to say I snapped them up.

Basically what I'm trying to say is keep your eye out on the 2nd hand market. I got everything I needed for close to the same price as a new atik 383l 

[Gina]

I shall be buying an ASI1600MM-Cool when I have the funds (hopefully in a couple of weeks) for DSO imaging.

[Davey-T]

Will it work with 1.25" filters ? could get very expensive if not.

Dave

[Gina]

Yes it will provided the filters aren't too far from the camera.  I worked out the formula for my scopes and filter wheel - Atik EFW2 with 9 1.25" filter spaces and 1000mm and 400mm FL scopes of about f5 focal ratio.

[Davey-T]

13 minutes ago, Gina said:

Yes it will provided the filters aren't too far from the camera.  I worked out the formula for my scopes and filter wheel - Atik EFW2 with 9 1.25" filter spaces and 1000mm and 400mm FL scopes of about f5 focal ratio.

Look forward to seeing your results Gina, some folks seem to have problems with the Atik383 and EFW2, the QSI683 with integrated FW manages with 1.25 filters although some nervous folk use unmounted ones.

Dave

[ollypenrice]

I can't comment on the new technology ASI cameras, being old-school CCD! However, you really do not need a motorized filter wheel, especially for narrowband. They are great for LRGB because you can ask them to scroll LRGB,LRGB and be sure to get at least something if the clouds come. But in NB you would never do that. You go for very long (multi night?) runs on just one filter. Manual wheels are fine for this. And more reliable!

Olly

[Andyb90]

I'm also looking at a potential upgrade from my DSLR to this camera.

From what I understand 16 bit can be achieved by stacking if enough frames are used.

I'd suggest having a look at the Cloudy Nights forum. There are plenty of posts on there with images. The results are very encouraging and new pics are being posted regularly.

Andy.

[uhb1966]

OK! I went for a used color ccd to start the learning curve... get sth to laugh here :

 

 

[Stub Mandrel]

That 12-bit/16-bit difference is probably very profound. When I went from a Canon 10D (121 bit) to 450D (14 bit) it was a huge increase in faint detail. In effect you get two stops more exposure for any given ISO as you can stretch the image four times more.

[Jokehoba]

On 20/07/2016 at 21:55, Stub Mandrel said:

That 12-bit/16-bit difference is probably very profound. When I went from a Canon 10D (121 bit) to 450D (14 bit) it was a huge increase in faint detail. In effect you get two stops more exposure for any given ISO as you can stretch the image four times more.

I reckon the improvement you've found is not down to bit-depth alone. The 10D has a 6M sensor, the 450D is 12M (so pixel scale is different). Plus the 450D pixels have a 50% increase in QE. 

Craig Stark's paper on the effect of stacking on bit depth is worth a read (http://www.stark-labs.com/craig/resources/Articles-&-Reviews/BitDepthStacking.pdf). 

I've just ordered an ASI1600MM-Cool...

[Stub Mandrel]

7 minutes ago, Jokehoba said:

I reckon the improvement you've found is not down to bit-depth alone. The 10D has a 6M sensor, the 450D is 12M (so pixel scale is different). Plus the 450D pixels have a 50% increase in QE. 

Craig Stark's paper on the effect of stacking on bit depth is worth a read (http://www.stark-labs.com/craig/resources/Articles-&-Reviews/BitDepthStacking.pdf). 

I've just ordered an ASI1600MM-Cool...

I hadn't considered either of those, but won't the twice-as-many therefore smaller pixels more or less balance out the increased QE?

The Stark paper is interesting, he seems to have independently discovered the principle of dithering (digital, not moving the camera) combined with oversampling to increase the resolution of a digital sampling system.  The advantage of bit-depth is that less noise is required and therefore fewer frames are needed to achieve a given result. In principle a bit-depth of 1 and a sufficiently noisy environment could provide full 16-bit resolution by stacking enough frames.

Greater bit depth both reduces the number of frames needed and thereby reduces overall noise.

[Jokehoba]

12 minutes ago, Stub Mandrel said:

I hadn't considered either of those, but won't the twice-as-many therefore smaller pixels more or less balance out the increased QE?

Quite possibly. Didn't want OP to think that bit-depth is the only factor. 

[Stub Mandrel]

I've given more thought to the Stark article. It's based on a false premise that output bit depth is limited by noise. Clearly teh more subs you stack teh less this is rue, plus if you 'zoom in' on his 16 bit graph compared to the 8-bit it would eventually look just as 'stepped'.

The points he misses are three-fold:

First, you need exploit greater bit-depth full you need to stack to a bit depth greater than the original data by 2^n where n=the number of subs; obviously if the subs don't exploit the full dynamic range you can get away with less than this.

Secondly, you need to stack more high bit-depth subs in proportion to the noise. So if you go from 10 to 14 bit and the noise doesn't change, you need four times the subs to fully exploit the increased bit depth.

Finally, he is ignoring (especially with his experiments) that you will stretch the data.

If he had used an image that only used a quarter of the dynamic range then stretched the final result to full dynamic range after stacking and stacked to 15 bits (to accommodate the full dynamic range of stacking 100 8-bit images)  I am sure the difference between his 8 and 4 bit stacks on the final page would be significant.

He does make many valid points, but I think he over-estimates the level of noise produced and therefore sets the point at which increased bit-depth brings additional benefits too low. All these points apply equally to digital oscilloscopes and digital audio and I think most users will confirm that both of these applications benefit considerably from increased bit depth.

It's worth noting that DSS and other stacking programs produce 32-bit stacks, and this is also why a pre-stretch of that 32-bit stack can show more detail than working with an unstretched 16-bit version of the stack.