Lucky imaging of DSOs using a ASI 178MM, C6 and AZ-GTi?

[vlaiv]

52 minutes ago, Ags said:

Yes, darks, lights and flats are all shot with the same Gain and Exposure Length. I'll try and get some FITS files for you (they are currently buried in SER files).

Temperature should be the same. Flats are shot indoors immediately after the session, so I don't think the camera chip warms up in 7 minutes. Flats are shot by putting a sheet of paper over the front of the scope and pointing it at a room light. The histogram peaks at 50% with no clipping according to Sharpcap. In contrast my lights are so underexposed I could use them as darks. Could the big exposure difference between lights and flats be an issue in itself?

Ah, now I see - you don't have set point cooling - that is regular camera. I did not pay attention to that detail.

Flat calibration can fail for a number of reasons - mismatch in gain or offset of exposure length or problem with darks or problem with flat darks or light leak - but all of these reasons boil down to single thing - there is some signal that should not be there. Flats work when only signal present is light signal that came "down the tube" - if there is unfocused light it will cause problem.

Issue with non cooled camera is that you can't guarantee that dark current will be the same between subs. In principle once camera reaches ambient - it should be fairly stable - but any change in ambient is going to reflect on dark current.

Only about 6 degrees C can double dark current and camera body is made out of aluminum - so it can quickly reach ambient. This goes both ways - it will cool faster, but it will also warm up faster. Powering camera on is going to heat it quite fast regardless of the fact that it was in cold just few minutes ago. Try shooting separate flat darks at the same ambient temperature as lights (if you shoot flats indoors - run camera for few minutes so that it stabilizes in temperature, then record set of flats and right afterwards record set of matching darks).

 

[Ags]

I will try shoot darks outside in the same sequence as lights and darks. I just need to find a good light source and a windproof way of holding my diffuser to the front of the scope.

[Ags]

Just to write a happy ending to this thread, I got darks working, and also found that lucky imaging / lucky tracking works much better at F6.3 than at F10!

[Annehouw]

Congratulations with this result!

I have tried lucky DSO imaging and failed miserably. Here's my story, maybe there is something in it for you.

 

After seeing the magnificent results of a French guy under the name exaxe on Astrobin, I decided to try this myself. 

A bit of research told me that to beat the seeing you need to image somewhere between 10 and 100Hz. With those short exposure times you need a telescope that can collect a lot of light and a camera with low read noise. Read noise is getting lower every year with modern CMOS cameras. I decided upon a ZWO 290M (uncooled) camera. To get more light in early experiments I decided to image at f 3/3. This gives an imaging scale of 0.6 arcsecond per pixel. And that is way better than my typical seeing of 2-2.5 arcseconds.  Enough to get this starting.

 

The plan of attack was as follows:

- Select a bright subject

- Take bursts of movies in SER (raw) format, with 10 fps (or higher, signal levels allowing)

- Generate darks with the same approach (they are essentially bias frames at these short exposures)

- Process these movies ( I created about 150 Gbyte in a single session!) into an end result, using about 30-50% of the data (only the sharpest subs)

- Take conventional color images of the same object with my OSC camera at a scale of 1 arcsecond per picture (smeared by the seeing, but that is OK as this is just the color data).

- Combine in a LRGB approach

- Done 😀

I found SIRIL to be an excellent piece of software to handle this https://www.siril.org/

 

The result? A rather sharp image, but with horrible, horrible raining noise.  Unusable.

To try and improve this, I would have to start dithering while making the movies.  Now this is doable, but all in all it would take so much time to perfect the technique that I decided to use the not so many clear skies around here to image with my well known set-up. Maybe I will try again sometime in the future. It looks a lot like planetary lucky imaging, but with new challenges.

 

For further inspiration, here is one of the few practical write-ups I know of http://astrophoto17.eklablog.com/acquisition-du-ciel-profond-en-mode-rapide-c28772848

 

Edited May 24, 2020 by Annehouw

[Ags]

I am not able to get down to 100 ms frames. I am looking at an F4 reducer lens suggested on another thread by @vlaiv, which would get me closer the 10 Hz frontier. 

https://starizona.com/store/night-owl-4x-sct-reducer-corrector

I stacked this image with a low frame rejection rate (80% of frames accepted into the stack), so I wonder if I reduced that to 50% or less I would get more detail in the core and tighter stars in general. The trouble is noise would go up...

Another thought is to increase gain from from 310 to 360 on my ASI 178 - the theory is this will spread the weak signal so I will end up with a richer spread of levels after stacking.

 

Edited May 24, 2020 by Ags

[vlaiv]

19 minutes ago, Ags said:

Another thought is to increase gain from from 310 to 360 on my ASI 178 - the theory is this will spread the weak signal so I will end up with a richer spread of levels after stacking.

I don't see that happening.

Only thing that you can benefit in going from 310 to 360 is maybe reduction in read noise.

According to this:

difference will be minimal - both are about 1.35e give or take.

[Ags]

My signal is skimming the bottom of the histogram - if a pixel has a value of 5 or 6 in individual stacks, then after combining them I am still only left with 5 or 6. But if I increase the gain so the same signal gives a pixel value of 10 or 12, then after combining all the sequences I should have 10, 11, or 12, more closely reflecting the average of the sequences?

Edited May 24, 2020 by Ags

[vlaiv]

46 minutes ago, Ags said:

My signal is skimming the bottom of the histogram - if a pixel has a value of 5 or 6 in individual stacks, then after combining them I am still only left with 5 or 6. But if I increase the gain so the same signal gives a pixel value of 10 or 12, then after combining all the sequences I should have 10, 11, or 12, more closely reflecting the average of the sequences?

No, provided that you use required precision.

If you use fixed point integer math - you should add your subs. In case you use floating point math - do simple average.

(5+6+5+6+5) / 5 = 5.4 and not 5

Btw - with stacking addition and average is the same from SNR perspective - both increase SNR in the same way.  "Interpretation wise" - addition is the same as having one long exposure - signal "accumulates over time", while averaging is having more precision in that particular exposure.

Simple multiplication with a constant is all that is separating the two.

[Ags]

2 hours ago, vlaiv said:

No, provided that you use required precision.

Did I miss another setting in AS!3?  I don't seem to be able to sum, I can only increase SNR.

[vlaiv]

8 hours ago, Ags said:

Did I miss another setting in AS!3?  I don't seem to be able to sum, I can only increase SNR.

If you are using AS!3 - it should handle that for you. In fact - I believe there is 32bit "scientific" precision setting somewhere that you can enable if you want floating point precision.

I would personally advise all to always use that level of precision - especially when doing lucky / planetary type imaging and stacking thousands of subs.

[Ags]

I think maybe my problem is i am not checking "mormalize stack" so perhaps the floating point values gt collapsed back to the same integer scale. I will try repeat the above image with "normalize stack" set to 75%.

BTW, there is definitely no 32-bit option or "scientific stack" option in AS!3. I assume it just uses 32 bit precision all the time?

Edited May 25, 2020 by Ags

[vlaiv]

4 hours ago, Ags said:

I think maybe my problem is i am not checking "mormalize stack" so perhaps the floating point values gt collapsed back to the same integer scale. I will try repeat the above image with "normalize stack" set to 75%.

BTW, there is definitely no 32-bit option or "scientific stack" option in AS!3. I assume it just uses 32 bit precision all the time?

Could be that it uses 32bit precision all the time, but look at this:

If you select fits output - it will save as 32bit per channel

[Ags]

Firstly found that normalize stack is not available for stacks using surface stabilization. But the FITS format does solve the issue - especially as I now see GIMP can open FITS directly. I did a quick stack and stretched the data to the extremes, but at no point the the histogram turn into a comb, so my problem is solved!

 

Edited May 25, 2020 by Ags