ASTAP vs DSS

[almcl]

I have been stacking my images in DSS and recently tried ASTAP as well having heard some positive comments about it.

ASTAP seems to give slightly less noisy images and although I don't find the interface particularly intuitive, it does have some useful tools (CCD inspector, HFD contour diagram &c).

One thing I have noticed is that it is SLOW.  Yesterday as an experiment I stacked 40 lights, 30 flats and 30 bias frames from an ASI 2600 MC in DSS and timed it.  It completed in just under 7 minutes.  The same stack in ASTAP took in excess of 38 minutes to complete.

Does anyone else find ASTAP similarly slow, or could I have got one or more of the settings wrong?  

Edited April 8, 2021 by almcl

[bottletopburly]

I’ve been using Astap recently, I’m not too bothered about speed , I only need it to stack , I like you can calibrate flats on its own too and master dark before loading lights  , you don’t need to tinker too much with the setting , how are you loading files in Astap are you loading bias as darks if dithering  , are you calibrating flats with bias , 

Edited April 8, 2021 by bottletopburly

[almcl]

Thanks for the thoughts.

I am not over bothered about speed, either,  but this seems excessive just to stack the lights , flats and bias.  I don't use darks; on the DSLR they didn't help at all and on the ASI 2600 they don't appear necessary, but I do dither every frame and always take more than 20 lights to allow Sigma Clip a good chance.  So lights get loaded in the lights tab, flats in the flats tab and bias in the Dark Flats tab.  Unfortunately this doesn't correct very well at the moment and I get much better vignetting removal in DSS but only if I use Flats, Dark Flats and Bias frames.  ASTAP, of course doesn't allow for that.

Slightly curious to know what advantage you get from calibrating flats by themselves?  

Edited April 8, 2021 by almcl

[bottletopburly]

If you load dark flats , analyse  and in flats analyse and calibrate to leave a master flat calibrated with your DF frames try loading your bias frames in your darks tab analyse and calibrate to leave you with a master dark , see how that turns out once lights stacked , I only do each step separately as a work flow leaving lights last to analyse and stack .

[festoon]

I've found the speed of stacking in ASTAP to be similar to DSS.

I'm still to find a way of applying different flats to datasets from different nights using ASTAP. Using DSS this can be done by grouping data. To me this is an advantage of DSS, unless it can be done in ASTAP.

Edited April 9, 2021 by festoon

[bottletopburly]

35 minutes ago, festoon said:

I've found the speed of stacking in ASTAP to be similar to DSS.

I'm still to find a way of applying different flats to datasets from different nights using ASTAP. Using DSS this can be done by grouping data. To me this is an advantage of DSS, unless it can be done in ASTAP.

I’m assuming the stacked fits can be stacked together from multiple nights of the same target maybe @han59 can  help .

[festoon]

8 minutes ago, bottletopburly said:

I’m assuming the stacked fits can be stacked together from multiple nights of the same target maybe @han59 can  help .

Yes, good point, you can stack stacked results in ASTAP. Not sure of the science of this, but if you have subs n1,n2,n3,...,nn with master dark d1, master flat f1 from night one and m1,m2,m3,....mm subs with master dark d2, master flat f2 from night two - is the resulting signal to noise of a stack of ((n1-d1)/f1+(n2-d1)/f1+(n3-d1)/f1+....+(nn-d1)/f1+(m1-d2)/f2+(m2-d2)/f2+(m3-d2)/f2+....+(mm-d2)/f2) subs equal to the signal to noise of a stack of ((n1-d1)/f1+(n2-d1)/f1+(n3-d1)/f1+....+(nn-d1)/f1) with ((m1-d2)/f2+(m2-d2)/f2+(m3-d2)/f2+....+(mm-d2)/f2)

[almcl]

13 hours ago, bottletopburly said:

try loading your bias frames in your darks tab 

That's an idea I haven't tried yet!

Thanks for the thought!

[almcl]

12 hours ago, festoon said:

I've found the speed of stacking in ASTAP to be similar to DSS.

That's interesting.  Is the time  taken of the six minute variety by both DSS and ASTAP or does it take a bit longer in both?

[han59]

20 hours ago, festoon said:

I'm still to find a way of applying different flats to datasets from different nights using ASTAP. Using DSS this can be done by grouping data. To me this is an advantage of DSS, unless it can be done in ASTAP.

There is no clear way at this moment.

Would the flats be very different?  I have a fixed setup, so I would use the same flat for lights of different dates.  The only difference could be a dust particle added between the image sessions. I have added lights from several years together and didn't find experience any problem.

At the moment you can classify the flats on filter used. So the red flat is used for the red lights and so on. What I could envision is a "classify" on flat date.  An other way would be to stack each night separately and then stack the different nights  again together, but then sigma clip will not work so well.

So a classify on a date for a flat could be an idea, but is this really required?  I make flats only once in a few months. An what would be the time span to classify, one Julian day, one week, one month? Or classify them by selection. That would be tricky to realise.

Han

[han59]

When your imaging with a mobile setup and disassemble the setup at the end of each night it could be better to take flats each night.

I could modify ASTAP to look for all compatible flats in the flat tab. If two compatible flats are found (same dimensions, and filter if selected), take the flat with the closest date. That would not require an new menu option and will work automatically. The user  has just to add the flats to the flat tab.

At the moment it just takes the first flat found. Same action could be applied for darks.

Han

[han59]

What you could do now is to calibrate only. The dark and flat-dark master is applied to the light and the light is stored as calibrated.  For each night series you do that with the appropriate dark and flat. Then just stack the calibrated lights.  The calibrated files and up in the results tab and can be copied to the lights tab again for stacking.

 

[festoon]

On 10/04/2021 at 14:38, han59 said:

What you could do now is to calibrate only. The dark and flat-dark master is applied to the light and the light is stored as calibrated.  For each night series you do that with the appropriate dark and flat. Then just stack the calibrated lights.  The calibrated files and up in the results tab and can be copied to the lights tab again for stacking.

 

Thanks for the suggestions @han59. If I understand correctly, what you are suggesting here is for for each nights work in ASTAP calibrate the subs with the correct dark and flat then save the result. Then for different night repeat the same procedure. The stack all the results together at the end. Is that correct?

[han59]

Yes that is correct and is one option.

Note that I will have later today a new version which will select the most suitable master flat and master dark based on the time in the FITS header. So you have to a prepare master flat for each night. So it is a little different then DSS where selection is folder/directory based. ASTAP looks only to the FITS header(s)

[han59]

The first new ASTAP version 0.9.526 is ready. 

1) Master flat and master dark can be created based on date.  So you can add all flats to the flats tab and per day a master flat will be created. This is activated by option "classify by date for master creation"

2) For the light the master dark and flat with the nearest date will be selected.  This will happen if more then one compatible flat or dark is found.

Download developments version: http://www.hnsky.org/astap_setup.exe  Feedback is appreciated.

Han

 

 

[festoon]

6 minutes ago, han59 said:

The first new ASTAP version 0.9.526 is ready. 

1) Master flat and master dark can be created based on date.  So you can add all flats to the flats tab and per day a master flat will be created. This is activated by option "classify by date for master creation"

2) For the light the master dark and flat with the nearest date will be selected.  This will happen if more then one compatible flat or dark is found.

Download developments version: http://www.hnsky.org/astap_setup.exe  Feedback is appreciated.

Han

 

 

Thats super cool - I'll give that a go @han59

Thank you for including this

[han59]

Unfortunately there is a bug in the flat&dark selection in this development version. I will fix it this afternoon.

Han

[almcl]

While many of the comments and Han's information are very interesting, could I, for a moment, drag the thread back to the original question, and ask if anyone else finds ASTAP taking over half an hour to stack 30 light frames with associated calibration frames?

[han59]

Problem fixed download at: : http://www.hnsky.org/astap_setup.exe  Feedback is appreciated.

 

almcl,

30 minutes is very long. Especially colour  and images with large dimension take time.  Stacking 46 mono images  2328 x  1760 pixels, method Sigma Clip takes about 115 seconds. I have them on a solid state disk. Have a fast disk system helps. ASTAP doesn't keep the images in memory but loads them from disk. That is a design choice to avoid memory problems. 

Han

 

 

[festoon]

Hi @almcl apologies for that...hopefully this helps

DSS (72 light frames, 1 master dark, 1 master flat)

Register = 2 mins

Stack = 4 mins 40 secs

ASTAP (72 light frames, 1 master dark, 1 master flat)

Analyze and Organize = 1 min

Stack (Sigma Clip) = 5 mins

So as I suspected in my previous message, there is hardly anything in it

[bottletopburly]

ASTAP

23 lights 

1 master  dark 

sigma clip average 

bilinear interpolation 

945 sec

i 5 processor  ssd drive 

nikon d5300 raws 

analyse organise and stack 

[han59]

Maybe a coincidence, but I made ASTAP a little faster. I noted a flickering progress indicator at the Windows taskbar and removed an unnecessary action. ASTAP v09.526b : http://www.hnsky.org/astap_setup.exe

I would love to make the application faster but that is not easy. Most of the time the program is for stacking simple shifting data around. Multitasking that in separate tasks in not easy. ASTAP is a single core application. But the speed test of @Festoon indicates that it is performing  okay against DSS which is a multi core application. I'm a fan of efficient coding rather the brute force solutions.

For sigma clip average method it has to load each image three times. That is not so efficient but the program memory usage is as a consequence  very moderate. Only for ctrl-Z action it keeps up to 3 images in the memory. Stacking 1000 images or more is not problem.

And I'm programming just for the fun. 

It is clear sky here so time to get the equipment running.

Han

 

[almcl]

1 hour ago, festoon said:

Hi @almcl apologies for that...hopefully this helps

DSS (72 light frames, 1 master dark, 1 master flat)Register = 2 mins

Stack = 4 mins 40 secs

ASTAP (72 light frames, 1 master dark, 1 master flat)

Analyze and Organize = 1 min

Stack (Sigma Clip) = 5 mins

So as I suspected in my previous message, there is hardly anything in it

That's most interesting, Festoon, I suspected something was wrong with my set up or work flow and that pretty much confirms it.  I'll have to start digging...

 

[almcl]

33 minutes ago, bottletopburly said:

ASTAP
945 sec
i 5 processor  ssd drive 

 

So my i7 + ssd drive is no where near that - clearly I've got something mis-configured.  Thanks for the comparison.

[almcl]

32 minutes ago, han59 said:

Maybe a coincidence, but I made ASTAP a little faster. I noted a flickering progress indicator at the Windows taskbar and removed an unnecessary action. ASTAP v09.526b : http://www.hnsky.org/astap_setup.exe

I would love to make the application faster but that is not easy. Most of the time the program is for stacking simple shifting data around. Multitasking that in separate tasks in not easy. ASTAP is a single core application. But the speed test of @Festoon indicates that it is performing  okay against DSS which is a multi core application. I'm a fan of efficient coding rather the brute force solutions.

For sigma clip average method it has to load each image three times. That is not so efficient but the program memory usage is as a consequence  very moderate. Only for ctrl-Z action it keeps up to 3 images in the memory. Stacking 1000 images or more is not problem.

And I'm programming just for the fun. 

It is clear sky here so time to get the equipment running.

Han

 

Han, please no criticism was intended of your software, it's free and it works, and that you do it for fun is wonderful.  Rather I was trying to find out if my experience is the same as other people's and it clearly isn't.  

If you are ever able to consider a multi-threading version that would be marvellous, but even without, the performance others report is pretty good.

Hope your clear skies work well!