BSOD with siril + drizzle + starnet++ removing stars

[TiffsAndAstro]

anyone have any clue why my PC blue screens of death when i use siril + drizzle + starnet++ and try removing star?

im assuming its because the image is 4 times the size it would be if i didn't drizzle. what would be a good way to downsample (?) it back to my sensor resolution of 5184 x 3456 ?

i can crop it down to 5184x3456 and obviously reduce the fov, but im not keen on that, even though sn++ then seems to work ok. 

4 core 4 thread 6600k with 32GB ram. should be ok performance wise.

[Elp]

Have you looked up the blue screen error code? Although the spec seems decent, performance between generations of chips goes up magnitudes. My 4 core laptop is useless for CPU intensive tasks compared to my 8 core one for example.

[Elp]

Try without drizzling as it creates huge megapixel data

[TiffsAndAstro]

17 minutes ago, Elp said:

Have you looked up the blue screen error code? Although the spec seems decent, performance between generations of chips goes up magnitudes. My 4 core laptop is useless for CPU intensive tasks compared to my 8 core one for example.

its a windows_stop code, and the little pictogram thing just takes me to the usual vague ms webpage

if i get really bored i can try cropping down incrementally and testing each one to find a reliable size/tile size.

also pc is good enough for this stuff. worst that could happen is it takes longer than a faster 6, 8 or more core pc

Edited April 17 by TiffsAndAstro

[TiffsAndAstro]

14 minutes ago, Elp said:

Try without drizzling as it creates huge megapixel data

its fine without drizzling (usually) but i dithered, so id like to drizzle

ill have a look at gimp's re-sizing/downscaling and give it a go after that too

[Elp]

You can search for the stop code the one with lots of zeros then a number. I suspect it might be memory allocation related despite you having the RAM, sometimes even a drop in power within components can do it which is why it's always wise to have a very good PSU.

As you've dithered, you don't really need to drizzle, I believe it's better for when you're imaging targets in high resolution (IE long focal length). I suspect you want to see the difference, so by all means try it, maybe on less number of light frames.

Edited April 17 by Elp

[TiffsAndAstro]

22 minutes ago, Elp said:

You can search for the stop code the one with lots of zeros then a number. I suspect it might be memory allocation related despite you having the RAM, sometimes even a drop in power within components can do it which is why it's always wise to have a very good PSU.

As you've dithered, you don't really need to drizzle, I believe it's better for when you're imaging targets in high resolution (IE long focal length). I suspect you want to see the difference, so by all means try it, maybe on less number of light frames.

as a youtube trained expert, i was under the impression i needed to drizzle as im undersampled and because i dithered. will have another go sans drizzle then ty

[Elp]

In 3 years of doing this with various optics, short and long FL and cameras small and large pixel sizes I've never seen the need to. Depends on the images you're getting.

Edited April 17 by Elp

[PhilB61]

If you get it to work I'd be interested to see the results. I've avoided drizzling mainly because of the massive increase in file size and resultant impact on processing speed etc. It is my understanding that it's mainly beneficial where you are undersampled in relation to your seeing conditions, and that in addition to dithering, you need a lot of subs to counter the increased noise that drizzle introduces. If you are only marginally undersampled and don't have many subs it may have a negative effect on your images, in addition to the obvious strains on your processing pc.

[Elp]

That's one of the main reasons for me, the image sizes it produces is a magnitude larger, drizzling dslr data will create likely over 100MB drizzled images per image.

[TiffsAndAstro]

22 minutes ago, Elp said:

In 3 years of doing this with various optics, short and long FL and cameras small and large pixel sizes I've never seen the need to. Depends on the images you're getting.

Honestly, comparing (just now)  a basic osc processing/stacking to a drizzled one, i can't see any difference. except the image less pixels and smaller file size. which makes things quicker.

ill give drizzling a rest for the time being. 

[TiffsAndAstro]

22 minutes ago, PhilB61 said:

If you get it to work I'd be interested to see the results. I've avoided drizzling mainly because of the massive increase in file size and resultant impact on processing speed etc. It is my understanding that it's mainly beneficial where you are undersampled in relation to your seeing conditions, and that in addition to dithering, you need a lot of subs to counter the increased noise that drizzle introduces. If you are only marginally undersampled and don't have many subs it may have a negative effect on your images, in addition to the obvious strains on your processing pc.

im undersampled, but i can't work out what my seeing is - i assume its 2" ish. Didn't notice a massive increase in noise but not many subs so quite noisy anyway.

i'll leave off the drizzle for the time being i think. ty for the advice.

[Clarkey]

11 minutes ago, TiffsAndAstro said:

i assume its 2" ish

What is your sampling rate and what kit are you using? To make drizzling worthwhile you really need to be quite under sampled. If you are getting seeing of 2 you are doing very well for the UK. I consider anything near 2 pretty exceptional and I have never got < 2 when imaging.

[TiffsAndAstro]

1 hour ago, Clarkey said:

What is your sampling rate and what kit are you using? To make drizzling worthwhile you really need to be quite under sampled. If you are getting seeing of 2 you are doing very well for the UK. I consider anything near 2 pretty exceptional and I have never got < 2 when imaging.

How do you know what your seeing is? Sky meter?

I said 2" because the mf summit calculator used it as an average default value.

I have a evostar 72ed 600d flattener.

[TiffsAndAstro]

1 hour ago, PhilB61 said:

If you get it to work I'd be interested to see the results. I've avoided drizzling mainly because of the massive increase in file size and resultant impact on processing speed etc. It is my understanding that it's mainly beneficial where you are undersampled in relation to your seeing conditions, and that in addition to dithering, you need a lot of subs to counter the increased noise that drizzle introduces. If you are only marginally undersampled and don't have many subs it may have a negative effect on your images, in addition to the obvious strains on your processing pc.

I think I didn't have enough frames to tell either way. Possibly I was at the cusp, so didn't see any worsening of the noise, but the noise was reasonably high anyway.

I can post a drizzle and non drizzle versions if you want to see? Just tell me jpg or tiff or fit etc 

It's only about 100 mins total 120subs on m51

Also have similar on iris nebula of it would be interesting to compare?

Isuspect I need way more total integration time for it to be even worth comparing though.

Edited April 17 by TiffsAndAstro

[Clarkey]

6 minutes ago, TiffsAndAstro said:

How do you know what your seeing i

You can tell from the analysis of the star sizes. I run an auto-focus routine which gets the best focus point. This also gives the HFR (or FWHM) figure. Good seeing for me is 3-4 arc secs. Very occasionally < 3. I'm not familiar with Siril, but I suspect it will analyse a light frame and give you a FWHM or HFR value. (When close to optimum focus they are to all intense and purposes the same).

Your set up gives 2.48 arcsecs/px or 2.11 arcsec/px depending on whether you are using the standard Skywatcher flattener/reduced or a straight flattener. Given the size of your scope, you will not be able to resolve much more than this value - around 1.6 arcsec/px. For the type of images the 72ED is good for (widefield) this sampling rate will be fine in my opinion.

[TiffsAndAstro]

17 minutes ago, Clarkey said:

You can tell from the analysis of the star sizes. I run an auto-focus routine which gets the best focus point. This also gives the HFR (or FWHM) figure. Good seeing for me is 3-4 arc secs. Very occasionally < 3. I'm not familiar with Siril, but I suspect it will analyse a light frame and give you a FWHM or HFR value. (When close to optimum focus they are to all intense and purposes the same).

Your set up gives 2.48 arcsecs/px or 2.11 arcsec/px depending on whether you are using the standard Skywatcher flattener/reduced or a straight flattener. Given the size of your scope, you will not be able to resolve much more than this value - around 1.6 arcsec/px. For the type of images the 72ED is good for (widefield) this sampling rate will be fine in my opinion.

Wow Ty for this. Dawes limit for my scope is 1.6 but I'm not too sure how relevant that is these days with ccds and software processing, blur exterminator etc, but I am v noob.

From what I understand, drizzle is exactly for under sampled set ups but how atmospheric seeing effects it I dunno.

As Hubble doesn't have seeing issues, I'm guessing seeing will effect it.

It seems to be a reasonably balanced setup, but that might not be the case if I replace the dslr with an imx585 sensor at some point, cos like you said about  resolving below 1.6. 

Also Nina does a star anylisis as my sequence progresses, but I'm really not sure what I do to rectify any issue I even noticed, apart from the focusing or noticing cloud.

Edited April 17 by TiffsAndAstro

[TiffsAndAstro]

2 hours ago, Elp said:

That's one of the main reasons for me, the image sizes it produces is a magnitude larger, drizzling dslr data will create likely over 100MB drizzled images per image.

lol try 800MB