The "No EQ" DSO Challenge!

[Nigel G]

58 minutes ago, SilverAstro said:

I do like a good experiment, well done Nige ! Awwww go on, heheee.

Actually I am going to propose another experiment, Ian and me and no doubt many others were surprised at how those 5x90 made such a difference, so one wonders if the other 85x45 were needed ! perish the thought !! So, experimentally, what do the 5x90 look like by themselves stacked alone ?

So many variables so many foundations to wobble

 

I stacked the 5 x 90s,  with much the same result as trying to cheat with duplicates. Just not so bright, this has surprised me, not much detail loads of noise, or maybe the light from the nebula but no sharpness. 

Nige. 

[SilverAstro]

Ummm, another imponderable ?, are gradients better removed before or after ! I expect a batch command could be concocted in some softwares to do it before ? but anyway, yes, they are all low frequency components, , ,   but frame-wide nebulosity is a lurking problem !!

Edited May 10, 2016 by SilverAstro

[Stub Mandrel]

1 hour ago, Filroden said:

I thought I'd try and show what I meant about light pollution gradients with a graphic as I can't explain it well in words. One of the main issues we face as AltAz imagers is field rotation. Our images rotate over time as we track across the sky. The top three "photos" show the same group of 5 stars taken, lets say, 20 minutes apart and show considerable rotation. I've shown an even light pollution rising from the bottom that impacts the bottom part of each photo. So the stars rotate but the light pollution does not.

When you stack these images, the software registers the locations of the five stars and aligns the photos so they stack over each other. However, now the gradient is in a different location. The lower stacked image shows how the program de-rotates the images to align the stars but this now causes the gradient in each image to be rotated the exact opposite angle in each image and this is stacked - pollution being an additive.

It's much easier to remove the simple linear gradient from each image (but very time consuming) than remove the more complex, bow tie, gradient in the stacked image. I wonder how much this impacts on the final quality of our images. 

I'm not convinced. Why? because the gradients on my images are sometimes caused by vignetting (not entirely eliminated by flats) combined with light pollution and sometimes nearby bright stars and the moon so they can be quite complex - and they change over time with an EQ mount as stars typically move out of the LP in my bit of sky.

This means very complex patterns.

Removal programs don't seem to struggle with the resulting effects though.

So, although I agree with your point that rotation means anon-linear gradient, in practice I think its only a problem if you use a gradient based removal strategy.

 

[wxsatuser]

47 minutes ago, Nigel G said:

I stacked the 5 x 90s,  with much the same result as trying to cheat with duplicates. Just not so bright, this has surprised me, not much detail loads of noise, or maybe the light from the nebula but no sharpness. 

Nige. 

This is to be expected as SNR increases as the square root of the number of light frames.
Although you have longer subs you have many less than the shorter subs.

Have a look a Samir's experiment with short subs.

http://www.samirkharusi.net/sub-exposures.html

 

[The Admiral]

1 hour ago, nicks90 said:

you are right, the gradient will 'spread' across the bottom of the image and up the sides due to field rotation.

But... how many hours of subs in a single sitting are people taking here? Look at the overall angular rotation of something like Leo over 2 hours by watching it move in stellarium in fast forward... few degrees maybe?

Here's a plot I made of angular rotation rate with azimuth and altitude. Towards the South (or North) the rotation is maximised, whereas East and West it is theoretically zero. It also increases as altitude is increased (as shown by the different lines). Exposures are typically 1 - 2 hours I suppose.

 

Ian

[The Admiral]

1 hour ago, Filroden said:

I never know how much stock to put in the scores other than as a relative measure between subs. For different targets I get very different scores - where there are few stars, the scores are much lower. It's not something I really understand within DSS and only use it as a guide.

I'm never that sure about DSS scores (or the scores in Autostakkert either, for that matter), and I sort of agree with you, except that if they are relative to within a session, why isn't the maximum normalised to, say, 100%? Also, the DSS manual says:

What is the score, and what is its meaning?
The score is a measure of the picture quality.
To put it simply, the higher the score, the more round and not too big stars were found.

and

Is the score a measure of the absolute quality of a picture?
No. The score is a relative measure that is only used to sort the pictures of roughly the same area in order to keep only the best pictures for stacking.
If the raw development process settings are modified you must register the pictures again and a new score will be computed.

Ambiguous?

Ian

[SteveNickolls]

This is all getting very interesting!

Cheers,
Steve

 

Edited May 10, 2016 by SteveNickolls

[Filroden]

Back to the pretty pictures...

I've just reprocessed M13 using PixInsight. I'm getting a little faster but it's still a lot slower than using DSS. However, given I'm using the same subs I think the stacked images from PixInsight are much sharper than in DSS. Here's 21 x 45s ISO 1600 using Esprit 80 on the Evo mount and the Canon 60D. Bias and flats used for calibration but tested with no dark; stacked and processed in PixInsight with final tweaks in Photoshop and Lightroom. I think this is better than my original attempt though the difference isn't as pronounced as it was for NGC7000 and M46 from earlier in the week.

BEFORE

AFTER

[Nigel G]

Ken. There's a good improvement there, sharper , more details and better colour. 

Nice ☺ I'm going to look into pixlnsight with improvement like that.

Nige.

[Filroden]

9 minutes ago, Nigel G said:

Ken. There's a good improvement there, sharper , more details and better colour. 

Nice ☺ I'm going to look into pixlnsight with improvement like that.

Nige.

If you do then take a look at the tutorials here http://www.lightvortexastronomy.com

[The Admiral]

My experience with PI was not so positive, when I tried it on an M106 image (see this thread, 23rd Feb, and https://stargazerslounge.com/topic/264074-early-days-with-pixinsight-problem/). I found it very difficult to get any colour out of it, but it is probably down to my ham-fisted attempts. I came to the conclusion that it was quite hard work to get a decent output when the data is poor, unlike ST. Good luck with honing your skills.

Ian

Edited May 10, 2016 by The Admiral

[wxsatuser]

21 hours ago, The Admiral said:

That's a startling difference for a 12% increase in total exposure. Now my foundations are getting a bit shaken up, which I admit are a bit shaky anyway, because I keep coming back to this post, "To stack or not to stack: 30 x 1s = 1 x 30s?" (https://stargazerslounge.com/topic/245183-to-stack-or-not-to-stack-30-x-1s-1-x-30s/#comment-2668145 ). May be it has a lot to do with camera noise. I realise that longer subs do allow one to dig out the fainter objects, but in this case it's not as though you've a lot of longer subs in the mix. Puzzling.

Can I ask a few questions please?

1.     When you say "85 x 45s plus 5 x 90s", is that the actual number that DSS used for stacking, or what you loaded it with?
2.     What scores did DSS give to the 90s subs?
3.     If you were to stretch the bottom image more, do you get nasties showing up?

Ian

 

It's all about exposure.

It's possible to get close to 30 x 1s = 1 x 30s if we expose correctly, it won't be 100% but close.

We need to have a clear gap between the histogram tail and the histogram origin.

My normal method is to have the luminance histogram peak around 20>25% on the back of the camera.
This leaves a clear gap between the origin, plus a bit, so all the three colours clear the origin.

BTW
I learnt this from Jerry Lodriguss, but if you look on Samir Kharusi's minimal exposures page, he allows a 10%
gap between the origin and the histogram tail.
Samir's experiments are with older cameras, more modern ones may give more leeway.

 

[Nigel G]

1 hour ago, wxsatuser said:

It's all about exposure.

It's possible to get close to 30 x 1s = 1 x 30s if we expose correctly, it won't be 100% but close.

We need to have a clear gap between the histogram tail and the histogram origin.

My normal method is to have the luminance histogram peak around 20>25% on the back of the camera.
This leaves a clear gap between the origin, plus a bit, so all the three colours clear the origin.

BTW
I learnt this from Jerry Lodriguss, but if you look on Samir Kharusi's minimal exposures page, he allows a 10%
gap between the origin and the histogram tail.
Samir's experiments are with older cameras, more modern ones may give more leeway.

 

wxsatuser,

 To be honest I don't understand what you mean, ( histogram peak on the back of the camera  ) & ( clear gap between histogram tail and histogram origin ) 

At the moment this is foreign to me, but I need to understand.  I think I need to do some research ☺

Nige. 

[Nigel G]

I'm heading for West coast Scotland today for 2 weeks, so scopes packed away, 

Keep up the great work guys, I hope to see some more images posted when I return,

I m taking my camera and tripod just in case I get the chance to grab a few shots from a dark site.

Shame someone hasn't designed a goto mount which works on a heaving yacht ☺☺☺that would be something special. 

Nige.

[Filroden]

1 hour ago, Nigel G said:

wxsatuser,

 To be honest I don't understand what you mean, ( histogram peak on the back of the camera  ) & ( clear gap between histogram tail and histogram origin ) 

At the moment this is foreign to me, but I need to understand.  I think I need to do some research ☺

Nige. 

Here's a screenshot of one of my rates showing the histogram. This 45 second shot shows the peak blue starting around 30% in from the left (left is black and right is white on the scale). You want some gap between the start of your first peak and the left hand as lots a feint details are found here. If you don't expose enough, that detail is never captured. 

Of course, my histogram also shows my light pollution because the red is further right, I.e brighter. Again, you don't want to over expose so you also want a gap at the right. This one is harder to judge because stars can be easily over exposed and end up having no colour. 

So I aim for the peaks to be in the 30-40% from the left. This just happens to be about 45 seconds at ISO 1600 for my garden. 

Edited May 11, 2016 by Filroden
Fat fingers and auto correct

[wxsatuser]

2 hours ago, Nigel G said:

wxsatuser,

 To be honest I don't understand what you mean, ( histogram peak on the back of the camera  ) & ( clear gap between histogram tail and histogram origin ) 

At the moment this is foreign to me, but I need to understand.  I think I need to do some research ☺

Nige. 

Normally under the INFO button you should find the histograms.

Take test exposures so the white luminance peak is at least to 20%, first line on Canons.
Just double check that all colours are away from the origin.

Here is one of mine from a Canon 60Da, the extra blue is from the Astronomik clip filter.

[SteveNickolls]

Hi Mike, Thanks for these posts. I can understand some of the article but floundering elsewhere. I took a look on my camera at the RAW exposures I took a few nights ago with the histogram also showing. I found that the histogram, whether for brightness or RGB channels was showing well off from the LH side about 60-70% along not the 10% in the article. Now this was for 50 seconds exposures at ISO 1600. I presume it means the image is being swamped by light pollution (it's bad here) so I should reduce the exposure time or drop the ISO value? Or does it matter? Unfortunately I have no background in photography to draw upon, just learning from mistakes and successes.

Cheers,
Steve

[Filroden]

43 minutes ago, SteveNickolls said:

Hi Mike, Thanks for these posts. I can understand some of the article but floundering elsewhere. I took a look on my camera at the RAW exposures I took a few nights ago with the histogram also showing. I found that the histogram, whether for brightness or RGB channels was showing well off from the LH side about 60-70% along not the 10% in the article. Now this was for 50 seconds exposures at ISO 1600. I presume it means the image is being swamped by light pollution (it's bad here) so I should reduce the exposure time or drop the ISO value? Or does it matter? Unfortunately I have no background in photography to draw upon, just learning from mistakes and successes.

Cheers,
Steve

Here's a 90 second exposure at ISO 1600. It's at 80% but it does develop and I have star colours so it hasn't been over exposed. I guess some trial and error is needed to know what your conditions allow. 

[Nigel G]

OK ☺ now I see what you mean,  thanks for explaining ☺.

Nige.

[Stub Mandrel]

An anyone answer me a question - is the histogram x-scale linear or logarithmic scale?

i.e. if it has say 5 divisions are these:

1          2          3         4        5            -  linear

1          10        100      1000   10000     - log base 10

1          2          4          8       16           - log base 2

My guess is that it's a log base 2, which would fit with things like changing aperture by one f-stop, changing ISO by one step or doubling exposure would move the histogram by one division, this would also explain why a histogram shifted right by a longer exposure doesn't get noticeably wider.

Does anyone know what scale Canon use?

[Filroden]

12 minutes ago, Stub Mandrel said:

An anyone answer me a question - is the histogram x-scale linear or logarithmic scale?

i.e. if it has say 5 divisions are these:

1          2          3         4        5            -  linear

1          10        100      1000   10000     - log base 10

1          2          4          8       16           - log base 2

My guess is that it's a log base 2, which would fit with things like changing aperture by one f-stop, changing ISO by one step or doubling exposure would move the histogram by one division, this would also explain why a histogram shifted right by a longer exposure doesn't get noticeably wider.

Does anyone know what scale Canon use?

A quick search doesn't give me a clear answer. Your assumption about it being log base 2 looks broadly correct, but based on a test someone did and posted to flickr, it seems the histogram shows more than 5 stops, but with the additional stops bunched at either end of the scale. Not sure me using spaces in this does it correctly but the numbers 0, 4, 8, 16 and 64 line up with the five vertical lines, with the 2 and 32 being halfway between the first set and last set of lines. Note: this is specific to Canon.

0      2      4            8            16    32     64

Hopefully someone much more technical than me has a better answer!

[SteveNickolls]

2 hours ago, Filroden said:

Here's a 90 second exposure at ISO 1600. It's at 80% but it does develop and I have star colours so it hasn't been over exposed. I guess some trial and error is needed to know what your conditions allow. 

Thanks for this comparison, I might see what effect using ISO 800 has on the histogram and on images taken.

Best Regards,
Steve

[Herzy]

Let me get this straight - If I take a 30 second exposure at ISO 3200, my histogram peaks 30% from the left side. However, if I take a 75 second exposure at ISO 800 I get the same result. Which is better? 

One more question: I'm limited to 30 second exposures right now because I don't have an intervalometer to insure that I get the correct exposure length. Would it be beneficial to increase the ISO to increase the brightness of what i'm imaging?

Edited May 11, 2016 by Herzy

[Stub Mandrel]

I've read that with my Canon D450 and long exposures the lowest read noise is at 800 iso, and taht iso 1600 isn't significantly worse. Try it and see.

[Filroden]

ISO in digital cameras just amplifies the signal and noise. You still capture the same amount of photons at ISO 100 as ISO 1600 ( I think). You can achieve the same amplification in processing  

Exposure time is the important factor. 

There are benefits to higher ISO settings. I think read noise is lower at higher ISO and you also see your image quicker so you can see if you have a good sub.