Lee_P

37 minutes ago, CloudMagnet said:

Ok, just to check what exposure times are you using? It could be down to read noise creating a minimum floor that dithering cannot help with. This would be worse with multiple short exposures (as you are reading more often)

Looking at the specs on your camera, there is a step change downward in read noise from using >100 gain. What gain are you using?

Very interesting to see that dark flats are making no visible difference. Thanks again for putting this together

Lights are integrations of two-minute exposures. Flats taken using my ASIAIR PRO's auto flat exposure, which came out as 2.7s. Dark Flats 2.7s too, to match. Gain 100.

@Ouroboros @CloudMagnet I've made a few more noise tests: increasing the dithering, and trying Dark Flats. Nothing seems to make an appreciable difference in the end. So either I'm getting something wrong in the image acquisition / pre-processing stages, or my Bortle 8 skies are forever cursed to have this level of background noise. I've been staring at these images for hours though, so maybe some fresh eyes can spot something I'm missing! 🥴

On 04/05/2021 at 08:15, Ouroboros said:

As you say the pattern in the noise does seem to remain the same. By how much did you dither?  Should we conclude by not enough?  Or are these variations in light and dark real?  Possibly due to more distant galaxies and light absorbing dust clouds. 

I'm using an ASIAIR PRO, with the dither options being 1, 2, 3, 5, or 10 pixels. I've been using 3. Sounds like another test is called for, with a higher number.

The variations being real is something I hadn't even considered 🤯

39 minutes ago, Magnum said:

Hi Lee, Ive also been testing filters and on same target, but my goal was to see which one picks up the faint IFN in the background. So far I've used my standard UV/IR filter, my Idas P2, the Astronomik CLS ccd, though none of these are designed with LED lights in mind.

I see you went for the P3 but they also released a D3 at the same time specifically designed for LED LP. By the specs the new D3 would seam better as like the previous D2 it has a cut out for the brightest LED peak. Maybe you could ask to swap it for the D3? I would like to try one myself but they are too expensive to just buy on a Wim, I know FLO have the 30 day return policy but I only just retired another filter I was testing and its not really fair to them to keep buying things just to test.  Maybe as you've already bought the P3 it would make more sense to see if they would let you exchange for the D3 to add to your existing tests ?

https://www.firstlightoptics.com/idas-filters/idas-lps-d3-light-pollution-filter.html

Lee

P.S. im also a Lee P, P for Phillips what's your P for LOL

Ooh interesting, I didn't know there was a D3. Thanks for letting me know. I'll give it a go and will report back -- might be a while though, given all the cloud that's forecast!

p.s. my P stands for "Pullen" 😁

4 minutes ago, CloudMagnet said:

Might just be down to the Bortle 8 skies. You are just reaching the maximum limit of what that setup in that location is capable of. Someone in Bortle 2/3 would likely hit that signal to noise ratio in a fraction of the time. I'm sure someone with more knowledge can work out mathematically, but that would be my assumption. I have deliberately avoided the Iris nebula from my Bortle 6/7 skies as I’m sure I read that somewhere that it would take me around 40 hours to reach the same SNR as someone taking 4 hours in a proper dark sky zone.

You must just be tending towards the natural minimum noise possible and no more dithering can remove noise beyond that limit. Only easy option to improve further would be to run the camera at a cooler temperature. Anything beyond that is tweaking at the edges. Such as using more dark frames or using dark flats (if not already), looking at gain settings and exposure to minimise read noise.

Ah, I calibrated them with Flats and Darks, but not Dark Flats. Something else to read up about!

12 minutes ago, CloudMagnet said:

Very interesting! Huge difference from 2-4-6 hours then a steady drop in difference as we would expect. Think that the practical limit is still around that 10hr mark as before.

Main difference I can see from 10 hours onward is just around the very faint outer arms near the top left. Some aren't even visible in the early pictures but do show up later on. Might be helped by the steady darkening of the background-I assume this is the noise cancelling out with longer integration.

Well done putting that together, exactly what I meant by keeping the brightness of the galaxy the same for a better comparision

Having consistency in brightness was a good shout, it does make it easier to see the differences. I've tried to do the same with the original Iris Nebula data.

Any thoughts on why the noise tests a few posts above show no real improvements after four hours? That's got me scratching my head.

26 minutes ago, CloudMagnet said:

In terms of making the images the same brightness, either by eye would be ok or using the eyedropper tool in Photoshop that will give you a brightness reading of a pixel sample- just put this in the same place on every image.

I went down the eyedropper route. In for a penny, in for a pound!

Whaddya make of that?

1 minute ago, CloudMagnet said:

Were the images dithered? I would expect that would even out any noise with higher integration time (fewer frames occupying the same exact space).

In terms of making the images the same brightness, either by eye would be ok or using the eyedropper tool in Photoshop that will give you a brightness reading of a pixel sample- just put this in the same place on every image.

Yep, they were dithered.

I'll have a crack at making the galaxies the same brightness.

1 hour ago, CloudMagnet said:

Hmmm, my observation would be that there doesnt seem to be much change in background noise from 4-6 hours onwards. I thought that this would slowly improve but hits a limit much earlier than I would expect.

I think you're right actually. I'm confused, surely it should get smoother with more data? Even the pattern of noise looks the same. Have I messed something up in pre-processing? 🤔

1 hour ago, CloudMagnet said:

Thank you for putting this together. Would it be possible to have the galaxy the same brightness in each picture as well? Makes for a nice comparison for the finer detail in the core/dust lanes

I'd be happy to do this, but you'll have to give me some tips on how! In the way I've been doing basic edits on these tests, the galaxy naturally gets brighter.

On 01/05/2021 at 17:07, CloudMagnet said:

Thanks for putting this together, as you said looks like 10 hours onwards is the start of diminishing return. Definitely useful to see it in practice. Have you done anything similar for other types of targets like galaxies or emission nebula to see if the same integration time holds true?

Would be interesting to look at how different surface brightness could affect the "best" integration time as well.

As CloudMagnet suggested, I've repeated the experiment with an extreme crop of M81.

Looks like the same result to me: the first 10 hours are the most valuable. Data after that are still useful, but the impact certainly isn't as dramatic.

I'll repeat with L-eXtreme data on a nebula once galaxy season has passed!

2 hours ago, ollypenrice said:

With an LRGB image a good trick is to use only the RGB for the brightest parts. Often they are equivalent to short subs for this purpose and they'll carry more colour as well. I haven't processed this latest IKI set but that was how I did my own Iris and a good number of other images with bright parts needing control.

Olly

Good tip, thanks for sharing 😁

19 minutes ago, ollypenrice said:

That's great. No question.  Personally I'd have another look at the core, though. I suspect that, with a dedicated stretch carefully blended in, you could resolve the central star as a round star with bright nebulosity around it. As your stack has increased, so has the saturated area around the progenitor star, but that can be prevented in processing. It's well controlled as it is but I bet you could perfect it.

Olly

I think you're right, and it's funny you should say that because I've been editing the IKO Iris Nebula data trying to do exactly that 😁

15 minutes ago, ollypenrice said:

A helpful post. As your signal builds, so does the need for some differential processing to preserve the core while stretching out the dust. In PI you'd do this by masked stretching. I'd do it in Photoshop using layer masking. They key thing, though, is that it would be well to process the long set differently from the short.

Olly

Absolutely! That would make side-by-side comparisons rather difficult though. Not impossible I suppose, but too tricky for me 😅

FYI this is the final edited photo with all the data.

3 minutes ago, Dazzyt66 said:

I think that would be brilliant - I’ve had to trawl through loads of info with lots of trial and error as LP/lighter skies makes such a difference on what’s possible with AP (although I’m happy to try anything your thread on LP filters saved me some time AND money! 🙏) - and I do it mainly so my wife and kids can see what’s out there from our back garden too rather than me just looking through a lens 😄

Super, well feel free to suggest what you'd like to see on a website dedicated to OSC from a city, Knowing that the effort will be useful to at least one person might motivate me to do it!

2 minutes ago, Ouroboros said:

A very interesting project. Thanks for publishing your data, @Lee_P.  I am surprised you can get anything worth using at Bortle 8 with a colour camera and no LP filter.   I struggle at Bortle class 5 with a DSLR, but then I suppose the advantage of your ASI cmos camera is that it is both more sensitive and uses a shorter optimum integration time than a DSLR. There’s hope for us all in your example.  

Did you do much background subtraction/colour correction to your data or is this just stretched data? 

Thanks! The new-generation CMOS OSC cameras are very good. And work well with filters like the L-eXtreme. I'm increasingly coming to think that conventional wisdom like "don't use OSC from a city" and "you need a LPS filter" is outdated. 

To answer your question, for each image in this experiment I followed the same few steps in PixInsight. They were mostly to get a set of images that could be displayed side-by-side for fair comparison.

* Crop
* ColorCalibration
* AutomaticBackgroundExtraction
* EZ SoftStretch
* SCNR (remove green)
* Slight CurvesTransformation

10 minutes ago, Dazzyt66 said:

Another great Bortle 8 test from @Lee_P - I’m just getting into imaging (from Bortle 8 skies) so all this info is VERY useful and as I’m able to automate my sessions now 10 hours of subs seems doable (although I’d be pretty much happy with the images after 5! 😂).

Good work sir!

 

Great! I'm tempted to make a website dedicated to OSC imaging from city skies. We're a particular niche. Could be a project for the Summer...

9 hours ago, Avocette said:

Interesting experiment - thanks for making the results nicely accessible! 

Did you use any filter during image capture?

Thanks, I'm glad you found it interesting. No light pollution filter used. I haven't found them to be effective from my location. I posted another experiment, on this topic, here: 

Wow, I've never worked with such good data! I decided to focus in on the central nebula and try to bring out its details.

PixInsight:
* LRGB combine
* DBE
* ColorCalibration
* EZ SoftStretch
* Create a mask, apply LocalHistogramEqualization to add contrast to the central nebula
* Starnet++ to create starless version and starless mask
* Save starless version as a TIF and export to Lightroom...

Lightroom:
* Topaz DeNoise AI
* Boost Clarity
* Boost Saturation and Vibrance
* Save as TIF and export to PixInsight...

PixInsight
* PixelMath to add stars back in
* Save as TIF and export to Lightroom...

Lightroom
* 1x1 crop
* Boost Texture, Clarity, and a touch of Dehaze
* Little bit of Sharpening
* Colour Noise Reduction (I want the central nebula to be almost hypnotic, so this step was to desaturate the outer stars -- limiting the viewer's peripheral vision, focussing attention on the centre).
* Reduce Highlights over the central star

Just for fun, here's a comparison with the best data I've managed (Bortle 8, Askar FRA400, ASI 2000MC, no filters, 20 hours of integration). IKO wins 😂

12 minutes ago, SteveNickolls said:

Thanks so much for posting this experiment, very helpful to see how the final result varies with overall exposure time. Bortle 8 too. Have you by the way seen the IKI Observatory data for the Iris Nebula on SGL, few of us could aspire to the times pent imaging and in far better dark conditions?

Cheers,

Steve

Good shout, I'll download that and will have a play. Might make me a bit depressed about the quality of my data though 😅

35 minutes ago, DaveS said:

Having said which, I have several images / projects with more than 20 hours, and a couple with more than 30. Always a temptation to add just another few hours.

I'd be interested to see a similar experiment to mine, using your 30 hours!

5 minutes ago, CloudMagnet said:

Thanks for putting this together, as you said looks like 10 hours onwards is the start of diminishing return. Definitely useful to see it in practice. Have you done anything similar for other types of targets like galaxies or emission nebula to see if the same integration time holds true?

Would be interesting to look at how different surface brightness could affect the "best" integration time as well.

Well, I do happen to have 24 hours of data for M81 / M82, so perhaps I'll repeat the experiment for those! Once some nebulae are up in the sky again, I'd like to try using my L-eXtreme filter.

I thought it would be interesting to see the difference increasing integration times makes using my kit and sky conditions. So, I took 20 hours of data I'd collected on the Iris Nebula and produced images of different total integration times. Then I cropped in very close, and performed some simple edits in PixInsight, just to allow for comparisons. I thought I'd post the results here in case anyone could find them interesting or of use 😃

The earliest hours make the biggest difference, perhaps unsurprisingly! I'd say that the nebula's structure is there by 10 hours, and after that adding more time helps to smooth out the noise, but we're definitely into diminishing returns. It's hard to see much difference in the final few images, but there is a slight improvement. So, I'm thinking that 20 hours is a good target time for me to aim for.

Info about image acquisition:

* April 2021
* Bristol, UK (Bortle 8 )
* Telescope: Askar FRA400 f/5.6 Quintuplet APO Astrograph
* Camera: ZWO ASI 2600MC-PRO
* Mount: Orion Sirius EQ-G
* Guide: William Optics 32mm; ZWO ASI 120MM Mini
* Software: PixInsight, Photoshop, Lightroom
* ASIAIR PRO, ZWO EAF

On 27/04/2021 at 07:28, Dazzyt66 said:

I live in Bortle 8 (Doncaster) which is pretty much all LED now plus I live on a new estate which is definitely all LED so I’ve read this with interest - I was looking to maybe get a LP filter at some point, but now I’ve seen comparisons I don’t think they really make enough difference to justify. Great comparison! 😊

I think we're in the same boat. Maybe there's a filter out there that would help us with broadband targets, but for now I'm happy to just brute-force it with long integration times. I took the 20 hours of data from my tests, added in another four for good luck, and produced this image. It's not great, and a 400mm focal length scope isn't ideal for these targets, but at least we can get something even with our skies.

14 hours ago, Seelive said:

LED lighting seems to have such a wide bandwidth that I believe anything other than narrow band filters will let most of it through. If you are surrounded by LED lighting I have the impression that any of todays so-called LP filters will just give a minor improvement in the background sky pollution level but at the expense of a reduction in the image colour range.  (Bring back LP sodium lighting, these days it's surely easy to filter out the two D lines whilst leaving the rest of the spectrum intact?)

My tests seem to support this. I am assuming that it's LED lights swamping my skies though -- I don't know for sure that they're widely used in Bristol, but I wouldn't be surprised.

I recently saw FLO were selling a new light pollution filter: IDAS LPS-P3. I shoot OSC from Bortle 8, and figured this might help for broadband targets! (I use an L-eXtreme for narrowband, which works brilliantly). I already own an older-style IDAS D1, but don't routinely use it as I'm not sure of its effectiveness. Well, with these two filters in my possession, why not have a shootout and see the results?

For this test I used an Askar FRA400 f/5.6 72mm Quintuplet APO Astrograph and ZWO ASI 2600MC-PRO USB 3.0 Cooled Colour Camera. All shots were from Bristol (UK) city centre. 120-second exposures. The target was M81 / M82.

First up, I took a single sub using No Filter, D1, and P3. Each sub I edited slightly in the same way using PixInsight, just enough to make as fair a comparison as possible. Then I cropped very close into M82.

Hard to make out much difference! The D1 produces a colour cast that takes a bit of work to even out in post-processing. The P3 retains natural colours well, and perhaps gives slightly better definition than no filter. More testing needed... so next I gathered three hours of data for each filter:

Same again, to my eyes: all images very similar, but the D1 giving a slight colour cast.

Time for round 3. When imaging I aim for long integration times, so let's try that here. I'd need to cut one filter out just because there are only so many clear skies. So bye-bye D1! Now it's No Filter versus P3. 10 hours of integration time for each. Who will emerge the victor?

These are the stacks right out of PixInsight. The P3 has a bit of a colour cast, and a gradient -- to be expected actually, as the Moon was up more during the P3 test. Let's do some basic edits...

They look essentially the same to me. Some very slight differences in the background, but explained by my use of Pixinsight's DBE tool to remove the Moon's gradient. Let's have a tight crop on M81:

I really can't see any difference! These tests are leading me to conclude that the sources that make up Bristol's light pollution don't lend themselves to being suppressed by a filter. Maybe I've got lots of LED lights nearby.

It feels almost heretical to say, but is there any point in light pollution filters from a city centre? Sure, from towns or areas with particular street lights (e.g. sodium), but from the middle of a big city? It seems to me that long integration times are the only way to claw back some of that signal to noise ratio. This Iris Nebula photo was taken with the same kit as the above tests, and is 20.5 hours of integration time -- no filters at all. 

I'm probably going to return the L3, and put the D1 up for sale. Unless anyone can see something I'm missing?!