Jump to content

  •  

CNers have asked about a donation box for Cloudy Nights over the years, so here you go. Donation is not required by any means, so please enjoy your stay.

Photo

Triad ultra - process my data

  • Please log in to reply
40 replies to this topic

#1 fourfives

fourfives

    Lift Off

  • -----
  • topic starter
  • Posts: 19
  • Joined: 20 Jan 2020

Posted 22 July 2021 - 02:52 PM

I've ran into a processing wall,

 

I've processed osc data with pixinsight and dss for about 1.5 years at this point and have had plenty of good results with a normal broadband light pollution filter but I now have a triad ultra filter and I've spent more time than I'd like to admit trying to figure out how to process color correctly in pixinsight.  I've tried splitting the RGB channels up - liner fit, putting them back together or doing nothing at all and trying to process the image like I would a normal OSC image but I'm at a loss for how to get the reds to be red - the blacks to be black etc.  I've never had a problem with OSC broadband data ....been very easy to do but I'm completely at a loss with how to process with this filter.  Any tips would be appreciated.

 

Target: Elephant's Trunk Nebula

Scope: Radian Raptor with the reducer F4.5  275mm

Filter: Triad Ultra quad band

Lights: 34x300 sec subs - Gain 100 offset 0

Camera: ASI 533 OSC  @ -5c

No flats, No Darks, No Bias

Captured with Sharpcap pro

Bortle 9

EQ6-R pro, guided etc.

 

Data: (fitz files, Dithered every sub)

https://drive.google...f?usp=sharing  



#2 rj144

rj144

    Viking 1

  • -----
  • Posts: 764
  • Joined: 31 Oct 2020

Posted 22 July 2021 - 02:56 PM

I probably can't do anything, but I like to see if I can anyhow.  Do you have the unstretched stack too?



#3 fourfives

fourfives

    Lift Off

  • -----
  • topic starter
  • Posts: 19
  • Joined: 20 Jan 2020

Posted 22 July 2021 - 03:04 PM

Sure - I uploaded 'Autosave.fits' to the same link - its the dss stack but not drizzled to keep the file size small.


  • rj144 likes this

#4 rj144

rj144

    Viking 1

  • -----
  • Posts: 764
  • Joined: 31 Oct 2020

Posted 22 July 2021 - 03:44 PM

Sure - I uploaded 'Autosave.fits' to the same link - its the dss stack but not drizzled to keep the file size small.

Thanks.  Didn't see it at first.

 

Not that great, but a quick run with Siril and Affiinity (using the Astroflat plugin):

 

Autosavesira.jpg



#5 bobzeq25

bobzeq25

    ISS

  • *****
  • Posts: 25,030
  • Joined: 27 Oct 2014

Posted 22 July 2021 - 04:07 PM

Quick hack job in PixInsight.  Comments.

 

You really need flats and bias, that's very vignetted.  Wrecks processing, I cropped out the middle in an attempt to cope.  You might get away with no darks, but you're better off taking them until you see why you could.

 

Very overexposed.  That's why the stars have no color, when you max out RGB all you can get is white.  30 seconds (or less) would be more like it.  Also shoot 10X the number of subs, you don't want to reduce total imaging time.  Watch this video.

 

https://www.youtube....h?v=3RH93UvP358

 

I recommend Astro Pixel Processor.

 

anothers elephant.jpg


Edited by bobzeq25, 22 July 2021 - 04:07 PM.


#6 fourfives

fourfives

    Lift Off

  • -----
  • topic starter
  • Posts: 19
  • Joined: 20 Jan 2020

Posted 22 July 2021 - 04:36 PM

So I have darks and bias I just haven't bothered to stack them with this data because I can't get past the color issues.  I'll likely shoot flats tonight since I plan to add more subs anyway and I haven't touched the image train but unsure about the 30 second sub thing.  My other scope is a rasa 8 and I tend to stick around 60 secs for it on broadband filters and don't over expose there.  The histogram in sharpcap does not look overexposed when shooting 5 min subs - maybe its wrong - I have exclusively used it with the rasa 8 for 1.5 years

 

I tried hoo with APP - ripping out HA and O3 from the original stack and the colors are better but its still not where i'd like it to be.  Just don't get the same pop that I do with Rasa8 and broadband.  I know some of that is trying to make up the F4.5 to F2.0 time but still.  Feels like i'm missing something big on the post process side of the house.



#7 bobzeq25

bobzeq25

    ISS

  • *****
  • Posts: 25,030
  • Joined: 27 Oct 2014

Posted 22 July 2021 - 05:12 PM

So I have darks and bias I just haven't bothered to stack them with this data because I can't get past the color issues.  I'll likely shoot flats tonight since I plan to add more subs anyway and I haven't touched the image train but unsure about the 30 second sub thing.  My other scope is a rasa 8 and I tend to stick around 60 secs for it on broadband filters and don't over expose there.  The histogram in sharpcap does not look overexposed when shooting 5 min subs - maybe its wrong - I have exclusively used it with the rasa 8 for 1.5 years

 

I tried hoo with APP - ripping out HA and O3 from the original stack and the colors are better but its still not where i'd like it to be.  Just don't get the same pop that I do with Rasa8 and broadband.  I know some of that is trying to make up the F4.5 to F2.0 time but still.  Feels like i'm missing something big on the post process side of the house.

Watch the video.  Look at the chart about 51 minutes in.  For Bortle 9, F4.5, narrowband, he's recommending about 12 seconds.  I didn't want you to fall off your chair <smile> so I said 30.

 

You are missing some very big things.  Not in processing.  Here they are.

 

Flats (which need bias to work).  Not having them really messes up processing, the software gets _very_ confused.

 

Much less exposure (but keep the total imaging time up there with a compensating increase in the number of subs).

 

Those two are not close calls.  <smile>

 

In my Bortle 7, with a C8 RASA, broadband, I use 10 second subs.  Below you see what I get with 662 of them.  Still some white stars (10 seconds is a bit overexposed), but that's a tough (high dynamic range) target.

 

Better version here.

 

https://www.astrobin.com/t5173s/

 

Pleadies 2019 V3 small.jpg


Edited by bobzeq25, 22 July 2021 - 05:17 PM.


#8 rj144

rj144

    Viking 1

  • -----
  • Posts: 764
  • Joined: 31 Oct 2020

Posted 22 July 2021 - 05:29 PM

So I have darks and bias I just haven't bothered to stack them with this data because I can't get past the color issues.  I'll likely shoot flats tonight since I plan to add more subs anyway and I haven't touched the image train but unsure about the 30 second sub thing.  My other scope is a rasa 8 and I tend to stick around 60 secs for it on broadband filters and don't over expose there.  The histogram in sharpcap does not look overexposed when shooting 5 min subs - maybe its wrong - I have exclusively used it with the rasa 8 for 1.5 years

 

I tried hoo with APP - ripping out HA and O3 from the original stack and the colors are better but its still not where i'd like it to be.  Just don't get the same pop that I do with Rasa8 and broadband.  I know some of that is trying to make up the F4.5 to F2.0 time but still.  Feels like i'm missing something big on the post process side of the house.

Is my take closer to what you're to acheive though?



#9 fourfives

fourfives

    Lift Off

  • -----
  • topic starter
  • Posts: 19
  • Joined: 20 Jan 2020

Posted 22 July 2021 - 07:29 PM

Here is what I came up with in pixinsight - just processing as if it were normal broadband 

 

Pix.jpg



#10 cucubits

cucubits

    Messenger

  • *****
  • Posts: 417
  • Joined: 20 Jul 2020
  • Loc: Dallas, TX

Posted 22 July 2021 - 07:53 PM

Watch the video.  Look at the chart about 51 minutes in.  For Bortle 9, F4.5, narrowband, he's recommending about 12 seconds.  I didn't want you to fall off your chair <smile> so I said 30.

 

Question on this. I also made a mistake by choosing too long exposure times for a longer project. I'm on the edge of Bortle 9 so we can call it 8 to make me feel better. I'm shooting with a 533MC as well, with an L-Enhance dual narrowband. I have very little imaging time per night so once I choose some settings I sort of have to stick to that, but in this case I think I have to do some damage control.

 

I've spent the first few sessions with 180s subs which now I understand it's too much. What should I do here? Do I have to scrap to first few hours of exposure? Can I continue with 30s exposures starting now and then combine the two stacks separately? Side question, can PixInsights wbpp handle sets of light subs with various exposures (with correct calibration frames for each set)? I'd love to not have wasted the first few nights...



#11 jonnybravo0311

jonnybravo0311

    Surveyor 1

  • *****
  • Posts: 1,970
  • Joined: 05 Nov 2020
  • Loc: NJ, US

Posted 22 July 2021 - 07:58 PM

I've ran into a processing wall,

 

I've processed osc data with pixinsight and dss for about 1.5 years at this point and have had plenty of good results with a normal broadband light pollution filter but I now have a triad ultra filter and I've spent more time than I'd like to admit trying to figure out how to process color correctly in pixinsight.  I've tried splitting the RGB channels up - liner fit, putting them back together or doing nothing at all and trying to process the image like I would a normal OSC image but I'm at a loss for how to get the reds to be red - the blacks to be black etc.  I've never had a problem with OSC broadband data ....been very easy to do but I'm completely at a loss with how to process with this filter.  Any tips would be appreciated.

 

Target: Elephant's Trunk Nebula

Scope: Radian Raptor with the reducer F4.5  275mm

Filter: Triad Ultra quad band

Lights: 34x300 sec subs - Gain 100 offset 0

Camera: ASI 533 OSC  @ -5c

No flats, No Darks, No Bias

Captured with Sharpcap pro

Bortle 9

EQ6-R pro, guided etc.

 

Data: (fitz files, Dithered every sub)

https://drive.google...f?usp=sharing  

Offset 0? That's going to cause issues.

No calibration frames. That's going to cause issues.

 

Fix those two things then reprocess the data and post the unedited stacked image.

 

In the meantime, here's what I was able to do with the data provided:

 

med_gallery_347158_14910_2656910.jpg


  • bobzeq25 and Ibuprofen200mg like this

#12 bobzeq25

bobzeq25

    ISS

  • *****
  • Posts: 25,030
  • Joined: 27 Oct 2014

Posted 22 July 2021 - 08:32 PM

Question on this. I also made a mistake by choosing too long exposure times for a longer project. I'm on the edge of Bortle 9 so we can call it 8 to make me feel better. I'm shooting with a 533MC as well, with an L-Enhance dual narrowband. I have very little imaging time per night so once I choose some settings I sort of have to stick to that, but in this case I think I have to do some damage control.

 

I've spent the first few sessions with 180s subs which now I understand it's too much. What should I do here? Do I have to scrap to first few hours of exposure? Can I continue with 30s exposures starting now and then combine the two stacks separately? Side question, can PixInsights wbpp handle sets of light subs with various exposures (with correct calibration frames for each set)? I'd love to not have wasted the first few nights...

Couple of good possibilities.

 

How about a learning experience?  Try 30 seconds, and get an image (same total imaging time), and compare with 180s.

 

Some processing programs (like PixInsight) let you combine different exposures in a thoughtful way to increase dynamic range.


  • cucubits likes this

#13 klaussius

klaussius

    Apollo

  • *****
  • Posts: 1,279
  • Joined: 02 Apr 2019
  • Loc: Buenos Aires

Posted 22 July 2021 - 11:04 PM

If I

  • Fix vignetting with some synthetic flats (actual flats would be better)
  • Fix O3 landing on green channel only (sometimes I've seen this with ZWO filters, and that seems to happen here as well)
  • Play with color balance
  • Carefully do gradient reduction and try to boost O3 signal
  • Discard blue and process as HOO (blue seems to contain only H-beta - or too much pollution - and is too low SNR)

I get as close as I can with this dataset to "reasonable" colors. Still missing blue stars, but it may be the filter, and the stretching was perhaps a bit sloppy, and it could use a lot of NR, but:

 

Autosave.jpg

 

Also, I'll echo what has been said above:

  • Offset 0 sounds like trouble
  • Do some darks, this should not be so noisy. Maybe it would improve with darks. Maybe not.
  • Do some flats. It will definitely improve with flats.

If darks don't bring out some more signal, you'll need a lot more integration time. Strange, because the Triad Ultra should be narrow enough to provide decent contrast, but these subs have very low SNR. Maybe you're just bathed in light pollution (I'm at B9 and don't get it this bad on NB filters though). You'll need a lot more subs to compensate in any case.

 

Processing this as HOO is definitely a way to go about it, but you'll need better O data for that.

 

Edit: attached wrong version, fixed.


Edited by klaussius, 23 July 2021 - 12:15 AM.

  • bobzeq25 likes this

#14 Ibuprofen200mg

Ibuprofen200mg

    Sputnik

  • *****
  • Posts: 29
  • Joined: 16 Jul 2020
  • Loc: SC, USA

Posted 23 July 2021 - 12:46 AM

This is running in GIMP, decomposing the image to RGB layers, tossing the blue layer and combining as R-R, G-G, B-G. Hackjob on manual vignette removal and some color adjustments. 

 

rgb-compose.jpg

 

 



#15 fourfives

fourfives

    Lift Off

  • -----
  • topic starter
  • Posts: 19
  • Joined: 20 Jan 2020

Posted 23 July 2021 - 01:10 PM

Dssstack 100 gain, 70 offset, flats, bias, darks  (approx 1 hour, 5 min subs with triad)

https://drive.google...iew?usp=sharing

 

Dssstack 100 gain, 70 offset, flats, bias, darks  (approx 1 hour, 30 sec subs with no filter at all)

https://drive.google...iew?usp=sharing

 

Here is a side by side of the two after quick processing and starnet to ditch stars.  Left is the no filter image, right is the traid.  I realize that running narrowband mucks with the colors - just wish I could make the color on the right...more like the one on the left - just at a loss at how to do it in post processing.  The left side has odd spots due to starnet - the stars were overexplosed likely because I was not running IR/UV cut at all and probably should have gone even shorter exposures with zero filters on a nearly 100% moon.

 

 

SideBySide.JPG

 

 

 



#16 cucubits

cucubits

    Messenger

  • *****
  • Posts: 417
  • Joined: 20 Jul 2020
  • Loc: Dallas, TX

Posted 23 July 2021 - 01:13 PM

Dssstack 100 gain, 70 offset, flats, bias, darks  (approx 1 hour, 5 min subs with triad)

https://drive.google...iew?usp=sharing

 

Dssstack 100 gain, 70 offset, flats, bias, darks  (approx 1 hour, 30 sec subs with no filter at all)

https://drive.google...iew?usp=sharing

 

Here is a side by side of the two after quick processing and starnet to ditch stars.  Left is the no filter image, right is the traid.  I realize that running narrowband mucks with the colors - just wish I could make the color on the right...more like the one on the left - just at a loss at how to do it in post processing.  The left side has odd spots due to starnet - the stars were overexplosed likely because I was not running IR/UV cut at all and probably should have gone even shorter exposures with zero filters on a nearly 100% moon.

 

 

attachicon.gifSideBySide.JPG

Why did you take the filter off for the 30s subs run? It would've been interesting to see a comparison where only the subexposure times are different.



#17 jonnybravo0311

jonnybravo0311

    Surveyor 1

  • *****
  • Posts: 1,970
  • Joined: 05 Nov 2020
  • Loc: NJ, US

Posted 23 July 2021 - 01:56 PM

I don't image with OSC or multi-band filters, so forgive my ignorance here. Is the histogram and stats on the triad image considered "normal"? The red channel is WAY out to the right (median ADU is 11k). That's a value I'd expect to see AFTER doing a stretch, not before. Looking at the non-filtered image, the median ADU value for the red channel is over 6x less (about 1.7k). Something just isn't sitting right seeing these values.

 

Here's a side-by-side in PI. The Triad is on the left, the non-filtered is on the right:

 

med_gallery_347158_15202_6295888.png

 

I took the original images, split them into component R, G and B parts. Then I did a linear fit to the _lowest_ median ADU channel. Normally, if I run linear fit, it is to the highest median channel, but in this case, the median ADU of the red channel in the data from the triad was over 11k. I then recombined the channels and registered the unfiltered image to the triad one. Next, I cropped each to 2750x2750. At this point, I ran ABE on each, then I did a background neutralization and color calibration. Finally, I ran SCNR on each. The last thing I did was run an unlinked STF on each to get the above result.

 

You can clearly see the wonky star colors on the triad image. Not surprising, since the filter blocks a significant portion of the visible spectrum.



#18 unimatrix0

unimatrix0

    Viking 1

  • -----
  • Posts: 545
  • Joined: 03 Jan 2021

Posted 23 July 2021 - 02:29 PM

With an F5.6 telescope at 446mm focal range + dual narrow band filter on the asi533mcpro at Bortle 5 , depending how bad the sky is (wild fire smoke, sky fog) I take subs at 80seconds-160 seconds. 
.  Also that 0 offset is gonna hurt you,  for the ASI533, it likes to be at offset 50 or at least 30. 



#19 bobzeq25

bobzeq25

    ISS

  • *****
  • Posts: 25,030
  • Joined: 27 Oct 2014

Posted 23 July 2021 - 03:58 PM

Dssstack 100 gain, 70 offset, flats, bias, darks  (approx 1 hour, 5 min subs with triad)

https://drive.google...iew?usp=sharing

 

Dssstack 100 gain, 70 offset, flats, bias, darks  (approx 1 hour, 30 sec subs with no filter at all)

https://drive.google...iew?usp=sharing

 

Here is a side by side of the two after quick processing and starnet to ditch stars.  Left is the no filter image, right is the traid.  I realize that running narrowband mucks with the colors - just wish I could make the color on the right...more like the one on the left - just at a loss at how to do it in post processing.  The left side has odd spots due to starnet - the stars were overexplosed likely because I was not running IR/UV cut at all and probably should have gone even shorter exposures with zero filters on a nearly 100% moon.

 

 

attachicon.gifSideBySide.JPG

The UV-IR cut makes very little difference.

 

You're just overexposed, period.

 

Could you upload just one of each light (broadband and narrowband) and just one bias?  Then I'll run the numbers for you.



#20 bobzeq25

bobzeq25

    ISS

  • *****
  • Posts: 25,030
  • Joined: 27 Oct 2014

Posted 23 July 2021 - 04:03 PM

With an F5.6 telescope at 446mm focal range + dual narrow band filter on the asi533mcpro at Bortle 5 , depending how bad the sky is (wild fire smoke, sky fog) I take subs at 80seconds-160 seconds. 
.  Also that 0 offset is gonna hurt you,  for the ASI533, it likes to be at offset 50 or at least 30. 

You're Bortle 5, he's Bortle 9.  The proper subexposures are not _remotely_ comparable.

 

Look at these tables.  See how much proper subexposure changes with sky brightness.

 

https://www.cloudyni...olour-versions/

 

Thye only way to do this is to use some kind of method that incorporates light pollution level.  Without that, comparisons are utterly meaningless.



#21 fourfives

fourfives

    Lift Off

  • -----
  • topic starter
  • Posts: 19
  • Joined: 20 Jan 2020

Posted 23 July 2021 - 05:44 PM

Hist.JPG



#22 bobzeq25

bobzeq25

    ISS

  • *****
  • Posts: 25,030
  • Joined: 27 Oct 2014

Posted 23 July 2021 - 11:01 PM

It's not clear.  Is that supposed to be an unstretched light?  If so, the histogram is crazy.

 

Here's what the histogram of an unstretched light is supposed to look like.  And the light itself.  This was taken with an NBZ filter, similar to the Triad.  30" Bortle 7 F2.  This is the calibrated and debayered light.  Click on the picture to enlarge.

 

I know, you almost can't see the histogram.  That's what the unstretched histogram is supposed to look like.  In the next post I'll stretch it.  Note the midtone slider has been moved over, and the diagonal line (showing unstretched) is now curved (stretched).

 

a jellyfish light debayered calibrated.jpg


Edited by bobzeq25, 23 July 2021 - 11:15 PM.


#23 bobzeq25

bobzeq25

    ISS

  • *****
  • Posts: 25,030
  • Joined: 27 Oct 2014

Posted 23 July 2021 - 11:11 PM

The above light's histogram, when stretched.

 

Here's the final image, 239 of those subs calibrated, stacked, and processed.

 

https://www.astrobin.com/kis712/

 

If you upload an unstretched light and bias (straight from the camera) I'll run the numbers that will show whether the light was properly exposed.

 

a jellyfish light debayered calibrated streched.jpg


Edited by bobzeq25, 23 July 2021 - 11:18 PM.


#24 fourfives

fourfives

    Lift Off

  • -----
  • topic starter
  • Posts: 19
  • Joined: 20 Jan 2020

Posted 24 July 2021 - 10:28 AM

Sure, uploaded those 2 here:

 

https://drive.google...IBH?usp=sharing

 

If you look prior to debayer I have 3 spikes for each color 1/3'd of the way over on the historgram.  If you debayer everything is still less than 1/3rd. 


Edited by fourfives, 24 July 2021 - 10:35 AM.


#25 bobzeq25

bobzeq25

    ISS

  • *****
  • Posts: 25,030
  • Joined: 27 Oct 2014

Posted 24 July 2021 - 02:41 PM

The 533 is a 12bit camera, so you do the calculation in 12 bits.  All numbers below came from either PixInsight or the ZWO website.

 

The light has a median value of 403 ADU.  The bias, 25 ADU.  The corrected ADU of the light is 378 ADU.  Since you're where 1 ADU = 1 electron, that's 378 electrons.

 

The read noise of the camera, gain 100, is 1.5 electrons.  Squared it's 2.25 electrons.

 

The corrected median level of the light is 168X the read noise squared.  You'd like it to be 5-10X.  Your at least a factor of 16 overexposed.  300/16 is 19 seconds.  As I said, Glover (the author of Sharpcap) recommends approximately 12 seconds.  Close enough.

 

This is the method I (and others) have used to set subexposure for years.  You can see the results on my astrobin.

 

Glover's video explains the derivation of the method.  Another good explanation (recommending 5-10X read noise squared) makes up most of a chapter in this superb book.  Warning - math ahead.  <smile> .

 

"The Astrophotography Manual: A Practical and Scientific Approach to Deep Sky Imaging" by Chris Woodhouse

 

https://www.amazon.c...h/dp/1138055360

 

The virtue of this approach is that it includes all 3 of the most important factors.  Light pollution level, optical speed (included in the corrected mean of the light), and read noise of the camera.

 

30 seconds could work.  You'd need 340 subs.  Such is life in Bortle 9.

 

300 seconds is very overexposed.

 

Overexposure cannot be corrected in processing.  You can make a stab at correcting vignetting with some sophisticated methods, but taking flats (and bias, necessary for flats to work) is a far superior approach.

 

Bottom line.  You need less overexposure and flats to get good data.  You need good data to process it well.  There is no magic way to fix bad data in processing.

 

1/3 over is proper for the stretched histogram on the back of a DSLR.  A linear histogram for an astro camera does not extend far from the lefthand edge.  See post #22.  The corrected mean of that light was 35 electrons.  Read noise squared was 2.25.  A factor of 15X, a bit overexposed, but you don't have to get it exact.


Edited by bobzeq25, 24 July 2021 - 03:03 PM.

  • Rasfahan and cucubits like this


CNers have asked about a donation box for Cloudy Nights over the years, so here you go. Donation is not required by any means, so please enjoy your stay.


Recent Topics






Cloudy Nights LLC
Cloudy Nights Sponsor: Astronomics