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

NGC 7000 - looking for advice

  • Please log in to reply
13 replies to this topic

#1 Entr04y

Entr04y

    Lift Off

  • -----
  • topic starter
  • Posts: 5
  • Joined: 03 May 2022

Posted 23 June 2022 - 06:34 PM

Hi all, just getting started with astrophotography, and about to head out on a roadtrip taking us through the southwest and northwest (New Mexico, Arizona, Wyoming, etc) and I want to be sure I make the most of the dark skies.  I'm looking for advice on both capture and processing.

 

My Equipment is as follows:

iOptron GEM45
RedCat 51
Canon R5, unmodified
Aurora Flat Panel
N.I.N.A for telescope control and capture, shooting in raw, using freeimage to save.
PixInsight for processing

 

The other night I took about 200 frames of NGC 7000 from my backyard, transparency wasn't perfect, and some thin clouds rolled in at the end, so I wound up rejecting quite a few frames.  Ultimately I wound up with:

Bortle 7 skies, all shot at ISO 800, no filter
129 Lights @ 60s
25 Darks @ 60s
50 Flats @ 0.1s
25 Flat-Darks @ 0.1s

 

This is an initial attempt at processing:

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

 

It was processed as follows:
WBPP all the way through integration
Automatic Background Extraction: Division, then rerun as subtraction
Photometric Color Calibration
SCNR
Multiscale Linear Transform
Histogram Transform (just dragged the Screen Transfer Function to HT and saved the autostretch)

 

My processing skills aren't great as I'm just learning pixinsight, and as soon as I try to stretch I tend to make the images worse, so I'm looking to get pointers on all the linear stages prior to figuring out stretch/saturation/etc

 

I've put the raw data up if anyone feels like reviewing it / messing with it, and in the processed folder are some variations between ABE and DBE and running both through starnet.

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

 

Issues I'd like advice on:

1) I can't seem to get rid of vignetting in my images, it's relatively faint in this one, but it is still there, it's fairly obvious in the starless versions

2) Flat exposures.  I can't seem to make heads or tails of the statistics values in PIxInsight, the raw images are 14 bit and if I put the readout in 16bit in Pixinsight the max value is ~16k and the median is ~8k which is where I think I want it, but if I put it in 14bit, the max value drops to 4096 and the median to 2287.  What I'm not confident about is whether PixInsight is scaling the numbers up in 16bit or reporting the actual 14bit value and scaling down in 14bit mode (in other words, is my median really at ~50% or is it at ~25% and underexposed?)  I can't really see any vignetting in the flats unless I stretch the image way beyond anything reasonable.  NINA seemed to think the exposures were in the middle.

3) Is my Light exposure too long?  I feel like Dave in 2001 - "My God, its full of stars"  I'm not getting much structure in the nebula though.  I realize it's Bortle7 and a stock DSLR, but I've seen others post images with similar setups and much better results...

4) How did I do on the processing so far?  I can't tell if it's color noise or if there's just that much nebula around the entire frame.  It's hard to compare with the images I've found on-line because they are all in different color pallets and seem to have much better resolution than I'm getting.

5) I do have an Astronomik CLS clip in filter, but I see conflicting info on whether it will help or hurt when imaging DSO, so I decided to try without this time... anyone have any exprience / tips / when to or not to use it?

 

Thanks!  I appreciate the help!

 

 



#2 pedxing

pedxing

    Mercury-Atlas

  • *****
  • Posts: 2,548
  • Joined: 03 Nov 2009
  • Loc: SE Alaska

Posted 23 June 2022 - 07:22 PM

I'd say you are doing pretty good.

 

It's supposed to be full of stars, it's the milky way.

 

You could use more color in your stars, but I think that's processing, not over-exposure.

 

If you shot this in a Bortle 7, based on your image I wouldn't use the CLS - I think it would do more harm than good (although you are certainly free to experiment and decide for yourself).

 

A couple of processing ideas would be to increase the saturation (using a saturation curve) before you do histogram transformation and to do the histogram transformation in multiple steps instead of a single step with the STF auto setting. That will give you more control over the stretch and hopefully a bit more color/contrast.

 

All in all I would say you are off to a very good start!


Edited by pedxing, 23 June 2022 - 07:23 PM.


#3 idclimber

idclimber

    Mercury-Atlas

  • *****
  • Posts: 2,565
  • Joined: 08 Apr 2016
  • Loc: McCall Idaho

Posted 23 June 2022 - 07:44 PM

Here is your first flat as viewed with Histogram Transformation. 

 

Screen Shot 2022-06-23 at 6.37.02 PM.jpg

 

Your light is biased toward the blue. You are also close to max on that channel toward the center of the image. A warmer light source would help as well as slightly shorter exposures. As at minimum you  want the green where the red is now. Average ADU on blue is 54,371 and you have max pixels of 65k in all the channels. 

 

Screen Shot 2022-06-23 at 6.42.19 PM.jpg


Edited by idclimber, 23 June 2022 - 07:45 PM.


#4 Entr04y

Entr04y

    Lift Off

  • -----
  • topic starter
  • Posts: 5
  • Joined: 03 May 2022

Posted 23 June 2022 - 09:13 PM

I'd say you are doing pretty good.

 

It's supposed to be full of stars, it's the milky way.

 

You could use more color in your stars, but I think that's processing, not over-exposure.

 

If you shot this in a Bortle 7, based on your image I wouldn't use the CLS - I think it would do more harm than good (although you are certainly free to experiment and decide for yourself).

 

A couple of processing ideas would be to increase the saturation (using a saturation curve) before you do histogram transformation and to do the histogram transformation in multiple steps instead of a single step with the STF auto setting. That will give you more control over the stretch and hopefully a bit more color/contrast.

 

All in all I would say you are off to a very good start!

Thanks, I'll give playing around with the saturation before histogram a try.  In this case I just dropped the STF on the histogram to get a presentable jpg as I'm trying to get the kinks out of the preprocessing before I start working on post...  I'll have to play around with the CLS filter some more as well.



#5 Entr04y

Entr04y

    Lift Off

  • -----
  • topic starter
  • Posts: 5
  • Joined: 03 May 2022

Posted 23 June 2022 - 09:34 PM

Here is your first flat as viewed with Histogram Transformation.

attachicon.gifScreen Shot 2022-06-23 at 6.37.02 PM.jpg

Your light is biased toward the blue. You are also close to max on that channel toward the center of the image. A warmer light source would help as well as slightly shorter exposures. As at minimum you want the green where the red is now. Average ADU on blue is 54,371 and you have max pixels of 65k in all the channels.

attachicon.gifScreen Shot 2022-06-23 at 6.42.19 PM.jpg

How did you get to those values? If I take the file before debayering the statistics on my system look like this (the system is not letting me post jpgs for some reason,maybe because I am new here):

https://drive.google...iew?usp=sharing
_2022_06_19_01_50_26__38c_0_10s_0000
K
count (%) 100.00000
count (px) 44747433
mean 9172.9
median 9216.0
avgDev 2692.2
MAD 3312.1
minimum 4423.0
maximum 16383.0

And after debayering:
https://drive.google...iew?usp=sharing
_2022_06_19_01_50_26__38c_0_10s_0000_RGB_VNG
R G B
count (%) 100.00000 100.00000 100.00000
count (px) 44747433 44747433 44747433
mean 5462.012 11320.936 8588.675
median 5490.500 11462.678 8679.000
avgDev 169.134 380.124 296.112
MAD 146.142 253.049 221.648
minimum 4278.167 8617.000 6619.719
maximum 12829.000 16383.000 13352.000

and the histogram is on the left:
https://drive.google...iew?usp=sharing

Am I missing a step or two to evaluate flats? The color bias is in a different order as well.

Edit: it just occurred to me, there’s a setting in pixinsight for “pure raw” or something along those lines, if you don’t have that set PixInsight will debayer/stretch/process the raw image, maybe that’s what’s different?

Edited by Entr04y, 23 June 2022 - 10:44 PM.


#6 idclimber

idclimber

    Mercury-Atlas

  • *****
  • Posts: 2,565
  • Joined: 08 Apr 2016
  • Loc: McCall Idaho

Posted 23 June 2022 - 11:18 PM

It is possible that PI misread the file, but on my install it automatically debayered the image on opening RGGB. I do not work with Canon files much so I am not certain what happened. 



#7 idclimber

idclimber

    Mercury-Atlas

  • *****
  • Posts: 2,565
  • Joined: 08 Apr 2016
  • Loc: McCall Idaho

Posted 23 June 2022 - 11:45 PM

I opened the same file in Photoshop and it matches what I have in PI. The only difference is the histogram is further to the right because of the correction that the RAW plug in makes with a DSLR image. It definitely shows the same blue bias to the image once I back off on the exposure slider.  



#8 idclimber

idclimber

    Mercury-Atlas

  • *****
  • Posts: 2,565
  • Joined: 08 Apr 2016
  • Loc: McCall Idaho

Posted 24 June 2022 - 12:03 AM

After a bit of reading I believe the R5 has a variable bit depth depending on what mode it is in. I think what could be happening is your version of PI is incorrectly assuming it is 16 bit when it is actually lower. To confirm, what does the flat look like when it is viewed on the back of the camera. Does it look nearly white or does it look dark? How about the histogram if you view that on the camera? 

 

This is a relatively new camera so it may just be the raw converter on my Mac not handling the file correctly. 



#9 Ryou

Ryou

    Messenger

  • -----
  • Posts: 449
  • Joined: 10 Apr 2020

Posted 24 June 2022 - 01:20 AM

After a bit of reading I believe the R5 has a variable bit depth depending on what mode it is in. I think what could be happening is your version of PI is incorrectly assuming it is 16 bit when it is actually lower. To confirm, what does the flat look like when it is viewed on the back of the camera. Does it look nearly white or does it look dark? How about the histogram if you view that on the camera? 

 

This is a relatively new camera so it may just be the raw converter on my Mac not handling the file correctly. 

I have an R6 for normal photography and this shares a processor with the R5, though has a different sensor, so I'm not 100% sure if this is applicable... However I can indeed confirm on the R6 the bit depth changes depending on the mode. I think for the max bit depth you need to be in single shot mechanical shutter as the electronic shutter actually lowers the bit depth. Additionally the higher speed shooting modes (which may be tempting to use for flats) can also lower the bit depth. Given that they share an image processor/control chip I don't see why the same shifting scale wouldn't be true on the R5, however don't have the first hand experience to say this for sure.

 

Hope that helps


Edited by Ryou, 24 June 2022 - 01:21 AM.

  • idclimber likes this

#10 ngc7319_20

ngc7319_20

    Mercury-Atlas

  • *****
  • Posts: 2,681
  • Joined: 25 Oct 2015
  • Loc: MD

Posted 24 June 2022 - 05:48 AM

As far as the lack of nebulosity goes, the IR block in the stock camera certainly is not helping...  H-alpha is near the IR cutoff, so there may be much variation between camera models and even cameras.  The stock R5 images I've seen were mostly bright M-objects and star fields.



#11 Entr04y

Entr04y

    Lift Off

  • -----
  • topic starter
  • Posts: 5
  • Joined: 03 May 2022

Posted 24 June 2022 - 10:01 AM

Thanks all, I wasn't aware of the R5 using different bit depths based on shutter / shooting mode.

 

On the RAW formatting issue, from what I understand from several of the pixinsight tutorials (primarily Adam Blocks' fundamentals course) setting PixInsight to Pure Raw is necessary because the default settings will perform much the same functions as Adobe RAW and other plugins, namely debayer, stretch, apply white balance, and some other manipulations that make the images non-linear on import.  That being said, I loaded the above flat on the SD card and the cameras histogram does show the peak almost all the way to the right... of course this is after it applies its JPEG manipulation to it.  NINAs representation as I was shooting it was that it was exposing near the middle, so it must be working from the linear, non-debayered image.  I guess the question then is do I expose flats for the values while the image is still linear or for the "back of camera" version?

 

Regarding bit depth, the Lights and Darks were shot in bulb mode, which uses electronic first curtain mode and the Flats and DarkFlats were shot in manual mode with electronic first curtain.  From the specs, mechanical and electronic first curtain both result in 14 bit images, hf+ gives 13 bit images, and full electronic shutter provides 12 bit images, so the CR3 files from my flats *should* be 14bit.  I'm not sure how to definitively identify the bit depth in the file.  The max values seem to make sense for a 14 bit file being read into pixinsight as a 16 bit file without scaling. (no values over ~16k)  The only actual place I could find from Canon where it describes the bit depth is on the European site under file type.

 

I expect the H-alpha performance to not be spectacular, but looking at RNClarks R5 gallery, he's got some pictures of the Veil Nebula with much shorter integration time that show quite a bit of detail in the nebula.  Looking at this article, the author has got nearly the same image as I do, but again, there's more detail in the nebula with lower integration time... both do have longer subs though, so that might be part of that issue.  It could also probably be down to post processing and color manipulation though.  It's also hard to get good transparency here during the summer, so there's definitely some haze in my LIghts. 

 

At this stage, I want to primarily be sure my data capture is sorted... I can always re-process good data, but if the data is bad, there's not much we can do with it!


Edited by Entr04y, 24 June 2022 - 10:16 AM.


#12 ngc7319_20

ngc7319_20

    Mercury-Atlas

  • *****
  • Posts: 2,681
  • Joined: 25 Oct 2015
  • Loc: MD

Posted 24 June 2022 - 12:02 PM

Thanks all, I wasn't aware of the R5 using different bit depths based on shutter / shooting mode.

 

I expect the H-alpha performance to not be spectacular, but looking at RNClarks R5 gallery, he's got some pictures of the Veil Nebula with much shorter integration time that show quite a bit of detail in the nebula....

 

Yes, but the structure of the Veil and North American are very different....  Probably stock R5 is not optimal for North American...



#13 idclimber

idclimber

    Mercury-Atlas

  • *****
  • Posts: 2,565
  • Joined: 08 Apr 2016
  • Loc: McCall Idaho

Posted 24 June 2022 - 12:17 PM

I use a mono camera so I have not spent any time optimizing PI for OSC cameras. I am also currently using a copy of PI on my new Mac and not my dedicated PI Linux machine. I was able to make some changes my RAW setup and open the image. 

 

The mean displayed this way indeed shows 9172 on 16 bit scale. It should be closer to 30k so is underexposed. foreheadslap.gif Oddly when I apply the debayer script it shows green as the brightest channel just as yours did. There is definitely something wrong with the way PI was importing this file using the default raw plug in. 

 

Ultimately the flats need to be calibrated by a master bias. They should be in the same exact 14 bit format. If the bit depth changes with longer exposures which I think it does, then you would want the darks to match them. 

 

It would be better if the camera could be set so all the frames are in the same 16 bit mode. It sounds like that short exposures are one mode and long exposures are in another.  Probably the best way to double check bit depth without digging into the actual data would be file size. 

 

Getting this to work may require some trial and error. I would suggest taking a sample set of flats at various exposures using a fixed light panel. Using Nina make note of there average ADU it reports. Then import the set to PI looking for the variant that gets you closest to 30k (16bit mode). Then you know what average ADU as reported by Nina gets you the correct middle exposure. To analyze lights you ignore what Nina says and simply import into PI for analysis.

 

Another idea related to the vignetting is to take a couple sets of flats at various exposures and see what exposure calibrates your lights the best. You could even skip dark calibration for this test. You will need matching bias though. 


Edited by idclimber, 24 June 2022 - 12:18 PM.


#14 Entr04y

Entr04y

    Lift Off

  • -----
  • topic starter
  • Posts: 5
  • Joined: 03 May 2022

Posted 24 June 2022 - 02:58 PM

Ok, a little more information... I used adobe bridge to read out what it thought the bit depth of the CR3 files is and it says all of the files (LIghts, Darks, Flats, DarkFlats) are 16 bit.  I opened one of each in PI and looked at the statistics and they all have a maxium value of 16383.0, with the exception of the dark flat having a maximum of 12559.0.  Given 16383 is the max value for a 14bit integer and 12559 is larger than a 13 bit number and still in the 14bit range, my conclusion is that the camera is writing unscaled 14bit values into a 16 bit container, which would mean I'd want to target something in the 8k range to hit the midpoint, assuming that my flat exposure targets are supposed to be based on the raw, undebayered, linear image and not the "back of camera" histogram.

 

On another thread elsewhere, someone suggested that the exposure time for my flats might be too short as they had issues with rings/vignetting even at the correct exposure values at high shutter speeds and I should try diffusing / dimming to get the exposure time closer to .5 or a full second to avoid that issue.  My next imaging run I'll collect some LIghts, then do a suite of flats / flat dark runs and see if I can figure out which exposures work best.


  • idclimber likes 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