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

Under/over sampling

  • Please log in to reply
14 replies to this topic

#1 gfunkernaught

gfunkernaught

    Mariner 2

  • -----
  • topic starter
  • Posts: 212
  • Joined: 10 Feb 2020
  • Loc: New York

Posted 04 March 2021 - 05:34 PM

I have a Canon EOS Ra with a pixel size of 5.36 microns.  I am looking to get a refractor for larger targets.  I use this tool  https://astronomy.to...ccd_suitability to figure out which glass is best for my camera and typical seeing conditions, which are my bortle 7-8 backyard.  I saw a scope that has a 559mm focal length that I am interested in.  According to the tool, it will have some undersampling under good seeing conditions.  I've read that undersampling leads to pixelated stars and overall lower res images, which can be dealt with using drizzle.  But the thing is I took a pic of Andromeda using a 60mm guide scope with a focal length of 240mm and didn't notice any pixelated stars, not the best quality of course, since it is a guide scope.  Take a look at the image below.  I had to squash it so it would fall within the upload guidelines.  So my question is:  How much undersampling is too much undersampling?  The scope I am looking at is the Zenithstar 81 APO.  The field of view is what I am looking for.

Attached Thumbnails

  • Andromeda_60mm_2.jpg

Edited by gfunkernaught, 04 March 2021 - 05:36 PM.

  • *skyguy* and boxcorner like this

#2 *skyguy*

*skyguy*

    Skylab

  • *****
  • Posts: 4,042
  • Joined: 31 Dec 2008
  • Loc: Western New York

Posted 04 March 2021 - 06:22 PM

I image using my 12" LX200 scope at 1920mm (f/6.3) with an old SBIG ST-9E camera producing undersampled images (2.15 arc-seconds/pixel). The seeing in my area is around 3 arc-minutes, sometimes a little more and sometimes a little less. 

 

I usually use Drizzle stacking with a 2x upscale factor along with image processing to to restore resolution.  I'm happy with my results ...

 

NGC 4565

 

NGC4565_web_160mL_30mR_30mG-50mB.jpg

 

 


  • zakry3323, DJL and boxcorner like this

#3 BQ Octantis

BQ Octantis

    Skylab

  • *****
  • Posts: 4,467
  • Joined: 29 Apr 2017
  • Loc: Red Centre, Oz

Posted 05 March 2021 - 04:58 AM

For some reason, over- and undersampling relative to the 1-2 arcec per pixel rule seems to be a hot topic this year. Not to be a naysayer, but I think for most imagers this is a red herring. If you've made it far enough to where it's the shapes of your stars due to sampling that are limiting your image quality at full sensor scale (vice your tracking, your capture, or your processing), then maybe this would be a factor. But it's almost always SNR that limits the usable image size to well below the full sensor size. Using the fastest aperture possible is what will get you to usable images at full sensor resolution. Then it's aperture size that gives you better sampling.

 

I shoot a lot at 200mm f/2.8 with my 600D/T3i with 4.3µm pixels, which is grossly undersampled (4.4"/pixel). But the flux at f/2.8 is such that I can get very usable data at 100% scale in just 2-4 hours with an unmodded sensor. And here's what I can get with 200% bicubic upsampling in Photoshop:

 

(Click for full size @ 200%)

gallery_273658_12412_119055.jpg

 

In a high light pollution environment, higher flux also gives you plenty of light to filter…

 

BQ



#4 gfunkernaught

gfunkernaught

    Mariner 2

  • -----
  • topic starter
  • Posts: 212
  • Joined: 10 Feb 2020
  • Loc: New York

Posted 05 March 2021 - 09:48 AM

From my yard, seeing quality can vary greatly, mostly ok, sometimes poor, and very rare occasion good. Is there a way to determine or quantify seeing quality based on temperature and humidity? I think that scope will be good for my yard, but my concern is taking it to a dark site where seeing is either good or exceptional which, according to the tool, exceptional seeing quality could lead to undersampling. I don't want to buy another camera, so finding the right glass for it is crucial. I currently have a C8 with the f6.3 reducer which works great with my camera. Larger targets are harder to do with the scope, I'd have to make mosaics, which sounds like fun. I wonder how much detail I'd lose going from 1280mm to 559mm with the same 5.36u pixel size.

#5 BQ Octantis

BQ Octantis

    Skylab

  • *****
  • Posts: 4,467
  • Joined: 29 Apr 2017
  • Loc: Red Centre, Oz

Posted 05 March 2021 - 09:58 AM

Use Meteoblue. They use vertical temperature and wind speed profiles to generate a seeing prediction.



#6 gfunkernaught

gfunkernaught

    Mariner 2

  • -----
  • topic starter
  • Posts: 212
  • Joined: 10 Feb 2020
  • Loc: New York

Posted 05 March 2021 - 10:26 AM

Wow that site is better than timeanddate.com! Thanks for that.

#7 BQ Octantis

BQ Octantis

    Skylab

  • *****
  • Posts: 4,467
  • Joined: 29 Apr 2017
  • Loc: Red Centre, Oz

Posted 05 March 2021 - 04:47 PM

No worries! Note that I only use it for deciding whether to bother doing planetary imaging any given night. It's mainly the jet stream that I watch—if it's above 50 m/s, there's little point in even setting up—especially if the two seeing indexes are also poor.


  • tim53 likes this

#8 gfunkernaught

gfunkernaught

    Mariner 2

  • -----
  • topic starter
  • Posts: 212
  • Joined: 10 Feb 2020
  • Loc: New York

Posted 05 March 2021 - 06:08 PM

Just noticed your avatar pic of Mars, very nice!  My best pic of mars looks like something you'd see in a filtered texture from a 90s game :(

 

Where on that site do you see seeing index?



#9 BQ Octantis

BQ Octantis

    Skylab

  • *****
  • Posts: 4,467
  • Joined: 29 Apr 2017
  • Loc: Red Centre, Oz

Posted 05 March 2021 - 06:18 PM

Here:

 

Screen Shot 2021-03-06 at 8.45.56 AM.png

 

Not sure what city in New York you're in, but here's the prediction page for NYC:

 

https://www.meteoblu...america_5128581

 

With seeing like that, I definitely would prioritize flux over sampling!

 

BQ

 

P.S. The avatar pic was day 2 of the dust storm that we tracked on CN last November. I shot it with my lowly 600D/T3i through a Fujiyama orthoscopic eyepiece on my 7-in Skywatcher Mak:

 

post-273658-0-82061700-1605271005.png

 

I've been meaning to change it…


Edited by BQ Octantis, 05 March 2021 - 06:27 PM.

  • tim53 likes this

#10 gfunkernaught

gfunkernaught

    Mariner 2

  • -----
  • topic starter
  • Posts: 212
  • Joined: 10 Feb 2020
  • Loc: New York

Posted 05 March 2021 - 06:38 PM

Cool thanks.  This is what I got with my C8 and EOS Ra...severe meh:

 

 

Attached Thumbnails

  • Mars_Tv12500s_20000iso_1344x896_20201009-01h24m36s_000001_pipp.jpg


#11 BQ Octantis

BQ Octantis

    Skylab

  • *****
  • Posts: 4,467
  • Joined: 29 Apr 2017
  • Loc: Red Centre, Oz

Posted 05 March 2021 - 06:51 PM

That's just too big. It's unrealistic to get static details at this scale from an 8-in aperture. The Airy disk in my image is there for a purpose: the size of any detail smaller than that can't be trusted because it is below the point spread limit of the aperture. And any detail 10× smaller than that is either noise or a processing artifact. And if you didn't sample at that size, Drizzle isn't going to save you—especially for such a large blowup.

 

Looking at your image, the detail tips over at between 6 and 8% scaling. Here it is at 7.75%—not bad at all:

 

post-324943-0-58643000-1614987535.jpg

 

BQ



#12 gfunkernaught

gfunkernaught

    Mariner 2

  • -----
  • topic starter
  • Posts: 212
  • Joined: 10 Feb 2020
  • Loc: New York

Posted 06 March 2021 - 11:53 AM

For some reason, over- and undersampling relative to the 1-2 arcec per pixel rule seems to be a hot topic this year. Not to be a naysayer, but I think for most imagers this is a red herring. If you've made it far enough to where it's the shapes of your stars due to sampling that are limiting your image quality at full sensor scale (vice your tracking, your capture, or your processing), then maybe this would be a factor. But it's almost always SNR that limits the usable image size to well below the full sensor size. Using the fastest aperture possible is what will get you to usable images at full sensor resolution. Then it's aperture size that gives you better sampling.

 

I shoot a lot at 200mm f/2.8 with my 600D/T3i with 4.3µm pixels, which is grossly undersampled (4.4"/pixel). But the flux at f/2.8 is such that I can get very usable data at 100% scale in just 2-4 hours with an unmodded sensor. And here's what I can get with 200% bicubic upsampling in Photoshop:

 

(Click for full size @ 200%)

gallery_273658_12412_119055.jpg

 

In a high light pollution environment, higher flux also gives you plenty of light to filter…

 

BQ

Finally got a chance to see this in hi res, pretty good!  The resolution and detail is similar to my 240mm guide scope and 5.36 pixel size, which is undersampled.  I'm thinking the scope I want which is 559mm focal length is the right path for my camera.  Maybe a bit longer too, like 700mm.  If I use a reducer/flattener with 559mm I'll end up undersampled further.



#13 gfunkernaught

gfunkernaught

    Mariner 2

  • -----
  • topic starter
  • Posts: 212
  • Joined: 10 Feb 2020
  • Loc: New York

Posted 06 March 2021 - 11:55 AM

That's just too big. It's unrealistic to get static details at this scale from an 8-in aperture. The Airy disk in my image is there for a purpose: the size of any detail smaller than that can't be trusted because it is below the point spread limit of the aperture. And any detail 10× smaller than that is either noise or a processing artifact. And if you didn't sample at that size, Drizzle isn't going to save you—especially for such a large blowup.

 

Looking at your image, the detail tips over at between 6 and 8% scaling. Here it is at 7.75%—not bad at all:

 

attachicon.gifpost-324943-0-58643000-1614987535.jpg

 

BQ

Yeah I shouldn't have drizzled it lol.  Looks fine as a thumbnail I suppose :)  I imaged that 2.5x magnification.  Any more or less was just unusable.  Also used an ADC for that.



#14 BQ Octantis

BQ Octantis

    Skylab

  • *****
  • Posts: 4,467
  • Joined: 29 Apr 2017
  • Loc: Red Centre, Oz

Posted 06 March 2021 - 04:54 PM

Ironically, I grossly oversample my planetary captures. I shoot at 3.4 to 3.8× over prime through my f/15 Mak, which works out to f/52-f/58. That's 0.08 - 0.09 arcsec/px. For a static image I'll typically scale that down to 50-80%, but for an animation I can actually use that and more—I can use in-browser zoom to push to another 1.7× or 0.05 arcsec/px. For instance, here's my best Jupiter from last year:

 

https://www.cloudyni...86_12553308.png

 

Use your browser to zoom to 170%; then stare at the animation to find little details that persist across the disk. Once you've had your fill, use your browser to scale down to 50 - 80% to enjoy the smoothness of the image. Finally, zoom to 30% to see the recommended sampling size.

 

BQ


Edited by BQ Octantis, 06 March 2021 - 04:58 PM.


#15 BQ Octantis

BQ Octantis

    Skylab

  • *****
  • Posts: 4,467
  • Joined: 29 Apr 2017
  • Loc: Red Centre, Oz

Posted 06 March 2021 - 06:07 PM

As to which DSO scope to get, it always helps to have a reference. Pixel angular resolution is a trivial calculation (literally just p/f ; maybe the units conversion is a little challenging for those unfamiliar with π radians), but flux equivalence varies as the area of the entrance pupil and the surface area of each pixel—so the square of the focal ratio and the square of the pixel pitch. As with many things, I find it helpful to build a spreadsheet with the basic calculations to understand the what-if's of the trade to be made. For me, time is the most important dependent variable—my tripod can only handle 30 sec subs at 200mm unguided without stars going oblong. And my time in the outback is limited, so I want the shortest total integration time possible.

 

Comparing the EOS Ra on a Zenithstar 81 apo to my 600D/T3i on a 200mm f/2.8, I get the following comparison of factors:

 

Screen Shot 2021-03-07 at 8.19.36 AM.png

 

Line 10: The apo's diffraction limit is 1.7 arcsec.

Line 13: The Ra sampling is at 2 arcsec/px.

Line 18: The flux from the Ra-apo is 1/4 that of my 600D-200mm

Line 19: The Ra-apo would need 4× longer subs and 4× total integration for equivalent SNR at full sensor scale.

Lines 24-25: The field of view of the Ra-apo is roughly half that of the 600D-200mm.

Lines 26-27: These are the dimensions of M31.

Lines 28-29: This is how much of the sensor is taken up by M31. M31 would take up 81% of the width and 42% of the height of the sensor on the Ra-apo setup.

 

The Ra has far better Ha response in the red channel (4× per what I've read) than my unmodded 600D, so the red channel flux on Ha targets would be similar between the two setups. But the green and the blue sensitivities of the two cameras are the same, so non-Ha targets would need 4× longer with the Ra-apo than the 600D-200mm. You can always scale the image down—SNR at the image pixel level scales with the inverse square of the image scaling (a crude form of binning). But the extra pixels from the Ra would still come at a processing and storage price of 1.7× over the 600D (Line 31).

 

BQ


Edited by BQ Octantis, 06 March 2021 - 06:16 PM.



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