Go Back   IceInSpace > Equipment > Astrophotography and Imaging Equipment and Discussions

Reply
 
Thread Tools Rating: Thread Rating: 5 votes, 5.00 average.
  #41  
Old 24-06-2015, 07:34 PM
Atmos's Avatar
Atmos (Colin)
Ultimate Noob

Atmos is offline
 
Join Date: Aug 2011
Location: Melbourne
Posts: 6,980
Quote:
The 814 didn't come up so well. I guess in good seeing though it would be hard to beat as it is giving .66 arc secs/pixel which I thought was ideal under good conditions. Roland goes for arc sec /pixel based on average seeing of around 3 arc secs. I think my seeing at home is more 2-3 so .66 arc secs to 1 arc sec would be ideal with 3X sampling. The 8300 sensor gives you .96 arc secs hence he goes for that.
The problem that we don't want to get caught up on is trying to have the most "efficient" system... there is no perfect system... you always have to make compromises. With the same telescope, smaller pixels will ALWAYS mean longer imaging time to get the same SNR. Conversely, larger pixels will NEVER give better resolution unless you oversample too much.

It is difficult to decide between 0.67 and 0.8 resolution. I personally would go for the 0.67 as my DSLR on my 10" F/10 has the same resolution. On a good night I can get about 3 FWHM. With a CCD as opposed to a Bayer Matrix, that would drop even further, as a guess, much closer to 2 which is perfect sampling (what I am aiming towards).

As Ray mentioned, all of this is a good way to give an indication of how long your exposures need to be to compare one optical system to another. In the real world though, it is a little pointless when trying to calculate what camera/telescope to use. Personally, I believe that your pixel scale should pretty much be the determining factor. The reason for this comes down to what you're wanting to do.

I personally am more interested in trying to get as close to possible to perfect sampling, it's the scientist coming out in me. I would prefer to be able to resolve every star that I possibly can within a globular cluster than shorter exposure times or wider FOV. All comes down to what you want though
Reply With Quote
  #42  
Old 24-06-2015, 08:34 PM
Shiraz's Avatar
Shiraz (Ray)
Registered User

Shiraz is offline
 
Join Date: Apr 2010
Location: ardrossan south australia
Posts: 4,918
Quote:
Originally Posted by Atmos View Post
As Ray mentioned, all of this is a good way to give an indication of how long your exposures need to be to compare one optical system to another. In the real world though, it is a little pointless when trying to calculate what camera/telescope to use. Personally, I believe that your pixel scale should pretty much be the determining factor. The reason for this comes down to what you're wanting to do.
Overall agree with your philosophy, but think that you are missing the primary real-world driver in the direction of efficient imaging. With weather like we are currently suffering, opportunities to image are few and far between and having the most sensitive system (with appropriate resolution) becomes critical in getting good enough data on desired objects. Since relatively minor changes in system geometry can have really significant impacts on imaging time, I think that a tool that allows one to choose between different options on the basis of sensitivity may serve a valuable purpose.

Last edited by Shiraz; 24-06-2015 at 08:55 PM.
Reply With Quote
  #43  
Old 24-06-2015, 09:23 PM
Atmos's Avatar
Atmos (Colin)
Ultimate Noob

Atmos is offline
 
Join Date: Aug 2011
Location: Melbourne
Posts: 6,980
Oh I totally agree with you Ray, don't get me wrong! For most of the imagers on here, the 0.8-1.8 pixel scale is the best fit for them. They are really good image scales for all round sky images.
I however am a researcher, I care more about resolution
Reply With Quote
  #44  
Old 27-06-2015, 02:57 PM
rmuhlack's Avatar
rmuhlack (Richard)
Professional Nerd

rmuhlack is offline
 
Join Date: Nov 2011
Location: Strathalbyn, SA
Posts: 916
Can this calculation be used for comparing bin 1x1 with bin 2x2 ? Is it simply a matter of doubling the pixel size for a 2x2 binned image ?

Similarly, how about a 2x drizzle? Is this equivalent to using a camera with half the pixel size?
Reply With Quote
  #45  
Old 27-06-2015, 05:24 PM
Atmos's Avatar
Atmos (Colin)
Ultimate Noob

Atmos is offline
 
Join Date: Aug 2011
Location: Melbourne
Posts: 6,980
Quote:
Originally Posted by rmuhlack View Post
Can this calculation be used for comparing bin 1x1 with bin 2x2 ? Is it simply a matter of doubling the pixel size for a 2x2 binned image ?

Similarly, how about a 2x drizzle? Is this equivalent to using a camera with half the pixel size?
When binning with a CCD it does work that way as it does a signal amplifier when it bins while keeping the same read noise. With DSLRs though, from what I have heard it doesn't work so well, that's just because binning isn't a good idea with them in general.

As for drizzle it doesn't make a difference because the calculations are for each individual image. It is a calculation of how much sky each pixel covers. As dizzling is a group of images with a slight offset, each individual frame is is what is being calculated.
Reply With Quote
  #46  
Old 27-06-2015, 05:54 PM
rmuhlack's Avatar
rmuhlack (Richard)
Professional Nerd

rmuhlack is offline
 
Join Date: Nov 2011
Location: Strathalbyn, SA
Posts: 916
Quote:
Originally Posted by Atmos View Post
When binning with a CCD it does work that way as it does a signal amplifier when it bins while keeping the same read noise. With DSLRs though, from what I have heard it doesn't work so well, that's just because binning isn't a good idea with them in general.

As for drizzle it doesn't make a difference because the calculations are for each individual image. It is a calculation of how much sky each pixel covers. As dizzling is a group of images with a slight offset, each individual frame is is what is being calculated.
Very interesting. If drizzle does not affect sensitivity... then does that mean that you can bin 2x2 (with a CCD) to get a 4 fold increase in sensitivity and then follow that with a 2x drizzle to recover the lost resolution. Two bites of the cherry as it were...?
Reply With Quote
  #47  
Old 27-06-2015, 07:49 PM
Stefan Buda
Registered User

Stefan Buda is offline
 
Join Date: Jul 2013
Location: Melbourne, VIC
Posts: 820
Great thread Ray! I love your formula. Haven't seen the sensitivity issue summed up so nicely before.
Reply With Quote
  #48  
Old 28-06-2015, 12:37 AM
Atmos's Avatar
Atmos (Colin)
Ultimate Noob

Atmos is offline
 
Join Date: Aug 2011
Location: Melbourne
Posts: 6,980
Richard,

As I have never actually tested it myself I cannot be 100% sure but I would say not. Once you lose resolution you cannot get it back because the pixels (now the super pixels) take up a portion of the sky. What drizzling does is increase SNR and tightens up stars, allows for greater sharpening on galaxies and the like. Technically it doesn't give back the resolution, only allows you to get more out of the resolution you're getting.

In science, you cannot have your cake and eat it too Physical laws that govern the universe tend to get in the way of fun haha
Reply With Quote
  #49  
Old 28-06-2015, 12:36 PM
rmuhlack's Avatar
rmuhlack (Richard)
Professional Nerd

rmuhlack is offline
 
Join Date: Nov 2011
Location: Strathalbyn, SA
Posts: 916
Quote:
Originally Posted by Atmos View Post
In science, you cannot have your cake and eat it too Physical laws that govern the universe tend to get in the way of fun haha
No kidding That's kind of what I was alluding to with my question.

Perhaps a more specific example is called for here, to help tease out what i'm trying to get to. If I capture 50 x 10min subs, both taken with my ST10 (6.8 micron pixels, average QE = 0.55) and my VC200L (f6.4, FL=1278mm, optical efficiency = 0.65)

a) Bin 1x1
b) Bin 2x2, followed by 2x drizzle integration

in example (b) the sensitivity is increased 4-fold as we have doubled the pixel size. Drizzle then spreads 1 binned super pixel out over four drizzle pixels. Does this then mean that we have produced an equivalent result to example (a). If not, then how specifically is the end result different?

Drizzle has the effect of increasing SNR.
Reply With Quote
  #50  
Old 28-06-2015, 01:20 PM
Stefan Buda
Registered User

Stefan Buda is offline
 
Join Date: Jul 2013
Location: Melbourne, VIC
Posts: 820
Quote:
Originally Posted by rmuhlack View Post

Drizzle has the effect of increasing SNR.
Richard,

That is an incomplete sentence. It should continue like this: ..at the expense of resolution, and resolution, once lost, it cannot be regained.
Reply With Quote
  #51  
Old 28-06-2015, 01:59 PM
gregbradley's Avatar
gregbradley
Registered User

gregbradley is offline
 
Join Date: Feb 2006
Location: Sydney
Posts: 17,871
Also 2x2 binning efficiency varies with sensors. The KAF8300 for example has a flaw in that it can't bin 2x2 properly resulting in a much lower advantage in SNR than it should and is more like 2X increase rather than a 4 times increase. The KAF16803 though is close to perfect and does bin well with a large gain.

What I have noticed though is 2x2 loses rez even with only colour and if you can afford the time investment 1x1 is better. Another advantage of a fast scope is the ability to afford 1x1 binning on colour as well as luminance without the time hit resulting in a more vibrant and detailed image.

Greg.
Reply With Quote
  #52  
Old 28-06-2015, 02:10 PM
rmuhlack's Avatar
rmuhlack (Richard)
Professional Nerd

rmuhlack is offline
 
Join Date: Nov 2011
Location: Strathalbyn, SA
Posts: 916
Quote:
Originally Posted by Stefan Buda View Post
Richard,

That is an incomplete sentence. It should continue like this: ..at the expense of resolution, and resolution, once lost, it cannot be regained.
Apologies, you are correct in that it is an incomplete sentence - it shouldn't have been there at all as it was a fragment left over from a previous draft of the post.

That said, my understanding of drizzle is that its purpose is to recover information that is lost due to undersampling (as described here). In my example (b) from my previous post, the image scale of a 2x2 binned image with a ST10 on my VC200L is 2.2 arcsecs per pixel (ie clearly undersampled). If a 2x drizzle when applied to my example (b) does not "recover information that is lost due to undersampling" then what is it doing? is not the "recovered lost information" referring to recovered resolution. I thought that was the point of drizzle.

(this is a genuine question btw)
Reply With Quote
  #53  
Old 28-06-2015, 05:25 PM
Stefan Buda
Registered User

Stefan Buda is offline
 
Join Date: Jul 2013
Location: Melbourne, VIC
Posts: 820
Richard,

One can get into deep water answering this question but the simple answer is that a drizzle algorithm isn't likely to recover anywhere near as much resolution as it is lost through 2x2 binning.

In fact in a case of gross udersampling and perfect guiding it will not produce significant improvement. Dithering is a must for drizzle to work.
Reply With Quote
  #54  
Old 28-06-2015, 06:17 PM
Atmos's Avatar
Atmos (Colin)
Ultimate Noob

Atmos is offline
 
Join Date: Aug 2011
Location: Melbourne
Posts: 6,980
My understanding is that it doesn't really increase "resolution". What happens is that you just get the most out of the resolution you're working with when you drizzle. When you drizzle what happens is that you smooth out some of the imperfections, tightening up stars (reducing airy disk a little) which allows you to see more BUT this isn't increasing resolution.

Resolution is the ability to resolve individual stars from one another, seeing fine detail in structure. Even when you are doing a 1x1 bin, if you drizzle that you'll help as well.

Basically, drizzling increases the SNR and helps sharpen up the image. This is why it appears to increase the resolution, because it does definitely improve the image. Once you lose resolution though, that can NEVER be reclaimed.
Reply With Quote
  #55  
Old 28-06-2015, 06:29 PM
Shiraz's Avatar
Shiraz (Ray)
Registered User

Shiraz is offline
 
Join Date: Apr 2010
Location: ardrossan south australia
Posts: 4,918
interesting questions Richard.

I think that I understand how drizzle improves resolution in dithered undersampled images, but I have been doing some experiments to see what happens to SNR under drizzle and do not yet understand the results. Drizzle certainly makes the pixels smaller, but am not sure what that means to sensitivity or how to use the equation in the presence of drizzle. Will post again when I get my head around what is going on.

Thanks for the encouragement Stefan.

regards Ray

edit: Drizzle definitely improves resolution by extracting slightly different high spatial frequency information from each of a set of dithered frames and then combining this information in the resulting stack - ultimately, more detailed information is available in the stack than is available from any single input frame. It doesn't change the optics/atmospheric resolution, but it does restore (partly) information that is lost through undersampling. It only works on images that have been dithered and that are undersampled.

Last edited by Shiraz; 28-06-2015 at 09:02 PM.
Reply With Quote
  #56  
Old 28-06-2015, 08:07 PM
Atmos's Avatar
Atmos (Colin)
Ultimate Noob

Atmos is offline
 
Join Date: Aug 2011
Location: Melbourne
Posts: 6,980
As a picture is worth a thousand words, I thought I would get out the trusty white board and draw a picture as to what is actually happening. When under-sampling a star could take up an entire pixel irrespective of the rest of the background sky that is also falling onto that single pixel. This is the very centre square. The other squares around that are that same pixel as it is being dithered.
There are two other pixels to the NW and SE that also interact with that star, so these will also have the star taking up the entire pixel, albeit with differing amounts of flux to the centre one due to containing more sky background.
To the further NW, SW and NE the pixels don't contain that star so they put boundaries on the actual size of that.

Dithering/drizzling helps sharpen up and add "resolution" to an image by reducing some of the incorrect aspects of the image, this is the apparent size of everything within any single frame (this is the definition of over-sampling).

DISCLAMER:

I'll be honest, I never even knew that something like this worked until yesterday when it was brought up. It was something that I was thinking about testing when I had some clear nights with my new "old" CCD that I got a few weeks back. Weather hasn't really even allowed me First Light yet!! Terrible!
Anyway, 7.4 micron pixels on a 900mm FL give about 1.7 arcsec/pixel which is over-sampling a bit more than I would personally like.
Attached Thumbnails
Click for full-size image (Drizzle.JPG)
75.2 KB14 views
Reply With Quote
  #57  
Old 28-06-2015, 09:44 PM
rmuhlack's Avatar
rmuhlack (Richard)
Professional Nerd

rmuhlack is offline
 
Join Date: Nov 2011
Location: Strathalbyn, SA
Posts: 916
Colin, your whiteboard sketch certainly depicts what is happening for a single pixel during dithering of undersampled sub-images. But does your example also address what then happens (in terms of "resolution" and SNR) when a drizzle is subsequently applied to the set of dithered sub-images? [This paper describes the drizzle process in more detail.]

Last edited by rmuhlack; 28-06-2015 at 10:18 PM.
Reply With Quote
  #58  
Old 28-06-2015, 10:33 PM
Atmos's Avatar
Atmos (Colin)
Ultimate Noob

Atmos is offline
 
Join Date: Aug 2011
Location: Melbourne
Posts: 6,980
Quote:
Originally Posted by rmuhlack View Post
Colin, your whiteboard sketch certainly depicts what is happening for a single pixel during dithering of undersampled sub-images. But does your example also address what then happens (in terms of "resolution" and SNR) when a drizzle is subsequently applied to the set of dithered sub-images? [This paper describes the drizzle process in more detail.]
If you haven't drizzled between each image then when you run them through a dithering process there is no difference between all of the pixels, won't make any difference. I think I am understanding you correctly; dithering a batch of un-drizzled images?
Reply With Quote
  #59  
Old 28-06-2015, 11:05 PM
rmuhlack's Avatar
rmuhlack (Richard)
Professional Nerd

rmuhlack is offline
 
Join Date: Nov 2011
Location: Strathalbyn, SA
Posts: 916
Quote:
Originally Posted by Atmos View Post
If you haven't drizzled between each image then when you run them through a dithering process there is no difference between all of the pixels, won't make any difference. I think I am understanding you correctly; dithering a batch of un-drizzled images?
You've got your terms mixed up there. Dithering is when you move the telescope between each sub-frame - its a process that is performed during image capture. After calibration and registration of your subs it allows you to stack signal only, and not fixed pattern noise. Drizzle however is an integration process, where pixels from registered sub-frames are mapped onto a smaller pixel grid. eg if the original sub-frames are width and height W x H, then the stacked 2x drizzle final image will have width and height of 2W x 2H.

It is dithering between sub-frames that is required for drizzle to work.

What I am trying to get at is what is the effect (with respect to both Ray's original formula and to the resulting resolution of the final image) of performing:
1) a 2x drizzle integration on dithered sub-frames that were captured with 2x2 binning, compared with
2) integration (without drizzle) of dithered sub-frames that were captured with 1x1 binning
Reply With Quote
  #60  
Old 29-06-2015, 09:22 AM
RickS's Avatar
RickS (Rick)
PI cult recruiter

RickS is offline
 
Join Date: Apr 2010
Location: Brisbane
Posts: 10,584
Hi Ray,

Late to the party (I've been away for a few weeks) but your formula is effectively the same as one I've been using in a spreadsheet for comparing scope/camera combinations

I don't see that Drizzle is relevant to the calculation as it is something that happens after data capture and it doesn't violate the TANSTAAFL principle. For those interested in the theory: http://www.stsci.edu/hst/HST_overvie...34.html#385457

Cheers,
Rick.
Reply With Quote
Reply

Bookmarks

Thread Tools
Rate This Thread
Rate This Thread:

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

Forum Jump


All times are GMT +10. The time is now 01:03 PM.

Powered by vBulletin Version 3.8.7 | Copyright ©2000 - 2024, Jelsoft Enterprises Ltd.
Advertisement
Bintel
Advertisement
Testar
Advertisement