Reducer Myth: On Point

[vlaiv]

Just now, Rodd said:

You miss my point.  If the reduced crop has a higher SNR than the unreduced original.....that is contrary to the myth, which states that the SNR will be the same once the images are size equalized

Ok, I'm missing the point since I don't know what the myth is and what would be contrary to it. I can only argue what would happen from perspective of theory of imaging and signal processing.

[ollypenrice]

8 minutes ago, Rodd said:

I call it the myth because that is its name.  You should ask Olly about it--I did not name it.  The myth came around because people believed that if you use a focal reducer on a galaxy say, to increase SNR you collect more data in a shorter period of time.  Then you can crop the image and enlarge and you will have the equivalent image captured faster.  The myth says this is not true, that the cropped and enlarged image will be of lesser resolution--quite a bit less.  But I think I have shown that the difference in resolution between the unreduced and reduced is so slight that it is, in fact, true enough.  that is the moral victory.

Not at all. What you have shown is that you could not exploit the higher resolution of your unreduced image because you didn't have the guiding or the seeing or whatever to do so. This isn't remarkable. Nor is it reprehensible. It's just reality. But what happens if you resample your unreduced image downwards so that it's the same size as your reduced? How much difference is there?

Note: I didn't invent the term 'F ratio myth.' I am persuaded by its authors but I am not one of them. I am very much a latecomer to this debate but if consider this: You have a C11 and want to image M33. At F10 you need 100 minutes for 'x' signal. At F2 with a Hyperstar you need 4 minutes for the same SNR. So, in fairyland, you just need a Hyperstar to turn a hundred minutes into four minutes. Dang, I wish I lived in fairyland!!!

Olly

[Rodd]

4 minutes ago, ollypenrice said:

Not at all. What you have shown is that you could not exploit the higher resolution of your unreduced image because you didn't have the guiding or the seeing or whatever to do so. This isn't remarkable. Nor is it reprehensible. It's just reality. But what happens if you resample your unreduced image downwards so that it's the same size as your reduced? How much difference is there?

Note: I didn't invent the term 'F ratio myth.' I am persuaded by its authors but I am not one of them. I am very much a latecomer to this debate but if consider this: You have a C11 and want to image M33. At F10 you need 100 minutes for 'x' signal. At F2 with a Hyperstar you need 4 minutes for the same SNR. So, in fairyland, you just need a Hyperstar to turn a hundred minutes into four minutes. Dang, I wish I lived in fairyland!!!

Olly

But...the reduced crop has a higher SNR than the unreduced crop.  If I use a reducer on M33 with the FSQ 106, I am likely to crop and enlarge it so it is the same size as an unreduced image of say the TOA 130--that is the reason the focal reducer was used PER THE MYTH--to increase SNR so that when you enlarge you have a image collected faster of the same target.  This, as far as I am able to understand is contrary to the myth.

[ollypenrice]

9 minutes ago, Rodd said:

But...the reduced crop has a higher SNR than the unreduced crop.  If I use a reducer on M33 with the FSQ 106, I am likely to crop and enlarge it so it is the same size as an unreduced image of say the TOA 130--that is the reason the focal reducer was used PER THE MYTH--to increase SNR so that when you enlarge you have a image collected faster of the same target.  This, as far as I am able to understand is contrary to the myth.

No, it's not contrary to the myth. You're trading resolution for signal. The myth is irrelevant if you can't exploit the longer FL signal.

Olly

[Rodd]

1 minute ago, ollypenrice said:

No, it's not contrary to the myth. You're trading resolution for signal. The myth is irrelevant if you can't exploit the longer FL signal.

Olly

So your saying that the reason the resolution is not better in the unreduced image is because of sky conditions (my guiding was excellent TYVM).  And if these images were shot on the same night-the resolution would be, would have to be,  markedly better in the unreduced image.  Correct?  

Rodd

[vlaiv]

6 minutes ago, Rodd said:

So your saying that the reason the resolution is not better in the unreduced image is because of sky conditions (my guiding was excellent TYVM).  And if these images were shot on the same night-the resolution would be, would have to be,  markedly better in the unreduced image.  Correct?  

Rodd

This is where you are wrong, in a single word. Resolution is better in unreduced image but it is not markedly better. It is perceivable on close scrutiny, but on a casual look one would be hard pressed to see the difference. Same is true for x2 binned image, it has slightly more detail loss than reduced image, but again, one might say that on casual look it looks the same as unreduced version.

Just how much impact is detail loss going to have on a particular image depends both on level of detail loss (expressed as high frequency cut off / attenuation) and how much high frequency components there were in the first place. Stars are particularly good candidates for examination, and as you've seen in gif I posted in unreduced version they look sharper. In the rest of the image there is just couple of places where one might notice detail loss - and that is "hats" and "rims" of two nebula features. Rest is smooth as is (not much high frequency components to begin with) so loss of high frequency components is not going to be noticeable.

[Rodd]

2 minutes ago, vlaiv said:

This is where you are wrong, in a single word. Resolution is better in unreduced image but it is not markedly better. It is perceivable on close scrutiny, but on a casual look one would be hard pressed to see the difference. Same is true for x2 binned image, it has slightly more detail loss than reduced image, but again, one might say that on casual look it looks the same as unreduced version.

Just how much impact is detail loss going to have on a particular image depends both on level of detail loss (expressed as high frequency cut off / attenuation) and how much high frequency components there were in the first place. Stars are particularly good candidates for examination, and as you've seen in gif I posted in unreduced version they look sharper. In the rest of the image there is just couple of places where one might notice detail loss - and that is "hats" and "rims" of two nebula features. Rest is smooth as is (not much high frequency components to begin with) so loss of high frequency components is not going to be noticeable.

Then you need to fight it out with Olly--because this is what he is saying.  If you are right--then he is wrong and I have shown something contrary to the myth.

[ollypenrice]

7 minutes ago, Rodd said:

So your saying that the reason the resolution is not better in the unreduced image is because of sky conditions (my guiding was excellent TYVM).  And if these images were shot on the same night-the resolution would be, would have to be,  markedly better in the unreduced image.  Correct?  

Rodd

Clearly a given pixel scale, if supported by the optical resolution, guiding, seeing etc, gives a certain theoretical resolution. Does the sky/guiding/whatever deliver this resolution? If it doesn't then why not use a reducer? No reason I can see. The F ratio myth exponents don't urge you to try to image at scales you have no hope of realizing.

Your higher theoretical resolution captures are either really higher resolution or they're not. If they're not I guess you should use the reducer.

Olly

[Stub Mandrel]

8 hours ago, vlaiv said:

Increase contrast of the image - it will be perceived as sharper. Add noise to the image and it will be perceived as sharper. We don't even perceive color the same way always, nor the brightness.

Absolutely, the images need to be matched for dynamic range to judge detail.

[knobby]

Maybe not the right place , but hey ho !

So does a reducer help with guiding ? Is it really making your focal length less ?

[Rodd]

1 hour ago, knobby said:

Maybe not the right place , but hey ho !

So does a reducer help with guiding ? Is it really making your focal length less ?

Its not so much that it helps with guiding as that it hides some of the errors you might get with less than perfect guiding.  Or, in other words, your guiding does not have to be as precise when shooting with a reducer compared to shooting without one .  The errors are still there--but some may be below the threshold of the resolution of the system.  Anyway--that's my take

Rodd

[ollypenrice]

6 hours ago, Rodd said:

Its not so much that it helps with guiding as that it hides some of the errors you might get with less than perfect guiding.  Or, in other words, your guiding does not have to be as precise when shooting with a reducer compared to shooting without one .  The errors are still there--but some may be below the threshold of the resolution of the system.  Anyway--that's my take

Rodd

That's also my take. As you say, a reducer in the imaging scope has no effect on the guiding itself but reduces the precision needed. A rule of thumb which I follow says that your guide RMS in arcseconds needs to be half (or less) of you pixel scale in arcsecs per pixel. If it is, then the errors will not affect the final image. I can't vouch for the correctness of this rule of thumb but it is widely circulated when the topic comes up.

Olly

[Stub Mandrel]

1 hour ago, ollypenrice said:

That's also my take. As you say, a reducer in the imaging scope has no effect on the guiding itself but reduces the precision needed. A rule of thumb which I follow says that your guide RMS in arcseconds needs to be half (or less) of you pixel scale in arcsecs per pixel. If it is, then the errors will not affect the final image. I can't vouch for the correctness of this rule of thumb but it is widely circulated when the topic comes up.

Olly

From practical limitations of my setup, rather than choice, my rms error generally comes out pretty much equal to my pixel scale.

I don't get weird-shaped stars when processed, but I don't think my images are as sharp as they could be. This crop should appear at 1:1 so you can see what guiding @ 1:1 looks like when processed (42 5-minute subs).

 

On a single sub you can see the stars aren't 100% perfect (and just how poor my skies are!):

[ollypenrice]

1 hour ago, Stub Mandrel said:

From practical limitations of my setup, rather than choice, my rms error generally comes out pretty much equal to my pixel scale.

I don't get weird-shaped stars when processed, but I don't think my images are as sharp as they could be. This crop should appear at 1:1 so you can see what guiding @ 1:1 looks like when processed (42 5-minute subs).

 

On a single sub you can see the stars aren't 100% perfect (and just how poor my skies are!):

As I'm sure you know, round stars don't indicate perfect guiding. Random but equivalent errors on both axes will give round stars but resolution is still being lost. When fine tuning the first Mesu's guiding parameters we saw exactly this phenomenon. Stars were round but the image was soft. Once PHD was set correctly the mount could produce an RMS of 0.3 and we were good to go at 0.63"PP.

Olly

[Stub Mandrel]

1 minute ago, ollypenrice said:

As I'm sure you know, round stars don't indicate perfect guiding. Random but equivalent errors on both axes will give round stars but resolution is still being lost. When fine tuning the first Mesu's guiding parameters we saw exactly this phenomenon. Stars were round but the image was soft. Once PHD was set correctly the mount could produce an RMS of 0.3

Indeed, but even so an rms of about 1 pixel is an improvement over unguided for my EQ3! My achievable best seems to be  0.84, over 2 sometimes and typically 1.4-1.7. The key is getting the dec just right, it needs regular tuning.

TBH I think I will have to move to a bigger tripod to raise my game.

[Rodd]

3 hours ago, ollypenrice said:

s I'm sure you know, round stars don't indicate perfect guiding. Random but equivalent errors on both axes will give round stars but resolution is still being lost. When fine tuning the first Mesu's guiding parameters we saw exactly this phenomenon. Stars were round but the image was soft. Once PHD was set correctly the mount could produce an RMS of 0.3 and we were good to go at 0.63"PP.

In that case I would expect the stars to have been larger than expected...round, but bloated?

Rodd

[ollypenrice]

1 hour ago, Rodd said:

In that case I would expect the stars to have been larger than expected...round, but bloated?

Rodd

Yes, that's what we saw. The stars were not terribly bloated but they got smaller as we tuned PHD.

Olly

[Rodd]

37 minutes ago, ollypenrice said:

Yes, that's what we saw. The stars were not terribly bloated but they got smaller as we tuned PHD.

Olly

I wonder if it would be possible to create an app or script that uses a particular star field (based on coordinates), FOV and focal length (Pixel scale, etc), so that when your subs download, there is an overlay that has small circles that define what size the stars should be for a particular pixel scale you are shooting at.  It would just be a visual depiction of FWHM used to calculate star size--But it would be allot easier to use the overlay as a template as apposed to looking at FWHM values and doing it your head.  The script could have variables like seeing, declination, etc, even specifics about the optical system used to accurately depict targeted star size.

Rodd

[vlaiv]

1 minute ago, Rodd said:

I wonder if it would be possible to create an app or script that uses a particular star field (based on coordinates), FOV and focal length (Pixel scale, etc), so that when your subs download, there is an overlay that has small circles that define what size the stars should be for a particular pixel scale you are shooting at.  It would just be a visual depiction of FWHM used to calculate star size--But it would be allot easier to use the overlay as a template as apposed to looking at FWHM values and doing it your head.  The script could have variables like seeing, declination, etc, even specifics about the optical system used to accurately depict targeted star size.

Rodd

I don't know how useful that would be since you can't really change focal length on the fly when imaging so I'm not certain that such information would be useful at that time.

What I do know, and I'm working on it, is that FWHM can be used to decide on optimum resolution to stack images at. This sounds strange, because no software that I'm aware of is doing it now, but there is stacking algorithm (actually aligning part of stacking) that can alter stack resolution to specific value.

My idea is that one should use high resolution as a base - close to resolution that one would use under best conditions for given equipment to sample the image, and later on in stacking phase decide what resolution to go for based on FWHM values. There is a way to determine appropriate sampling resolution based on FWHM and single parameter - we can call it detail loss coefficient or something like that. If you examine my previous posts you will see that a bit more details for your particular images was lost when binning x2 versus using focal reducer which is equivalent of binning x1.4something. Depending on FWHM and your preference for detail loss - stacking process can in align phase produce low resolution (or high resolution for that matter - similar to drizzle but not quite the same) result with increased SNR.

[Rodd]

5 minutes ago, vlaiv said:

I don't know how useful that would be since you can't really change focal length on the fly when imaging so I'm not certain that such information would be useful at that time.

What I do know, and I'm working on it, is that FWHM can be used to decide on optimum resolution to stack images at. This sounds strange, because no software that I'm aware of is doing it now, but there is stacking algorithm (actually aligning part of stacking) that can alter stack resolution to specific value.

My idea is that one should use high resolution as a base - close to resolution that one would use under best conditions for given equipment to sample the image, and later on in stacking phase decide what resolution to go for based on FWHM values. There is a way to determine appropriate sampling resolution based on FWHM and single parameter - we can call it detail loss coefficient or something like that. If you examine my previous posts you will see that a bit more details for your particular images was lost when binning x2 versus using focal reducer which is equivalent of binning x1.4something. Depending on FWHM and your preference for detail loss - stacking process can in align phase produce low resolution (or high resolution for that matter - similar to drizzle but not quite the same) result with increased SNR.

It would be useful; because it would tell you immediately by visual inspection that your stars were too bloated and that you need to refine your guiding.  If the stars are bigger than the circle you need to tweak.  I think that would be eminently useful, and as I intoned, much quicker than looking at numbers.

Rodd

[vlaiv]

15 minutes ago, Rodd said:

It would be useful; because it would tell you immediately by visual inspection that your stars were too bloated and that you need to refine your guiding.  If the stars are bigger than the circle you need to tweak.  I think that would be eminently useful, and as I intoned, much quicker than looking at numbers.

Rodd

Yes, but what is star "base" or circle enclosing the star? It also depends on length of exposure and SNR. You can only distinguish that part of star base that has high enough SNR to present it self as filled circle versus surroundings. In short exposures star "wings" might not have enough signal and it would appear as star being smaller in "circle" - but the actual PSF is characterized differently and associated blur. It also depends on type of stretch you use to view image.

[vlaiv]

Have a look at these:

Only way that you could have consistent display of "circles" is to use measured sigma of gaussian fitting or calculated FWHM value being displayed as a circle, but both of those values (their accuracy) depends again on SNR. It makes sense to use them for comparing individual frames of "full" exposure. Comparing frame and focus vs full length exposure is not quite the same because of differences in SNR.

[ollypenrice]

But don't we know from experience what our best FWHM values are? I bear them in mind and if, on a new project, they are high I don't shoot luminance or any NB data carrying fine detail. I'll shoot colour instead.

Olly

[vlaiv]

+1 what Olly said, just pay attention to usual FWHM values under same conditions - either in PHD2 or frame & focus 2s exposure or in regular sub (use the same source always).

[Rodd]

7 minutes ago, ollypenrice said:

But don't we know from experience what our best FWHM values are? I bear them in mind and if, on a new project, they are high I don't shoot luminance or any NB data carrying fine detail. I'll shoot colour instead.

Olly

Some of us are not as experienced as others.  

 

5 minutes ago, vlaiv said:

+1 what Olly said, just pay attention to usual FWHM values under same conditions - either in PHD2 or frame & focus 2s exposure or in regular sub (use the same source always).

I am surprised in this contrivance loaded endeavor there is so much resistance.  All parameters can be uploaded to the software--an estimate of noise, etc.  I did not say it would be foolproof--I said it would be useful, and for more visual oriented people (with limited experience) it may prove very useful.  

Rodd