Reducer Myth revisited

[Rodd]

2 minutes ago, vlaiv said:

That is absolutely correct - and is the origin of F/ratio myth.

There is in fact one thing that is correct in statement that 1" scope at F/5 is equal to 100" scope at F/5 - and that is: if you use the same camera, don't care about resolution loss, then in principle they will be "equally" fast on signal that they can equally record.

That is true in daytime / regular photography (or nearly true) - because it works in light dominated regime (plenty of light so no one cares about all the noise sources for the most part) and resolution is not severely impacted. It is also often used to characterize different lens on same camera - so pixel size does not change. In that world - it holds. It is only problematic when you try to extrapolate reasoning to cases when other factors start to dominate - like in astro imaging.

lets cut to the chase, shall we.  You know my gear--its in my signature for the most part.  I have reducers for all 4 scopes.  2 cameras. If it was you, which setup would you use to create the "best" image.   Your answer may point to teh reason why I am always unsatisfied.

[Thalestris24]

1 minute ago, Rodd said:

lets cut to the chase, shall we.  You know my gear--its in my signature for the most part.  I have reducers for all 4 scopes.  2 cameras. If it was you, which setup would you use to create the "best" image.   Your answer may point to teh reason why I am always unsatisfied.

Depends on the target - imho!

Louise

[Adam J]

16 minutes ago, Rodd said:

It appears I was wrong....the following screen shot of Craig Stark's article seems to reiterate your position.

 

FRC 300 at f/7.8 and f/5.9. I ran a DDP on the data and used Photoshop to match black and white points and to crop the two frames. Click on the thumbnail for a bigger view and/or just look at the crop.


Here is a crop around the red and yellow circled areas. In each of these, the left image is the one at f/7.8 and the right at f/5.9 (as you might guess from the difference in scale. Now, look carefully at the circled areas. You can see there is more detail recorded at the lower f-ratio. We can see the noise here in the image and that these bits are closer to the noise floor. Again, the point is that it’s incorrect to say that the f-ratio rules all and that a 1” scope at f/5 is equal to a 100” scope at f/5, but it’s also wrong to say that under real-world conditions, it’s entirely irrelevant.

What you want it the lowest F-ratio possible for imaging at your chosen focal length / image scale. That also translates to the largest aperture possible for you chose focal length.  

This all comes back to the comparison I previously made between your FSQ106 and a WO ZS61, both similar foacl lengths so a good comparison in terms of relative F-ratio. 

If you can use a reducer and achieve your desired focal length while having a larger aperture then you win. But what you should not do is compare scopes of different focal length in terms of f-ratio alone. 

If you are over sampled then a focal reducer is always a good idea. 

There is no f-ratio myth, just people who don't understand optics. 

Here is a good example of re-sampling. If I take a ASI1600mm pro and put it on a ESPRIT 80 @F5 and a ESPRIT 100 @F5.5 and re sample the Esprit 100 image in software to match the Esprit 80 at 1.91 arcseconds per pixel then the 100 image will have a lower SD then the 80 image at a smaller FOV, now if my target fits within the FOV of the 100 then why would I want to use the 80 despite it being a lower F-ratio. So in that case what matters is a larger aperture for my desired fixed image scale as oppose to the F-ratio of the optics themselves. 

Adam 

Edited October 18, 2019 by Adam J

[vlaiv]

8 minutes ago, Rodd said:

No way....if there is a star in the upper left hand corner, some photons collected by the scope will go into that star--or another galaxy, or nebula.  If teh galaxy only fills 1/4 of the FOV, and there are other things in the FOV, most certainly all the photons collected by the scope WILL NOT go into that galaxy!

You are right - things that are not part of the galaxy will not contribute to light emitted from the galaxy - but read carefully what I've said.

13 minutes ago, vlaiv said:

All the photons collected by the scope that originate from the galaxy will go into "image" of that galaxy regardless of the fact how large it is in FOV (except when it extends beyond FOV).

I was not talking about additional photons at all - they will be focused on a different place by telescope optics. They will come to aperture at a different angle and therefore will be at different place on the image.

However - all the light from the galaxy will end up in the image of that galaxy - no more, no less. There we agree. Only thing that I was pointed out was spread of that light over pixels and consequent level of signal per each pixel - or SNR in the end.

[Rodd]

1 minute ago, Adam J said:

If you are over sampled then a focal reducer is always a good idea. 

This makes sense.......at .78 arcsec/pix I am way oversampled.  But--software binning would be even better according to Vlaiv. Both together may be too much

[Rodd]

6 minutes ago, Thalestris24 said:

Depends on the target - imho!

Louise

Shouldn't--not with the idea of cropping and upsampling as explained by Vlaiv.

[Adam J]

5 minutes ago, Rodd said:

This makes sense.......at .78 arcsec/pix I am way oversampled.  But--software binning would be even better according to Vlaiv. Both together may be too much

A focal reducer will only allow you to redice by fixed factor, you can re-sample to any scale you want to in software or by using a combination of the two. See my esprit example that I added above. 

[andrew s]

Maybe one way to look at this is via extreme. If the focal ratio is effectively 0 then all light goes to one pixel. Max SNR min resolution. Extremely large focal ratio minimum SNR max resolution.

Same diameter scope.

Regards Andrew 

[Adam J]

2 minutes ago, Rodd said:

Shouldn't--not with the idea of cropping and upsampling as explained by Vlaiv.

The target matters if the target wont fit into your FOV. Hence so long as your target fits in your FOV larger optics re-sampled and or reduced are always better. 

[Rodd]

9 minutes ago, Adam J said:

What you want it the lowest F-ratio possible for imaging at your chosen focal length / image scale. That also translates to the largest aperture possible for you chose focal length.  

This all comes back to the comparison I previously made between your FSQ106 and a WO ZS61, both similar foacl lengths so a good comparison in terms of relative F-ratio. 

If you can use a reducer and achieve your desired focal length while having a larger aperture then you win. But what you should not do is compare scopes of different focal length in terms of f-ratio alone. 

If you are over sampled then a focal reducer is always a good idea. 

There is no f-ratio myth, just people who don't understand optics. 

But this not what we are talking about.  We are talking about using a reducer on a scope--so aperture is fixed.  We are not talking about changing scopes.  A person has 1 scope--he can shoot at native or with reducer.   He has teh choice of 1 aperture, 2 focal lengths and 2 focal ratios.  that's it.

[vlaiv]

5 minutes ago, Rodd said:

This makes sense.......at .78 arcsec/pix I am way oversampled.  But--software binning would be even better according to Vlaiv. Both together may be too much

I would like to add to this that software binning is not always the best option - it is case dependent. If one is way oversampled at native FL - use of focal reducer might not result in optimal sampling rate. For example .78"/px with x0.7 reducer will result in 1.12"/px. If conditions (sky, scope, mount) don't allow for this resolution either - it is then better to bin as 1.56"/px will be closer to optimal sampling rate.

14 minutes ago, Rodd said:

lets cut to the chase, shall we.  You know my gear--its in my signature for the most part.  I have reducers for all 4 scopes.  2 cameras. If it was you, which setup would you use to create the "best" image.   Your answer may point to teh reason why I am always unsatisfied.

If we go down that path we will change direction of this discussion quite a bit. I'm ok with that.

First I would like us to define what is the "best" image. Then we can start discussing what is needed to create the best image, and after all of that we can see what combination of scope camera fits requirements for best image.

[Rodd]

1 minute ago, Adam J said:

The target matters if the target wont fit into your FOV. Hence so long as your target fits in your FOV larger optics re-sampled and or reduced are always better. 

Only if all else is equal--and things are never equal,.  craig stark shows very convincingly in teh article that a 4" APO beats an 8" Newt i n his example.  the image of teh HH from teh 4" apo was clearly more detailed and brighter than teh image form the 8" newt. 

[Adam J]

Just now, Rodd said:

But this not what we are talking about.  We are talking about using a reducer on a scope--so aperture is fixed.  We are not talking about changing scopes.  A person has 1 scope--he can shoot at native or with reducer.   He has teh choice of 1 aperture, 2 focal lengths and 2 focal ratios.  that's it.

Its all tide together in the same argument. All a reducer is is a tool to expand the FOV while reducing resolution and gaining SNR. In software you can reduce resolution and gain SNR too, but you cant increase the FOV like a reducer can. 

[Adam J]

5 minutes ago, vlaiv said:

I would like to add to this that software binning is not always the best option - it is case dependent. If one is way oversampled at native FL - use of focal reducer might not result in optimal sampling rate. For example .78"/px with x0.7 reducer will result in 1.12"/px. If conditions (sky, scope, mount) don't allow for this resolution either - it is then better to bin as 1.56"/px will be closer to optimal sampling rate.

In software you don't have to re-sample by a factor of 2 you can do it by any amount you like and the algorithm will still work to increase SNR (if thats even the best way to describe it). So you could re-sample to 2"/pix or 1.2"/pix with the corresponding level of improvement in image quality. 

You can then combine both focal reducer and software to get the desired samping while also getting the benefit of a wider field from the reducer. 

Edited October 18, 2019 by Adam J

[Adam J]

6 minutes ago, Rodd said:

Only if all else is equal--and things are never equal,.  craig stark shows very convincingly in teh article that a 4" APO beats an 8" Newt i n his example.  the image of teh HH from teh 4" apo was clearly more detailed and brighter than teh image form the 8" newt. 

Did he re-sample both images to the same image scale?

 

[Rodd]

Just now, vlaiv said:

I would like to add to this that software binning is not always the best option - it is case dependent. If one is way oversampled at native FL - use of focal reducer might not result in optimal sampling rate. For example .78"/px with x0.7 reducer will result in 1.12"/px. If conditions (sky, scope, mount) don't allow for this resolution either - it is then better to bin as 1.56"/px will be closer to optimal sampling rate.

If we go down that path we will change direction of this discussion quite a bit. I'm ok with that.

First I would like us to define what is the "best" image. Then we can start discussing what is needed to create the best image, and after all of that we can see what combination of scope camera fits requirements for best image.

I am tire of this discussion anyway.  I need to think about it a bit.  i will refine my question and get back to you.

 

1 minute ago, Adam J said:

Its all tide together in the same argument. All a reducer is is a tool to expand the FOV while reducing resolution and gaining SNR. In software you can reduce resolution and gain SNR too, but you cant increase the FOV like a reducer can. 

I agree with all that.  A good place to end.  But the real question, which Vlaiv says yes, and Craig stark seems to agree with, is can you use a reducer to image a small target, then crop the image and enlarge the target with less exposure time than for the equivalent image taken at native.  Not all agree with this. At least the last I heard.   Olly doesn't (or didn't the last we spoke), and I can't fully dive into the "I agree" pool with Olly taking an opposing view.    

[vlaiv]

2 minutes ago, Adam J said:

In software you don't have to re-sample by a factor of 2 you can do it by any amount you like and the algorithm will still work to increase SNR (if thats even the best way to describe it). So you could re-sample to 2"/pix or 1.2"/pix with the corresponding level of improvement in image quality. 

You can then combine both focal reducer and software to get the desired samping while also getting the benefit of a wider field from the reducer. 

That is in fact true, and only thing that I would add to that is:

- binning produces known improvement in SNR

- binning is the form of resampling

- depending on type of resampling you can get couple of things happen to your data - loss of resolution due to pixel blur and pixel to pixel correlation and improvement in SNR.

- with regular binning you get pixel blur, but you can bin your data in such way that there is no increase in pixel blur.

- most other resampling methods don't provide known increase in SNR, and often have poorer characteristics than binning (pixel to pixel correlation / improvement in SNR).

- there is in fact fractional binning method that should provide benefits of integer binning and resampling. It introduces very small correlation, keeps pixel blur to a minimum and has predictable improvement in SNR (equal to "scale" of binning, or square root of binned surface). Only software that I've seen it implemented in is StarTools - and I'm guessing how it's done, and I believe there is a better way to do it ).

[Rodd]

3 minutes ago, Adam J said:

Did he re-sample both images to the same image scale?

yes he did--read the article--its only a page with pictures--not very long.  Just google Craig Stark and F ratio myth.  makes me glad I have a 4" apo!

[vlaiv]

3 minutes ago, Rodd said:

I agree with all that.  A good place to end.  But the real question, which Vlaiv says yes, and Craig stark seems to agree with, is can you use a reducer to image a small target, then crop the image and enlarge the target with less exposure time than for the equivalent image taken at native.  Not all agree with this. At least the last I heard.   Olly doesn't (or didn't the last we spoke), and I can't fully dive into the "I agree" pool with Olly taking an opposing view.  

I did hear (or better say read) once on Olly's position on this and yes as far as I remember he was not in agreement with this to an extent, but not sure it is still his position on the topic.

But instead of hear/say - we can ask @ollypenrice what is his stand on all of this?

[Adam J]

3 minutes ago, Rodd said:

I am tire of this discussion anyway.  I need to think about it a bit.  i will refine my question and get back to you.

 

I agree with all that.  A good place to end.  But the real question, which Vlaiv says yes, and Craig stark seems to agree with, is can you use a reducer to image a small target, then crop the image and enlarge the target with less exposure time than for the equivalent image taken at native.  Not all agree with this. At least the last I heard.   Olly doesn't (or didn't the last we spoke), and I can't fully dive into the "I agree" pool with Olly taking an opposing view.    

Yes, but if you then took the up reduced image and software re-sampled it to the same image scale as the 0.8x reduced image you would be back to almost parity

So Craig stark is correct, but you can almost achieve the same thing in software without the need to use a reducer. I say almost because there are some mathematical advantages to a reducer over software re-sampling, however in the end after you software re-sample the none reduced image you would have a hard time seeing the difference in SNR visually. 

[vlaiv]

Here is interesting PDF by Craig Stark - it contains much of what we discussed here but there is even more - some very good examples of images and effects - it shows how one can have the same image by shooting reduced and then enlarging if image does not contain information in the first place.

It also shows how much detail you loose by undersampling (depending on detail in the image) - I think it is very informative and sheds light on what is to be expected (and I'm guessing many will be surprised by subtle differences for example between 1"/px and 2"/px) - I also mentioned this above.

http://www.stark-labs.com/craig/resources/Articles-&-Reviews/ImageSampling_Fratios_SNR_RTMC.pdf

[Adam J]

10 minutes ago, vlaiv said:

That is in fact true, and only thing that I would add to that is:

- binning produces known improvement in SNR

- binning is the form of resampling

- depending on type of resampling you can get couple of things happen to your data - loss of resolution due to pixel blur and pixel to pixel correlation and improvement in SNR.

- with regular binning you get pixel blur, but you can bin your data in such way that there is no increase in pixel blur.

- most other resampling methods don't provide known increase in SNR, and often have poorer characteristics than binning (pixel to pixel correlation / improvement in SNR).

- there is in fact fractional binning method that should provide benefits of integer binning and resampling. It introduces very small correlation, keeps pixel blur to a minimum and has predictable improvement in SNR (equal to "scale" of binning, or square root of binned surface). Only software that I've seen it implemented in is StarTools - and I'm guessing how it's done, and I believe there is a better way to do it ).

Well different methods aside I know if I software re-sample my 130PDS from 1.24" / pix to 2" / pix (my scale for presenting nabula) there is a significant increase in image quality / perceived image smoothness and gives me a nice balance between detail captured and smoothness when viewed at full scale while keeping integration times managable. Thats just using the re-size function in Photoshop nothing fancy.   

Edited October 18, 2019 by Adam J

[vlaiv]

1 minute ago, Adam J said:

Well different methods aside I know if I software re-sample my 130PDS from 1.24" / pix to 2" / pix (my scale for presenting nabula) there is a significant increase in image quality / perceived image smoothness and gives me a nice balance between detail captured and smoothness when viewed at full scale. Thats just using the re-size function in Photoshop nothing fancy.   

I already linked to a thread about software binning, so you can have a look on difference between resampling and binning in terms of SNR improvement. It is best done while data is still linear and probably best approach in your case would be to bin x2 to get from 1.24" up to 2.48" and then upsample back to 2"/px.

[Adam J]

5 minutes ago, vlaiv said:

I already linked to a thread about software binning, so you can have a look on difference between resampling and binning in terms of SNR improvement. It is best done while data is still linear and probably best approach in your case would be to bin x2 to get from 1.24" up to 2.48" and then upsample back to 2"/px.

If my SNR is already acceptable though would that not result in reduced detail as you cant get detail back by up samping? 

My current heart project is four frames with the 130PDS and ASI1600mm pro. I find that I need very little noise reduction after re-sampling so am recording true details at the pixel level, hence over four frames I end up with an image that will print to a 60cm x 60cm size on Alu very nicely. 

Would I be better off getting a 70mm F5 and getting it in one frame but four times the integration in that single frame....not sure but if you only have a hammer every problem is a nail.  

Adam

[Laurin Dave]

Just read Craig Starks note..  my reading is that the f ratio myth would be a myth if it were not for read noise..  in my recent experience taking Ha at the same time through Esprit 150 SX-46 and Esprit 100 asi1600 the Esprit 100 stacks are better,(which is a trifle annoying..  added together though they are even better)  even though the Esprit 150sx-46 gets 1.5 times more photons per pixel.. read noise though is 9 vs 1.7 on the asi1600..  on LRGB the 150 combo is streets ahead..