Why are my stars so big? NGC891

[Chris49]

2 hours ago, michael8554 said:

If your stars are smaller and round in the subs, then your guiding is "adequate", so isn't it your processing that is to blame for stars that are too big?

I see what you are getting at, but I come back to the HFR values that I get for the stars in the subs with Sequence Generator Pro - these range from 3.5 to 6.  When I watch videos of people using SGP, they get values as low as 1.2 so my stars are large at that stage.  I suspect they then get bigger still with my attempts at processing, probably because I have stretched them too much.

Your comments are certainly making me think!!

Thanks

Chris

[wornish]

5 hours ago, vlaiv said:

Please don't continue to keep that way of thinking alive - focal length has nothing to do with performance of the mount.

Put two identical scopes (in terns of size and weight) on the same mount - one with 200mm FL and other with 4000mm FL (I know, such scopes don't exist, but let's imagine some crazy CN 212 that can be both hyperbolic newtonian and folded maksutov type scope in the same package). How on earth would mount know the difference between focal lengths of the scopes and why would it behave differently because of that?

Again, it's not focal length that is the issue - it is sampling rate (which in part depends on focal length - hence origin of the myth, but you can control it via other means - like using focal reducers, binning, using different camera with different pixel size, etc ...).

I use OAG and never understood all the "it's difficult to use" hype... Never had issues with not being able to find guide star and with guiding itself.

I am struggling with this discussion.  Like the OP I would love to go for smaller targets, but have held back from going for a longer focal length scope.

Surely the longer the focal length the smaller the image scale for a given camera. (that is defined by the cameras pixel size)

I agree that a mount has no idea what scope is on it.

Mounts can guide to a given tolerance depending on :

a. The mounts physical design and constrruction

b. The guide camera focal length

but when it all boils down given a reasonable mount and good guiding what matters is

c. The Seeing. 

 

If the seeing in your location is average ( i.e. you are not on Hawaii, in the Atacama desert etc) then seeing is the major factor that defines what you can resolve.

Isn't it simply wasting your money if you think you can some how magically fix the physics, despite what all the brochures promise?

What am I missing?

 

 

 

Edited December 20, 2019 by wornish

[Chris49]

2 hours ago, cfpendock said:

It seems to me that your subs are quite short.

With longer than 90 sec subs, the stars start to become sausages.  This is I believe the result of flexure between the main and the guide scope.  I have now improved the guide scope rigidity by putting it in rings on top of the main scope, but I haven't had a chance to try this yet - part of the back garden was underwater this morning. 🙁

Your NGC891 is stunning - thanks for showing me what can be done.

Chris

[vlaiv]

1 hour ago, wornish said:

I am struggling with this discussion.  Like the OP I would love to go for smaller targets, but have held back from going for a longer focal length scope.

Surely the longer the focal length the smaller the image scale for a given camera. (that is defined by the cameras pixel size)

I agree that a mount has no idea what scope is on it.

Mounts can guide to a given tolerance depending on :

a. The mounts physical design and constrruction

b. The guide camera focal length

but when it all boils down given a reasonable mount and good guiding what matters is

c. The Seeing. 

 

If the seeing in your location is average ( i.e. you are not on Hawaii, in the Atacama desert etc) then seeing is the major factor that defines what you can resolve.

Isn't it simply wasting your money if you think you can some how magically fix the physics, despite what all the brochures promise?

What am I missing?

 

 

 

My point with that post was - don't think of your focal length when you think if mount is capable of delivering. Think your sampling resolution.

Why do I say that? Because it is ultimately the sampling resolution that needs to be matched to level of detail, and while sampling resolution depends on focal length of the scope - it is not only factor. There are other things that go into sampling resolution "equation" - and some of those things you can control.

What does that mean? Someone might give you following recommendation - don't use 2 meter focal length scope because you have average seeing - you won't be able to guide /track it properly with your mount. But that is just simply flawed reasoning.

Let's say that your mount and sky conditions on particular night support 1.5"/px resolution (there is in fact direct link between optimum resolution and star FWHM on a given night and it goes like FWHM / 1.6, so if your FHWM is 2.4" then you are ok with 1.5"/px, but if your FWHM is closer to 3.2" then matching resolution is 2"/px).

Now you have 2000mm focal length and you want to sample at 1.5"/px. You need something like 15um pixel camera and you will get that sampling rate. You can't just take camera for every possible resolution, but you can get rather close - for example if you have cmos sensor with pixel size of 3.8um then you can get very close to 15um by simply binning x4 - that will give you 15.2um pixel size or resolution of 1.57"/px - that is close enough.

Another way to do it is to use smaller scope - let's say scope with 500mm focal length and pair that with 3.8um camera - you will also have sampling rate of 1.57"/px that way.

You can achieve proper sampling rate with both long focal length scope and short focal length scope if you manage your data in certain way (by binning for example) - and both those scopes will record properly what seeing and tracking/guiding has to offer on particular night.

Advantage of long focal length scope will be aperture, but disadvantage will be narrow FOV. If target fits narrow FOV - than that is not disadvantage at all for that target.

You are right - you can't beat the seeing nor can you fix mount issues in software or by choice of scope, but you can "adjust" your data to conditions on a given night and still have your stars look nice and tight, but on a smaller image simply because there is no detail to "zoom in" more.

On a night of a good seeing and when conditions are good - then it pays off to have that much focal length  - it is very easy to go high resolution with long focal length scope, but almost impossible to go with short focal length scope (you need a barlow / focal extender for that and that won't change aperture of small scope so it will still gather same small amount of light).

Btw, just as a comparison - original image in this post had resolution of about 2"/px. This is why stars looked ok when I scaled it to 25% of original size. With same aperture scope under good conditions and when your guiding is better than 0.5" RMS you can get stars like this:

This is sampled at 1"/px by the way, so if we bring it down to 2"/px like I did in my example, these stars look really tiny and pinpoint:

And this was shot with 1600mm of focal length on Heq5 mount. So this is not out of the reach - one just needs to pay a bit attention to guiding and if night is good - stars will be tight - regardless of focal length as long as you know how to match sampling rate to conditions.

[newbie alert]

17 hours ago, vlaiv said:

Please don't continue to keep that way of thinking alive - focal length has nothing to do with performance of the mount.

Put two identical scopes (in terns of size and weight) on the same mount - one with 200mm FL and other with 4000mm FL (I know, such scopes don't exist, but let's imagine some crazy CN 212 that can be both hyperbolic newtonian and folded maksutov type scope in the same package). How on earth would mount know the difference between focal lengths of the scopes and why would it behave differently because of that?

Again, it's not focal length that is the issue - it is sampling rate (which in part depends on focal length - hence origin of the myth, but you can control it via other means - like using focal reducers, binning, using different camera with different pixel size, etc ...).

I use OAG and never understood all the "it's difficult to use" hype... Never had issues with not being able to find guide star and with guiding itself.

How can fl have nothing to do with it when it controls your sampling, it's part of the digits that give you the sampling..

Yes you're right to say the mount has no idea what it's controlling on top but all mounts have limits of performance.. otherwise no-one would have issues trying to image long focal length scopes ..the job is far easier with a wider fov

A mesu will handle it better than a heq5 and neither will no what it's carrying on top

[vlaiv]

2 minutes ago, newbie alert said:

How can fl have nothing to do with it when it controls your sampling, it's part of the digits that give you the sampling..

Yes you're right to say the mount has no idea what it's controlling on top but all mounts have limits of performance.. otherwise no-one would have issues trying to image long focal length scopes ..the job is far easier with a wider fov

A mesu will handle it better than a heq5 and neither will no what it's carrying on top

Indeed, FL goes into sampling equation, but so does pixel size, and this second part gives you so much flexibility as you can choose how much you want to bin your image. In fact you can set target sampling rate and bin to that exact figure. It involves fractional binning, and there is rather nice way to do it when combined with stacking.

Mount error is in absolute units and not relative units - it is measured in arc seconds. So is seeing. Focal length maps those errors to certain number of pixels. This turns it into relative measure - per pixel. So if your error is some arc seconds in size - you can choose to make it one pixel large, two pixels large or less than one pixel large. Point is that you have fixed focal length (or maybe small flexibility by adding focal reducer or barlow) but you can in principle change your pixel size by manipulating data. It is easier to join pixels than it is to split pixels in math, and that is why it is easier to use longer focal length to get coarser sampling then it is to use short focal length to get finer sampling (involves drizzle algorithm and it won't quite work well in amateur setups).

What I was trying to say with this, and again you are right - mesu will be better than heq5 for example, but that does not mean that you need to use long focal length scope on mesu only - you just need to "accommodate" pixel size to the level of precision achieved depending on focal length you are using. That and the fact that using longer focal length scope will not make mount error larger in arc seconds (it will in pixels if you leave pixel size what it is).

[wornish]

Thanks Vlaiv for the explanation.

You should create a blog or perhaps a separate section on here that explains all this and gives the formulas to use.  I am sure many members have the same struggle to understand all the factors involved and the trade offs.

I think you might have just convinced me to spend money on a longer focal length scope.

[Davey-T]

21 minutes ago, wornish said:

You should create a blog or perhaps a separate section on here that explains all this and gives the formulas to use.

Good idea you could call it " Do my stars look big in this " 

Dave 

[vlaiv]

3 minutes ago, wornish said:

Thanks Vlaiv for the explanation.

You should create a blog or perhaps a separate section on here that explains all this and gives the formulas to use.  I am sure many members have the same struggle to understand all the factors involved and the trade offs.

I think you might have just convinced me to spend money on a longer focal length scope.

Do bare in mind that "/px is not the whole story. There is also issue of FOV and sensor size.

It seems that some people also don't understand image size in relation to how it's displayed on computer screen - fact that you can display it at 1:1 (screen pixel to image pixel) or 100% zoom level or you can display it to fit the screen (like when you open it in picture viewing app or here on SGL forum).

I've often seen that people oversample their images but are quite happy with the way image looks on their screen - because they view it on screen size. For example, they shoot image with high pixel count camera, something like 4000 x 3000 or similar, but they view that image on computer screen that is 1920x1080 or similar. That image gets reduced x2 for display by application used and consequently pixel scale is changed. If original image was good to be sampled at 2"/px but was sampled at 1"/px instead it will look ok at screen size when it is reduced to 50% of its size. It will not look ok on 1:1 pixel or 100% zoom - it will look blurry.

In my view, it's best to aim for proper sampling and make your image look at 1:1 or 100% zoom (or here stands to indicate that it can be named either or, and not to imply these are two different things - it is the same thing but someone might call it 1:1 pixel mapping while others would prefer 100% zoom) - then it will look good on both 1:1 and screen size, and I think there are people out there that don't just look at image as a whole, but also enjoy to zoom in as far as image will let them and observe all the small details in object, or even background galaxies or such small features that are spotted on 100% zoom - so yes, your image will be viewed like that by someone.

To recap (and explain why is that important when choosing focal length):

- make your image look good at 100% by properly sampling (and binning afterwards if you need to)

- make your image at least 1500x1000 in pixel count. Today's display devices will make image look too small if you don't have at least that much pixels in your image and people might be tempted to hit zoom button - and that is just going to make your image larger but blurrier. You want to keep your image with enough pixels so there is no need to hit the zoom button.

These two things are important when choosing focal length to be matched to certain sensor. Take for example ASI1600 that you have - it has 4500px in width, this means that you can use bin up to x4 and should not really use more than that. If your mount and skies can produce images that are 2"/px in quality (stars with FWHM of 3.2") then you should not exceed 1500mm of focal length.

Why is that? 1500mm of focal length gives you sampling rate 0.52"/px, so you can bin up to x4 and with x4 binning you will have 2.08"/px, but it also means that your fov will no longer be 4600 x 3500 but rather something like 1150 x 870 and that is already smallish image.

Now, you can go ahead and use even 3000mm focal length, but in order to get image according to above (I believe sensible, but not set in stone criteria) you will need to do mosaic, as each piece of mosaic will give you something like (when binned x8) 550x430 px and you will need 3x3 mosaic to compose image that is 1500 x 1200.

Hope that makes sense. There is one more consideration with focal length - FOV, it's best to keep fov such that intended targets fit inside - otherwise you will need to do mosaic regardless.

[wornish]

On a reasonably good night using the gear described in my signature I am able to get most stars around 2" to 2.5" FWHM .  I am more than happy with the star size it produces.

I was thinking of going for either the iOptron RC8 which has a focal length of 1624mm.  Tempted by the RC10. which has a FL of 2000 so that would need a reducer. It is also quite heavy and would be pushing the weight limit of the mount so probably not a good idea.

Would my guidecam attached to the 50mm/160 FL finderscope be up to the job or would an off axis guider be better?

 

Sorry for the never ending questions 

 

 

 

Edited December 21, 2019 by wornish

[vlaiv]

3 minutes ago, wornish said:

On a reasonably good night using the gear described in my signature I am able to get most stars around 2" to 2.5" FWHM .  I am more than happy with the star size it produces.

I was thinking of going for either the iOptron RC8 which has a focal length of 1624mm.  Tempted by the RC10. which has a FL of 2000 so that would need a reducer. It is also quite heavy and would be pushing the weight limit of the mount so probably not a good idea.

Would my guidcam attached to the 50mm/160 FL finderscope be up to the job or would an off axis guider be better?

 

Sorry for the never ending questions 

 

 

 

That means that on reasonably good night your sampling rate with current gear is somewhere between 1.2"/px and 1.5"/px.

Which of the two scopes is really down to size management. While RC10 has more aperture (hence better in that sense, and corrected over larger circle) it's going to be heavier and slower to cool down (larger mirror, but it is open design so it cools reasonably quickly).

With RC8 and ASI1600 (btw, that is combo that I have), you can get following sampling rates:

- native ~0.5 (0.48 but I'm rounding it up)

- bin x2 - 1"/px

- bin x3 - 1.5"/px

- Using FF/FR like Riccardi x0.75 (which works good with this scope and with RC10 version) will give you 0.64"/px without binning

- 1.28"/px with bin x2 (this would probably be best option for high resolution work as nights when you can go below 1.2"/px are rare)

- 1.93"/px with bin x3

Using ff/fr will give you reasonably wide FOV (well, it's anything but wide, but it's not unusable in size):

This was created on astronomy tools Fov with x0.73 reducer (for some reason there is no option for x0.75 in drop down list).

RC10 will be just a bit "shifted", meaning it will give you options with FF/FR that are same as RC8 without one (but again, you can calculate).

In the end, I'm working on software that will be able to do various fractional bin modes to help get closer to ideal sampling rate for given data (one can stack their data - measure FWHM and then decide to bin subs in certain way and restack binned like that to get good sampling rate and best SNR).

[newbie alert]

8 hours ago, vlaiv said:

Indeed, FL goes into sampling equation, but so does pixel size, and this second part gives you so much flexibility as you can choose how much you want to bin your image. In fact you can set target sampling rate and bin to that exact figure. It involves fractional binning, and there is rather nice way to do it when combined with stacking.

Mount error is in absolute units and not relative units - it is measured in arc seconds. So is seeing. Focal length maps those errors to certain number of pixels. This turns it into relative measure - per pixel. So if your error is some arc seconds in size - you can choose to make it one pixel large, two pixels large or less than one pixel large. Point is that you have fixed focal length (or maybe small flexibility by adding focal reducer or barlow) but you can in principle change your pixel size by manipulating data. It is easier to join pixels than it is to split pixels in math, and that is why it is easier to use longer focal length to get coarser sampling then it is to use short focal length to get finer sampling (involves drizzle algorithm and it won't quite work well in amateur setups).

What I was trying to say with this, and again you are right - mesu will be better than heq5 for example, but that does not mean that you need to use long focal length scope on mesu only - you just need to "accommodate" pixel size to the level of precision achieved depending on focal length you are using. That and the fact that using longer focal length scope will not make mount error larger in arc seconds (it will in pixels if you leave pixel size what it is).

Quite agree Vlaiv.. and your comment about its easier to join pixels( binning) than to split them I've never thought of..

[vlaiv]

19 minutes ago, newbie alert said:

Quite agree Vlaiv.. and your comment about its easier to join pixels( binning) than to split them I've never thought of..

That is sort of something that I gathered from thinking about drizzling and analyzing that approach. You can easily split pixel - just make it two adjacent pixels with same value, but when you do that and stack such subs even if you dither each sub, do you really recover high frequencies / detail in similar sense as you recover SNR with binning and lower resolution? I'm not sure.

Technical side of splitting pixels is easy, but getting true resolution out of it - that is probably rather hard.