Guiding RMS What gives?

You're forgetting about the seeing.

People typically say to divide the seeing by 3.33 to calculate the correct pixel scale to sample the seeing correctly.

For 2" seeing that's 0.6"/px  and 0.310"/px samples 1" seeing.   That's why guys with the big scopes spend big money and send them to Chile. :-)

Checkout this website to give you idea of how much extra blurriness various guiding RMS values give you over perfect guiding:  https://www.astrowithroro.com/guiding

I just went through this entire exercise last fall.   I ended up buying a premium AP Mach2 mount for big money because of it's Sky Modeling and Absolute Encoders.  The idea being that going unguided with a proper sky model creates "perfect" tracking that ignores the effects of seeing and then you get all the resolution the sky offers each night in your subs.

However, if you go to that website you'll see you save a boat load of money if you accept just 5% increase in your subs FWHM by buying a less than premium mount like the EQ6-R which can guide all day at 0.5" RMS.

For example:

Thank you everyone for the replies I'm not sure I totally understand still. First of all thanks for sharing that website That's a heck of a lot easier than doing the math myself. 

Just so I can kind of understand I'm going to say that I live in a normal neighborhood and have normal seeing in my backyard let's say that it's 3

Does that mean my guiding could be 3/3.33 (.9 ) So are we saying  if seeing is 3, that anything better than 0.9rms isn't going to be useful anyway?

I like that idea, but those numbers don't line up in my mind.

I think I am mixing guiding performance and imaging sampling too much,  and it is making me confused. I am going to write what I now understand and hopefully anyone who understand this better than I can correct any misunderstandings. 

Image sampling, specifically for imaging, and less so for guiding however that could be important too...... Not discussed below

You can really only capture as much resolution/detail as the sky permits, regardless of your optics.

If your optics are good enough (usually this is a combination of aperture/focal length/optics quality/sensor size/pixel size. ..... Then your limiting factor is often "seeing"

Example: Lets use a scenario where seeing is limited to 3, then it would optimal for the image to match this as closely as possible, while trying to avoid over-sampling or too much under-sampling 

Although seeing is variable all the time,  for the sake of example, lets say with current seeing of 3 we roughly have "seeing" of 1 arc second per pixel. 

If we have an imaging resolution (finer) than this we aren't receiving that signal anyway due to the seeing, and if we have imaging resolution higher (less fine) then we could be over-sampling and getting star bloat among other issues. 

Until this string started, I had always looked at required guiding numbers for RMS being directly related only to the imaging train sample/resolution.

I now recognize a couple obvious things. 
1. You cant guide any better than you can see
2. Guiding and Imagining is likely limited by seeing more often than not
3. The delta of having an RMS that is not optimal/perfect vs optimal/perfect RMS numbers is not nearly as big as I had originally feared.  
(This site actually helped me sleep better last night https://www.astrowithroro.com/guiding )


So  I am going to take the same camera sensor with 3 different focal lengths while we have a seeing of (3) as examples 

To calculate image scale it is generally accepted that that 206.3*pixel size / focal length gives the scale, ASI2600MC camera

300mm Focal Length   =  Pixel Scale of 2.58  (meaning the suggest half of this for target RMS guiding would be 1.29)
800mm Focal Length   =  Pixel Scale of 0.96  (meaning the suggest half of this for target RMS guiding would be 0.48)
2500mm Focal Length =  Pixel Scale of 0.31  (meaning the suggest half of this for target RMS guiding would be 0.155)  

General suggestions are that the RMS for guiding should be half of the Pixel Scale, which would mean the 300 and 800 seem reachable if you have a mid to high level mount, but the 2500mm version seems almost unreachable for many amateur AP tools. 

But.... You only can guide as good as the sky/seeing permits, and being "off" of the target RMS does not mean the sky is falling (yes that's a small pun)

Example... If seeing is 3, then the real achieved Pixel Scale arriving to the sensor is more around 1 arc second per pixel (pixel scale of 1.0), this means optimal guiding in a seeing of 3 never needs to exceed 0.5 because... you can only see what you can see.

The seeing limited Pixel Scale doesn't care that the focal length 2500 or 25000 or anything else, because its seeing limited to around 1.0 pixel scale.  I should note that if you are under-sampled you are throwing away details but doesnt have anything to do with guiding, its just the downside of having pixels that are bigger than needed as compared to your focal length, and this is pretty common with shorter focal length setups. 

Continuing on....  Lets pretend I have a mid grade mount that still cant achieve 0.5 RMS with my guide scope/camera combo but I want to use a big bad 10 inch SCT with 2500mm focal length.  

All is not lost, first remember my seeing is stuck at 3, so I only need guide as good as 1.0 divided in half so a target RMS of 0.5

And based on the fabulous calculator on https://www.astrowithroro.com/guiding I could get an RMS of 0.95 and this would only negatively affect my image/star size by an estimated 5%. So even with 0.95RMS I can take reasonable images. 

Not for this post, but I would argue that blurxterminator can fix that 5% in its sleep. 

What I have learned for this gives me a lot of optimism for using my current gear to still get longer focal length images, and it also helps me stop obsessing about guiding numbers. 

I can see where the lucky guy on top of a Chilean mountain with perfect seeing would start worrying about pixel scale and getting ridiculously low RMS numbers to take advantage of the sky seeing, but for regular peeps like me, I think seeing is between 2-3 even when I drive an hour out into the desert. 

Thanks everyone for kicking in info on this.

I can see your point of view, I dont know that I have enough experience to say that Oversampling or Undersampling is better or worse. I dont know that modern AI tools seems to be able to do more with oversampled data. 

But I am content, with what I have discovered about guiding in this forum and can go back to imaging vs tweaking my guide settings that likely arent going to make much of a difference anyway.

Recently it’s been noticed that blurx does a better job if you have more pixels sampling across the blurry star ‘spot’

how much better and how many more than the traditional 3.33 is up to your experimentation.

lastly, the smaller the square in the sky each pixel sees the less signal will be captured by said pixel so you compensate by taking longer subs, but then you have to worry whether or not your jiggly mount can keep said pixel still enough against the sky to support the longer exposures.

the only way to get more signal into that pixel is to 1) expose longer, 2) concentrate more light onto said pixel with a bigger mirror, 3) a little of both.

in general you pick an image scale (how much sky each pixel sees), then buy the biggest telescope you can that your mount can hold still enough.   This means for a fixed image scale the bigger the mirror the faster it’s focal ratio has to be.

also don’t forget that you can always downsize your image in PixInsight if you sample too fine.  Conversely you can drizzle during stacking if sample too coarsely.

its all a balancing act , enjoy the ride that is astrophotography.

Clayton, 

Perhaps it is me that is not understanding the premise — and I will gladly stand corrected.  I'm not sure I agree with the premise that Pixel Scale divided in half results in an "RMS." 

In two dimensions (represented by RA and DEC), the guiding RMS is calculated as the square root of RA² + DEC².

So, if one's Pixel Scale is 2.585, and the Guiding RMS is .70 RMS; all other factors aside (focus, seeing, wind, etc.), the guiding "error" will never be greater than the main camera's Pixel Scale.  The photons being collected will "never" move off of each Main Camera pixel, and the result will be "round stars" and detail as good as possible.

And, the guiding "error" is constantly being corrected by the mount.

Again, if I am not understanding the issue here, please let me know.

Jay

Jay Hovnanian:
Clayton, 

Perhaps it is me that is not understanding the premise — and I will gladly stand corrected.  I'm not sure I agree with the premise that Pixel Scale divided in half results in an "RMS." 

In two dimensions (represented by RA and DEC), the guiding RMS is calculated as the square root of RA² + DEC².

So, if one's Pixel Scale is 2.585, and the Guiding RMS is .70 RMS; all other factors aside (focus, seeing, wind, etc.), the guiding "error" will never be greater than the main camera's Pixel Scale.  The photons being collected will "never" move off of each Main Camera pixel, and the result will be "round stars" and detail as good as possible.

And, the guiding "error" is constantly being corrected by the mount.

Again, if I am not understanding the issue here, please let me know.

Jay

I dont think we are saying anything different but maybe I can clarify

It was suggested that a "good" RMS number would be that Pixel Scale divided in half, so this would be a reasonable target RMS goal. 

I agree with your pixel scale analogy. If your Pixel Scale is 2.585 and your RMS is 0.70 you can easily guide and retain the same round stars, 

But to retain a Pixel Scale of 2.585 at a focal length of 2500mm, you would literally need 25um pixels, so the issue at hand was 

What kind of RMS do I need if my pixel scale is 0.31? 

But the reality is that regardless of the pixel scale, most seeing provides a pixel scale of around 1.0, so... stressing over the 0.3 scenario is best left to someone in better seeing and with better equipment budgets.

Alex Nicholas:

Tony Gondola:
sampling is better than under.

Seeing in most places is highly variable so if you shoot shorter subs and cull as needed you can improve your average seeing at the cost of taking longer to capture data.

IMO, guiding as it presently is done, degrades high frequency information in your image the same way seeing does. Just like seeing effects, it averages out in stacking so you get round stars.

The bottom line is: guide as well as your hardware and conditions will allow no matter what the rules tell you is good enough.

I think this is a really important point that is easy to overlook when you're considering what your system 'needs' to guide at, but its also REALLY important to not get obsessive with your guiding numbers unless you want to spend most of your imaging time fine tuning the guiding over and over.

My rig gets 0.9"/px at f/7 and 1.14"/px at f/5.6, with the average seeing in my area being between 2 and 2.2". Given that, Guiding around the 0.9" rms is all I really NEED to achieve. I have tuned my setup down to about 0.4"~0.5" RMS... Some nights it runs a little worse at ~0.7" rms, some nights its a bit better at 0.3" RMS, but its always (even at its worst) producing better guiding than my system needs to achieve sharp round stars. This allows me to basically start imaging and simply ignore the guiding - knowing that it will perform at the level I need it to.

That's great advice and frankly the way we got to this whole point was I had just spent three nights tweaking guide settings. Now I can just go back to imaging

Jay Hovnanian:
So, if one's Pixel Scale is 2.585, and the Guiding RMS is .70 RMS; all other factors aside (focus, seeing, wind, etc.), the guiding "error" will never be greater than the main camera's Pixel Scale.  The photons being collected will "never" move off of each Main Camera pixel, and the result will be "round stars" and detail as good as possible.

This isn't quite correct.   Even the smallest "jiggle" in your camera pixel against the sky will result in that pixel seeing a bigger portion of the sky over time than the calculated image scale. 

If the "jiggle" is evenly distributed in both RA and DEC then that pixel will sample a "square in the sky" equal to the image scale plus the guiding RMS error.   If that random motion is uniformly distributed (a big, even smear) then it acts like you have a larger image scale (larger pixel), but if the random-ness is normally distributed (a bell curve) then it will sample the center of the area more than the edges according to the the classic normal bell shape of a normal distribution.

I haven't quantified the type of randomness in a mount's guiding error, but you can get a good idea by looking at PhD2's "target bullseye" and setting the history count there to 400.   The distribution of red dots will graphically show you the type of randomness your mount is producing over those 400 samples.

Ideally you'd like to see most of the red dots clustered in the center and then the count gradually diminish as you get further and further from the center of bullseye.  That's like a normal distribution.   However if it's just a big, even red smear then it's more like a uniform distribution.

I don't know how Rob's website calculates the increase in blurriness for each seeing/guiding RMS but I suspect he assumes a normal distribution.

Clayton Ostler:
Thank you everyone for the replies I'm not sure I totally understand still. First of all thanks for sharing that website That's a heck of a lot easier than doing the math myself. 

Just so I can kind of understand I'm going to say that I live in a normal neighborhood and have normal seeing in my backyard let's say that it's 3

Does that mean my guiding could be 3/3.33 (.9 ) So are we saying  if seeing is 3, that anything better than 0.9rms isn't going to be useful anyway?

I like that idea, but those numbers don't line up in my mind.

As a rule of thumb, and so as not to get to lost in the maths and physics and then not enjoy the hobby, take your average seeing of 3”/pixel, (you are seeing limited) then in theory approx 2.0 total RMS would be ok, but for the sake of errors take 1.0 rms, which gives a bit of headroom, and you will be good, but you want your RA and DEC rms errors to be as close as possible to still get nice round stars.
i image at 1.5”/pixel, and my average seeing is around 2.5 - 3, so I am seeing limited, and not limited by my optics, so as long as I keep the RMS under approx 1.5 then I am happy as long as the two axis are as close as possible, I usually get between 0.4 and 0.9 with my mount so I never even bother about the guiding numbers, as they are always well under what I need.

in your second example above, you will be seeing limited as your optics can resolve more detail than the seeing will allow, so you need to take that into account, just because you are imaging at 0.3”/pixel, does not mean you can see that much detail, because the seeing won’t allow it, so no point in chasing really low guide RMS numbers, it just ain’t gonna happen.