Is My Guide Scope Image Scale Too Small?

[Andyb90]

Hi Everyone,

I've been reading up on imaging and guiding image scale and I'm wondering if my guiding image scale may be too small.

WIth my Evoguide 50ED and ASI290mm mini it is 2.47 "/pixel.

I've read a couple of posts on forums suggesting the guide scope image scale should ideally be around half the typical seeing. Looking at Meteoblue for my location that would be 1-2".

When I plug in the above values in the ccd suitability page below it does say my combination of kit gives significant under sampling.

https://astronomy.tools/calculators/ccd_suitability

So I'm thinking could this cause issues with guiding? I do check my stars by zooming into subs in SGPro as they come in. There doesn't seem to be any real issue. Just wondering if increasing my guiding image scale could get stars tighter. I'd be grateful for any advice on this and it would be useful to hear of anyone elses experience.

Andy.

[Stub Mandrel]

My guidescope/amera combo is 4.3" per pixel and seems to guide my 590mm imaging rig OK.

The real question is what rms in arc-seconds do you actually get.

[newbie alert]

Too many worry about sampling rate..

Now it's spread to guidescope sampling..

For peace of mind lots undersample  with their imaging setup..

[michael8554]

Hi Andy

I think somebody has confused you, the under and oversampling discussion is solely for the imaging camera, and is a can of worms.

The trend in DSLRs and Astro cameras is smaller and smaller pixels, so often oversampling is unavoidable with longer focal lengths.

Listen to newbie alert.

There is a more pertinent discussion as to what ratio of imaging scale and guiding scale is suitable for guiding.

Michael

[Andyb90]

On 13/03/2019 at 19:17, Stub Mandrel said:

The real question is what rms in arc-seconds do you actually get.

I usually get around 0.5" rms. 

21 hours ago, michael8554 said:

There is a more pertinent discussion as to what ratio of imaging scale and guiding scale is suitable for guiding.

This was my thinking too. I found the below thread on the Cloudy Nights forum. But within the thread there is this post where under sampling comes up. 

https://www.cloudynights.com/topic/617386-imaging-vs-guiding-image-scale/?p=8567366

 

[Stub Mandrel]

41 minutes ago, Andyb90 said:

I usually get around 0.5" rms. 

I very much doubt that anything you can do will improve on that, and it's hard to see why you might need to.

[ollypenrice]

0.5" RMS would support imaging at 1"PP which many might consider to be the likely seeing limit on a good night anyway. I don't think you have a problem.

Also your mount is the excellent Avalon Linear Fast Reverse which may well have a 'real guide trace' which is better than the recorded guide trace suggests. The recorded trace between guide inputs is only an estimation because the software has no information on where the mount is really pointing. Avalon claim, convincingly in my view, that their mount returns to the correct position more quickly because it has no backlash and so it is on target for more of the interval between corrections. When I had one of these mounts I did feel that it seemed to give a better result than our EQ6 for the same recorded guide trace and I know other owners who feel the same.

Olly

[michael8554]

I didn't read all of the CloudyNights link but what I think was raised is:

If your guide image scale is say 8arcsec/pixel, and a star is only 0.3 pixels wide, then the star can wobble +/- 0.35 pixels or about +/- 3arcsecs without leaving the pixel that PHD2 is monitoring ?

Now PHD2 has "sub pixel guiding" capability, so somehow PHD2 knows the star is moving within the pixel ?

Perhaps PHD2 auto-chooses a star that is as big as the pixel ?

I can't quite visualise this I'm afraid. 

Michael

[Stub Mandrel]

6 hours ago, ollypenrice said:

Avalon claim, convincingly in my view, that their mount returns to the correct position more quickly because it has no backlash and so it is on target for more of the interval between corrections.

LOL!

What sort of maths did they prove that with.

The true RMS error is proportional to the area under the curve, it doesn't really matter if between points it comes back to the line sooner or later. A fast return is just giving longer for it to wander off target before the next correction.

[Stub Mandrel]

4 hours ago, michael8554 said:

I didn't read all of the CloudyNights link but what I think was raised is:

If your guide image scale is say 8arcsec/pixel, and a star is only 0.3 pixels wide, then the star can wobble +/- 0.35 pixels or about +/- 3arcsecs without leaving the pixel that PHD2 is monitoring ?

Now PHD2 has "sub pixel guiding" capability, so somehow PHD2 knows the star is moving within the pixel ?

Perhaps PHD2 auto-chooses a star that is as big as the pixel ?

I can't quite visualise this I'm afraid. 

Michael

The star is bigger than a pixel. If it's not perfectly centred the pixel to the left (say) will get more signal than that to the right. PHD2 shows this as  curve which is a 'cross section' through the star. Its possible to accurately calculate the centre point to sub-pixel accuracy by assuming the curve is symmetrical.

[michael8554]

7 hours ago, Stub Mandrel said:

The star is bigger than a pixel.

Yes for the star to have a FHWM curve it must spill over onto adjacent pixels.

So don't believe everything you see on CloudyNights ...........

Michael

[ollypenrice]

On 15/03/2019 at 16:51, Stub Mandrel said:

LOL!

What sort of maths did they prove that with.

The true RMS error is proportional to the area under the curve, it doesn't really matter if between points it comes back to the line sooner or later. A fast return is just giving longer for it to wander off target before the next correction.

Not sure I agree with this. Any delay in the implementation of the correction is lost 'accurate time.' With the fast reverse this lost time is avoided (or greatly reduced) and would allow the imager to reduce the guide interval to give the next input before too much wandering off had happened. I found I got best results with inputs every 1.5 seconds or so, though Avalon recommended longer.

Olly

[ollypenrice]

On 15/03/2019 at 16:56, Stub Mandrel said:

The star is bigger than a pixel. If it's not perfectly centred the pixel to the left (say) will get more signal than that to the right. PHD2 shows this as  curve which is a 'cross section' through the star. Its possible to accurately calculate the centre point to sub-pixel accuracy by assuming the curve is symmetrical.

...which may be why Craig Stark suggests slightly soft focus for guiding. The stellar image will then cover more pixels, aiding the centroid calculation and (possibly?) making the image rounder. I don't know about the second point.

Olly

[Stub Mandrel]

5 hours ago, ollypenrice said:

...which may be why Craig Stark suggests slightly soft focus for guiding. The stellar image will then cover more pixels, aiding the centroid calculation and (possibly?) making the image rounder. I don't know about the second point.

Olly

I use a rather short guidescope (183mm) so about 1/3 the length of the main scope. My FWHM is typically quite large at ~4.5 pixels. Interestingly the 'star graph' usually shows a small 'peaklet' either side of the main peak, which I assume is a vestige of an Airy disc although not really visible to the naked eye. This would suggest the curve isn't gaussian but I can't see any reason why (if fully plotted instead of just being relatively few points) the curve wouldn't be symmetrical. I suppose if it isn't because of coma, field curvature, astigmatism or other aberration it doesn't really matter as long as the 'centre point' is consistently calculated as the same point on the curve, but fitting a curve to only a small number of points would be a lot easier if you assume symmetry. (Definitely shooting from the hip here...)

[AKB]

14 hours ago, Stub Mandrel said:

doesn't really matter as long as the 'centre point' is consistently calculated as the same point on the curve, but fitting a curve to only a small number of points would be a lot easier if you assume symmetry. (Definitely shooting from the hip here...)

AFAIK, the calculation is simply for the centroid which is very robust and low computational cost, not curve-fitting?

[Stub Mandrel]

46 minutes ago, AKB said:

AFAIK, the calculation is simply for the centroid which is very robust and low computational cost, not curve-fitting?

I haven't been able to turn up an explanation of the calculation.

I have learned that it's worth using dark frames and that the advice to use a slightly out-of focus guidescope has been changed as the chance of a star actually fitting all in one pixel is vanishingly small.

Found this:

http://www.stark-labs.com/help/blog/files/PHDSubpixelAccuracy.php

Also PHD2 documentation states curve fitting is required for FWHM calculation but not half flux diameter. As PHD2 displays both it must do a curve fit, even if it doesn't use it for the star location calculation.

[Andyb90]

I hadn't noticed until now but the star profile in PHD2 shows the measured star size in arc seconds. I just checked a screenshot from my last session and it shows 10 arc seconds for the auto selected star, which is well above my guider image scale. The Cloudy Nights post assumes much smaller guide star sizes.