Analyzing PHD2 Guiding Results and PHDLogViewer

[+Derbyshire Dave]

I know that most of us spend a great deal of time wondering what the heck is going on with PHD2. I found this really useful tutorial, which can be downloaded as a PDF file to keep.

 

https://openphdguiding.org/Analyzing_PHD2_Guide_Logs.pdf

 

The PHDLogViewer program is just great.  You can download it here..

 

https://openphdguiding.org/phd2-log-viewer/

 

It gives you an overview of the log, not just a single screens worth, in my example below there were six guiding sessions in the evening (see top left). You can expand the graphs to look at them in more detail, see the correction pulses etc.

 

They say that you should aim for an arcescs/pixel of less than one, so you can see that here I am just outside that . You can also see the cyclical nature, I'm slightly overcorrecting - work to be done there!

 

The white line is the SNR (Signal To Noise Ratio), and you can see it drop off top right as the clouds rolled in.

 

Hope you find this useful guys....

 

 

 

[Bottletopburly]

Cheers Dave will have a gander later I usually have the white line showing as you say you can see the seeing conditions deteriorate or improve, I also have the algorithm predictive pec on Ra  I think it’s called that ,seems to work better as it learns your worm turning thingy mabob usually takes half hour to start kicking in , also guiding scope good focus helps too something I need to tweak 

[Clive]

Yes, it’s a really useful tool, it was my second download after PHD2 and proved to be invaluable in my initial setting up of PHD2 with my mount.  It's a shame that it cant produce the frequency response that PHD can do so as to diagnose where the RA tracking errors arise (at least I haven't found a way to do it).

 

" ... They say that you should aim for an arcsec/pixel of less than one, so you can see that here I am just outside that ... "  Your actual guiding error is 2.56 arc-sec/pixel (1.15  / 0.45).   

 

The guiding accuracy given by PHD can only be in terms of the guide scope pixels (since that is all it can measure from the guide camera data), and that can only be related to arc-sec if you supply the correct pixel pitch of the guide camera and guide scope focal length.  So the guiding accuracy given by PHD2 can give wildly varying imaging results in practice depending upon the guiding / imaging set-up.

 

Consider a setup using a guide camera with a pixel dimension of 5 um used with a guide scope with a focal length of 250 mm, then the scaling shown by PHD will be (approximately) 4 arc-sec / pixel.  So, if the guiding accuracy shown by PHD is 1 arc-sec RMS  (= 0.25 pixel RMS) then, with an imaging camera with a pixel dimension of 5 um and a telescope with a 250 mm focal length, the resulting image will also have an imaging accuracy of 1 arc-sec RMS (= 0.25 pixel RMS).  But, if the imaging camera has a pixel size of (say) 3 um and the imaging telescope has a focal length of (say) 750 mm, then the guiding error seen by the imaging camera will be (750 / 250) * 1  arc-sec  = 3 arc-sec and the guiding error in terms of the imaging camera pixels will be (3 / 1) * (5 / 3)  =  5 pixels.

 

[+RonC]

I found it interesting and had a look at some recent sessions. I managed on average 1.18  and am happy with that.

Guiding is a means to an end, the end product being a clean, star rounded image with no trailing etc.. so as long as 

I get that I'm happy, I'm not going spend a lot of time checking PHD graphs it can lead to astro-paranoia!! 😄

[Clive]

1 hour ago, RonC said:

I found it interesting and had a look at some recent sessions. I managed on average 1.18  and am happy with that.

Guiding is a means to an end, the end product being a clean, star rounded image with no trailing etc.. so as long as 

I get that I'm happy, I'm not going spend a lot of time checking PHD graphs it can lead to astro-paranoia!! 😄

I totally agree Ron, it can be a useful tool if you are having problems guiding, but, as you say, if you're getting round stars there's no need worry - just expand the Y axis scaling when guiding and everything will look great 😁

[+Graham]

On 01/10/2020 at 17:28, Clive said:

The guiding accuracy given by PHD can only be in terms of the guide scope pixels (since that is all it can measure from the guide camera data), and that can only be related to arc-sec if you supply the correct pixel pitch of the guide camera and guide scope focal length.  So the guiding accuracy given by PHD2 can give wildly varying imaging results in practice depending upon the guiding / imaging set-up.

 

Consider a setup using a guide camera with a pixel dimension of 5 um used with a guide scope with a focal length of 250 mm, then the scaling shown by PHD will be (approximately) 4 arc-sec / pixel.  So, if the guiding accuracy shown by PHD is 1 arc-sec RMS  (= 0.25 pixel RMS) then, with an imaging camera with a pixel dimension of 5 um and a telescope with a 250 mm focal length, the resulting image will also have an imaging accuracy of 1 arc-sec RMS (= 0.25 pixel RMS).  But, if the imaging camera has a pixel size of (say) 3 um and the imaging telescope has a focal length of (say) 750 mm, then the guiding error seen by the imaging camera will be (750 / 250) * 1  arc-sec  = 3 arc-sec and the guiding error in terms of the imaging camera pixels will be (3 / 1) * (5 / 3)  =  5 pixels.

 

I am with Ron on this one.

I still use the original PHD so the only thing I need to consider is the live graph and the final result.

Cannot be doing with worrying about all these numbers, logs ect. 

 

Clive as you have a keen grasp on the mathematics of this I want to pose a question.

 

Unless you are using an OAG most imaging scopes are going to be faster than the guide scopes.

The imaging camera's pixel size will also probably be different to the guide camera ( including that for OAG ).

 

My question is this.

Would the error be less in the imaging scope than the guide scope thus resulting in Ron's round stars despite the 1.18 error.

 

 

[Bottletopburly]

Best Rms was 0.25 average just over 1 

[Clive]

10 hours ago, Graham said:

 

I am with Ron on this one.

I still use the original PHD so the only thing I need to consider is the live graph and the final result.

Cannot be doing with worrying about all these numbers, logs ect. 

 

Clive as you have a keen grasp on the mathematics of this I want to pose a question.

 

Unless you are using an OAG most imaging scopes are going to be faster than the guide scopes.

The imaging camera's pixel size will also probably be different to the guide camera ( including that for OAG ).

 

My question is this.

Would the error be less in the imaging scope than the guide scope thus resulting in Ron's round stars despite the 1.18 error.

 

 

 

Graham, I think the answer to your question is no.  In terms of pixels, the error in the imaging system likely to be higher than that of the guide system since the guide scope focal length tends to be less than the imaging scope focal length.  But in terms of arcsecs, the guiding error is identical in both the guide and imaging systems.

 

I assume the guiding error quoted by Ron is 1.18 arcsec (rather than 1.18 pixels), so the imaging system will also see the same guiding error.  Ron’s imaging system is a 150PDS and a 183C (shout up if I’ve got it wrong Ron) so the imaging system focal length (assuming no reducers are used) is 750 mm and the imaging system pixel size is 2.4 μm.  Assuming 1 x 1 binning is used, each imaging camera pixel therefore represents 0.66 arcsec, so a 1.18 arcsec guiding error equates to 1.8 pixels, which isn’t a great deal but would be visible as slightly elongated stars if ‘pixel peeping’ short exposure images.  However, the guiding error generally varies randomly about a central point and this can be seen with the scatter plot in the log viewer:

 

 

So, providing you have a sufficiently long total integration time, I would suggest that the stars will still look perfectly circular, but the diameter of the stars will be 'bloated' by the guiding error.  The stacking process, particularly if using a Kappa Sigma method, will remove the outlier points shown in the scatter plot from the final image.  The scaling of the scatter plot not given but the red circle appears to represent 1 arcsec, and the total guiding error for this run was 1.26 arcsec.

[+Graham]

Thanks Clive

That makes sense 👍

[+Derbyshire Dave]

On 01/10/2020 at 17:28, Clive said:

Yes, it’s a really useful tool, it was my second download after PHD2 and proved to be invaluable in my initial setting up of PHD2 with my mount.  It's a shame that it cant produce the frequency response that PHD can do so as to diagnose where the RA tracking errors arise (at least I haven't found a way to do it).

 

" ... They say that you should aim for an arcsec/pixel of less than one, so you can see that here I am just outside that ... "  Your actual guiding error is 2.56 arc-sec/pixel (1.15  / 0.45).   

 

The guiding accuracy given by PHD can only be in terms of the guide scope pixels (since that is all it can measure from the guide camera data), and that can only be related to arc-sec if you supply the correct pixel pitch of the guide camera and guide scope focal length.  So the guiding accuracy given by PHD2 can give wildly varying imaging results in practice depending upon the guiding / imaging set-up.

 

Consider a setup using a guide camera with a pixel dimension of 5 um used with a guide scope with a focal length of 250 mm, then the scaling shown by PHD will be (approximately) 4 arc-sec / pixel.  So, if the guiding accuracy shown by PHD is 1 arc-sec RMS  (= 0.25 pixel RMS) then, with an imaging camera with a pixel dimension of 5 um and a telescope with a 250 mm focal length, the resulting image will also have an imaging accuracy of 1 arc-sec RMS (= 0.25 pixel RMS).  But, if the imaging camera has a pixel size of (say) 3 um and the imaging telescope has a focal length of (say) 750 mm, then the guiding error seen by the imaging camera will be (750 / 250) * 1  arc-sec  = 3 arc-sec and the guiding error in terms of the imaging camera pixels will be (3 / 1) * (5 / 3)  =  5 pixels.

 

 

On 02/10/2020 at 14:00, philjay said:

 

Thanks for pointing out my schoolboy error Clive, appreciated, and also appreciate yiur understanding and great example.

Agree with Ron though, all we all want small round stars. Mine are pretty good but will be using this tool to try and understand and improve things if I can..