Cause of jagged PHD2 graph?

[vlaiv]

Trying to sort out my guiding for some time, but I don't have much luck with it. I did not take screen shot of my guide graph, but will try to give best description of it and circumstances.

I'm trying to guide on really high res - 0.64"/pixel, imaging rig, with 240mm FL guide scope (60mm F/4 refractor).

Guide resolution is 3.22"/pixel - this is x5 imaging resolution, so I guess it is within acceptable bounds. Mount is Heq5, I did belt mod on it, and tweaked it for backlash (there is now just a bit of it). I pay attention to balance (slight disbalance - east heavy). I might be pushing the mount a bit more than recommended with 11-12kg stuff on it, and about 1.2m long scope on it. I do have small ST102 frac that I'm going to try with next - just to observe the guide graph and see if weight has anything to do with this.

Still I get really jagged guide graph, so I presumed it is due to seeing - I switched off guiding in DEC and observed the graph - I get +/- 1" peaks with no guiding in DEC. There were no wind, no vibrations of any kind, no cable snag - nothing that I could think of that would cause this. I did take darks (16 for each exposure) for guide camera - which is QHY5IILc, I tried guide stars with SNR in range of 30-50, I tried different exposure lengths (0.5s up to 4s), as I suspected the seeing but, each time graph behaved more or less the same (2" peak to peak error, almost like oscillatory motion but random enough not to be sine wave - this is DEC without guiding that I'm talking about). So I thought to myself - maybe it is due to undersampling, so I tried to guide with main camera (resolution 0.64"/pixel - ASI185 at prime focus) - or rather just to observe how would DEC graph behave - I got almost the same result - peak2peak was this time just a bit less than with guide scope - but same jagged graph.

So by this time I was rather frustrated, I took off cameras, rebalanced scope and decided to have a quick peek at Saturn and Mars. I usually observe with this scope on dob mount (it is SW 200 dob tube), but recently purchased 6mm baader ortho and apm barlow and wanted to try them out on EQ mount because using those on dob is not quite as good as some wider field combinations (high mag, narrow FOV - lot of pushing ). Then I noticed something that really confused me. Seeing was actually pretty good! I could crank up mag up to x400+ and still see good image (not perfect, but really really good) - at this kind of mags Cassini division was readily visible and pretty sharp. Mars was also good, a bit more affected by seeing but I could clearly see shape of dark regions and all (have not been observing Mars so close before - it was always red dot before ).  Then it struck me, seeing was decent enough and all of the time I thought that guiding problems were due to bad seeing.

So the question goes - what could be causing such a bad guide graph? I'm particularly concerned with the fact that DEC graph looked bad with no guiding at all - this means that guide stars were jumping around on their own - not caused by bad setting or something like that. I did not mention RA graph - it looked much the same, except a bit worse in sense that it looked like "chasing the seeing" lot of overshooting and overcorrecting, sometimes RA would spike to 2"-3" on one side ... Images on other hand look sort of ok, it is hit and miss affair, sometimes I get really nice and tight stars, sometimes there is elongation in RA (most of the time), if I turn on DEC guiding (either auto on resist switch or I pick correct side my self), there will be some distortion i DEC as well from time to time. What is even stranger, if I turn off guiding and do exposure (30s - 1m range) I get the same results (most of the time elongation in RA, but sometimes really nice tight star - depending where on the PE curve mount is at the time). I should also mention that I don't do PEC, but I had previously done PE recording and I have quite PE range (35"+ p2p) - this was before belt mod, but I suspect that worm period did not change when doing mod - I was hoping that high frequency components would be reduced so I could guide better but did not check since.

So that is the story. Does anybody have any idea what might be going on? Any help would be highly appreciated.

[daz]

The log files from PHD would be useful, if you still have them?

 

[vlaiv]

Well no :D, at least I don't think so, don't know if I had the logging turned on, and if it's turned on by default. I did some research in the mean time, and concluded that it might be the seeing after all.

Found this very informative paper on the subject, so I have a couple of ideas what to try next - basically ignore the graph, or rather it's jaggedness, do the PEC before guiding and calibration, and minimize response (aggressiveness) to avoid any idea of chasing the seeing in hope that I could live with uniform guiding errors (not so tight stars). Maybe even try lowpass algorithm for RA. Need to experiment to get the feel for what works best now that I know where I've been wrong in my assumptions (I presumed that it cannot be the seeing, because it did not make sense to me that the same behavior was present in range of exposures and seeing was fine visually afterwards - but it turns out that sometimes under particular circumstances - like imaging from populated area / towns one might need up to 10s to stabilize the seeing).

Here is the paper, it contains some really nice ideas and ways to think about guiding:

http://acp.dc3.com/McMillanAutoguiding11-2005.pdf

My idea goes a bit against ideas presented in this paper. Author suggests to try to minimize number of corrections. I will first try following approach: use shortest exposure time / highest correction rate possible and really lower aggressiveness - in this way I hope to achieve following: many small corrections that will not have big smearing effect on star size (for example if p2p of star position is +/-1" and I use 10% of correction I will end up with +/-0.1" introduced error - guess it is acceptable on 0.64"/pixel resolution) that will add up to correct following of star position (although random in nature as seeing is, they will follow trend of star position). Also due to short exposure I expect to have better responsiveness than I would if using very long exposure to average the seeing. So I will basically trade some FWHM of stars for quicker response time.

So I'm not stuck without ideas any more, and have a lot to experiment. Given current weather patterns (somehow moon and rain tend to anti correlate these days ) this will keep me busy and experimenting for some time in future.

 

[kens]

Try using the PHD2 guiding assistant to work out what is going on.

[Notty]

Hi, I use an HEQ5, and with my ED80-DS +DSLR +ST80 guidescope I get acceptable guiding but when using my 200PDS reflector+DSLR even with my new super lightweight QHY guidescope the guiding corrections go mad with RMS totals 1.0+/-0.1.  I haven't weighed the two setups separately but though the big reflector is certainly at the limit of the mount's design I reckon, it may be the sheer dimensions of the thing.  Do you have a lighter scope you could try it with?

[vlaiv]

I've gathered the list of things to try out next, now just waiting for clear skies (moon or not )

- Try out combination of PEC / Guide assistant / small corrections with minimum aggression on present setup

- Try out ST102 with this scope - lightweight setup / use guide camera at both guide scope and at prime focus (check guiding resolution) to see what will guide performance be like

[johnrt]

I think your issues are that the mount is overloaded for imaging, imaging 1.2 meters and 0.67"/pixel is also very demanding and would require a mount capable of handling the weight with ease, yours is struggling. 

[Pompey Monkey]

I hate to be the bearer of bad news, but as good value as it is, in my opinion a HEQ5 is just a HEQ5. It won't perform miracles.

You sound like you are set up as well as you could be (apart from a little bit too much weight) and have probably reached the best performance that you could get for the mount.

You don't say what your rms guiding errors are, but +/-1" peak to peak in declination sounds pretty darned good to me for this mount. It is very similar in performance to the best guiding that I get with my AZ-EQ6, and I'm perfectly happy with that.

Of course, I'll be delighted if your continuing experiments show me to be wrong!

[ollypenrice]

I entirely agree with Paul above. Also...

If you set a very short exposure time and so generate lots of small and frequent corrections there is a danger that you will create a good guide graph without creating good guiding. The guide graph does not show deviation from the position of the guide star, it shows deviation from the position of the image of the guide star, and if this image is dancing around due to the seeing you will chase it. If you take longer exposures the seeing will produce a blurred-out image whose calculated centroid will be more likely to lie where the star really lies behind the seeing.

I have imaged at 2.5 metres FL/0.66"PP very successfully but 1) I was using a Mesu 200, not an HEQ5 and 2) I was using an OAG.* I really don't think you have much hope of guiding effectively with a separate guidescope at the resolution you're looking for. An OAG would also save you weight.

Olly

*By the way, I recently did a quick test of a 10 inch Meade ACF using my regular 400mm guide scope. This normally guides our dual Tak FSQ rig with excellent results. Using it on the Meade I got my usual 0.5" guide trace but significantly oval stars. This confirmed my belief that long FL reflectors just do need OAGs.

[tony210]

I agree with above - I use RCTs to image and had enormous problems with a guide scope - once I tried an OAG never looked back- dramatic improvement in star shape even when the PHD graph is like crocodile teeth!- I suspect reflectors have subltle mirror shift issues and you need OAG with these as opposed to more stable refractors - good luck  Tony

[vlaiv]

There is something definitely wrong with the mount mechanics, I'll have to stripe it and redo lubrication and adjusting. Last night I wanted to measure PE with light load, but forgot that PERecorder does not work on 64bit windows, so after setting everything up, I ditched the idea of recording 2 hours of PE, and played with MetaGuide and PHD2 instead, and then decided to do kind of imaging / guiding session (with achromat on the night of the full moon ) to record logs and see how would mount behave under really light load. I used ST102 and guide scope (TS60mm f/4) so total weight on mount must have been lower than 5kg (I used only one counterweight). So I did 2 hours guided imaging and I noticed that I still get graph that I'm not satisfied with. There is sudden jump in RA every 100s or so to the one side - this might be due to over tightening something, guide star is all over the place (+/- 1.5") - seeing was pretty good I believe (order of 2-2.5" FWHM), and I noticed strange pattern - there seems to be a decent amount of correlation between RA and DEC error - this might indicate orthogonality problem with calibration but I reviewed cal data before guiding and it appeared to be pretty orthogonal to me. Either bad calibration or there is something in the mount mechanics that causes vibrations in both RA and DEC. I image from the lawn so don't think it's external to the system, I also make sure to stab legs into the ground instead just placing them on grass.

Here is guide log from phd2 if someone is willing to review it. First hour I used pretty much basic settings, second hour I switched to different guide star and just started experimenting with different settings (guide exposure length, aggressiveness, min motion, etc) but without any meaningful effect.

Some parameters that you may need if inspecting the log:

Guider pixel size: 3.75

Guider FL: 240mm

 

PHD2_GuideLog_2016-05-21_225330.txt

[Pompey Monkey]

1 hour ago, vlaiv said:

There is something definitely wrong with the mount mechanics, I'll have to stripe it and redo lubrication and adjusting. Last night I wanted to measure PE with light load, but forgot that PERecorder does not work on 64bit windows, so after setting everything up, I ditched the idea of recording 2 hours of PE, and played with MetaGuide and PHD2 instead, and then decided to do kind of imaging / guiding session (with achromat on the night of the full moon ) to record logs and see how would mount behave under really light load. I used ST102 and guide scope (TS60mm f/4) so total weight on mount must have been lower than 5kg (I used only one counterweight). So I did 2 hours guided imaging and I noticed that I still get graph that I'm not satisfied with. There is sudden jump in RA every 100s or so to the one side - this might be due to over tightening something, guide star is all over the place (+/- 1.5") - seeing was pretty good I believe (order of 2-2.5" FWHM), and I noticed strange pattern - there seems to be a decent amount of correlation between RA and DEC error - this might indicate orthogonality problem with calibration but I reviewed cal data before guiding and it appeared to be pretty orthogonal to me. Either bad calibration or there is something in the mount mechanics that causes vibrations in both RA and DEC. I image from the lawn so don't think it's external to the system, I also make sure to stab legs into the ground instead just placing them on grass.

Here is guide log from phd2 if someone is willing to review it. First hour I used pretty much basic settings, second hour I switched to different guide star and just started experimenting with different settings (guide exposure length, aggressiveness, min motion, etc) but without any meaningful effect.

Some parameters that you may need if inspecting the log:

Guider pixel size: 3.75

Guider FL: 240mm

 

PHD2_GuideLog_2016-05-21_225330.txt

Apologies if this is going to be asking the bleedin' obvious, but were you dithering between subs by any chance?

PHD Lab shows the deviations, but does not report them in the RA error either.

EDIT: - Your guide log does not seem to have any dither commands, so I'm at a loss here. Over to the experts!

[vlaiv]

No, no dithering that I know of .

That thing that you mention - 2.8minute jumps - I suspect to be from to tightened worm or something - it looks like friction thing - I'll probably sort it out with lubing / adjusting everything.

Other thing that I'm concerned is correlation between RA and DEC in jitter and amplitude of jitter itself - I still can't bring myself to accept it is the seeing (if it really is). I even switched to 8s guide exposure - still random +/- 1-1.5" deviations.

[Tommohawk]

Couple of thoughts - not an expert but have had similar PHD2 issues. Sorry if some of this mentioned already.

East heavy is fine, but DEC doesnt wnat to be too finely balanced - its better to be leaning fairly well to one side so it doesnt easily flop from one extreme of backlash to the other.

Consider defocusing guide star a bit - this seems to be preferred.

As Olly says, bear in mind you may be getting "reported" shift rather than actual shift - due eg to atmos turbulence.

I think youre saying you get DEC shift associated with RA shift ie the RA and DEC peaks seem to follow each other - I get this too. It could happen because of turbulence, ie guide star "seems" to move, or because if RA moves due to friction, PE or anytihng really, you might get a secondary wobble in DEC simply because the movement may cause a more general wobble. 

With DEC guiding off, there should be a general drift in DEC, unless polar align is perfect, but I get a series of peaks and troughs. This simply should not happen, but it does. Presumably due to reasons above ie "perceived" guide star movement or secondary wobble. 

Also, default option for DEC is "resist switch". This allows considerable movement before correction kicks in - I have found setting to "hysteresis" is better.

Good luck - I'd be interested in any conclusions you reach.

[vlaiv]

Thanks all for help and suggestions!

I almost completely solved the problem

I say almost completely since now I get <1" RMS in RA (range of 0.89" - 0.99") and <0.45" RMS for DEC.

DEC is now due to seeing, no doubt about it, since it changes with exposure length (1s - 4s tested). With longer exposures DEC RMS drops to <0.3".

With RA, the main problem now is of periodic nature with very short period (order of 10s ) - I suspect that it is stepper motor gear mesh period of 13.6s. I did a belt mod and now just have to figure out how tune it down/out. I get about 2" - 2.5" p2p in single period if not guiding and above mentioned RMS values with around +/- 1" peaks if guiding.

And yes here is the main culprit of my ordeal:

One of the RA worm shaft bearings was cracked, and in general there were a couple of bearings that needed replacement. I think this was the cause of vibrations - some of which were transferred to DEC axis as well.

[Tommohawk]

7 hours ago, vlaiv said:

Thanks all for help and suggestions!

I almost completely solved the problem

I say almost completely since now I get <1" RMS in RA (range of 0.89" - 0.99") and <0.45" RMS for DEC.

DEC is now due to seeing, no doubt about it, since it changes with exposure length (1s - 4s tested). With longer exposures DEC RMS drops to <0.3".

With RA, the main problem now is of periodic nature with very short period (order of 10s ) - I suspect that it is stepper motor gear mesh period of 13.6s. I did a belt mod and now just have to figure out how tune it down/out. I get about 2" - 2.5" p2p in single period if not guiding and above mentioned RMS values with around +/- 1" peaks if guiding.

And yes here is the main culprit of my ordeal:

 

One of the RA worm shaft bearings was cracked, and in general there were a couple of bearings that needed replacement. I think this was the cause of vibrations - some of which were transferred to DEC axis as well.

Well done - sound like youre sorted! And interesting to hear the effect of longer exposures and seeing. I get more probs nearer celestial equator - I had put this down to the greater mount rotation at that level, but now I wonder if its partly due to worse seeing at lower altitude.

 Re periodic error, I did a belt conversion and TBH it didnt make much difference to extent of PE. Much smoother though and probably makes for better control of errors. There seems to be mixed reviews re PEC - the main issue seems to be if you  lose your calibration position you have to re-do it all. I think if  PE isnt too horrendous the guiding should sort it - but thats just my impression, others may differ.

In my case most of the problems are in DEC so obviously PEC wont help that, although I guess it would minimise any transfer from RA to DEC.