Autoguiding - tuning my setup

[JohnH]

OK, so I have got my system almost doing what I want...attached is a crop of a 10 min test run on the Keyhole region. My stars are not round but little triangles (no its not a collimation problem). I use PHD to guide and the corrections were applied on both axes with DEC in one direction only. Exposure was 3 secs and I set the erorr window to .05. The error log shows corrections only being applied in one direction and typically taking 2 small pulses to re-centre the star.

So I think my problem is lack of sensitivity to the errors but I am not sure I can go below .05 (ie guiding to less than on twentieth of a pixel?) so I need to barlow?

My Guide scope is a WO 66ED - 388mm fl with an Orion Starshoot on the back (3.43" per pixel), imaging is the VC200L - 1800mm fl with the 20D on the back (0.74" per pixel).

Using those numbers I thought I would be ok without a barlow???

[Astroman]

is PHD correcting both axis at the same time? so it can correct Dec aswell as RA at the same time, might be that PHD corrects for one but misses a command to correct for the other so it catches up by then the star has moved too much in RA or Dec? Only a guess mind you, I know of some software that only being issued with one command at a time. I havent actually used Autoguiding on a mount yet, always used Pulse guiding.

[JohnH]

Yes PHD can guide on both axes at the same time, yes it uses pulse guiding via ASCOM and the TCPIP to the Starbook...

[[1ponders]]

Is that crop from the center of the image or more towards a corner?

[JohnH]

Paul, pretty close to middle - the stars do not vary in shape though no mater where I look on the image. Attached a 50% reduced full frame so you can see the issue...

[[1ponders]]

Ok IC. I had similar triangular stars in a series of images, but it boiled down to a Focal Reducer issue. Using the Meade 6.3 on an Orion ED80, the stars out near the corners all looked like little triangles.

In your secondary, do you have screws to hold the mirror to the secondary mount and screws to adjust the collimation or just collimation screws?

[[1ponders]]

It almost looks like mirror pinch. The VC200L has a fixed mirror doesn't it?

[JohnH]

But I am sure this is a guiding problem, here is a short, unguided, exposure (30s) - no hint of triangular stars - just a little tailing due to PEC/poor polar alignment....and yes the VC200L mirror is fixed.

[JohnG]

Have you tried exposures of 1 to 1.5 seconds, your system might be overshooting slightly with the 3 second exposure.

Cheers

JohnG

[Dennis]

Hi John

Chasing down auto guiding problems can be real fun eh – not!

I once solved one of my problems by turning down the aggressiveness of the guiding correction. I think this reduced the magnitude of the full correction to say, 50%, as I was finding that the correction was too aggressive and overshooting, so the next one would then overcompensate in the opposite direction and so on, leading to a oscillation effect.

This was in CCDSoft which has an aggressiveness setting option – I’m not sure if PHD does?

Good luck – hope you get it sorted.

Cheers

Dennis

[JohnH]

Dropped the exposure time to 2s, set aggressiveness up to 100% (was at 70% previously). This is a crop of a single 600s exposure at full scale - calibration and levels only. I think this is close to the limit for my setup, pixels are .7" so seeing would make a star blur out over a 3x3 area correct?

Zooming in on a faiter star seems to show perfomance is close to that...

[[1ponders]]

Looking much better John. :thumbsup:

The brighter stars still look just a little bit blobby (egg shaped) at about the 6:30 position on the stars. Maybe try a deconvolution on it at see how it turns out.

[JohnH]

OK, here is a processed stack of 8*10 mins of the same region, I think the stars are round enough...

[iceman]

Lovely!

[[1ponders]]

Nicely done John. :clap:

[EzyStyles]

i had this problem with my 8" F/4 that all the stars are in triangular shape. Came to the conclusion that my collimation wasn't correct and that my EQ6 mount tripod legs were not tight enough. great image btw john.

[JohnH]

Thanks guys!

I knew my gear was better than it's operator and the seeing/lp in my back yard - but I am very :D that I can now extract close to the best out of the conditions I have to work with, well, some of the time anyway.

Autoguiding is a black art indeed, not helped by no PEC on the Sphinx...

[JohnH]

What I am seeing is mostly very small errors/corrections. The attached plots show dx/dy in pixels at a scale of about 3.7" per pixel with an exposure time of 3s.

But are these large spikes normal - would it be seeing related? They do not appear to be periodic, also if I review the log files I see the jump is reported in position and then corrected - normally in a single cycle....

PHD Guide 1.5 pre2 -- 03/14/0722:50:12
Guiding begun
lock 476.0 328.0, star 476.0 328.0
RA angle 1.5, rate 0.0019 aggr 1.0
DEC angle -0.1, rate 0.0035, Dec mode 1
Frame,Time,dx,dy,Theta,RADuration,R ADistance,DECDuration,DECDistance
1,3.324,-0.02,0.01,2.7,0.0,0.0,0.0,0.02
2,6.809,-0.02,0.01,2.6,0.0,0.0,0.0,0.02
3,10.304,-0.03,0.01,2.7,0.0,0.0,0.0,0.03
4,13.789,-0.02,0.02,2.5,0.0,0.0,0.0,0.03
5,17.284,-0.03,0.03,2.4,0.0,0.0,0.0,0.03
6,20.769,-0.03,1.00,1.6,480.6,-0.90,44.4,0.15
7,24.785,-0.03,0.00,3.1,48.0,-0.09,0.0,0.03
8,28.320,-1.00,0.00,3.1,0.0,0.0,0.0,0.99
9,31.805,-0.03,0.01,2.9,0.0,0.0,0.0,0.03
10,35.300,-0.03,0.01,2.7,0.0,0.0,0.0,0.03

I had thought to put my dead band higher - but I cannot set it to 1 pixel and anything less would not alter this behaviour - is it real, is my seeing likely to move a centroid by 4" for 3s or more or is this spurious. Can it be wind loading - I do not have an observatory by my location is pretty sheltered...

If this is not seeing related where to look? It seems odd to me that it is always ~1 pixel and impacts both axes by the same amount - it is almost like a noise issue -perhaps I should see if it is still present if I run all gear off batteries. Perhaps I should consider chokes on the cables going to the mount/cameras?

The impact of the spikes is worse on shorter subframes as the excursions form a larger %age of the exposure time thus my 10 min subs have rounder stars than my 2min subs....attached is a 300% zoom of one of my Canon images at 300s - the stars are distorted in x and y...

Odd/confused (normal state of mind for me!)

[JohnH]

Thought I would share the journey with you. My latest experiments with guide parameters - faster slews, shorter slews, backlash adjustments etc etc have not yielded any significant improvement in guiding accuracy. I still get egg shaped stars....I did a run last night with the guiding outputs turned off.

The result is attached. As you will readily see I did not bother with accurate polar alignment for this run - so there is significant DEC drift.

I ran the mount off a battery and did not use TEC on the guidecam, I only had the guideport on the mount connected to the PC and only one USB connection - all to eliminate PC load and electical noise.

There are two things worrying me here. The DEC motor was off and yet there are step changes in the DEC trace and the RA trace shows jumps also. Exposure was 2s.

Is this slippage somewhere (clutches) or is it backlash. DO I need to unbalance in DEC as you do in RA to keep the gears loaded?

Your advice/input would be appreciated - I am runnig short of things to tweak...

[richardo]

Hi John,
I owned a VC200L for over ten years, 4-5 of those years I used it for imaging with a webcam and HX516 the latter working at around .85 arcsec/pix. with a fOV of 7x9.3.. like looking down a drain pipe you might say! Super critical. The HX916 when I got it was a little more forgiving.

First of all, what sort of mount are you using?? Good gearing and worm with little PE is essential.
Next you have to be pretty well spot with your polar alignment. Drift alignment is the only way.... guiding at these fL's is super critical for success. Many people think because their auto guiding, rough enough will be good enough... wrong!
Software is also an issue.... I've used Astroart and Maxim because their guide sub exposures are at a sub pixel level and it is what's needed at these critical levels imo..... not like the freebees available on the net, you don't get all the program flexibility where compensating for your mount type and drive speeds are available, heaps of other stuff as well etc. I tried many availably at the time when I was working through these issues.
Unfortunately, this hobby's demands are pretty heavy for good achievement with long focal lengths, unfortunately spending money is a big part of it.
If your mount can do consistent unguided exposures out to 2min, then 2 secs on your guided exposures, if only 1min consistent unguided exposures, then 1 sec... etc. With a Canon DSLR, things will be a little less critical than what I was experiencing with my stand alone astro cameras.


Secondly, how are you fixing your guide scope to the mount.
DON'T fix it to the VC200L as it is!!
The tube is quite flimsy. Even if you have a rail from the top of the primary and secondary housing, this will still make the tube flex.... maybe not when you are imaging with the scope perfectly upright (90*), but when it's starting to get on its side, (90*+/-) it will be pulled out of shape and warp.
Also the stock (native) dovetail rail that is used to fix the scope to the mount is very inadequate for such critical imaging purposes. Just look at it and think about it.... grab your scope and put some pressure on the tube, watch it flex... only has to be a little, that's enough to spoil any guiding at this FL.
The solution here is to spend some money on heavy duty tube rings. Forget the tube rings from the Vixen RSS200, they are still too flimsy!
"Parallax" are the best around and will do the job that you need your setup to do!! Because the tube dia on the VC is a non standard size, you'll find Parallax will be able to custom a set for you.... but it will cost you!
Now you'll be able to fix a guide scope to the top of the Parallax tube rings without any fear of differential flexure or flexure/ warping from the VC200.

When I had my GPDX (long gone) all tuned in using PEC, I was able to get out to 1.5- 2min exposures @ F9, I had no dramas with any sort of flexure, triangular stars. But when I went down setting it up for guiding, this is when my hair started to get pulled. I persisted for a couple of years
So all of the above was my findings with this scope.
Good results can be achieved however, you'll just need to remove all flexure!!
Here's a link to one of the best VC200L users around, check out how he attached his guide scope.... that's why he has great success, rigid tube rings. Look at his images..
http://members.inode.at/wasshuber/equipment/equipment.htm

Hope this is all some good food for thought, believe me, I've been down this track that you are venturing upon!
Here's a couple of so so images I took with my VC... still had egg shape stars though!
http://www.baytop-observatory.com/ccdimages/galaxies/m83f9.htm
http://www.baytop-observatory.com/ccdimages/galaxies/ngc5128f9.htm

Hope you have some success.
Cheers
Rich

[JohnH]

SO I could do another test last night.

It has been suggested to me the steps I am seeing in the error logs could be due to poor centroid calculations in the software. To test for this I set up my mount (it is a Vixen Sphinx btw Rich) with the power off except for the camera.

I just logged the star position as it drifted over the CCD using the PHD software. The results are attached. As far as I can see from these plots there is NO sub pixel position data calculated.

Also it seems that the speed of the camera is an issue - I had exposures set to 0.2s (to collect as much data as possible) but the camera returned only 1 exposure every half sec approx, I assume this is due to the limits of USB2 and a full colour 16 bit image (752x582).

The big spikes correspond to moments when the software reported NO STAR, the on screen image seemed excessivly noisy for those few frames, the wild spikes at the end of the run are where the star left the guide box and when it does that the log shows 0 as the position.

My conclusion from this is rather depressing - this camera/software combination is not suitable for the task I have set it - sub pixel guiding. Unfortunately the OSS is not widely supported in other software and is no longer in manufacture so this is unlikely to improve.

I would love to be wrong of course but I cannot see any other explanation for the steps seen in the trace.

I have tried defocusing and altering the gain/exposure etc to change the behaviour but to no avail.

Next test - revert to the LPI as a guidecam and repeat the experiment, clear skies all.

[JohnH]

First of all, what sort of mount are you using?? Good gearing and worm with little PE is essential.

A Vixen Sphinx - a bit better than the GPD s othey say...

Next you have to be pretty well spot with your polar alignment. Drift alignment is the only way.... guiding at these fL's is super critical for success. Many people think because their auto guiding, rough enough will be good enough... wrong!

Yup - I use WCS to get this close

Software is also an issue.... I've used Astroart and Maxim because their guide sub exposures are at a sub pixel level and it is what's needed at these critical levels imo..... not like the freebees available on the net, you don't get all the program flexibility where compensating for your mount type and drive speeds are available, heaps of other stuff as well etc. I tried many availably at the time when I was working through these issues.

Think you have something here, seem my other post.

Secondly, how are you fixing your guide scope to the mount.
DON'T fix it to the VC200L as it is!!
The tube is quite flimsy. Even if you have a rail from the top of the primary and secondary housing, this will still make the tube flex.... maybe not when you are imaging with the scope perfectly upright (90*), but when it's starting to get on its side, (90*+/-) it will be pulled out of shape and warp.

Hmm, I have an ADM rail connected to the front and rear housings - perhaps not perfect but if I go for rings and plates I will exceed the capacity of the mount. I do not think this is the issue at the moment ($$$ = denial)

Also the stock (native) dovetail rail that is used to fix the scope to the mount is very inadequate for such critical imaging purposes. Just look at it and think about it.... grab your scope and put some pressure on the tube, watch it flex... only has to be a little, that's enough to spoil any guiding at this FL.

I do not think I have got to the flexure issue yet (I am sure there is some it is just I have not got the precision to notice it yet!) - the reason I think that is that I can see the tracking errors in the guide logs, if it was flexure the guide logs would show nothing and I would still have out of round stars...

When I had my GPDX (long gone) all tuned in using PEC, I was able to get out to 1.5- 2min exposures @ F9, I had no dramas with any sort of flexure, triangular stars.

No PEc on the Sphinx means I can only go to 30s without guiding.

Look at his images..
http://members.inode.at/wasshuber/equipment/equipment.htm

Wow - now I have hope!

Hope this is all some good food for thought, believe me, I've been down this track that you are venturing upon!

Here's a couple of so so images I took with my VC... still had egg shape stars though!
http://www.baytop-observatory.com/ccdimages/galaxies/m83f9.htm
http://www.baytop-observatory.com/ccdimages/galaxies/ngc5128f9.htm

Wow again - makes my 5128 look like a joke - so this VC200L ain't bad is it?

Hope you have some success.

Me too! And thanks againn for the help Rich.

[JohnH]

Well I reverted to Maxim last night in light of the advice from Rich and others.

Maxim puts the OSS into 2x2 binning mode effectively turning it into a mono camera with higher sensitivity and lower resolution - but and here is the good bit sub pixel guiding works.

The results were a revelation - well to me anyway - 20x10 min shots with only 2 fails...much rounder stars in all the subs in fact acceptable at 100% res.

I have also established that guiding at .5x siderial gives a better result - which means a move from the ASCOM/TCPIP setup to ST4 and a GPINT...which means moving the shutter cable from parallel to USB which means...more wires,,, (thanks Hap!)

Still as Morris would say I am heading in the right direction!

Lesson - One shot colour cams do not make good guidecams and sub pixel centroiding claims may not be all you hope for with some of the software/camera combos out there, if in doubt do the little test and then you will know - I certainly was not aware that this was my problem and went down many rabbit holes.

[richardo]

That's good news John!
Glad things have worked out fairly simply without too much drama!!
Getting auto guiding up and running successfully with limited funds can be some what tedious with all the things one has to look at.

Cheers
Rich

[xelasnave]

My confidence to be able to auto guide has been absolutely destroyed John:lol: :lol: :lol: .
I have an eq6 hoping to work with a SN 750mm f/l using an lpi to guide:) . via an 80x400mm ar or a 70x600ar

This has been an excellent thread I have learned so much and would like to thank you for starting it off and to those who added great input... you dont make house calls do you;)
I have had the mount since November and are yet to give it a go because of time and the weather not co opperating:sadeyes: .
thanks again
alex

[JohnH]

Autoguiding is a fun challenge, your correspondant is blonde!



Plenty of folks seems to be doing well with the EQ6 - it was my second choice after the Sphinx due to weight/portability/precision. I think 750mm should be a snap for and EQ6. I am pushing the limits at 1800mm.

The LPI is not very sensitive, works well with Guidemaster but you will be retsricted to bright guide stars...~ Mag 6 limit probably.



Good, that was the point - to share my journey - I spent way too long with my mount in neutral balance when it needed to be heavy in the East - I really was unaware of the significance of that. I woried about DEC backlash to much as well when I should have had DEC guiding off and dealt with balance/alignment issues until RA tracking was good. That is hindsight of course when you start out it is difficult to separate the issues - I honestly thought precision alignment was not critical and my balance was good enough...wrong!

I thought switching from the LPI/Guidmaster to the OSS/Maxim/PHD would make life easier as I had a more sensitive camera - the fact it was a one shot colour camera and my lack of understanding of the implications of that led me to a centroiding issue that was completely unknown to me. Ouch what a learning curve.



Hmm, weather, family, equipment failures, jobs all can get in the way, then I spent most of the remaining clear nights trying various combinations of s'ware/camera etc trying to get a reliable setup. Last decent shot I took was in Novemrber too! Clear skies!

[xelasnave]

thanks John :thumbsup: :thumbsup:
I am already more confident as in truth I was going to give up on the 750 approach and go for a good 80mm short f/l triplet but I will stay on the course I have set:thumbsup: .
Thank you very much for taking the time to comment I very much appreciate it:thumbsup: .
The way I will approach the lpi guiding is if I can find a star to do the job I will photograph whatever is nearby.. even ëmpty regions.. and call those deep field.. just like Hubbles;)
alex:) :) :)

[JohnH]

The attached is a 400% zoom in of a cropped single, unprocessed frame (except for jpeging) from a test run on the Jewel Box last night. The exposure was 600 secs at F9 and iso 400 on the Canon 20D.

Here is my analysis of this image:

My guidecam/scope combination has a pixel size of 3.7" approx but as I now have to bin 2x2 my effective pixel is ~7" and tracking errors of <0.2 pixels (1.4") are therefore not detectable/correctable. My stars would therefore be out of round by 2-3 pixels on the Canon Frame (0.75" per pixel approx) at best.

To avoid backlash issue I guide DEC in one sense only (opposite drift) so I will should get excursions of 2-3 pixels in DEC in one sense only and in RA 2-3 in each direction.

Seeing is 2-3" in Sydeny on a good night so a stary will cover a grid of 4x4 at best.

My faint stars (not saturated so no blooming hopefully) are about 8x8 so that seems to be about right being 4x4 plus the tracking errors. The out of roundsness error size does seem to be in line with my calculation and is not symetric due to the unidirectional DEC corrections.

Any comments - seems I must either get a longer fl guide scope or get guiding to work at 1x1 if I am to get any better than this.

Question for guiders out there - is my analysis correct? Is this result in line with what I should get with this equipment? I do not want to kill myself trying to tune out errors that I cannot reasonably expect to get rid of without changing the mount. I would rather reduce my imaging FL - it is cheaper! Is there a way in image processing to make the stars round again?

BTW I have reported all the above to Craig Stark (PHD author) he is having a look into it for me. I must say he is very responsive and friendly so I have hope for an improvement there...

[jase]

PEC for the Sphinx mount should be available towards the end of April. This will greatly assist in the quantity of guiding corrections. Last I heard, it will be released as a firmware update. If the mount is ASCOM compatible, I recommend you use PEMPro (http://www.ccdware.com/products/pempro/) to program your PEC, not just autoguider adjustments. PEMPro will significantly reduce your PE.

IMO, 3.4 arcsecs per pixel is a little high, especially when considering the focal length of the main scope (.74 arcsecs per pixel). Increase the focal length of your guidescope or reduce your pixel size.

There should be no issue with the Sphinx handling the focal length of the VC200L, though you should try to match your cameras pixel size for deep sky objects. Dennis explains this extremely well in the post - http://www.iceinspace.com.au/forum/showthread.php?t=15250

[JohnH]

I can push my gear a little harder then, but need " per pixel on the guidescope, the good news for me is Craig just sent me a fixed up version of PHD so this will allow me to guide without binning I hope. It is cloudy now of course.

BTW If this IS the fix then it was not a centroid calc issue but incorrect offset (too high). Will do more tests and post the reults here asap.

[JohnH]

Another clear night let me test PHD 1.5 RC1a - the sub pixel math now works - unfortunately this is at the cost of stability - this version looses guidestar lock and/or mistakes them for hot pixels even when not saturated or tightly focused and so will only guide for a short period before failing....back to Maxim for now...

[JohnH]

The latest version of PHD worked well - guiding to sub pixel accuracy with the OSS. See the attached log/chart from a run last night.

Unfortunately it was clouding over while this test was done and only the brightest stars showing by the time I was set up, I was not able to do a polar alignment (just carry/plonk) so I do not want to read to much into the results - it should be possible to do much better in better conditions - so the only odd thing I can see in the data is occasional jumps in DEC that look like big overshoots - not sure what would cause those - perhaps my balance was off perhaps it was due to poor alignment. It is still way better than it was.

It is now forecast cloudy for the next few nights so I will not be able to do more tests for a while but I will re-do the mount off test and determin the detectable min errors, then I will also re-test guiding with a good polar alignment. Then perhaps I can do some long fl imaging and stop boring you all with this thread!

[JohnH]

This is (I hope) the penultimate post on this journey. I repeated the drift test with the relased version of PHD 1.5 - and found that the staircase effect was back. To put it another way sub pixel guiding still does not work with this combination, I have shared that information with Craig Stark and he has some ideas but not time to impliment them as yet. So I still have insufficient resolution to guide accurately with PHD. To confirm this I reverted to Guidemaster and the LPI and got good results, then I added a barlow and it got better (fitting a barlow to a WO66 has a thread all of its own...).

Of course this way I am limited to very bright field stars and have a tiny fov - not good really and the reason I bought the OSS in the first place.

The good news is it worked well, encouraged I tried the LPI with PHD and that worked well too both allowing the LPI to be used in long exposure mode. Still it is not sensitive enough.

Having found and confirmed my issue as insufficient resolution in the guide system I faced a choice - new guide scope (more appature and greater FL) or a new camera (bigger chip, more pixels, mono). Having thought about it before I re-examined the Qguider and this time took the plunge and ordered one. I was very happy to see that it will be supported in Guidemaster v2.

So while I am waiting for that to arrive and the moon is close to full I reverted to the OSS on the Barlow and tried defocussing to help PHD with centroid calcs but it did not improve the situation.

I finally reverted to the Maxim EE software which does centroid well - but bins 2x2 - the barlow reverses that in effect giving me 3.43" / pixel but guiding to about 0.7" which is what I wanted in the first place. Of course I have fov issues but the cam is sensitive enough (just) in this config to give me a guidestar most places with a 2-3s exposure. The remaining problem is the horrid Maxin EE software that insists on selecting the guidestar for you - and will of oten choose a hot pixel - but the reults were most encouraging - my focus was off at bit but I did a run of 3 10 min shots with no failures and round stars at full res of the 20D with no FR in place ...did a bunch of 3 and 5 min shots too, all good. Attaached is a tottaly untouched crop of the middle of one of the 10 min shots at 100% res shot in Large/fine iso200 full moon, the bright star is Alp TrA - I got equally good results on the ecliptic too.