PHD tip - set Min Pixels = 2.5 arc seconds

[g__day (Matthew)]

Playing round with this tonight - taking 10 minute shots of M16 and looking at how well the faints double stars are resolved - I confirmed something the Jim McMillian's notes on guiding posited; chasing seeing errors swamps all other sorts of guide errors (especially at longer focal lengths).

PHD defaults to about 0.15 pixel movement before a guide pulse is sent. Now 2*2 binned at full focal length on a C9.25 works out to be each pixel on my guide camera sees 2.2 arc seconds of sky. So I have to set pixel movement before I guide pulse is sent to 1.15 pixels.


1. Drift is likely < 0.10 arc seconds / second
2. PE is likely < 0.067 arc seconds / second
3. Seeing < 2.0 arc seconds / second

So Jim walked through most guide errors are likely to be chasing the seeing guide errors – typically indicated by periodic oscillations on your guiding graphs. Assuming your guide star is bright enough, and that you take shots between 2-3 seconds to optimise Signal / Noise in guiding shots, Jim had the following recommendations:

1. Set RA Aggressiveness relatively low to significantly reduce errors
2. Set RA Maximum move as low as possible
3. Set minimum move high enough to ignore corrections smaller than seeing corrections

So this implies many corrections are chasing the seeing. I typically set minimum drift before a pulse is issued to 0.2 pixels (but at my imaging 2.2 pixels an arc second – Jim would be advising my setting to be more like 0.8 – 0.9 pixels).

So I tired tonight varing minimum pixels from 0.6 up to 1.2 pixels - the absolutle sweet spot was 1.15 pixels - stars were beautifully round and resolved.

Too I set RA Aggressiveness to 50%, RA Hystersis to 25%, Max Dec duration to 80 msecs, and took 3 second guide shots (on a Meade DSI II Pro mono). The OSC index crept down as the min pixels before a guide is issued raised, but the RMS (v1.10.8) moves randomly between 0.80 - 0.90.

A give way should have been when you first select a star - prior to calibration - if it jumps alot - that's seeing - and you shouldn't try and guide that out!


Matthew

PS

Watching a star for 30 secs - without guiding - leet me see this much stare wobble (atomspherics)

My data from just watching a star (not guiding)

Star wobble - not guiding

Time X Y
3 309.3 287.5
6 309.4 286.8
9 311.8 286.3
12 309.2 286.4
15 310.8 285.9
18 309.7 285.7
21 309.3 285.5
24 311.2 285.8
27 309 286.3
30 308.4 285.7

Mean 309.8 286.2
Std Dev 1.03 0.58



That Std Dev should give me a hint to my pixel size before guided pulses are sent. The default of 0.15 pixels on my rig - chasing that seeing - would kill good guiding!

[gbeal]

Matthew,
darn this is good, I need help with mine. Some nights it can be right on the money, others well not so.
There is a "how to" somewhere?
Thanks again for the tips.
Gary

[g__day (Matthew)]

Gary,

There is basic information in the help from memory, but I think you need to spend some time understanding what it is doing. I think Craig Stark images at short focal lengths and has vey tight stars in his guide scope / guide CCD combo - judging from this image of his results he posted

http://files.getdropbox.com/u/521680/PHD1108atest.jpg

I know that using a single guide star and judging by size and frequency of corrections - PHD is at least as good as MaximDL 4.56. I am unsure how it compares to MaximDL whenyou have multiple guide stars selected (this may even out seeing errors).

Craig has such low RMS 0.2 compared to mine 0.8 I ponder is this sky quality (seeing) and/or focal length. Unless we highlight the challenges of longer focal length imaging I ponder if he mainly tunes and defaults PHD settings to optimise short focal lenght performance.

PHD can work very well - you just need to experiment some to get its parameters right. I do think sitting with Excel and charting the performance - plus by the human eye ranking the tightness of your stars is the best, most systematic approach to tuning PHD.

[Dennis]

Here are some graphs showing the periodic error in my mount. I used K3CCDTools to record the PE over a period of 1500 seconds (approx 25 minutes) on 4 consecutive runs on the same evening, with the same set up, at a focal length of 2160mm. I think the worm period on the EM200 is 480 seconds?

The runs sampled the star position at 1, 2, 3 and 5 second intervals and these are the results from importing the K3 logs into Excel, RA data only. You can easily see the effects of seeing.

The graphs are “polluted” by the effects of declination drift due to less than perfect polar alignment.

Cheers

Dennis

[g__day (Matthew)]

I think those graphs illustrate a very clear message - try guiding out seeing and you will jitter around a star. For jitter I imagine only adaptive optics might help. Elsewise - don't set the sensitivity of your guiding below that of your seeing!

[JohnH]

Matthew - thanks for that - I must admit I have been non-plussed by the mysteries of autoguiding more often than I care to think about - sometimes it works sometimes not and here is a qantative way to tune your guide paramters to the seeing - fantastic.

I have been thinking (until now) that I could compensate for bad seeing by increasing the expusure time (ie the guide image is then the average of the star posn over n secs - typically I used 2-5 secs). BUT this approach is better, I need to do more tests but I have already seen a big improvement in DEC stability and think I can get there on RA as well. Attached is a trace from last night of my guide errors from Guidemaster, I set the min error to 1" (I had previously used 0.5 or less), exposures were 1s and max guide pulse was 500ms at 0.5x.

I have plotted the correction signals as well as the errors - I am pretty happy with DEC but RA is still a little unstable leading to slightly out of round stars (m8 test shot attached shows this). I am not sure how far I can go with this I would like to get to +/-1" in RA as my imaging scale is 1.75"/pixel and such a small error would therefore be close to invisible. I might need a little more offset in RA balance to calm down the jitter I will try next that if the weather permits.

[Dennis]

Hi John

Nice graph and excellent M8! At long focal lengths (1500mm and above) I made significant inroads into guide corrections when I adjusted my guide correction rate down, from the default of 0.9x Sidereal to 0.5x Sidereal.

So, when an RA correction is required there is less tendency to over shoot then reverse the overshoot and so on, leading to a yo-yo situation.

Cheers

Dennis

PS – what mount do you have?

[JohnH]

Dennis,

Sorry I should have said - mount is a G11, image scale is 1.4" / pixel (127mm F7.5 ED with Opticstar 145m CCD), exposure was 4 mins (1 full worm roatation) x 20 with Ha filter, flats darks and levels in AA4.

I estimate my RA error has resulted in stars being 4x5 pixels instead of 4x4. I suppoose there must be limit here somewhere as DEC error can be 0 (or the seeing limit) if perfectly aligned but RA will always be bigger as tracking can never be perfect...I think I must reduce my image scale a little.

John

[g__day (Matthew)]

I ponder now - sometimes I change the binning on my guider (if the stars seem tight I reduce the binning), which changes the arc seconds per pixels. During calibration I probably should adjust the minimum motion per pixel to correspond to the new seeing conditions - and possibly even recalibrate when I change binning. Must explore this!

[JohnH]

Some more tweaks and twiddles to the GM settings yielded the attached tracking log...I am happy with it but wonder if there is room for more improvement. The offest in DEC is deliberate - I let the error build to 1" and correct from there - it prevents crossing the 0 line and introducing to possibility of backlash.

The amount of drift in this run was high as my polar alignment was off (I used Alignmaster instead of the drift method and was not greatly impressed) so I think I can get a better result with less drift - my images showed round near the guide star but also significant field rotation see attached m8 shot...very rough processing I am afraid - mismatched darks and flats - cloudy night etc but the purpose was only to gauge tracking - the guide star was one of the embeded NGC6530 cluster members...

[JohnH]

After about a week or so of trial and error I think I am getting a good result. My findings have been:

1. Gemini PEC does not yield significant improvement without adjustments which are hard to make with just the controller - curve smoothing, drift compensation and phase adjustments may make it work but it is not easy.

2. PEMpro or equivalent software lets you see what is going on and therfore give far better contol of PE - in theory however my results have be inconclusive as to the merits of PEC on the G11.

3. Guide parameters and alignment are FAR more important to get right than PEC. For a mid range mount like the G11 corrections should be small and a dead band should be set to prevent chasing the seeing. How you do this varies dependanty on the software you use but in may case I find exposures of 0.5 to 1s, slews limited to 250ms max at 0.5x and a min move of 0.5" for RA and 1" for DEC works well. Do not think you can compensate for bad seeing with longer exposures - I did and it does not work, reduce the exposure/gain to get a good guidestar (point) that is not saturated. Measure your PE and make sure the sware cannot send corrections bigger than the max PE possible in the guide/exposure cycle. Finally adjust aggressiveness (if you have to) so corrections do not cause overshoot.

Now having said all that my result in not perfect - I seem to have round stars only in a pert of the image - I think the remaing distortions are as a result of a combination of field curvature combined with lack of rigidity in the focuser (127mm ED scope with the stock focuser) but I would value other views. Basically I am happy if I am beginning to see other sources of error than the mount.

[Dennis]

Another nice and very useful write up John, not to mention those lovely M8 images you are posting as well – great work!

I’m finally getting round stars across the whole field, albeit on the small 765x510 chip of the SBIG ST7 CCD camera, with my 180mm Mewlon at F9.6 (fl of 1760mm) for 10 x 15 minute subs, with no field rotation. EM200 mount.

It’s taken me a while to get there, as my previous investigations were few and far between so I could never really consolidate any findings and apply them to the next session to see if they had worked. That is, had the changes I made to cable runs, balance, parameter changes, etc produced a permanent and reliable fix to a particular problem.

The recent spell of good weather has allowed me to string several nights together where I have been able to verify that certain changes were effective and persistent in my self guided configuration. I am yet to investigate my side-by-side issues!

Here is an example; the “new” PN in Sgr, 10 subs at 15 mins exposure each, self guided with the ST7.

Cheers

Dennis