Active Optics vs Guiding - an experiment

[RickS]

I recently purchased a Starlight Xpress AO-LF active optics unit and thought it would be interesting to see how the results of using it compared to auto-guiding. So, I took advantage of clear(ish) brightly moonlit skies last night to do an experiment...

Equipment:


Mount: Astro-Physics AP900
Scope: GSO RC10, FL 1984mm (according to Pinpoint plate solve)
Imaging train: Atlas focuser, SX AO-LF, SX Filterwheel, OAG/Lodestar, SXVR-H18 camera (0.56 arcsec/pixel image scale)

The scope was in an exposed position and there were strong gusty winds, so I was unable to make long guided exposures without wind gusts ruining every one of the subs. I ended up using 60 second exposures.

The experiment: I ran a series of 20 auto-guided and 20 AO-guided Luminance subs, each 60 seconds long. I actually interleaved groups of 5 subs alternately auto-guiding and AO-guiding to try to keep conditions approximately the same for both in case variations in temperature or seeing affected the results. Auto-guiding was done using 5 seconds subs and AO-guiding was done at a 4Hz rate (approximately).

The results:


the big difference was how well wind gusts were handled. Of the 20 auto-guided subs, 4 were badly ruined by wind. Only one of the 20 AO-guided subs was noticeably affected (in that case the wind gust was strong enough to knock the guide star out of the tracking box completely).
AO-guiding produced consistently rounder stars. I threw out the 4 bad auto-guiding subs and the 1 bad AO-guiding sub and then measured the remaining ones with CCDInspector. The average aspect ratio for the auto-guided subs was 13% (standard deviation of ~4.6) compared to 6.5% (standard deviation ~1.72) for the AO-guided subs (lower percentages are better). After integrating the auto-guided and AO-guided subs separately I also did a DynamicPSF analysis in PI which gave a roundness of 0.926 for the auto-guided image and 0.952 for the AO-guided image.
Auto-guiding produced stars with a FWHM a few percent smaller than AO-guiding but the variation in FWHM between subs was bigger with auto-guiding. Once again considering the good quality subs only, CCDInspector gave an average FWHM of 2.22 arcsec/pixel (standard deviation ~0.115) for auto-guiding compared to 2.34 (standard deviation ~0.084) for AO-guiding. The FWHM for the integrated images was 2.29 for auto-guiding and 2.37 for AO-guiding which is a little closer.
visual inspection of the integrated subs showed that the AO-guided image contains some very dim stars that the auto-guided image does not, but the effect was very subtle.

I don't understand why auto-guiding produced better FWHMs than AO-guiding. It certainly wasn't what I expected. I will probably do some further experiments on a still night when I can make much longer guided subs to see if I get the same result.

I was using a very good mount and had good polar alignment. I expect that AO-guiding would show greater benefits on a poorer mount!

I did cheat a little bit and didn't remove the AO unit from the imaging train to do the auto-guided subs - it was still present but not operational. The extra optical element may have had some effect on the auto-guided subs, but I don't think it would have been significant.

Overall, I am very pleased with the performance of the AO unit. Last night after running these tests I was able to take 15 minute Ha subs despite the wind. This would have been completely impossible with auto-guiding. If only the AO unit was also able to deal with the clouds that appeared around 11pm...

Cheers,
Rick.

[RobF]

Fascinating experiment Rick. Thanks for writing up so well.
Any chance of gawking on some example pics? :D

[RickS]

Thanks, Rob. I'll put a couple of images up tomorrow. As I said, the differences are pretty subtle. Hopefully, they would be more noticeable with longer exposures and more total exposure time... or a crappier mount :lol:

[Mighty_oz]

For that u'll need the laser addon unit Z1A2P-EM, i'm sure that will be developed soon :)
Interesting read, thanks for that.

Marcus.

[Poita]

Easily fixed, I'll swap you for my EQ6 :P

[gregbradley]

Great writeup Rick.

There is a lack of comparison data on the net between AO and non AO performance.

Thanks.

Greg.

[RickS]

Thanks, Marcus. I'm first on the list for the laser addon!



No thanks, Peter :lol:



Thanks, Greg. Yes, it's hard to find info on AO performance at all apart from the odd anecdotal comment. There don't seem to be that many people using them, although there are a few producing excellent results like Leonardo Orazi.

I asked about the increase in FWHM on the Starlight Xpress group and got a response that suggested that it was due to chasing poor seeing. That's certainly plausible, but I wanted to be correcting at a fast rate to deal with the wind. I will have to play with the parameters and learn how best to make use of the AO unit. It may require different strategies for different conditions. No surprise there!

Cheers,
Rick.

[cfranks]

Excellent write up Rick and very timely as I am about to start the process of learning to use my AO-LF. I assume your AO-LF has the Lodestar attached whereas my Lodestar is attached to a QSI 683wsg and the AO doesn't have the Guider attachment.
What is your Imaging/Guiding software?

Charles

[Moon]

Rick,

Very interesting - thanks for sharing.


The sign of a good experiment!

James

[Paul Haese]

I looked around the net some time ago for AO write ups and found it to be a bit thin too. I know that Martin Pugh was using one on his gear and the results were very good. I have considered getting one for the RC at some point myself and will be watching this thread keenly for anything you might come up with Ric.

[RickS]

Charles,

I'm using the OAG which is part of the filterwheel/camera assembly rather than the one that came with the AO-LF. I think your configuration should work fine. I'm using MaximDL for guiding and image acquisition.

If you have problems communicating with the AO-LF over the serial connection try plugging in to the guide port instead. On my unit the labels were the wrong way around. I was pulling out my hair for a couple of hours until I tried that as a last resort!

Cheers,
Rick.

[gregbradley]

I asked about the increase in FWHM on the Starlight Xpress group and got a response that suggested that it was due to chasing poor seeing. That's certainly plausible, but I wanted to be correcting at a fast rate to deal with the wind. I will have to play with the parameters and learn how best to make use of the AO unit. It may require different strategies for different conditions. No surprise there!

Cheers,
Rick.[/QUOTE]

That doesn't make sense as the whole point of an AO unit is to overcome the effects of the seeing.

Perhaps it is more how many hertz you can get the thing running at.

For example I can get an ST402ME or a Lodestar or a Starfish autoguider running at .5 second exposure times. But by the time the download is done and the corrections are done it is more like 1 second or a tad more.

So the AO would have to beat that lag. I thought that was the whole theory of how they work. The mirror can shift faster than a correction in the mount motors can be executed.

So the whole game would be to try to achieve the highest hertz (number of times per second) you can and that would mean getting the brightest guide star.

Perhaps the ST402 might be better for it as it is more sensitive and cleaner and therefore able to produce brighter guide stars that can be guided off.

One thing I am not sure of - I assume the hertz means the guide camera takes that many exposures per second? Like 3 hertz means 3 camera exposures per second. That then would requite a fast guide camera and fast download time.

Perhaps an SBIG STi may be better as it is a CCD plus it has a reducer lens built in lowering the F ratio making brigher stars.

Greg.

[RickS]

Here's a couple, Rob. They are small crops of IC2220, very lightly processed (DBE, slight MMT, and stretch). Don't expect too much from fewer than 20 x 1 minute exposures at f/8 ;) The first was AO-guided and the second was auto-guided.

I hope it's not my imagination, but I reckon the loop of nebulosity at 10 o'clock is better defined in the AO version and that most of the dimmer stars look slightly better defined in the AO version too.

Cheers,
Rick.

[RobF]

Thanks Rick. See what you mean about FHWM and fainter stars. Certainly seems to be lots of interest in your AO exploits.

Actually, I'm drooling over your stars in both images....;)

[RickS]

My next experiment is to determine the effect of drool on FWHM :lol:

[Terry B]

Dear Greg
My AO unit can sample up to about 8hz. It does this by only downloading a small box around the guide star- about 20 pixels square. The problem is getting a bright enough star to sample at that rate. I find that even 1/2sec exposures give very good results and this gives many more choices for guide stars


That doesn't make sense as the whole point of an AO unit is to overcome the effects of the seeing.

Perhaps it is more how many hertz you can get the thing running at.

For example I can get an ST402ME or a Lodestar or a Starfish autoguider running at .5 second exposure times. But by the time the download is done and the corrections are done it is more like 1 second or a tad more.

So the AO would have to beat that lag. I thought that was the whole theory of how they work. The mirror can shift faster than a correction in the mount motors can be executed.

So the whole game would be to try to achieve the highest hertz (number of times per second) you can and that would mean getting the brightest guide star.

Perhaps the ST402 might be better for it as it is more sensitive and cleaner and therefore able to produce brighter guide stars that can be guided off.

One thing I am not sure of - I assume the hertz means the guide camera takes that many exposures per second? Like 3 hertz means 3 camera exposures per second. That then would requite a fast guide camera and fast download time.

Perhaps an SBIG STi may be better as it is a CCD plus it has a reducer lens built in lowering the F ratio making brigher stars.

Greg.[/QUOTE]

[RickS]

Terry,

My opinion is that an active optical system like the AO-LF (which is much simpler than observatory grade adaptive optical systems) probably corrects for polar alignment errors, mount periodic error and other physical stuff like wind pretty well. I suspect they can help to some extent with seeing, but only a limited amount. That's for two reasons: seeing can almost certainly change faster than the system can react at times and also seeing is unlikely to be consistent over the whole FOV so a simple tilt mechanism can't correct over the whole FOV (adaptive optical systems that I've read about can actually make localized changes to the shape of the primary mirror to make different corrections in different regions of the FOV).

And yes, the "Hertz" is the frequency of guide exposures and tilt adjustments per second.

Cheers,
Rick.

[troypiggo]

Dumb question - how can you have 8 cycles per second with 0.5s guide exposures plus download tmes?

[RickS]

You can't. With 0.5s guide exposures you'd be getting a little under 2Hz on my setup. In my tests I had 0.25s guide exposures producing an AO guide frequency around 3.8Hz.

[Terry B]

Sorry my previous post was a little garbled. I only wrote the top bit. My fruit based not so smart phone garbled the quote.
I find that the AO corrects tracking problems well but won't help much when it is windy enough to rock the scope.

[Peter Ward]

Only picked up on this thread recently... hopefully this may give some more insight to the topic.

Amateur AO systems (and they are *adaptive* as they correct the overall wavefront tilt ) are not all equal. They vary in:

1) maximum tilt
2) slew rate
3) optical quality
4) positioning accuracy
5) tracking accuracy
6) jitter (internal noise)

I have used both mirror and refractive based systems, sadly due to calibration difficulties, mirror based systems were discontinued. More the pity as they were *fast* , as with a bright guide star they could easily guide at 30Hz.

The practical limit on AO systems nowdays seems to be about 15Hz

SBIG's implementation is quite strong on points 4 & 5 above as the system is looking at a patch of sky *very* close the what the imaging sensor sees.

Rick, your contradictory FWHM's may not be a problem. Use of AO will often give much taller stellar profiles, albeit with a slightly wider sigma spread at the base.

That said, when the seeing is poor, tip-tilt AO systems don't help much, as wavefront errors easily swamp any positional error. When the seeing is good however, AO can and does bring out very faint stars and structures
that will otherwise get lost in the noise.

Just my 2 cents worth.....

[cfranks]

Well, my AO-LF test wasn't too successful as I discovered one of the internal motors wasn't working. Terry from Starlight-Xpress has offered to send me a motor to see if that will fix it. Can't but praise their customer support!

Charles

[RickS]

Thanks for the comments, Peter. I've been intending to do some more testing but there was a severe shortage of clear skies and now I'm away on holidays for a while (I'm sending this post from a warm hotel room in a very chilly Oslo!)

Sorry to hear of your problem, Charles. Hope it gets resolved quickly.

Cheers,
Rick.

[DavidTrap]

Hope you see the Northern Lights - saw an article today saying "Biggest solar storm since 2005 pummels Earth"

DT

[RickS]

Me too, thanks David! We're heading up north to Alta tomorrow. Looks like we just missed a spectacular display, but with luck there will be more over the next couple of weeks while we're here.

Cheers,
Rick.

[Logieberra]

Rick, sorry to dig this up from early 2012- but did you persist with the SX OAG? What were your conclusions? Cheers

[RickS]

Hi Logan,

I moved onto big sensors and haven't used the AO with them because I was expecting vignetting and issues with flexure. I probably would have persisted if my mount was less capable. I think AO is of most benefit for improving the performance of a weaker mount. The AP900 might benefit a little but it's already doing a pretty good job...

Cheers,
Rick.

[gregbradley]

Good news Logan. I have just gotten an SX AO unit and will be implementing into a Proline 16803 soon. I'll let you know how it goes.
It may be a week or two.

Greg.

[Eden]

Hi Greg,

I'll be interested to see your findings. Do you have the large format or standard version?

[alpal]

Simpler than Adaptive optics?


I have a good idea which is so simple yet I don't see anything available.

A shutter on the camera set up to close whenever the guide graph goes over a certain preset limit e.g.
more than half a pixel or division.
I wish PHD guiding would have such a feature to control a shutter.

Sometimes when I am guiding a gust of wind might hit my mount -
if I could close the shutter automatically I wouldn't have to
discard the sub frame - just open the shutter when the mount had stabilised.

Is there any system like that?

cheers
Allan

[PRejto]

Allan, that is certainly an interesting idea. I can think of 2 issues straight away.

1. When the shutter normally closes the image downloads. So, to use the camera shutter one would need to do a modification of the camera electronics. Probably the easiest way would be to have a separate shutter somewhere in the image train.

2. The total light time would vary, though perhaps only by a matter of seconds. I guess darks/flats would work OK, but the signal would vary...perhaps like thin clouds reduce values in some subs.

Peter

[alpal]

Yes Peter,

1) Yes - the shutter in the camera could be modified - to have an external control or another shutter could be fitted.

2) If you wanted say a 300 second exposure & there was a bit of wind around you could set a 400 second exposure but download it when only 300 total seconds had expired.
That would give you up top to 100 seconds where the shutter could be closed.

It would be nice if it was all automatic - that's where the software writers like PHD could help.

Why isn't this done already?
I think it's such a good idea -
you could also have the shutter closed whenever the guide graph passed a pre-set limit.

cheers
Allan

[Logieberra]

Thanks Ric. Your initial write up was very good. If you revisit, please keep us posted.

[Logieberra]

Thanks Greg. Look forward to it.

[Peter.M]

While a good idea in theory, its application would be limited I think. If you are imaging broadband data, say luminance, by the time your guider has picked up that the star has exceeded the tolerance the frame is already smeared. If you were aiming to get the sharpest data, the aim would be to not have the star go outside your tolerance in the first place.

[alpal]

Of course that would be better.
It's a question involving timing - however -

if I look at a guide graph I can see a sharp excursion starting from the normal graph
& if I had a shutter button I could close the shutter & wait till
the excursion had reached it's peak & started heading towards zero before opening it again.
Surely the image would be better?

We're looking here for any improvement without spending megabucks.

To get good results in a closed loop system you need what's called a " nested loop. "
One loop could be at 1 second every second e.g the guide camera &
another much faster at 10 times a second such as adaptive optics.
Yes you'll get better results that way as you have a nested loop.
The loop at 1/10th of a second refines the first loop to increase the accuracy.

cheers
Allan

[ericwbenson]

Hi Allan,

I swear I had this conversation in an astro group a decade ago...I just can't find the email trail or yahoo group thread. As I recall the answer was it might help a bit with resolution, however the many artefacts it could be produce might nullify the gain.

A few obstacles I can think of off the top of my head:
-funny diffraction spikes caused by the shutter
-changes to the flat field i.e. star/bgd in middle gets more light than near the edges, unless using an sbig type sweeping shutter, which is btw usually slower than the iris types, which leads to the next point:
-in highly variable seeing your shutter open/close overhead time will seriously eat into your duty cycle.
-dark current accumulation (and the associated noise). If your effective duty cycle ends up being 50%, you dark current wrt the light signal just doubled.

Best,
EB

[alpal]

Thanks EB,
There are bound to be a few problems but I think the overall result
would be an improvement.

If your total exposure time was identical the lights would be perfect.
However the dark frames would be not matching perfectly as they
would be a little longer than a frame where no shutter was used.
It would be better to have a proper Iris shutter than the sweeping type found in the QHY9m.

I wrote to Craig Stark & he suggests I write to this forum here:
http://openphdguiding.org/

He thinks that there is some merit to it & it's actually an old idea that was never implemented as most people would just get a better mount.

I have yet to write to that site but I will soon.

cheers
Allan.

[ericwbenson]

Not quite, the lights and darks could still match in length (e.g. 10 min total exposure) i.e. the dark signal accumulation would be 10 min for the dark and 10min for the light frame (dark signal accumulates regardless of shutter state), but the object signal accumulation is only for 5 min (50% duty cycle on shutter let's say), so the relative size of the dark signal relative to the object signal is twice as big, that's all. This is no problem for bright objects, where you have lots of signal to throw away, but for faint small stuff, not so much.

BTW this is the same strategy as lucky imaging with zero read noise cameras... throw away all the bad frames and stack only the good ones, but if your duty cycle is low because your seeing criteria is tight then it takes too long to record the faint details.

EB

[alpal]

No - I am thinking that the software would be smart enough that it
would give you say a 10 minute subframe even if the shutter was closed for say 3 minutes.

In that case the dark information would be 13 minutes long & the light subframe 10 minutes.
The light frames would therefore not match perfectly to the dark exposure
meaning that a 10 minute dark frame that you later took to cancel out the 13 minute frame would not match.

That's the problem.