What is the best software for long duration auto-guiding?

[g__day (Matthew)]

What’s your preferred Auto-guiding software once you move past 10 minute shots?

I must admit I am struggling with auto-guiding right now, yes it works – but not with the precision I’d like on long duration shots.

Having used PHD for many months now I am about to give MaximDL a try – to see if it can give me better control and diagnostics over what is going on.

I find anything up around 6 minutes at 2.3 metre focal length to be fine (even unguided) – at around 10 minutes stars have a 2-3 mm streak (mostly in RA). At 20 minutes stars have a 4-5 mm (say a 1-2 arc minute streak). The surprising thing I am finding is that guiding the way I am doing it is not significantly improving images over my unguided shots!

My mount is on a permanent pier and aligned within 30 arc seconds of the SCP. I image off a 9.25” Celestron into a Canon 400D. When I visual sighting on a star thru the Canon’s viewfinder – and place it on the central focusing dot – the star stays there for 10 ~ 20 minutes as far as my eye can tell.

I have yet to do a PE correction from my main scope – so far I have being doing PEC using the MAK. Neither have I averaged runs from PEMPro – I am still learning how to get it all to work consistently and getting my image scale correct.

So given I have no or minimal drift in DEC, my challenge is to eliminate drift in RA. Therefore my corrections have to distinguish between seeing errors and tracking errors.

Now I have use a Meade CCD to guide, and have tried first a 480mm focal length 80mm Megrez, then a 127mm MAK with focal length 1500mm. Recently I added a focal reducer to bring the MAK to a focal length of 750mm into the DSI – to give me tighter, brighter stars.

PHD is a blessing (and its free), and Craig Stark is really brilliant at answer questions on his Yahoo group. I am now at a point where I need to enhance the performance I am achieving. I have tried using PHD analyser and reading raw PHD log files to determine what is actually occurring. I presume I am partly chasing seeing and am trying to understand where the guiding error is occurring – is the software too aggressive or too passive in how it adjusts for tracking errors?

Unfortunately there are several interacting variables to tune. Including:

1. RA aggressiveness (0.1 – 2.0 sidereal) on the hand controller (default +0.5 sidereal )
2. RA Aggressiveness 0 – 100% on PHD (DEFAULT (95%)
3. RA Hysteresis 0% - 100% on PHD (default 10%)
4. Minimum motion before pulse guide is sent 0.01 – 2.00 pixels (default 0.2)
5. Frame duration 0.01 – 10 seconds (default 1.5 – 2.5 secs)

I am now going to descend into the very bowels of PHD logs to see if I can work out what is happening, and now to fix it – wish me luck – and/or give me hints!

[jase (Jason)]

Matt,
Its hard to beat the capabilities of MaximDL when it comes to guiding and guiding analysis. The flexibility and features offered is impressive. Items such auto-darks, automatic star selection, calibration timers, min and max move settings etc, make it a very complete package. Most of the advanced features I leave alone.

However for the ultimate, I recommend using MaximDL in combination with ACP. ACP sits on top of MaximDL and exploits MaximDL's guider scripting features. It builds upon MaximDL's features and provides "Adaptive Autoguiding" - http://acp.dc3.com/autoguiding.html

I haven't had a guiding problem when using 20 or 30min subs, but with the fast FSQ at F/5, I typically image filtered luminance using 15min subs. Ha is the only time I'll go deeper to boost S/N ratios.

I should note that there are no automated tools to solve all your auto guiding woes. You'll still need to take the time to analyse guider graphs and correlate worm cycle position etc. Once you've got general settings under control, only then will additional software enhance your auto guiding method.

At 2300mm focal length, you're probably on the boarder line for on axis guiding. Are you doing this? This was the only way I could successfully guide my previous C11 @ 2800mm. Mirror flop etc was simply too hard to counteract when using a separate guide scope. As I used the STL11k with integrated guide chip, it wasn't too difficult to guide on axis. You may need to look at an off axis solution with a "pick off" prism. I suspect only then will reach your desired goal. Keep in mind that guide star selection for on axis solutions at such long focal lengths can be frustrating - field of view is small and stars a typically dim. You'll probably want a camera that can be binned to boost sensitivity.

[g__day (Matthew)]

Jase - huge thanks for you in-depth and knowledgable reply!

Basically I really want to perfect tracking so I can go on to target other challenges. I'm pouring through PHD logs to see the raw data (look like a get a consistent 1 pixel error in DEC every 12 seconds or five pixels a minute (about a star diameter) - so that's in line with being a small fraction of SCP alignment in DEC - Maxpoint says under 30 arc seconds.

That is larger than I suspected - looking at 6 minute unguided shots stars look round to me - at 10 minutes I see a 2-3 mm star trail on a 4" * 6" shot - with less than 1 mm in DEC error.

As to mirror shift and mirror flop - I have (maybe very ignorantly) ignored these because I only need to re-focus my SCT about once every 3 months. I have a Meade motor focus so I never adjust the mirror position with the knob on the back of the SCT. Focus on a carbon fibre tube seems so consistent I have read this as the mirror isn't moving at all so I have no worries... Does not having to re-focus indicate very little mirror shift or flop? Or asked the other way how do I best tell how much mirror shift / mirror flop I have and how to be compensate for it?

Yes I am getting into the bowels of PHD logs. My gut is maybe my MAK is going out of focus, maybe stars are still to large on the CCD for PHD to get the centeroid correct - its just my gut its the software and ghardware taking images from the guide CCD that is introducing errors.

MaximDL based guiding is my next to do, I will see if it gives me better results. I foresee that some time next year I will advance to on axis guiding. I've been spreading my purchases around - CCD is the next major purchase, then a larger APO refractor and then hopefully only more advanced software remains.

Autoguiding seems a bit like chess - not to hard to get the hang on the basic - very challenging to master!

[rogerg (Roger)]

I use CCDSoft for subs up to 40min no problem, but I don't think that helps you at all because CCDSoft doesn't support the DSI.

I've never had the same level of success with off-axis guiding as I have with my dual chip ST7. There's many factors that could be making the difference, haven't bothered working through them (not important to me), but I do wonder if long exposure off-axis guiding is practical at long focal lengths with normal amateur gear. There has to be a reason SBIG have gone to so much trouble in their new "coming soon" cameras to support that off-axis guiding with false star.



Roger.

[g__day (Matthew)]

Roger,

I too get the feeling that S-Big are onto something when it comes to auto-guiding long duration, long focal length imaging. No that I have transfered all my gear off Celestron dovetails and onto much sturdier Losmandy bars throughout I'd say rigidity has at least doubled - and pointing is all the better for it. MaxPoint confirms this too!

What I do wish was better analysis of where I am loosing the auto-guiding war. My thoughts are review the PHD logs looking for either:

1) sudden, large movements in RA
2) large movements in RA that then reverse shortly thereafter

If I can see a pattern I can work to posit a cause and effect. One thing I do want to try with PHD is shooting say NGC 2070 on the same guide star for two full worm periods (say 16 minutes) with the frame being taken every evry 0.5, 1, 1.5, 2., 3.5, 5 and 10 seconds - so after these 7 runs I'd like to see what sort of trails I get - where sweet spots occur and what sort of errors I get chasing seeing.

Much to learn here!

[jase (Jason)]

Originally Posted by g__day

As to mirror shift and mirror flop - I have (maybe very ignorantly) ignored these because I only need to re-focus my SCT about once every 3 months. I have a Meade motor focus so I never adjust the mirror position with the knob on the back of the SCT. Focus on a carbon fibre tube seems so consistent I have read this as the mirror isn't moving at all so I have no worries... Does not having to re-focus indicate very little mirror shift or flop? Or asked the other way how do I best tell how much mirror shift / mirror flop I have and how to be compensate for it?

There is still focus shift with a carbon fibre tube, be it minimal. This shift typically does not affect the primary mirror riding on the internal baffle. Thus, if you have a secondary focuser (which you do), it is unlikely that this will be contributing to the mirror movement. Its important to note that the mirror can shift/flop by very small amounts depending on where you are pointing the telescope. For example, the shift maybe ~10 pixels when pointing east and ~4 pixels in the west. This of course can impact your pointing model as the flop is not always consistent. During guiding, I'd imagine that there would be minimal (I'd like to say, almost non existent, but in reality...)

How I combated mirror movement (the easy way) was to rack the course focuser all the way out. Thus the primary mirror was snug at the end of the baffle at the rear of the OTA. I then built the imaging train around this with the appropriate spacers so I could reach focus. As luck would have it, I could use the Lumicon GEG, JMI NGF with Robofocus, a 10mm spacer, then finally the STL11k. This combination reached focus exceptionally well. Though I could only operate the GEG in a few focal reducing modes. I think F/4 and F/5.5 if I recall correctly. Now the hard way of reducing movement is drilling holes into the OTA rear cell and inserting three threaded plastic or Teflon screws to push against the mirror. You need to make sure you turn each screw in a methodical fashion otherwise problems will arise (as you can imagine).

Of course, you can not worry about either of these and try auto guiding on axis (SBIG style) or off axis through pick of prism/mirror. The latter seems the most logical step in my opinion. Just connect your current guide camera up to an off axis guider. I'd be interested to see what you experience then. You've already mentioned PE reductions. Make sure you also focus on this. The lower the PE, the less guiding corrections will need to be made.

I think you mentioned focusing. Yes. This is another important piece of the puzzle, though not as much as one would think. Software like MaximDL can find the star centoid through different algorithms (predominately HFD), even if focus is soft. If you have a desire to calibrate manually - you can do this through a MaximDL plugin - http://winfij.homeip.net/maximdl/manualcalibration.html though I doubt you'll need to go this deep. Its important to get a good S/N ratio on your guider output. Hence taking darks for your guider can assist. This will also allow you to guide on much fainter stars.

You'll soon obtain pin point stars once you get all components working harmoniously together. As you put it. Understanding the basics is the easy part, but to make it work efficiently and consistently is difficult.

[ballaratdragons (Ken)]

Even though I have MaximDL, I still prefer to use Guidemaster for my guiding.

I can't fault its accuracy even over very loooong periods. If it ever starts letting me down, then I'll switch to guiding ith MaximDL, but I'm very happy with Guidemaster. Very easy to use too.

Guidemaster accuracy at present is +/- 0.5 arcseconds on nights of even mild wind.

[jase (Jason)]

Providing it works for you Ken, that's all that matters. 80% of the time, its not the software at fault, but the equipment set up. Balance, flexure, etc (items I mention in the above posts). What software brings is simply improved flexibility to "tame" the equipment.

I should be careful of the terms I use to ensure I don't confuse people. I refer to "on axis" guiding as defined by a self-guiding camera. This is guiding in its purest form along the same optical path as the main imaging chip. "off axis" guiding is that performed by the rigid coupling of an in optical path device that diverts/splits a small amount of off axis light to another guide chip.
Some "off axis" guiders can be seen here;
http://www.astrodon.com/documentatio...n.cfm?DocID=13
http://www.rcopticalsystems.com/dbs.html - advanced version of an OAG
http://www.celestron.com/c2/product....=50&ProdID=357

Off axis guiding can be performed successfully, but it comes with its own inherent problems.

[Bassnut (Fred)]

Peter of ATS tells me SBIG started life developing guiding software, and its CCDOps software is the best hes for guiding, its one one of the very few that has "adaptive tune". It comes with SBIG cams and is very basic in other features, so I guess many dont use it. Its a bit academic I guess, as it only works with SBIG cams, interesting tho ;-).

[jase (Jason)]

Interesting insight Fred. Though I think you're talking about SBIG CCDTrack for the ST4. The software used a Hysteresis algorithm that took into account the RMS error of the object (star) in pixel X and Y for past 16 guider cycles. I guess you could call it adaptive tuning! I suspect this was also incorporated into CCDOps, but not sure about CCDSoft considering its a collaboration between SBIG and Bisque. I feel certain many other camera control applications now use similar algorithms.

[g__day (Matthew)]

Slightly off subject - but does binning help if the stars are slightly large - will gluing pixels together say 2 * 2 help at all or is no real use?

PS

Really good information in this thread!

[jase (Jason)]

Matt,

Binning will allow you to guide on much fainter stars as the camera's sensitivity increases. It does not make a difference if the star is large or small. What is important however is that the guide star does not saturate the well depth too quickly. This is unlikely considering guiding exposures are short (on average). MaximDL will automatically select a star through analysing its brightness and proximity to other stars in the FOV. The latter is important as ideally you want to the star to be a reasonable distance from others in case auto guiding mysteriously chooses the other star instead (it does happen). It's interest to watch what star MaximDL chooses when going through the calibration routine. It is certainly not the star I would have picked in the field (which is usually the brightest one! Obviously brightest isn't always the best).

In the MaximDL manual you find the following;
"Important Note: Some cameras use a TC-237 guide sensor instead of the original TC-211 guide sensor. These sensors have much smaller pixels; it is strongly recommended to operate these sensors at 2x2 or 3x3 binning."

The above actually applies to any chip you are using to guide that has small pixels. The TC-237 has 7.4u pixel size. If you were coupling this to a c9.25 @ F/10 it would deliver an arcsec/pixel value of .65. This is not practical for a guider. At .65 arcsec/pixel, the guiding is going to be heavily compounded by seeing conditions - "aka chasing the seeing" Hence the reason why binning the guider is more efficient. A 2x2 bin gives a more realistic 1.3 arcsec/pixel value - probably a better match to your seeing conditions on an good night. The only time you probably wouldn't bin is when operating at much shorter focal lengths.

[Bassnut (Fred)]

Jase, very OT, but Ive seen a few esoteric industrial PID controllers that offer "fuzzy logic" adaptive control (its a 90s wank I suspect, another term for "adaptive tune") that offer constant imrovement. It would be very interesting to see such a device used in auto guiding, could make all the difference in a modelable mechanical system (other than PEC).

[Terry B]

Originally Posted by ballaratdragons

Even though I have MaximDL, I still prefer to use Guidemaster for my guiding.

I can't fault its accuracy even over very loooong periods. If it ever starts letting me down, then I'll switch to guiding ith MaximDL, but I'm very happy with Guidemaster. Very easy to use too.

Guidemaster accuracy at present is +/- 0.5 arcseconds on nights of even mild wind.

I agree. I use both guidemaster and PHD but get consistently better tracking with guidemaster. It is much simpler to change the aggressiveness than PHD and can be done on the fly. You then watch the result on the little 2 arcsec graph.
I only use 5 min subs and I think that flexure is going to be a problem with longer subs. I don't see a lot being achieved with longer subs with the exception of film cameras. Much more likely to waste time when a plane or satellite ruins a sub. The difference in noise would be hard for me to notice.

[jase (Jason)]

Also make sure you've got your auto guider calibration times correct. These are not critical, but will help. You can calculate this using the formula;

(number of pixels * pixel size * 206)
----------------------------------- = X seconds
(15 * guide rate * focal length)

number of pixels = is the desired number of pixel you want to move the star during calibration.

You can also set the min and max movements using this calculator - http://www.ccdware.com/resources/autoguidercalcv4.cfm

Finally, check out the attached slide presentation for general autoguiding information. Slide 22 has some suggested auto guiding calibration times - a good starting point.

[jase (Jason)]

Originally Posted by Terry B

I don't see a lot being achieved with longer subs with the exception of film cameras. Much more likely to waste time when a plane or satellite ruins a sub. The difference in noise would be hard for me to notice.

Deviating OT here, but for CCD work, longer subs are the only way to go. As I mentioned in a previous post, the signal to noise ratio of a CCD increases linearly with the square root of the exposure time. This means that quadrupling the exposure time doubles the signal to noise ratio. So take the longest effective exposure you can. This applies to all forms of CCD imaging, but is especially important for narrowband imaging where filters restricted the quantity of light reaching the chip.

----
That's an interest thought Fred. "Fuzzy logic" could be making a come back. If there was a reference point or sample over a long period it would be possible to use such technologies for auto guiding.

[Terry B]

Originally Posted by jase

Deviating OT here, but for CCD work, longer subs are the only way to go. As I mentioned in a previous post, the signal to noise ratio of a CCD increases linearly with the square root of the exposure time. This means that quadrupling the exposure time doubles the signal to noise ratio. So take the longest effective exposure you can. This applies to all forms of CCD imaging, but is especially important for narrowband imaging where filters restricted the quantity of light reaching the chip.

This is true but I find I get to a stage of diminishing returns. The problem is time available vs usable pics. As the exposures get very long in my camera the stars saturate and bloom. This degrades the images more than the readout noise.
There is an interesting article about this at http://www.licha.de/astro_article_ccd_combining.php

[ballaratdragons (Ken)]

Originally Posted by jase

Providing it works for you Ken, that's all that matters. 80% of the time, its not the software at fault, but the equipment set up. Balance, flexure, etc (items I mention in the above posts). What software brings is simply improved flexibility to "tame" the equipment.

Good point Jase

My gear must be running very good

I also always have the counteweight slightly heavier off balance. It improves the guiding markedly! I don't mean off balance by lots. It hardly moves when testing for balance, but it is favoured towards heavier on the counterweight side

[jase (Jason)]

Originally Posted by Terry B

This is true but I find I get to a stage of diminishing returns. The problem is time available vs usable pics. As the exposures get very long in my camera the stars saturate and bloom. This degrades the images more than the readout noise.
There is an interesting article about this at http://www.licha.de/astro_article_ccd_combining.php

Thanks for the link Terry. Interesting read. Read out noise is less of a problem. What you are trying to do is overwhelm the read out noise with sky noise by taking longer sub exposures. If the subexposure times are too short, the sky background noise will be minimal, thus resulting in the image being dominated by read noise. In contrast, if the subexposure times are long enough, sky background noise will be larger compared to the read noise, therefore overwhelm the effects of read noise. When the subexposure duration read noise contributes to less than 10% of the total noise in the image, there will be no major advantage to prolonging the duration of the subexposure further. As CCD's are linear in light collection, it's important to go long if you want to pick up the faint details. Though if you've got a high QE chip, you'll pick up the details in a much shorter exposure time. Note, I'm excluding aperture and focal ratio out of the equation as this is another discussion. This is the reason why I stated "Take the longest *effective* exposure you can". In your case, the longest effective exposure is limited by blooming. I wouldn't consider this noise. Its just a side effect of the chip characteristics and not terribly difficult to deal with. In this instance, you'll need determine how fast the well depths will fill based on exposure duration for your particular imaging train set up. Keeping in mind that different filters such as narrowband will take longer for saturation to occur due to the reduction in light they cause.

There are a few resources available to calculate "optimal" exposure times based on sky noise, read noise etc.
http://www.ccdware.com/resources/subexposure.cfm
and the attached spreadsheet (zipped) - (credit Steve Cannistra)

I'm happy to discuss this in more depth, but suggest we start another thread to ensure we don't further shift from the original threads focus - autoguiding.

[g__day (Matthew)]

Well I've worked out that on 20 minute duration shots at 2.3 metre focal length and seeing about 1 arc second drift per minute in RA.
My set up is as follows:

Mount Vixen Atlux with SkySensor2000-PC v2.05 on a heavy duty permanent pier in my astrolab,
Dovetails Losmandy G11 side saddle bars
Main imaging OTA Celestron 9.25" Carbon Fibre tube
Main imaging camera Canon 400D
Focuser Meade Motofocuser
Main autoguide OTA - 5" Skywather MAK and/or 80mm Megrez refractor
Main autoguide camera Meade DSI colour with 0.5 focal reducer
Main scope control program Software Bisque The Sky6 Professional -> ASCOM -> Maxpoint -> SkySensor2000-PC
Main autoguide program stark-labs PHD (may try MaximDL next)

Seeing conditions in suburn Sydney are only poor - average, and my favourite targets are planetary nebulae.

My South Celestial Polar alignment is +/- 30 arc seconds from the SCP according to Maxpoint with a 32 star model. I had to carefully balance my rig (loaded with 20 kgs) so RA was perfect. Each night I simply turn on my mount, recall last settings - slew to a star - re-sync if its more than 5 arc minutes off and I'm away imaging.

I find six minute guided or unguided shots look great. But between 10 to 20 minute shots I am seeing stars with a diameter of 1 mm on my 6" * 4" DSLR have a 1mm drift (which is 20 arc seconds). I calculate this from the Moon wills the full 6" shot so 150 mm = 30 arc minutes 3mm = 1 arc minute.

See piccy here

http://www.iceinspace.com.au/forum/a...3&d=1197160980


As soon as weather improves I'll try MaximDl for guiding and Binning on PHD V 1.7

* * * * * * * * * * * *

This site recommends firming up the guide scope - as differential flexure might be the cause of my woes!

http://astronomy.mdodd.com/astronomy-obs.html#Telescope