Autofocus Backlash Problem

With my new Celestron RASA 11 scope I have run into a problem that I never noticed before with any of my prior equipment configurations. The RASA architecture moves the primary mirror when focusing and that causes an image shift problem. The RASA has the camera in front of the scope so the only practical focusing mechanism is the provided primary mirror focusing. My focuser backlash compensation setting is IN for step size +1000. Temperature Compensation value is -110 steps/degree (F). These numbers seem high but the autofocus increment that works well is 350 steps. The gearing architecture requires these settings.

Everything works fine when the temperature is dropping steadily during the night, at lease steadily enough to keep the temp sensor from registering a false momentary upward tick. This causes the focuser to rack out as the temp drops which is what we want. No backlash compensation required.

So the problem occurs when the temperature increases a tiny amount during the night. Which happens regularly since even if the temperature is dropping steadily throughout the night, it is pretty common for the temperature sensor to register a brief uptick, even if other sensors nearby might not. This causes a swing in the focuser up by 1000+x then down by 1000, which totally ruins the image it was taking. Double stars over the entire image, with nice lines connecting each pair. Quite annoying. One example was a night when I was taking 4 minute exposures. Approximately 1 out of every 4 or 5 images was ruined. Temperature dropped about 12 degrees over the course of the night, starting at the highest value, and ending at the lowest.

Am I missing some setting that would let me set a minimum temp increase before the temp compensation move kicks in? If not, could this be added? One easy initial fix for this could be to just automatically set the minimum reverse focus move to some fraction of the autofocus routine’s step size. In my case a cumulative reverse of 175 steps (350/2) or about a 1.5 degree temp increase would be required to actually trigger a focuser movement. This would rarely occur unless a real warming trend occurred. And being off up to half a step size in focus would probably not be noticeable.

SGP only runs temperature compensation between images. We do not move the focuser during images. Do you have temperature compensation enabled elsewhere like in the driver?


This is only true if the total time required to complete the entire backlash cycle is less than about 5 seconds. In my case the in move takes 14 seconds then the out move takes 14 seconds, for a total of 28 seconds. For the initial 23 seconds of my image the focuser is moving. The backlash compensation process is started and is quickly followed by starting the image capture, which occurs without any check of the focuser backlash loop to confirm that the focuser has finished its moves.

The reason that this has not been a widespread problem for users is that most scope/focuser combinations will complete this cycle in less than 5 seconds. Because of the RASA geometry, and I suspect every other Celestron that focuses by moving the primary mirror, the numerical steps required is quite large because of the gearing Celestron uses. In the attached log you will see that I was using a backlash compensation value of 2000 which takes 28 seconds per cycle. I have since decreased that to 1000, for a cycle time of 14 seconds. Which still overlaps my image capture by 9 seconds. The 1000 is somewhat adequate, but not totally satisfactory.

I should note that all other scope/focuser setups I have used have worked with a focuser increment of 30-60 and a backlash compensation of up to 100, which allowed for a very quick backlash adjustment cycle. My Rigelsys focuser gear connects with the coarse adjustment knob on the RASA, so that’s the fastest it can move. And I have set Rigelsys to move at its fastest rate.

Here is central part of image BAD_IC2177 Eagle which started capture at 9:51:57.

and log

Two other examples occur in this log:
Image B_0013 at 11:12:04
Image B_0014 at 11:14:29


You may want to check out this thread

Yes, definitely the same issue. Thanks.

A somewhat related issue:
Seems to me for long exposures, say 15 to 30 minutes, it is probably important to do autofocus moves several times during the capture to stay in really good focus. Temperature and focus can move a lot over a 30 minute period. Clearly this would only involve moves that don’t invoke backlash compensation.

Maybe this should be a feature request.


Looking at the RigelSys web site, it appears their focuser’s ASCOM driver supports temperature compensation directly. If you enabled that, the ASCOM driver would make focuser adjustments during the exposure. My Optec focuser has built in TC which I use all the time. I do not enable TC in SGP. The Optec focuser maintains excellent focus throughout a 30 minute NB sub. However, if your focuser’s movement is not rock solid, making moves during the exposure could be a problem.

Also, looking at my own SGP logs, it appears that SGP does ask the ASCOM driver if the focuser has reached the commanded position. Its possible that the RigelSys ASCOM driver is reporting completion before it really has.

If your Rowe-Ackermann has the Feathertouch focuser, you might want to look into Optec’s products:



Did you go thriugh the SGP logs jmacon provided and see the backlash timeouts?

Moving mirror systems do not work very well trying to focus in the middle of an exposure especially if backladh is required.

Thanks for the input Charlie. My RASA does have the Feathertouch focuser, but I don’t think the optec focuser would make any difference. The problem seems to be the moveable mirror, which is independent of the focuser. If he optec actually changed the effective gearing to give me faster moves, that would help.

I have thought about using the Rigelsys built-in TC, so I may shift over to that. If I set the temperature averaging interval to something like 120 seconds, it should prevent really small temp increases from triggering a backlash compensation. The image I referred to above triggered the backlash compensation for a temperature increase that was so small the log indicated the change was 0.0 degrees. Calculating backward from the step change of 1.6 steps I get a temp increase of .015 degrees.

What I really need is a focuser that will sit in front on the RASA central obstruction and connected directly to the QSI camera so I can lock down the mirror. Not any bigger than the central obstruction with a backfocus requirement with camera attached somewhere around 10mm. Anyone know where I can find one of these?

1 Like

I have a few comments on this because some of it is related to focusing with the primary at f/10.

First - if you are at f/2.2 I don’t know why you would need long exposures - beyond perhaps 5 minutes or so. In that case I would never focus during the exposure.

Second - if you find that the tendency over the night is for the focuser knob to turn clockwise as things cool down - and if you really want to do temp. adjustments during the exposure - then I would set backlash the other way so that during the night the mirror is “falling” with gravity and backlash compensation is not invoked.

When I focus EdgeHD11 with the primary, I have a geared down robofocus motor driving the focuser knob and it takes about a minute to wind out backlash. So your 14 seconds is relatively fast.

I have done hyperstar imaging at f/2 with the EdgeHD11 and the sgp automatic focus works fine on the primary focuser knob with Rigel - but that is using periodic autofocus between images. There are problems with the autofocus routine for me but only with certain cameras - and the asi 1600mm works fine.

I don’t use any temperature compensation at all - and I just refocus periodically. Have you tried that? My hyperstar fwhm’s are in the low 2" with about 2m exposures. And even with narrowband there is little need to go beyond 2m exposures at f/2 - particularly since you need a wide bandwidth filter at f/2 - such as the Baader fast filters I use.


Your comments are right on Frank. The longest exposures I do on the RASA are 4 minutes, as you say, f/2.2 is very fast, although my QSI6120 has 3.1 micro pixels, so that requires longer exposures. My interest in the TC during longer exposures is geared toward my TV100 at f/5.8 which has a full set of NB filters, and has a very well defined temperature dependence. And I do use the Baader fast f/2 filters for Ha and OIII, which I currently have mounted with RGB in my 5 station filter wheel on the RASA. And I did indeed turn off the TC and started running the AF routine every 30 minutes. That has solved my TC problem on the RASA. My biggest focus problem at the moment is that my RASA does not seem to maintain a consistent focus. There is not any real correlation with temperature. Floats all over the place. Any ideas on that? Should I perhaps tighten down the locking knobs a bit?

And this is even more of a mystery to me because I am getting very nice looking focus curves. Its just that maybe the next one I do 30 minutes later might be shifted down several hundred steps, far more than any temp change would warrant. Then the next one might be back up again.


I would make sure that the autofocus curves are repeatable - otherwise you may not be reaching good focus at all. So - make sure that if you get a nice curve, if you repeat it immediately you get the same curve and same optimal focus value. If you get something different - that suggests either 1) the focuser is moving too fast and it is losing steps 2) the backlash compensation needs to be much larger or 3) there is something mechanically slipping in the focus mechanism.

Otherwise I can’t explain why the focus value would change so much or float around. I never got a good behavior of temperature with focus with an sct - and I view it as a fairly complex problem in terms of thermal changes and expansion - but at the same time I do get very repeatable autofocus curves with the primary focuser once it is set up right.

I don’t use locking knobs at all with EdgeHD. Are you sure you need them? I leave everything loose and then the periodic autofocus will reseat the mirror against gravity in a consistent way.

The RASA should be better behaved in terms of focus - so I expect you can get everything very consistent.



I looked at my own SGP logs and it showed an entry confirming the commanded position had been reached. I am not familiar with the ASCOM logic but I assume you can poll the focuser to see if the movement has completed and then poll to get the current position. Then compare the currently reported position to the position specified.


I believe I have this, and other, backlash related “wait” issues taken care of and they’ll be in the next beta. If that doesn’t seem to resolve things please also attach a time stamped log from the focuser application so we can correlate things with SGP.

Thank you,

Thanks Frank. When I get back to the obs I will loosen the lock knobs completely. I left them slightly snug which may be causing the problem. I do get repeatable curves taken in series. Maybe the shift I have seen is because I started a new target. I will start doing autofocus on each new target.

Frank, here is an update on the auto-focus results on my RASA 11", and results on my SVQ100 for comparison. The following graph plots the last couple of months of auto-focus results on the RASA. It shows a very repeatable temperature dependence that yields this linear equation: FocusPos = 15393 + 72 * temp. The standard deviation in FocusPos is about 250 steps, which is 70% of the 350 step increment I use in the auto-focus routine.

What I conclude from this is that, in general, the outlier focus points are probably a worse focus position than just using the point calculated from the best fit line.

Here is my plot over the same period for the SVQ100 refractor. It yields this linear equation: FocusPos = 17704 + 10 * temp. We see that it is a tighter plot, with a mean deviation from the line of about 14 steps, which is 35% of the 40 step increment I use in the auto-focus routine.
This is about half the deviation exhibited by the RASA, which is what we would expect.

I think these results support the value of the auto-focus routine providing some kind of focus quality measure so that if the results are unreliable, some alternative action can be taken, such as a focus re-run or simply using the best fit calculated value. This is already a feature request in the selected but not implemented section.

Personally, I think the best option is to allow input of the 2 values that determine the line: the intercept and the slope. The slope is currently implemented, it is the Temperature Compensation Steps value. All that needs to be added is the intercept value at a pre-defined temperature such as 0F. This would allow the program to completely control the focuser position as a function of temperature without ever running the auto-focus routine. Because of the larger random error in focus position exhibited by my RASA, I can’t really trust the focus routine to give me a better focus value than the linear fit line does. So my SOP is to set the initial focus position based on the formula, then let the Temperature Compensation feature control the focuser for the rest of the night, without ever running the auto-focus routine. Obviously you have to run the routine a lot to generate the data to determine the best fit line.

If this feature were to be added, the user could choose an option that would use the routine generated focus position ONLY if it was within a given tolerance of the best fit line value. In essence it would replace all outlier values with the best fit computed value.

For those who might argue that those outlier values are actually the best focus at that point in time, I know this is not the case. Because on many nights I have run the auto-focus routine 5 or 6 times in succession, and I get this same spread in values of about 300 or more steps. This is not really that bad, since my focus routine increment is 350 steps between points. I also can correlate many of those outlier points with really marginal looking focus plots.


That’s an interesting result and I don’t think I would have such a strong correlation of focus with temperature on my sct’s - but I haven’t checked recently.

You should be able to tell the impact of the variation by looking at the bowl shape of your autofocus curves. If you can’t get a nice bowl shape on a night with good seeing - then I would try to tweak the autofocus parameters. And if you have a sharp ‘V’ then I would decrease the step size so you can see a bowl. The bowl will let you estimate what tolerance in focus position you have before the HFR changes by perhaps 10% or something. You can derive that from a single, good curve.

If you then do repeat curves and you find that the scatter in the result is greater than that - you know you have a problem because it means the autofocus curve can’t reliably return to the true focus position after taking the curve.

But if the curve does repeat itself nicely, then you can compare the scatter in the plots you show relative to that tolerance. If the scatter is greater than the tolerance then I would say you would lose focus accuracy by just using the linear model.

For any of these scopes it’s hard to know the thermal state of the entire system from just one or even a few measurements.

But I can see the merit of your feature request - but I’m not sure how many people would use it. And if you are able to get a good focus curve - I would always use that.

For me - I’m just hoping that one day my scope will be able to center a target accurately so I can do multi target sequences while I sleep. As it is I can only set up a single target manually and use the first part of a good night - wasting the rest of it. The autofocus works great for me now - but I am waiting to be able to center a target


Another simple test you can do. Try doing autofocus and confirm you get a good curve. Then compare the focus value to your linear model based on temperature. If you find a big difference then take an image with the position you are currently at. And then set the focusedr to the value set by the model. Then compare the fwhms in the two images to see if one is better.