Characterizing Filter Focus Points

I’ve recently done several long autofocus runs on several different filters, in an attempt to properly characterize the focus point and the temperature compensation coefficients for my filters. The results are confusing me. My system is an fsq 106 running at F5 with a moonlite focuser with a step size of 0.26 microns. Using the new ciritical focus zone formula, my critical focus zone, using a 10% margin for error and 3" seeing, is ~59 microns, so ~30 microns, or 115 focuser steps on either side of perfect focus.

My autofocus runs, which virtually all hit 96% or higher scores in SGP, show focus points for a given filter at a given temperature that differ by up to 8,000 plus steps. The physical configuration of the system has not changed over the course of these runs. Graphic illustrating this below, as well as links to the focus run files.

Obviously, these results are inconsistent, and I must, at times (if not always), be substantially away from optimal focus. This clearly frustrates any attempt to determine focus set points, but more than that, it make me question the focus results even when I force focusing during an imaging session.

Any ideas on what I may be doing wrong or on how to drive convergence/consistency in the results?

For ease of interpretation:
Black = Lum
Brown = Ha
Pink = SII
Red and Blue = Red and Blue

https://1drv.ms/u/s!AsHXUvMw3K77grd-pkJaJt_Nt0Dy9g?e=D7Ptlb

Hi,

For what it’s worth I have been doing a similar exercise and obtained results that are not dissimilar, so I have also been trying to rationalise the results that I have seen.

Looking at your charts, with each filter you seem to have two or three reasonably linear sequences but with some quite significant jumps in FocPos between those sequences. This makes me think that you are probably using data sets taken on different nights or alternatively experienced a significant focuser slip. If this is the case then I would suggest separately loading the log-files for the different nights imaging runs, note the results , then average the sets of temperature coefficients and filter displacements that you obtain. Hopefully the different sets of results you get will be reasonably closely bunched so you can have good confidence in the resulting averages.

You may think differently but I think the important point is to discover as accurately as possible these different values as for each night’s imaging you will always do a focus run at the start of the evening (basically a calibration run) so if the relationships hold good you should then get good results for the remainder of the night.

Explaining the causes of the discontinuities between the different linear sequences is more difficult especially if they were not taken on different evenings as I suggested above. The most probable reason I suggest is some form of mechanical disturbance between your different runs. You noted that the range of FocPos values was 8000 steps at 0.26 micron or circa 2/1000 of a millimetre by my reakoning. If you are using a portably set-up then this small amount would be easy to rationalise but becomes more difficult but not impossible for a well mounted permanent observatory set-up, e.g. mechanical flexure, a minor knock or jolt, response to recent temperature changes, or possibly even ground vibrations.

I’m interested to hear other readers thoughts.

Regards

Mike

PS: These are the rambling thoughts of an old man with too much lock-down time on his hands. If your results were obtained in laboratory-like conditions using an artificial star as the target then I my amateur musings are totally off-beam.

PPS: Also, similar to my results, I notice that the responses to temperature change do not seem to be completely linear. In my case I have put this down to either a) the optical train being composed of several materials (by my recollection coefficients of thermal expansion are normally quoted for a single material) b) delayed response of optical train to earlier fluctuations of the downwards temperature gradient. What I have not been able to satisfactorily explain to myself is why the temperature change coefficient does not seem to be exactly the same for the different filters!

I use a Takahashi FSQ 106 at f/5 myself, in my case it is an older model, a non-ED one, and it has the original rack & pinion focuser. The focus shift of this type of refractor is caused predominantly by the temperature dependence of the refractive index of the glass utilized for the objective cell. This results in a change of the focal length: decreasing temperature -> shortening of the focal length. This is the reason why a negative temperature coefficient is observed with this scope.

There seem to be several issues here.

  1. Large deviations
    The large deviations (8000 steps * 0.26 µm = 2080 µm = 2.08 mm) apparently are caused by a not correctly adjusted focuser. Please specify what type of focuser is used. If it is a Crayford type, slippage might occur, resulting in large deviations between different nights. My rack & pinion focuser doesn’t exhibit slippage by design.

  2. Curve shape
    Some data points look reasonable, e.g. Brown = Ha or the lower Black = Lum, whereas others look dubiously. Generally it seems that the slope of the curves increased at temperatures below 0 °C. This again might indicate a not correctly adjusted focuser as well (more slippage at low temperatures). It is also possible that the adjustment of the focuser ia not stable and altered over time.

  3. Thermal equilibrium
    Heat flow and temperature equalization process need time and follow an exponential law. That means: the scope is characterizable by a half-value period. I tried to determine this value for my scope and found it to be about 35 min. Normally a period of 3 to 4 half-value periods is regarded sufficient for thermal equilibrium. Therefore it is important to let the telescope adapt to the ambient temperature long enough, placing it outside (in the shadow) before sunset. However, the change of the ambient temperature can vary considerably depending on the observation site or can be different from night to night. Often the ambient temperature decreases steadily and eventually reaches a stable plateau, but in some nights I experienced rapid temperature changes due to warm airstreams. In this situation, thermal equilibrium might be disturbed.

Therefore it is crucial where the temperature sensor is located: at best you want to measure the temperature of the objective cell. This is normally not feasible. In my view, the best location for the temperature sensor is: directly attached to the metal tube near the objective cell and thermally isolated from the ambient air (this is a general recommendation for refractors). Definitely you don’t want to measure the temperature of the ambient air (rapid temperature changes cannot be followed by the objective cell). You don’t want to measure the temperature in the focuser housing either (electronics and motor produce heat).

  1. Temperature coefficient
    For the blue filter you detemined k = 0.26 µm/step * -109 steps/K = -28 µm/K, but it is not reasonable to use data points which are that different. The best data points seem to be the ones for the Ha filter, and they correspond nicely to a subset of the data points blue filter and of the red filter with T > 0 °C. For the Ha filter I roughly estimate k = 0.26 µm/step * -200 steps/K = -52 µm/K. (For my refractor I determined -50 µm/K.)

Bernd

Bernd,

Thank you for the comprehensive and thoughtful response. In answer to some of your questions/observations:

  1. I use a moonlite nitecrawler focuser with absolute encoders. Slippage should not be an issue, and any physical slippage of the focuser would be reported in the focuser positions stats, unless there were a mechanical failure that resulted in both slippage and encoder inaccuracies. That is not impossible, but unlikely enough that I haven’t given it much thought.

  2. I agree, several of the curve shapes look reasonable. Ha in particular, which were all taken in one session. Generally, what I have noticed is that when I do a session devoted to focusing, the curve shapes for that evening look reasonable. When I combine data points across different sessions, they become incomprehensible.

  3. Thermal equilibrium may well be a factor here, as several of the nights I used experienced consistent drops in temperature for much of the session (the scope had been set up outside and shielded from the sun, and in any case, I generally save the focus runs for late in the evening after my imaging is done). But, since part of what I am trying to do here is determine the temperature compensation coefficient for my filters, that is inherent to the nature of the exercise.

  4. I have a temperature sensor that is connected to the focuser by a wire. The temperature sensor itself is attached to the exterior of the dew shield of the scope, near the end of the dew shield. Not sure how to thermally isolate it (perhaps cover it with painters tape)? Happy for ideas here. I have been thinking that it likely measures the temperature of the dew shield metal more than the temperature of the objective lens, but I don’t see what I can do about that.

Mike,

You deduce correctly! The data is for sessions taken on several different nights, and with a couple of exceptions, the data from any individual session are coherent and behave as expected (e.g. the Ha are all on one night, and the several Lum series behave correctly within each session).

It is a portable setup, but one I leave up for several nights at a time (weather permitting). I don’t change anything about the configuration, simply left the OTA off the mount and carry it 20 feet from my garden to a shelf in my office, then back again. Everything in the imaging chain is threaded.

The focuser itself is a moonlite nite crawler, with absolute encoders.

Unless I am doing the calculations incorrectly, the range of 8000 steps equals roughly 2000 microns, which is roughly 2mm, which while still small, makes me think that this is not likely due to mechanical differences.

I agree with your approach to calculating temperature coefficient slopes, given that they seem to be consistent between sessions for at least some of the filters, and have been playing with that. But my bigger concern at the moment is how to have confidence in the focusing results when the optimal focus position, at a given temperature can differ by 40x the size of my critical focus zone.

I did another session last night, without having moved the scope at all, so no mechanical disturbances to consider. The graphic below plots the results for two consecutive nights, equipment untouched:

As you can see, the slopes are reasonably consistent (so I am inclined to believe the temperature compensation data there), but based on this, the focus position for the Lum filter could be the same across a 6C temperature range, despite the fact that there is a relatively acute slope for temperature changes. Those two facts seems inconsistent. With the slope shown, it cannot be the case that optimal focus is found at 11,500 steps for both 1.47C and 7.5C.

I am a bit mystified. The ironic thing is that i haven’t noticed a problem in my imaging, but then I am relatively new to this hobby, so not sure I have a critical enough eye yet.

It seems like, for a given filter, SGP is finding multiple, significantly different focus points as equally optimal solutions at significantly different temperatures. That worries me, as it means I can’t rely on the accuracy of the initial focusing runs at the start of a session. Given the low likelihood of mechanical slippage in my system and the fact that this problem appears even when I do not move the system between sessions, I’m inclined to wonder if there is a problem in my SGP autofocus configuration.

I wonder if perhaps there is a problem with my settings? My sampling is 3.5" on this system and I am using 2x2 binning for focusing. Could that be creating discontinuities with the HFD calculation? Should I have a particular setting for minimum star size?

I use an auto focus step size of 650, calculated in the manner instructed in the help file, which consistently gets me a range of HFDs that is 3-4x the “in-focus” range. I could try a smaller number for step size, but that will being me under the 3x-4x top to bottom range recommended in the help file.

I have tried tightening the focus success criteria to 98%, but to my chagrin, the system can’t seem to achieve better than 96% with consistency. While that seems pretty good to me, perhaps that in itself is an indication of an issue, given that I am using a refractor with a very tight imaging chain?

Open to any and all help, as I would really like to figure this out.

If it helps, below is another graph with the results from the two sessions in the graph above, plus a couple of earlier dedicated focusing sessions. Each of the Lum lines is on a different night, as is each of the SII lines. The Ha lines and the Red line were each taken on one night. As you can see, apparently the optimal Lum focus position at 7C is 12,000, and that is also the optimal focus position at -2C. The slope of the Lum line in each session would indicate that the optimal focus points at those two different temperatures would be different by ~3,600 steps.

In my view it looks like the inconsistencies occur only when the focuser and the temperature measurement are shut off, whereas the results within one session appear plausible. This could mean: either the initial temperature or the initial focuser position (each in absolute values) is wrong, the relative values probably are correct.

A focuser with absolut encoders is expected to report the correct focuser position. You can check whether a focuser position at the beginning of a new session is the same as at the end of the proceeding session (without having moved the focusser to a “zero” position in the meantime).

Assuming that the position is correct, only the measured temperature values remain doubtful. So I also would check the temperature with a second measuring device. Probably it is sufficient to check the initial temperature.


As to the thermal insulation of the temperature sensor: the sensor on my refractor is located at the tube, from the dew shield few centimeters in the direction of the focuser (i.e. the location of the objective cell). The sensor has direct contact to the metal tube. I covered the temperature sensor with 2 layers of a thin (about 2 mm thick) foam and fixed it tightly with parcel tape. Thus wind cannot shift the temperature reading.

Bernd

Hi,

I completely agree with Bernd on this point. I had overlooked the the possibility of the temperature readings being significantly wrong but without knowing what level of positional accuracy is expected from your Moonlight focuser my money is still on some form of linear or rotational slippage.

I am thinking that if the relative movement between sessions is of the order of 2mm then this should be measurable with a ruler but certainly with a micrometer.

I cannot tell from your posts but if slippage is the issue then assuming your scope is parked with the camera dragging downwards then the FocPos should be moving inwards as the camera drags outwards. Is this the case? Maybe you could park your scope so the cameras weight is pressing inwards in which case if there is slippage the focusser movement would be inwards to compensate.

I don’t see why you think this is the case. I assume that your AF is resulting in very similar minimum HFD values so AF is working despite the FocPos / Temperature discrepancies. Also it seems to me that that the relative displacements for your various filters are consistent when measured in the same session so if, for example, you wish to use L filters for focusing then switch to NB this should not be a problem.

I feel your scope should be able to get 98% consistently unless seeing is really poor. I suggest checking:

a) Any signs of residual backlash on the AF curves?
b) Is the AF routine detecting a reasonable number of stars at the outer extremities of the curve. If fewer than say 5 then maybe reduce you step size a little and/or increase your AF exposure times so hopefully more stars will reach reach the qualifing SNR level to count as star detections.
c) Do you see better results when using L filter. If so increasing AF exposure time for say NB and Blue may improve matters.
d) SGP’s parabolic curve fit does not give best results it the AF curve is noticeably asymetric about the best focus position. Hopefully with only some minor tweaks your success rate will improve.
e) Do you see any improvement if your mount settle time is increased a little?

Not sure what you can do about thermal lag. I assume that the temperature sensor is likely to be a a thermocouple so I think this will always have a much lower thermal mass (and therefore respond more quickly to change) than the objective lens (assuming this is the most temperature critical component of the optical train). I guess the answer might be an infrared temperature sensor if these are available but I’m not aware of any for this purpose.

Regards

Mile

Mike,

I feel it shold be able to get 98% too!!

Seeing is usually okay (between 1.5"-2.5", so that shouldn’t be the problem with achieving a better than 96% score. What does residual backlash look like on the AF curves? I will go back over the last few sessions and try to identify it.

Having said that, with respect to slippage, unless the encoders in the focuser are misreporting, even if the focuser slips, the step position should be reported accurately. So, if I understand how things work correctly, returning to the proper focus position for a given filter at a given temperature should result in returning to the same step count, shouldn’t it? Nonetheless, I have a digital calipers, and so will measure at the end of a session and 24 hours later, to see if there is a change, and whether that is reflected in step count.

Generally yes, detecting 20 plus stars. No, even with Lum, I my highest repeatable score is 96%. It sometimes gets above that, but rarely and randomly. On these focus runs, I am continuously tracking a target, so don’t believe settle times factor into the equation?

Bernd,

That is quite logical. It explains why I get consistent curves across filters within a session, but why these see incompatible across sessions. I am at a bit of loss to explain though why the reported temperatures from the focuser would differ so much across sessions, and why the delta T itself would differ across sessions. And even if the initial temperature is a session was wrong, one would expect that over the course of 50 observations, it would eventually trend to accurate, unless the thermometer is just broken…

I will use a temperature gun to check the temperature on the surface of the scope near where the temperature sector is half a dozen time during a night and compare that to what is reported by the focuser, and will do that over three nights to see if the delta-T is consistent or jumps around.

A lop-eared appearance to the AF curve, see the screen-shot in this post, on the right-hand side.

You may be tracking but the movement of a heavy camera during the AF moves can cause a displacement which autoguiding (if running during AF) will endeavour to correct. But if you have a mount settle time specified this will be applied between AF exposures. You will see it in the SGP log.

No, If you have focus points defined in your filter definitions then if you change from say Lum to Red, SGP will calculate the relative movement between your Lum and Red definitions then move the focuser to the most recent Lum focus position plus/minus the delta. In my case I have the lowest deta (in my case Green) set at 0 then moving up to Ha at 46.

Mike - thanks again for continuing to help think through these issues.

Got it on the settle issue. I am not guiding, so that isn’t a factor, I don’t think.

I have not defined any focus points yet, so that isn’t a factor here.

I will run down whether I have dog ears like that.

Probably I missed the obvious: SGP stores the assignment of filter number (= index of the filter array, beginning with 0) to the filter names in each sequence. If that assignment was changed at some time, you could end up with sequences storing different filter assignments. The use of sequences with different filter assignments could explain the confusion that you observe.

In order to check whether this is the case, please check whether the filter assignment is consistent in all sequences and in the correspondent equipment profile. And of course the assignment must match the placement of the filters in the filter wheel.

Bernd

Fair question, but I have not changed the filter assignments since initially setting up SGP months ago.

I am sorry. Then I have no idea what causes your issue.

Bernd

Not at all! You have given me several ideas to track down. I just need the rain to stop so I can give them a go. It is vexing though.

Update.

I have run a 10 hour test, with the scope in park position (pointing at the celestial pole). Cycled the focuser on and off three times.

Every time, the focus position was reported at the same position (18,000 steps). When measured with a digital calipers, there appears to be no physical slippage of the focuser due to camera drag.

Each time, I checked reported temperature as well, and then cross checked that by using a temperature gun to measure the temperature of the surface of the scope at the location of the focuser temperature probe. Each time the temperature appeared accurate (albeit in a fairly narrow 3 degree range, as this was all done indoors).

So, it feels like I can reasonably rule out slippage, misreporting of starting position and misreporting of temperature?

By way of update, I have improved the focus score by reducing the autofocus step size from 650 to 475. Now consistently scoring 97%. I may try tweaking a little more to get to 98%+ with consistency.

Still no idea why the acceptable filter focus points appear to vary so much though.

This thread reminded me of your issue: https://www.cloudynights.com/topic/764960-focuser-position-drift-or/ Maybe it is helpful for you?

Bernd

I Did some imaging last night from roughly 9pm to midnight, and then a new set of focus runs (the first 8 focus events are from the imaging session. Events 9-63 are from the focus runs). I cycled through 7 focusing events for each of LRGB several times. Almost every focus event reported 97% or better scores. The results are displayed in the graph below, which continues to illustrate the very inconsistent outcomes the focusing seems to be reporting. I am a bit mystified as to what to make of this.

One trend did catch my eye, in the tabular data (see below). In the vast majority of cases, the focus position moves in with each and every focus event in each series of 7 events for each filter. This seems to be the case even when the temperature is stable or rising. I would have expected more of a random walk around the “ideal” focus position. Perhaps that is a clue to what is happening, though I haven’t been able to reason out what that might be.