I have recently moved my deep sky camera, the SBIG 16803 from a C14 EdgeHD to a C11 EdgeHD. I need to focus with the primary mirror and therefore to have substantial backlash compensation.
I hardly ever had issues with getting a nice autofocus curve with the C14. I used 9 steps with a step size of 400 and 2x2 Binning and 1500 steps of backlash compensation. This is a typical curve for Luminance in this case:
https://www.dropbox.com/s/kkn1ez4ff6...focus.JPG?dl=0
However, now that I moved the camera over to the C11, I have been having problems achieving autofocus with the major issue being the non-conformity of the curve shape. This was typical of what I got using my old settings:
https://www.dropbox.com/s/lpv1bthlty..._%231.JPG?dl=0
The flatness or "wing" as I've seen it described at right, causes the quality rating to drop below 90%, which is SGPro's cut-off and a failure to determine the focus results. You can see that the rest of the curve looks OK. The "wing" shape is, as I have read elsewhere, symptomatic of insufficient backlash compensation.
Firstly, I set out to redetermine the optimal step size. According to SGPro, the method involves determining the mean HFR at the best focus position, determined manually, until the HFR is 3-5X this. I carried out this exercised and developed the following table:
Step Size Multiplier
350 3.1
400 3.6
450 3.8
500 4.0
The focus curve began to flatten beyond this and the stars took on a donut shape due to the central obstruction of the SCT, so I didn't continue. I understand from other forums, that once the donut shape is encountered the HFR calculation becomes inaccurate. I decided to run with 500 steps.
The first trial, with the new step size and still 2x2 Binning gave this curve:
https://www.dropbox.com/s/4gti09ce0y..._%232.JPG?dl=0
Once again the shape of the right side of the curve was not correct and if anything the result was worse than with a step size of 400.
I increased the backlash compensation from 1500 to 2500 steps. Immediate success!:
https://www.dropbox.com/s/ljcauol1lh..._%231.JPG?dl=0
The quality figure came out at 92% so better than the threshold and the focus was determined. However, the shape of the curve to the right side was still somewhat concave down, instead of straight or concave up. This suggested to me that the HFR figures for the 2 right hand points were too low and still suggestive of inadequate backlash compensation. So I took the backlash compensation up to 3000 steps, with the following result:
https://www.dropbox.com/s/u3addx0ncv..._%231.JPG?dl=0
This looked better, with the quality rating going up to 94%. I also trialled 3500 steps but that gave much the same result, so I settled on a backlash compensation of 3000 steps. This worked very well for all the Luminance trails I made. I didn't have time to try R,G, or B but presume they would be fine too.
Since I had to add 1500 points to the backlash compensation, it makes sense to me that the two right hand points, spanning 800-1000 steps, were affected by the insufficient backlash compensation and also that they read too low. I presume there is a huge amount of hysteresis in the focuser moving the primary mirror and this requires a large compensation. Quite why the C11 seems to need more than the C14, I don't know.
I then tried Ha. Unfortunately, with these same settings, there was a large wing on the right side. The first HFR reading was actually lower than the 2nd, then it behaved itself, but that first point was enough to cause the algorithm to fail to find focus. Stupidly, I didn't copy that curve.
I had read, again in other forums, that the narrowband filters can suffer worse from the donut shaped stars at the extremity of the focus. The recommendation is to increase the binning in this case. I also sought to narrow the focus range by going with 7 points. This worked:
https://www.dropbox.com/s/gtc4z9vb0i..._%231.JPG?dl=0
I then set off an automated run with Lum, followed by Ha, OIII and SII in sequence, forcing an autofocus at each filter change. All the Luminance ones worked and all the narrowband ones failed.
I am fairly sure that the wing on the Ha curve is not caused by backlash, since once that was sorted for Luminance, it should be the same for all filters. Rather I assume the issue lies with the donut shapes of the stars when further away from the optimal focus point. I believe, therefore that staying with 3x3 Binning, giving an image scale of 3 x 0.67 = 2 arc secs per pixel, will help with HFR determinations when further away from the optimal focus point. I want to go back to 9 points, as I think that 7 is too few to give reliable results, so I am thinking that pulling the step size back to 400 or even 350, which will still meet SGPro's criteria, but will narrow the range of focus and reduce the likelihood of donut star shapes fouling up the HFR determination. In the field I was doing the work in there was a paucity of stars for the narrowband filters, so perhaps I need to increase the exposure time from 45 secs to 1 min.
Does anyone have any perspectives on this particularly in terms of the conclusions I have drawn and/ or the next steps?
CS Niall MacNeill
