So points 2 and 3 have a FWHM = -1. That means there were no stars that had FWHM curve fits. That is either because there were no stars detected or the curve fitting was unable to fit gaussians to any stars that were detected.

Points 1 and 4 have "large" error bars which means a very small number of stars / or gaussian fits. So marginally better than points 2 and 3 but probably the same issue.

This is actually a good example of why useWeights works well. The curve fitting has more or less ignored points 1-4 so the curve fits the other points quite well.

Looks like you don't use filters so won't be anything to do with that.

If you post a log with verbose and focus set I'll be able to find out more. If you didn't take one then put these settings on for next time.

The solution points are quite different. The 308k solution on the second curve is between points 6 and 7 on the first curve - clearly not near the minimum, so something looks wrong. You mention the temperature being quite similar. Were these runs taken at roughly the same point in the sky and how much time separates the 2 runs? How do the subs look for each Autofocus run? (roughtly in focus, or not in focus, etc)?

I suppose a simple test would be to run Autofocus 3 or 4 times consecutively. The solutions should be pretty similar.

Thanks again for your feedback.

So I may conclude that measurements 1 - 2 - 3 - 4 are (thanks to the weights) not or hardly usable and I am therefore not worried about it. And correct, I'm not using filters except a 3-band light pollution filter.

Both runs are taken on exactly the same point in the sky (more specifically centred to Merak) around 1 AM within an interval of 10 minutes. I'm going to do the tests again this night (or when the weather permits) with specific attention to the subs and logging enabled.

In the meantime, I have the following questions about the settings in the CFZ tab:

1. Wavelength

The manual says: This is the light wavelength to use. It is defaulted from the currently used filter. Remember to set this up in Filter Settings for your filters.

I'm not using filters except a fixed 3-band light pollution filter. Which value should I enter as λ?

2. Focal ratio

I'm using a 700/107 APO corrected by the software as being a 709 mm. And behind that is a flattener/reducer of 0,75.

Do I here enter 6.5 (f of the telescope) or 4.9 (f of the telescope + flattener/reducer)?

So there's something odd with those solves then if they were only a few minutes apart.

Q1. Put the midpoint of the pass-band. If there are multiple pass-bands then an average (even if the result doesn't lie in one of the pass-bands). It won't make a big difference to the CFZ.
Q2. Its the F/ of the optical train (telescope + flattener) that you need to enter. In the official 3.6.5 version its defaulting from just the telescope (there is a bug whereby the reducer in the Optical Train is not taken into account). There are 2 ways around this. 1) Just change it in the CFZ to 4.9. 2) Setup the Optical Train with the telescope Focal Length = 532 and flattener = 1.

The bug in the answer to Q2 is fixed now, so in the latest version 3.6.6 beta, the F/ defaulted from the optical train will take into account the reducer (i.e. for you it should default 4.9)

3.6.6 should be released shortly (beginning of August I believe).

Filter specs:



1st band: CWL495 nm - FWHM 35 nm
2nd band: CWL 654 nm - FWHM 12 nm

So λ average = 575 nm

Should I also have to do something with the (average) FWHM value ?
And/or is this related to the FWHM setting in the CFZ tab ?

575 should do fine.

The FWHM quoted for the filter is the width of the band-pass. Nothing to do with the seeing FWHM used by CFZ.

With your last feedback I managed to successfully focuss "4 in a row":









But ... I only get this result with binning 3x3 or 4x4.

When binning 1x1 with various exposure times and different gain, the following works of art are created:



It doesn't look like anything and I don't believe the R² value = 0,99.

Agree that R2 value looks wrong. Did you manage to get a log?

Binning 3 x 3, perfect focus:




Binning 1 x 1, crazy result:




Binning 1 x 1, another one:




Did enable verbose logging for Focus, but got a > 100.000 lines text file.

This is an extract:

"Linear: step 11, newMeasurement(277500, 8.04483 -> 8.04483, 29)"
"Starting LM solver, fit=hyperbola, solver=cholesky, scale=more, trs=levenberg-marquardt+accel, iters=10000, xtol=1e-06,gtol=6.05545e-07, ftol=1e-06"
"LM Solver (Hyperbola): Solution found after 0ms 10 iters (small step size). A=13458.3, B=2.56304, C=313789, D=0.768889"
"Linear: fit(11): 313789 = 3.33193 @ 277500 distToMin -36289"
"Linear: Solution #4: 313789 = 3.33193 @ 277500"
"Linear: 1stPass solution @ 277500: pos 313789 val 3.33193, min measurement 3.32139"
"Focusing outward by 36,289 steps..."
Restarting focus motion timer...
Abs Focuser position changed to 278940 State: Busy
Restarting focus motion timer...
Abs Focuser position changed to 286829 State: Busy
Restarting focus motion timer...
Abs Focuser position changed to 294718 State: Busy
Restarting focus motion timer...
Abs Focuser position changed to 307291 State: Busy
Restarting focus motion timer...
Abs Focuser position changed to 313789 State: Ok
"Focus position reached at 313789, starting capture in 0 seconds."
"Capturing image..."
"Image received."
"Detecting sources..."
"Detection complete."
"LM Solver (Gaussian): Guess perturbation=1, A=34, B=4.52527, C=4.09912, D=0.204045, E=0, F=0.204045, G=623.183"
"Starting LM solver, fit=gaussian, solver=qr, scale=more, trs=levenberg-marquardt+accel, iters=1000, xtol=1e-05,gtol=6.05545e-06, ftol=1e-05"
"LM solver (Gaussian): Failed after 0ms iters=1 [attempt=1] with status=11 [exceeded max number of iterations] and info=27 [iteration is not making progress towards solution]"
"LM Solver (Gaussian): Guess perturbation=0.9, A=30.6, B=4.07274, C=3.68921, D=0.18364, E=0, F=0.18364, G=560.865"
"Starting LM solver, fit=gaussian, solver=qr, scale=more, trs=levenberg-marquardt+accel, iters=1000, xtol=1e-05,gtol=6.05545e-06, ftol=1e-05"
"LM Solver (Gaussian): Solution found after 0ms 2 iters (small step size). A=23.83, B=4.37955, C=3.93824, D=0.17097, E=0.0218364, F=0.189536, G=623.585"
Star 0 R2= 0.518025 x= 534.525 vs 534.38 y= 753.099 vs 752.938 HFR= 1.8431 FWHM= 3.98148 Background= 623.183 vs 623.585 Peak= 34 vs 23.83
"LM Solver (Gaussian): Guess perturbation=1, A=28, B=4.33929, C=4.95148, D=0.22881, E=0, F=0.22881, G=623.183"
"Starting LM solver, fit=gaussian, solver=qr, scale=more, trs=levenberg-marquardt+accel, iters=1000, xtol=1e-05,gtol=6.05545e-06, ftol=1e-05"
"LM Solver (Gaussian): Solution found after 0ms 6 iters (small step size). A=28.6494, B=4.4163, C=4.72296, D=0.269479, E=0.0108649, F=0.195333, G=622.926"
Star 1 R2= 0.446987 x= 634.339 vs 634.416 y= 688.951 vs 688.723 HFR= 1.7405 FWHM= 3.2167 Background= 623.183 vs 622.926 Peak= 28 vs 28.6494
"LM Solver (Gaussian): Guess perturbation=1, A=23, B=3.784, C=3.97961, D=0.231974, E=0, F=0.231974, G=623.183"
"Starting LM solver, fit=gaussian, solver=qr, scale=more, trs=levenberg-marquardt+accel, iters=1000, xtol=1e-05,gtol=6.05545e-06, ftol=1e-05"
"LM Solver (Gaussian): Solution found after 0ms 2 iters (small step size). A=20.7365, B=3.80718, C=3.95927, D=0.217813, E=0.014975, F=0.188599, G=623.144"
Star 2 R2= 0.394481 x= 830.784 vs 830.807 y= 616.98 vs 616.959 HFR= 1.7286 FWHM= 3.60788 Background= 623.183 vs 623.144 Peak= 23 vs 20.7365
"LM Solver (Gaussian): Guess perturbation=1, A=27, B=4.01294, C=4.00175, D=0.277424, E=0, F=0.277424, G=623.183"
"Starting LM solver, fit=gaussian, solver=qr, scale=more, trs=levenberg-marquardt+accel, iters=1000, xtol=1e-05,gtol=6.05545e-06, ftol=1e-05"
"LM Solver (Gaussian): Solution found after 0ms 11 iters (small step size). A=24.4917, B=4.09605, C=3.78682, D=0.208198, E=0.0200591, F=0.23132, G=624.927"
Star 3 R2= 0.280434 x= 784.013 vs 784.096 y= 9.00175 vs 8.78682 HFR= 1.58067 FWHM= 3.60851 Background= 623.183 vs 624.927 Peak= 27 vs 24.4917
"LM Solver (Gaussian): Guess perturbation=1, A=19, B=3.98627, C=3.75603, D=0.338858, E=0, F=0.338858, G=623.183"
"Starting LM solver, fit=gaussian, solver=qr, scale=more, trs=levenberg-marquardt+accel, iters=1000, xtol=1e-05,gtol=6.05545e-06, ftol=1e-05"
"LM Solver (Gaussian): Solution found after 0ms 4 iters (small step size). A=19.3925, B=3.83587, C=3.55447, D=0.236587, E=-0.0765049, F=0.260118, G=623.727"
Star 4 R2= 0.639566 x= 538.986 vs 538.836 y= 208.756 vs 208.554 HFR= 1.43022 FWHM= 3.48041 Background= 623.183 vs 623.727 Peak= 19 vs 19.3925
Original Stars= 5 Processed= 5 Solved= 5 R2 min/max/median= 0.280434 / 0.639566 / 0.446987 FWHM= 3.57899 Weight= 23.475
Focus newFITS # 1 : Current HFR 1.66462 Num stars 5
"Linear: solution @ 313789 = 3.57899 (expected 3.32139) delta=0.257608 or 6.04734 sigma worse than expected. POOR result. If this happens repeatedly it may be a sign of poor backlash compensation."
"Linear: points=[(352500, 8.96653, 0.207497), (345000, 7.12019, 1.42183), (337500, 6.06117, 2.10879), (330000, 5.07755, 2.20561), (322500, 3.72227, 11.3833), (315000, 3.32139, 0.916498), (307500, 3.80415, 4.21213), (300000, 4.40165, 7.30139), (292500, 5.33991, 0.79437), (285000, 6.85037, 1.01413), (277500, 8.04483, 0.729416)];iterations=11;duration=67;solution=313789;value=3.57899;filter='';temperature=-127;focusalgorithm=3;backlash=2500;curvefit=1;useweights=1"
"Linear Curve Fit check passed R2=0.993049 focusR2Limit=0.95"
"Focus procedure completed after 12 iterations."
Autofocus values: position, 313789 , temperature, -127 , filter, "" , HFR, 1.66462 , altitude, 72.041
Stopping Focus

Thx, but I'm going to need the whole file. Can you attach it to a post to this thread? If its too big can you send it to me via some sort of file transfer process such as Google Drive?

Hi John,

In attach you'll find 2 logfiles, respectively with binning 3 x 3 (auto focus successful) and binning 1 x 1 (auto focus failed).

I've limited the screenshots to the curve of the failed autofocus with binning 1 x 1:



Thanks before hand to take a look at it...

Thanks for the logs.

I see you're using the All-Stars SEP profile. Is there a particular reason for this, as opposed to the Focus profile? Have you tried the Focus profile?

The All Stars profile by default (unless you've changed it) will try to pick up as many stars as it can, some will be very small and could be noise. When you bin 3x3 you'll counteract these very small stars, I suspect.

Looking at the FWHM Gaussian solving its working very well on the 3x3 binning (albeit on a smaller number of stars). Its doing OK (less well) on the bin 1x1, but its certainly solving enough stars to get a reasonable datapoint.

Can you post a screenshot of the All Stars SEP profile please. (If you click the button next to the All Stars label you'll bring up the editor with the parameters.)

Also which camera do you have?

Hi John !


Again thank you for your help !
The camera I'm using is a ZWO ASI 533MC at 0 °C.

Some new test graphs, switched from All Stars to Focus Default.


Binning 3 x 3 - Focus Default




Binning 1 x 1 - Focus Default




All Stars SEP profile:




Focus default profile: