View Issue Details

IDProjectCategoryView StatusLast Update
0001882CCdcielGeneralpublic18-03-17 18:58
ReporterThomas Jaeger Assigned ToPatrick Chevalley  
Status resolvedResolutionfixed 
Product Version0.9 
Target Version1.0 
Summary0001882: Suggestions for Autofocus
DescriptionHello Patrick and Han, I have some suggestions to improve the usability of autofocus for the casual user. Most of them apply to "Dynamic" method. I must apologize for my simple pencil drawings.

1. To use V-Curve and Dynamic there must be filled out HFD for "Autofocus tolerance" and "Near focus". This can be too complicated or missleading for most ordinary users. Especially if CCDCiel becomes more adapted, which will happen, sure.
How to messure, let the user simply record one value. Start with a focused star. Move the focuser that the star has 3 times the HFD and read how many ticks (from the focuser) it takes. It is n2-n1 in my drawing.

2. Represent "Dynamic" autofocus graphically at runtime. Bad influence from backslash, seeing, misconfigurion and so on is better shown in a graphic than in logfiles. Start with a dynamic curve (simplified with 5 points in my drawing), if the number of points is not sufficient automaticly expand it with more points (right). With the knowledge of the above (1), nearfocus and tolerance can be optained.
Even using V-Curve autofocus, a graphic at runtime would be a help.

3. Using more stars for autofocus instead of serveral images of one star. To speed up, 2x2 bin is fine for that in most cases.
With the new "Measure image" in the Star profile tool. Autofocus can messure a better HFD value using more stars and is more robust against hot pixels or bright nebula patches. Make a image overlay which stars are suitable.

4. For some telescope optics it is better to avoid messuring on the edge of the field. A circular border is too complex to manage, a rectangle as a border would be fine.

5. Handling of focuser backslash is not documentated, or I have not found it.

6. How is a star selected for unattended autofocus?

Documentation said:
You can check “Stay in place for autofocus” if you are sure that the center of the image field contain a suitable star for auto-focus.
If unchecked or if the in place focus failed, the telescope is first moved to a focus star of the magnitude defined in the auto-focus preference, and returned to the target after the operation.

A overlay with "all good stars" can help.

TagsNo tags attached.


Thomas Jaeger

18-01-18 16:07


CCDCiel - using more stars.jpg (593,587 bytes)   
CCDCiel - using more stars.jpg (593,587 bytes)   

Patrick Chevalley

18-01-21 15:48

administrator   ~0004395

Some king of auto-learning of the focuser parameters can be very useful but this is also difficult to implement because of the width range of available hardware.
For example n2-n1 can be anything from 20 to 1000 depending the focuser and controller model. Maybe start with small increment and increase until some HFD difference is measurable.

OK to show a graphic when running autofocus.
Yes I can increase the number of point in the case of your drawing, but not if the missing points are on the left.

At the moment this "Measure image" (now "Image inspection") function work reliably only for focused image. I make a test with defocused image and find the median HFD value very unreliable. This require more work to select only the stars with enough signal when defocused.

At the moment a star is automatically selected only if it is in a circular area of a diameter of half the image height. This select only stars near the center of the image.

The program use a backlash compensation for focuser that can only move by a relative movement but not to an absolute position. In this case you can set the backlash value that is added to the movement when the direction is reversed.
This not apply to absolute position focuser (as your) and any backlash must be compensated by the driver or focuser itself because it make no sens for a program to ask for an absolute position + backlash.

The same as after moving to the focus star, except we not move, and a bright enough star must be present.
A star is searched in the area indicated in 4. , the brightest pixel group is selected, a check is made this is not a double star and the SNR is high enough.
The best is to experiment on stars of different magnitude with the different exposure time and filter factor you can set. Then it is easy to check if there is a star of this magnitude near the center of the planned image.

Thomas Jaeger

18-01-21 16:24

reporter   ~0004399

Thanks for detailed answer, give me time to rethink it.

Thomas Jaeger

18-01-22 17:39

reporter   ~0004402

Hello Patrick and Han,

1. I did not thought on auto-learning. I suggest that the user should defocus a star eg 3-times and wrote down the ticks the focuser need. This would be very robust.

2. Yes why not, just move the focuser to that point? Maybe there are understanding problems on my own.

3. Ok I understand. In any case do show the user what CCDCiel does. A graphical representation is always good. The user can see, if there are hotpixels oder nebula patches were regcognized as stars.

4. If this a fixed parameter, please put it into documentation and/or put a circle line into the image at this process stage.

5. I do not agree with your statement, because the driver cannot feel or messure its own backlash. Even so, my moonlight focuser (clone) lacks possibility to do so.

Many Greetings,
by the way, I do not expect that you answer every point. If it is good, than close this thread.

Patrick Chevalley

18-02-15 22:48

administrator   ~0004446

This last days I implemented a few change on this subject but I am still waiting for Sourceforge to reactivate the source repository after their datacenter maintenance so I can build a new version.

The good point of this delay is this let me the time to update the documentation, probably the first time I do the doc before to commit the code :)

1. A new "Focuser calibration" tool set almost automatically all the autofocus parameters:

2. With the new calibration and backlash procedure it is probably no more necessary to add points because of wrong centering.
I add a graphic in the "star profile" tool that show the progress during autofocus:

3. To avoid hot pixel you can set a "Bad pixel map" from the File menu.

4. Yes the star detection area is fixed, it correspond to the square that circumscribe the external circle of the bull eye, so a star in the bull eye is OK.
Added to the documentation:

5. OK, you are probably not alone in this case. I change my backlash procedure so it can also work with absolute position focuser. Now you can indicate a number of steps for the compensation and the direction the focuser always finish to move:

Patrick Chevalley

18-02-16 17:21

administrator   ~0004450

Most of the change are in revision 792:

I am currently building version 0.9.24 that include all of this.

Patrick Chevalley

18-03-17 18:58

administrator   ~0004564


A new version 0.9.30 is now available.

This add the capability to use all the stars in the image center to measure the focus HFD, as in your "CCDCiel - using more stars.jpg" drawing.
This work best with the Dynamic method when you select "Stay in place", please look at the new options in the Preferences.

Please open a new issue on this subject if you have any problem or question.

Issue History

Date Modified Username Field Change
18-01-18 16:05 Thomas Jaeger New Issue
18-01-18 16:07 Thomas Jaeger File Added: CCDCiel - graph dynamic autofocus.jpg
18-01-18 16:07 Thomas Jaeger File Added: CCDCiel - settings for dynamic autofocus.jpg
18-01-18 16:07 Thomas Jaeger File Added: CCDCiel - using more stars.jpg
18-01-21 15:48 Patrick Chevalley Assigned To => Patrick Chevalley
18-01-21 15:48 Patrick Chevalley Status new => assigned
18-01-21 15:48 Patrick Chevalley Target Version => 1.0
18-01-21 15:48 Patrick Chevalley Note Added: 0004395
18-01-21 16:24 Thomas Jaeger Note Added: 0004399
18-01-22 17:39 Thomas Jaeger Note Added: 0004402
18-02-15 22:48 Patrick Chevalley Note Added: 0004446
18-02-16 17:21 Patrick Chevalley Status assigned => resolved
18-02-16 17:21 Patrick Chevalley Resolution open => fixed
18-02-16 17:21 Patrick Chevalley Note Added: 0004450
18-03-17 18:58 Patrick Chevalley Note Added: 0004564