Two stage autofocus

Sequence Generator Feature Requests
1

Description

Here is an updated question with some more background:

When making a HFR profile over the focusing range, the profile will normally look like a ‘saddle diagram’. Best focus is in the middle of the saddle. HFR does not make much sense outside of the saddle when stars get diluted.

There are three focusing situations, simplified:

a) Good observing conditions which gives well defined saddle diagrams when focuser scans HFR throughout the relevant parts of the focusing area. Autofocus works very well, and covers well variety of situations which occur especially with non parfocal equipment. This gives variations of saddle diagrams. I am imaging in a city and will get HFR in good conditions down to 1.5-1.8.

b) Poor observing conditions but good enough so that HFR diagram still resembles a saddle diagram. Focusing under such conditions is unstable. In this case, notes of focuser position from good seeing conditions help. The saddle diagram is in this case the best indication of the best focuser position. I would get HFR down to 1.9-2.3.

c) Conditions are so poor that you can’t register the saddle. Not point of imaging.

It is easy to register saddle diagram with the new autofocuser algoritm. The autofocuser really works impressive.

I don’t have a permanent site nor parfocal filters, I also change camera often, so there are a lot focusing profiles.

Since I work mostly under case b), autofocus handles well many situations but gets often stuck or can not come to conclusion.

It would be nice if autofocuser would, in a fashion similar to calibration which done by PHD2, make a saddle diagram (HFR profile) of the equipment configuration.

This would have several benefits. It will help to determin when the fine focus adjustment meningfull. It will help to run in poorer conditions based on the settings from good conditions. It will also provide another means to determine how good seeing conditions are.

So the suggestion is to add ability to store HFR profiles to autofocuser. Profile can be registered by step size, for example 200. The fine focusing can then be done based on position of the saddle in the HFR profile.
/Janez

GRAPH_AFID-002_POS-15236_HFR-370_Q-09-NoFit

/Janez

Link to Logs

Useful Info

OS: Microsoft Windows 10 Pro
Ver: 3.1.0.429
.NET: 4.8

2

If the data points between 15536 and 15836 are reasonably accurate, then you need to make the autofocus run stay in that area.
Do the following:

  1. set number of steps to 9
  2. set step increment to 35

You will get much better results with those settings. Both 200 and 10 will give you disastrous results.
To be sure this is really the best advise, I would need to see one of the plots that you say are working well for you.

Specifically to address your request for a two stage focus with two different sets of parameters: that would be guaranteed to give you worse results than the approach that gives the best results. The best result approach is to find the optimum values for step count and step increment. That optimum set of values is one that produces Max HFR / Min HFR ratio somewhere between 2.5 and 4. Usually 9 steps is the best trade off between good results and unnecessarily long run times. Just use 9 for step count and then find your best step increment value.

3

With my Celestron 9.25 Edge HD SCT, autofocusing fails if seeing and transparency are on the poor side. If I get good initial focus, then I record that number. If conditions deteriorate and autofocus fails, then I either go to the value I logged or give up.

One request for the HFR calculation is to improve it to account for scopes with central obstructions. Perhaps a telescope parameter could be added for an OTA with an obstruction that cause out of focus stars to become donuts. The HFR calculation is not robust enough to handle this, ergo the dog ears in the V curves when step sizes are too large.

Mark W