At "Imaging The Sky" in 2001, the same conference at which I gave my focus lecture in 1999, Steve Brady and Larry Weber presented their paper on a new automated focus method called FOCUSMAX. I will give a brief explanation below. For more details and the software itself (it is presently freeware) go to the FOCUSMAX web site.
Perhaps the biggest problem with automated focusing is finding a parameter that the software can examine to determine if good focus is present. Many have been tried such as peak value and FWHM, which we have already discussed. FOCUSMAX uses a new parameter called "Half Flux Diameter" (HFD) This is similar to FWHM except that instead of representing the star width at half of the peak value, it represents the diameter inside which half of the star's energy is contained. This measurement has proven much more stable and seeing resistant and has formed the basis for the FOCUSMAX method. A screen shot showing FOCUSMAX's display of HFD is shown below.
FOCUSMAX works by characterizing the behavior of your optical system as it goes in and out of focus. Prior to actually using FOCUSMAX for a final focus, you perform a number of "V Curve" runs using the software to generate this characterization. FOCUSMAX will run your system well past focus to a very large (40 pixels or so) HFD. It will then step toward focus and then past focus in increments, measuring focuser position and HFD at each position. This data is plotted on a curve called (for obvious reasons), a "V Curve" as shown below. HFD is on the vertical axis and focuser position on the horizontal axis.
The slope of the right and left lines is calculated using a best fit algorithm (blue lines). Once an accurate slope on each side of focus has been determined, focus is done by commanding the focuser to go away from focus by a bit and to accurately measure the HFD at that point. Then using the slope previously calculated, it can determine the point of best focus and move the system to perfect focus.
I have done limited testing so far but I am very encouraged by the results. Testing on my RoboFocus equipped Takahashi has shown focus to be accurate and repeatable. It should be noted, however, that accurate focus is also dependent on your hardware - no amount of software will allow cheap hardware to focus well. I suspect, however, that we will all focus this way in the very near future!
MaxImCCD version 3 uses a similar but less detailed version of this focus system. Other focusers besides RoboFocus are also supported such as the Optec TCF and RC Optical Servo Focuser.