Background: Development of the ST-4 auto-guider was the answer to correct minute errors in equatorial tracking due to any number of deficiencies. Image exposure and transfer times were too long on the primary imaging CCD, so, corrections created by use of a second CCD camera taking rather short exposures seemed appropriate. I am proposing this might not be necessary in this day and age.
Just as today's Polar alignment routine uses an accumulated correction model for proper slewing, guiding could be enhanced using a similar model designed for tracking. The proposed advantage of this system would be to eliminate use of a secondary camera, off axis items, or guide cameras.
The premise is based upon the idea that a perfect polar alignment, coupled with an exact tracking rate, and, precise gearing will maintain a centered image target for the period of 1 exposure. It further requires the primary ccd image field must contain at least 1 guide star. Ekos use of astrometry for slew to target will assure target centering and adequate guiding stars.
This methodology will not correct for transient events such as gearing defects, wind, or temporary atmospheric distortions.
1: KStars Ekos when set to take an imaging sequence, will automatically take a calibration image of a short but detectable length. Automatic astrometry imaging will have already handled proper position identification, target centering, and slewing corrections. Within that field, a guide star will be identified, either, by user or automatically. A second calibration image will be taken with tracking off. Offset for stellar drift will be calculated. A third image taken with an offset in declination, and again, appropriate offsets will be calculated to determine the needed motions for maintaining center of the desired target.
2: From the above data, (just as Lin_guider and others now do) appropriate guiding vector directions and amounts will be established. Appropriate guiding pulsing will be applied, and Ekos will spiral up the exposure time to take the next image. Corrections to the guiding model will be applied by comparing the somewhat longer exposure with the previous one.
3: Each subsequent longer image will be compared and corrections applied. The theoretical hope is, at the desired full exposure time, a single correction vector will produce adequate tracking to image the desired target.
This is a little similar to the idea of using phase detection for "virtual guiding" ot alignment and stacking of single frames. We have started implementation of such procedure but it is a bit complicated and it is in limbo right now. I plan to return to it. The idea is to just correct alignment of frames instead of moving the mount.
My understanding is that it was just regular guiding, but you don't need to select a star, but rather use the complete image. The "phase shift" would give you the offsets that you need to use to guide the mount.
I said "a little" . You are right with respect to the mode we were trying to implement - it was just different source of the guiding signal. You can also use this method for post-processing of large number of short expositions - to align and stack them together. That was the mode I was thinking about. But now I think that I was wrong about similarities with proposed approach. Actually I do not think the proposed "build a movement model" approach will work. The deviations of the mount movement are (apart from periodic errors) essentially random - and you cannot model randomness - by definition...
Interesting discussion starting here. I would not recommend using the entire image because of possible frame rotation causing errors. The automatic guide star should be a single star selected as close to the image center as possible and with adequate separation from other interferences.
If guiding errors were totally random during an entire exposure, the image would not show star trailing. Instead, blooming of the star would occur. It is because tracking errors are NOT random that guiding works.
Random errors such as winds pushing the telescope during the exposure can not be corrected by this method, but would be caught and corrected by a guider camera.