The internal gear train (3 gears) has all of the three gears with the same number of teeth.
I have locked the last gear in place for testing and have found that I have about 3 degrees of slop at the turret. That's about 1.5 degrees of slop at each gear mesh point (2 mesh points in a three gear train.
So, the stepper will locate the turret numerically accurately, but the turret can be +- 1.5 degrees when it stops. depending on how fast the turret decelerates, it can be + 1.5 degrees or -1.5 degrees. if the turret is not located +- .5 degrees of the indexing location, it will not lower down onto the locating dogs correctly.
My location problem turned out to be not having the A axis homed. now, the locating accuracy problem is purely mechanical backlash, as the stepper is going exactly where it should. I am positive that the issue (currently) is that even though mach 3 and the stepper locate the first gear in the train accurately, after 2 gear mesh points, enough slop is introduced to have the turret off just enough to only lower down onto the dogs about half of the time.
if I make a detent arm, the detent arm will snap into each of the detent locations (8 total) with accuracy which is entirely dependant upon how much accuracy the detent locations are machined to. This assumes that the detent arm does not itself have any slop (backlash)
Fortunately, cutting detent locations on my mill will be very accurate as I have a high resolution Peiseiler 4th axis. The detent arm can be made to have almost zero slop if I make it right. I plan on having the detent arm made with the spring tension being adjustable. I think the detent slots should be V shaped for positive location locking.
I hope all this makes sense.