I think the reasons for reduction are as follows:
#1, Camera can get a 20mm x 20mm chip in view, so with a larger chip, you have less resolution using a 200 step motor. Micro stepping without inertial load on stepper can be marginal at best.
#2, Run out. Steppers are not known for their shaft tolerances / run out. They are designed to be a driver, not a precision bearing assembly Hence the typical use of flexible couplings. Even a small amount of run out at the motor shaft translates into large problems once a nozzle coupling is added to the end.
#3, Compact assembly height. The TMV920 nozzle holder assembly is within (<5mm) of being vertically even with the front of the motor shaft, It shortens the length of the typical assembly by almost an 20 - 25mm. Shorter assembly translates to smaller run out errors if present.
#4, No need for hollow shaft steppers. Part of the problem with using small NEMA8 steppers for rotation is the limited torque produced, and the drag presented by the rotary coupling used for vacuum connection. This mitigates the problem by increasing output torque using the belt drive reducer, while increasing resolution as a byproduct.
Just my thoughts.
It is offset with a belt drive, with about a 3:1 reduction for more accuracy in rotational angles.
That is very interesting. I presumed more or less the same design, wrongly.
I wonder if that is really necessary. Without the belt we have 200*8 steps. For a 10mm chip that translates into 0.04mm shift if the angle is off by one full step. But theoretically it should not be off by more than 1/2 step if visual system works 100% reliably. If it does not then higher resolution for angles does not really help anything at all.
I'm just wondering if that makes any difference in some circumstances.