It's impossible to give a general answer. Some people have shown evidence the primary is the limiting element in their particular scope. Others have shown the primary baffle is clipping in their particular scope. Others may be clipping at the rear port in their particular scope. It also seems quite likely the design may have changed over the years, so keep that in mind.
Not trying to dodge the question, though. The math when the near/far measurements are equal -- i.e., tilt estimated at 0 -- is this:
uncertainty_pupil^2 = uncertainty_shadow^2 + uncertainty_tilt^2 * f^2 * (f/z-1)^2
uncertainty_shadow = 0.29 when the measurements are made to +- 0.5 mm
uncertainty_tilt = 133 urad (0.5 amin), when near/far measurements are made to +- 0.5 mm, and near/far separated by a focal length (1540 mm)
The hard part is, what value to use for z (distance from aperture stop to focus)? Assuming the rear baffle, perhaps z ~ 160 mm? (Just spitballing here, if you know more accurately, please advise.)
This gives the uncertainty in the pupil estimate when linear measurements are +- 0.5 mm and identical near/far measurements separated by one focal length (f=1540 mm) and assuming 160 mm from limiting element to focus:
uncertainty_pupil = +- 6 mm
To know any more precisely, you're going to have to do more work to eliminate elements from consideration. If you've verified -- somehow -- that the limiting element is not the primary baffle or anything at the secondary, you can perhaps consider the primary to be the limiting factor. (Reasonable when meniscus and primary are the same size.) Spitballing the primary location from focus at 920 mm from focus (again, please advise if you know a more accurate value), gives a pupil uncertainty of +- 1 mm.
I think much of the discussion over the years has been useful in a pedagogical sense. (I especially enjoyed reading the circa 2012/2013 threads as people's understanding of optical principles seemed to solidify. :-) I hope it is clear that while the pupil location (or even limiting element) does NOT figure into the BELT-estimate of pupil size (as Glenn has maintained), it most definitely does impact the uncertainty thereof (as Frank has maintained).
Edited by dakinemaui, 25 April 2019 - 11:07 AM.