What's new
What's new

Calibrating a penta-prism with only one autocollimator and mirror

Mark Rand

Diamond
Joined
Jul 9, 2007
Location
UK Rugby Warwickshire
I know the recommended NIST/NPL way of calibrating a pentaprism is to use two autocollimators and two mirrors like in this link.

Since I haven't got two autocollimators, is it reasonable to sit one autocollimator and the pentaprism on a surface table with the pentaprism directing the light up to a mirror mounted above it. Adjusting the mirror to get a datum reading. Then rotating the pentaprism and autocollimator by 180° horizontally. Then taking the indicated error of the mirror angle as twice the error of the pentaprism? This does assume that any adjustable feet on the autocollimator don't move during the process, of course.


Can anyone shoot this down in flames or offer a simpler method that I've missed?
 
That may work

But moving so much of the test setup does introduce uncertainty.

You might want to do an error analysis and calculate the effect of "tenths" on any data values you would otherwise attribute to defects in the optical square.
 
It depends on whether or not you want to be able to certify your measurements to NIST.

If you do, then you have to send it out to a cal lab. You can't certify it yourself.

- Leigh
 
It depends on whether or not you want to be able to certify your measurements to NIST.

If you do, then you have to send it out to a cal lab. You can't certify it yourself.

- Leigh


Nah! I just want to get a 0.5' one off EBay (missed a 90° 0' 0'' one by £20 today) and calibrate it a bit better to check the knee and top of a milling machine:crazy:
 
You can't calibrate something to an accuracy tighter than its specs. That's a fundamental law of metrology.

If you could, then everybody would buy Lego blocks and certify them as gage blocks. :eek:

- Leigh
 
Leigh

Considering the intent and usage, I would bet that just knowing the magnitude of any error would suffice. Calibration may be too strong a word ;-)
 
It's not the magnitude of the error that's of concern, it's the stability of the error over varying conditions, particularly temperature.

Also, the standard used for calibration must have a limit of error substantially less than the item being calibrated.

- Leigh
 
A Penta prism is a pretty stabile piece. All internal reflectances. About the only thing to consider is to let the piece sit at "room temp" a few hours before you get serious.

Angle deviation or pyramid errors are not likely to be introduced at any temperature that might be expected in an optical set up. Even a warm hand touch would very likely neutralize in a few minutes, glass not being the best conductor of heat and all. Now the fixtures that hold the optical square is another matter ;-)!

In the case described, the standard is the retracing of an optical path, and the refinement of preception and the precision of the collimator. (add the huge uncertainty of indexing the works 180)

Again, Not "calibration" but rather characterization of the given sample used to again characterize a machine tool. Calibration may be too strong a word. How about "Measurement with degree of certainty"? ( Insert error bars as needed)

Not really rocket surgery, but way better than a carpenter's level ;-)
 
You can't calibrate something to an accuracy tighter than its specs. That's a fundamental law of metrology.

If you could, then everybody would buy Lego blocks and certify them as gage blocks. :eek:

- Leigh


Actually, you can and it's done as a matter of course in metrology. When you make a stack of gauge blocks you add their calibrated lengths, not their nominal, labelled lengths if you are working to the greatest accuracy. Similarly, when using pressure transducers, thermoucouples, PRTs, flowmeters, CTs, VTs and Watt-hour meters for performance tests on steam turbines, we always corrected the indicated readings for the calibrated errors of the devices. Not doing so would increase the uncertainty to an unacceptable level.

In this case, the grade of the instrument is ±.5', which is equivalent to about 1.5thou in 10 inches. If I can quantify the error to better than ±.5', with an uncertainty of better than that, then I can take that measured error into account and work to the uncertainty of my calibration methods, not the uncertainty of the un-corrected device.


I was looking for gotchas on the technique of using a single autocollimator and mirror instead of two pairs, rather than the entire concept of calibration and applying corrections to devices.:D


:cheers:


For CalG, I think the uncertainty of rotating the pentaprism should be a second order effect, due to the nature of pentaprisms compared to 45° mirrors. Most of the uncertainty is going to come from moving the autocollimater around the surface table. I hope I can get that bit a lot better than the 1.5thou/10" of the uncorrected prism though..


Regards all for the comments.
 
When using the certified values, you must also add the uncertainties thereof. Each block has its own uncertainties.

The certified values are only valid at the temperature used during the certification.

- Leigh
 
Mark --

I think you idea is darn clever, and I think its "noise floor" can be down in the single-digit arcseconds if 1) you can be sure that the surface table doesn't move if the autocollimation mirror is independently supported, or 2) the autocollimation mirror is supported directly by the surface table, and 3) you can "zero" your autocollimator when its optical axis is parallel to the surface table. If the surface table and pentaprism are perfect, but the autocollimator zero isn't parallel to the surface table, turning the autocollimator and pentaprism on the table will shift the through-pentaprism return from the autocollimation mirror by twice the non-parallelism UNLESS you can deliberately induce an equal-and-opposite autocollimator parallelism error into your autocollimator zero.

Setting the autocollimator zero to parallel the surface table can be done in three ways (that come to mind): 1) zero the autocollimator on a mirror that is known to have its mirror surface perpendicular to its mounting surface, 2) zero the autocollimator at the average return from the two sides of a plano-parallel mirror, or 3) IF the autocollimator has a cylindrically-ground barrel -- preferrably meeting Aircraft Industries Association specifications -- set the zero parallel to the barrel cylinder (by reversing it in its mounting Vs) and then setting the barrel parallel to the surface table using a very precise dial indicator.

Best wishes with your project, and consider "publishing" your results. There's no reason that the "Moody method" of surface table calibration shouldn't have a sidekick called the "Rand method" of penta-prism calibration.

John
 








 
Back
Top