I had him run the test again. Same results: the 50-second wedge measurement was great, but the 10-second wedge result was still off. And it was noticeably different from the last time. Again, the raw data showed normal scatter for the 50-second wedge, but terrible variation for the 10-second wedge. It just didn’t make sense that the technician could be “proficient” when measuring a 50-second wedge, but not for the 10-second wedge.
An autocollimator of this type is basically a telescope with an illuminated reticule. It projects an image of the reticule to a mirror, and the position of the reflected image of the reticule with respect to its initial position indicates the angle of tilt of the mirror. Placing an optical wedge between the autocollimator and the mirror shifts the beam a constant angle.
This particular autocollimator has a 10 arc-minute range. You measure the angle by lining up a movable crosshair on the reflected image of the reticule using a lead screw attached to a micrometer drum. One revolution of the micrometer drum amounts to a 30-second angle deviation.
As I was looking over the raw data readings from the two sets of tests, it struck me that in all cases the technician took his readings centered almost exactly at mid-scale on the autocollimator, right around the 5 arc-minute mark. This is a fairly common practice. The assumption is that these instruments are most accurate in the center of their range.
Mostly as an act of desperation, I asked him to re-measure the 10-second wedge with the scale centered around three minutes instead. When he did this, the measurement came out good -- great, in fact. His readings showed normal scatter, and the overall result was within two tenths of a second of the value we measured at the primary standards lab.
I didn’t have access to his instrument to test it myself, but my theory is that after forty years of use, with the operators preferentially setting up their measurements at the center of the scale, the lead screw had become worn and sloppy in this area. The angle deviation of the 50-second wedge was large enough to be outside the worn-out region of the lead screw, but the 10-second wedge deviation was right in the worst part of it.
I recommended that they donate the instrument to a museum, but failing that, not to use it around the 5-minute mark on the scale.
This entry was submitted by Steven K. Smith and edited by Rob Spiegel.
Steven K. Smith is a musician, a poet, an artist, and a mechanical engineer who has worked for the Air Force Metrology and Calibration Lab for more than thirty years.
Tell us your experience in solving a knotty engineering problem. Send to Rob Spiegel for Sherlock Ohms.