In article , Jim Wilkins
wrote:

"Joe Gwinn" wrote in message
...

Following up to my own post, I performed the rods measurement using
five different test rods.

Details: Clamped the rest upside down in a vice, so the base is on
top. Cleaned machined surfaces with acetone and a razor blade.
Clamped a magnetic base to the flat, with arm and digital indicator
over the Vee groove, with indicator probe moving vertically and more or
less perpendicular to the plane of the flat. Installed the 0.500" rod
and zeroed the indicator. Took indicator measurements on five rods,
being 5/8, 1/2, 13/32/ 23/64, and 5/16 inch diameter. Fitted a line
to
the data, and also used a formula on measuring tapers using discs of
various diameters from Machinery's Handbook (27th edition, page 715,
in Angles and Tapers).

The consensus answer is 79.7 degrees included angle; one assumes that
the target was 80 degrees. One would assume that people don't sweat
getting Vee-groove angles exact on steady rests.

I'm not criticizing you, but here is a good place to mention this:

I was trained in analytical chemistry, where the goal is to find and
eliminate error sources and make one very accurate measurement with
the limited sample and time available. Later as an electronic test
engineer I practiced solving real-world examples, guided by a former
Keithley Instruments genius Ph.D. of the ponytail, sandals and VW bus
persuasion.

Two useful principles I picked up were that scattered data warns of a
procedural error and shouldn't be blindly averaged in, and to keep
track of the accuracy limit of each measurement. We converted accuracy
limits into parts per thousand to make correlating different types of
measurement easier.

The plotted data is quite straight. The standard deviation of errors
(deviations from the fitted line) for the five data points is 0.000,431
inches, which is reasonable, as the resolution of the indicator is
0.000,5", and only 5 measurements were made.

If one assumes a uniform distribution of errors over a 0.0005 inch
range, the expected standard deviation if one made many measurements
would be Sqrt[(0.0005^2)/12]= 0.0001443 inches, so the measured fitting
error is about three times that from indicator resolution alone. This
is most likely due to the accumulation of small errors, such as the
indicator axis not exactly perpendicular to the plane of the flat on
the rest base, and the low number of measurements.

But more to the point, it isn't obvious that added accuracy will tell
us anything more, as the ultimate intent is to tell which manufacturer
made the lathe for this stead rest.


For example if you measure with a 1" dial indicator the best you can
achieve is one part per thousand, half a tick at either end. If your
readings range over 0.1" your accuracy is at best ten parts per
thousand, or 1%, and the final result shouldn't have more digits than
the least accurate measurement validates.

The digital indicator has a rectangular LCD as its display, with
0.000,5 as the least significant digit. It is a Mitutoyo code 575-123
"absolute digital indicator", bought used.


A good sanity check of your height vs width measurements is to see if
they converge to zero at the bottom.

It won't go to zero because the indicator was arbitrarily zeroed on the
1/2 inch rod. The point of the V is not accessible, because the
V-groove has the typical slotted bottom where the Vee point would have
been.

PrecisionmachinisT makes the point that it's better to have the
included angle be slightly too small, as the jaws will spread slightly
and yield a firm fit, versus the wobble if the V is slightly too
obtuse, so the 0.3 degree delta may have been intentional. The old
lathe makes made these kinds of tweaks all the time.

Joe Gwinn