The rotating accuracy of a machine tool spindle directly affects the r
oundness of machined parts. Commonly, a precision arbor and one or two
probes are used to inspect the spindle axis error motion. When the sp
indle error motion is in the same order of magnitude as the accuracy o
f the reference arbor, it is desirable to separate the roundness error
of the reference arbor from the spindle error. One of the methods use
d is the three-probe method. This paper presents an exact geometric mo
del and error analysis for the conventional three-probe method. The ex
act model is used to show that there is an approximation error in the
commonly used governing equations of the three-probe method. To reduce
inaccuracy in the converted axis motion and arbor contour, the refere
nce arbor accuracy should be at least ten times better than that of th
e axis motion. It is also shown that the mounting error of the probes
should be less than one-fiftieth of the size of the axis motion and th
e arbor size. The exact geometric model developed in this paper can al
so be extended to analyze the accuracy of other spindle inspection or
roundness measurement methods. (C) 1997 Elsevier Science Inc.