Fiber geometry specifications and its relation to measured fiber statistics.

Fiber quality requirements can be specified in a number of ways; the most typical is to state a required diameter and give a tolerance band to allow for manufacturing variations. Actual fiber diameter statistics have been shown to be close to Gaussian for a certain range of fiber lengths. These fiber statistics are discussed showing the bounds of stationarity and their relation to fiber specifications. Thus the fiber statistics in terms of standard deviation and sample length can be related to the specification. Not only variations within one fiber are important, but also from fiber to fiber. Thus, the effect of mean-to-mean differences in fibers and their influence in meeting fiber specification for given standard deviations are considered. In considering these factors, the following conclusions are made. There is a bandwidth which can be related to the fiber drawing process, and measurement of fiber diameter for the purpose of characterization must be made at a rate greater than this bandwidth to guarantee fiber quality. Since fiber drawing is a predominately random process, specifications should be written which relate to a statistical description of the process. If a tolerance band is used for the specification of fiber diameter, it ought to comprise two parts. One part will allow for the random fiber diameter variations. The second part must provide for the deviations of the mean diameter of each complete length from the nominal.