Techniques for accelerated measurement of low bit error rates in computer data links

Measurements of very low bit-error rates for an optical packet switching node

“…The sinusoidal interference method has been advocated as an efficient and accurate methodology for extrapolating very low BERs [7]- [9]. The theoretical basis of this technique is the canonical estimation of the BER, assuming a Gaussian noise distribution, given as the probability of error for each bit (1) where is the complimentary error function, and and are the signal intensities and noise levels of the "1" and "0" bits, respectively, and is the decision threshold.…”

“…2. Plot of the normalized error between the approximation (6) and the exact expression (4), using the constants c from Chrostowski [7], [8] (--) and the authors' constants (-).…”

“…This result is utilized in [7]- [9], but the values given for the constants , as given in Table I, may be adjusted. A hypergeometric approximation is given in [10], but this function is difficult to fit into a form similar to (7) for linear regression.…”

“…A hypergeometric approximation is given in [10], but this function is difficult to fit into a form similar to (7) for linear regression. Additionally, the values from Chrostowski [7], [8] are not optimal for typical BER levels. In order to minimize the root-mean-square (rms) error between (4) and (6) over the range of -factor values from 6 to 9 (BER from to ), which are typical for the measurements described below, we present a new set of constants for (7).…”

“…Additionally, the values from Chrostowski [7], [8] are not optimal for typical BER levels. In order to minimize the root-mean-square (rms) error between (4) and (6) over the range of -factor values from 6 to 9 (BER from to ), which are typical for the measurements described below, we present a new set of constants for (7). The constants were obtained from a second-order series expansion of (4) specifically optimized for -factor values in the range of interest.…”