Tektronix 071-0590-00 user manual Summary

BERT Technical Articles

Make a plot of BER versus Attenuation for SNR0 = 10−10 by starting at BER = 10−10 for Attenuation = 0, and draw a straight line parallel to the dashed lines, as in Figure 7. We will use a smaller stress than in Example 3. An optical attenuation of 1 dB (equivalent to 2 dB of electrical attenuation) would raise the

BER to 2×10−7, or a stressed error rate of rs = 51840000 × 2×10−7 = 10.37 errors per second. So for 95% confidence, we stress the system with 1 dB of optical attenuation and test it for Ts = 3 / rs = 0.289 seconds. If there are no errors in that time, then we have 95% confidence that the stressed error rate is less than 2×10−7. By extrapolation, we are 95% confident the unstressed BER is less than 10−10. In this case stressing has reduced the test time by a factor of 2000.

Summary

The inaccuracy of an error rate measurement is not due to any fault of the measurement equipment. Rather the inaccuracy come from incomplete knowledge when too few errors are observed. Therefore the accuracy can be improved only by taking more time to do the measurement. A good rule of thumb is that the inaccuracy is the reciprocal of the square root of the number of errors counted. For example, the inaccuracy is 5% if 400 errors are counted.

Determining that an error rate is below some limit takes less time than determining the error rate accurately. If the system under test is error-free of a period T, then we can say the error rate is probably below some rate r. An easy rule to remember is that for 95% confidence that the error rate is less than r, the system must test error-free for T = 3 / r.

The time to measure a low error rate can be reduced by stressing the system with a signal attenuator. Measure the higher error rate with stress quickly, and extrapolate to the unstressed error rate. (This method assumes Gaussian noise and insignificant noise present before the attenuator.) Use the special graph paper provided at the end of this article help in the extrapolation; you are free to copy it. Take care to observe the guidelines for stressing in order to get good results.

It is worth taking the time to understand the properties of error rates. A small investment of time here can pay off in big savings in reduced test time.

================

GB1400 User Manual B-41