The stressed eye's crest factor
By Martin Rowe, Senior Technical Editor- June 1, 2010
Using an arbitrary waveform generator, you can apply low-probability, high-amplitude jitter at known points in a serial data stream. The problem of accurately measuring BER at 10–12 or 10–15 is that the test time required to properly evaluate the full effects of random events can range from 30 min to many days. High-amplitude jitter can occur at any time, so if you can apply jitter at sensitive points, you can cut measurement time and increase repeatability in your measurements. The CFE measurement technique can reduce test time to less than 1 min and is 100% deterministic in its ability to be reproduced from run to run.
To perform the BER measurement, apply a high-amplitude jitter signal to a bit transition where both the median ISI (intersymbol interference) and half the amplitude of the sinusoidal jitter specified in a serial-data standard occur. If the receiver fails to correctly identify the bit (BER > 10–12), then you can repeat the test and verify if the error is indeed caused by the receiver. You'll know if the failure was caused by ISI, periodic jitter, or random jitter.
Download Ransom Stephens and John Calvin's paper and slides used at DesignCon 2010.