Jitter Test in Production for High Speed Serial Links

1. The new definitions of jitter In the last few years, more and more serial link standards adopted the concept of separating jitter into deterministic jitter (DJ), periodic jitter (PJ) and random jitter (RJ). The old concept of histogram based peak-to-peak jitter has been replaced by the concept of total jitter (TJ), which is associated with a certain bit-error-rate for the serial link (typically 10). This makes the jitter test more meaningful with a direct connection to system performance. It is much harder to perform a proper jitter test directly in production, as it is specified in terms of DJ, PJ, RJ and TJ, rather than a simple peak-to-peak or rms jitter measurement. It means we need to decompose the jitter measured into various types, each having its own budget on the spec sheet. The trend of separating jitter into different types also applies to the jitter tolerance test. To conduct a jitter tolerance test, jitter must be deliberately injected into the data stream in a controlled fashion. That means we need to mix a certain amount of DJ (including some PJ) with a certain amount of RJ to perform this test to the spec. [1] In this panel discussion, we need to distinguish jitter test solutions for production between full standard compliant tests and tests providing only partial coverage.