Separating Random and Deterministic Jitters
The major test equipment manufacturers such as Tektronix and Agilent offer instruments or software that separate the various different jitter components. They also have application notes describing their algorithms in much more detail than I have space for here — if interested in seeing more, go to the Web site http://tektronix.jitter360.com/ or http://cp.literature.agilent.com/litweb/pdf/5988-9109EN.pdf.
If after reading these application notes, you are still hungry for more, conduct a Web search on the term ‘dual dirac’ and you will hit the mother lode of jitter analysis data. The mathematics involved in doing the separation of jitter can be quite complex, but the software available from the test and measurement vendors makes it easy — you just press the button.
Total jitter is an attempt to combine the deterministic jitter along with the random jitter so that you can arrive at a single number for jitter. Unfortunately as we discovered above, if you have random jitter, it is unbounded, so when you define the total jitter, you need to do so for a certain level of confidence.
Therefore, the way that you add random jitter to deterministic jitter will depend upon what your BER budget is. If you do not have a tight BER requirement, then you might want to specify the total jitter as being the sum of the deterministic jitter plus 6σ of the random jitter. As your requirement for BER gets more stringent, you will want to include more of the random jitter component.
Sources of Deterministic Jitter
OK, I’ve measured my jitter, and figured out my BER, and I have blown my budget — what do I do now? This is where most of us have found ourselves at some point or another. We have a system that has an unacceptably high bit error rate, and a manager who wants to know how you can get the system to meet its spec. There are a few things you can do to help reduce your jitter. The first is to not generate so much jitter in the first place. If your major issue is with the deterministic jitter, then you should try to reduce the source of the jitter. This will require analysis of the jitter to determine its source, and then acting to reduce the cause of that source.
If you have deterministic, pattern dependent jitter, this usually points to a bandwidth limitation within the system somewhere, and if you work to improve the bandwidth of a weak link, then you can reduce this jitter component. It could be that by moving to a faster cable driver, you can reduce ISI, and bring your system into compliance. Perhaps the bandwidth limitation is coming in with the cable that is being used for transmission, in which case the best strategy may be to shorten the cable runs, or to use a lower loss type of cable.
Another source of deterministic jitter is non-pattern dependent jitter; this may come from interference from an adjacent channel or signal. In this case, your total jitter can be reduced by improving the channel isolation through better shielding, careful board layout techniques, and greater physical separation from the aggressor signals.
NEXT: Impedance Mismatches, Cable Equalization and Sources of Random Jitter
Page 4: next page
