Subject: RE: SAS PHY teleconference 5/24/07, 10 am CDT: Definition of BUJ and T10/07-259r0 Date: Fri, 25 May 2007 10:51:07 -0600 From: "Jenkins, Mike" <Mike.Jenkins@lsi.com> To: "Larry McMillan" <Larry.McMillan@wdc.com>, <Alvin.Cox@seagate.com>, <t10@t10.org> X-Message-Number: 7819 Formatted message: HTML-formatted message Attachment #1: image004.jpg Attachment #2: image003.jpg Larry, I do appreciate your distribution of these citations. I will insert comments for each paper below. Overall, I think the disagreement boils down to whether periodic jitter (PJ) is or is not included in bounded uncorrelated jitter (BUJ). First of all, the division of jitter into unbounded (i.e., RJ), bounded uncorrelated (BUJ) and bounded correlated (i.e., DDJ) is a perfect taxonomy by the very names of the categories. It guarantees that all types of jitter are covered and any type belongs to only one category. I suspect the motivation for measurement equipment manufacturers in separating periodic jitter before processing the residue is that spectral peaks are numerically difficult to deal with when mixed with broad spectral components. But that is no excuse for asserting that a bounded jitter component which is uncorrelated to the signal is not bounded uncorrelated jitter. I would suggest a new name for BUJ with the periodic component(s) removed. How about "broad band BUJ" (BBBUJ - pronounced like BUJ, but with a stutter). I don't lightly choose to propose different terminology than what is in practice, but current practice is simply self-contradictory. I will attempt to roll this into a rev of T10/07-259. Regards, Mike ________________________________ From: owner-t10@t10.org [mailto:owner-t10@t10.org] On Behalf Of Larry McMillan Sent: Thursday, May 24, 2007 10:33 AM To: Alvin.Cox@seagate.com; t10@t10.org Subject: SAS PHY teleconference 5/24/07, 10 am CDT: Definition of BUJ and T10/07-259r0 Regarding the discussion in today's phy call regarding the definition of bounded uncorrelated jitter (BUJ): Here are a few references I found in just the 1st page of results from a quick web search. In skimming the various papers/articles I didn't see a single definition of BUJ that agrees with the definition put forth in T10/07-259r0. It seems to me the views of the rest of the world need to be seriously considered before we generate our own variant definition of BUJ. The links follow. Comments anyone? Larry McMillan Western Digital 949-672-7687 Cal Tech Paper differentiates between BUJ and DDJ: <http://www.its.caltech.edu/~hajimiri/pdf/Crosstalk_DDJ.pdf>; http://www.its.caltech.edu/~hajimiri/pdf/Crosstalk_DDJ.pdf <http://www.its.caltech.edu/~hajimiri/pdf/Crosstalk_DDJ.pdf>; I believe this paper and my presentation agree on definitions. I should probably use the (lately more common) term of DDJ rather than ISI. BERTScope paper differentiates between BUJ and other non-ISI DJ: http://www.ofcnfoec.org/materials/BERTScope_Testing.pdf I like their figure showing the taxonomy of jitter, except that they do not include PJ & SJ inside BUJ. The definition of BUJ provided in the paper does seem to include PJ & SJ, so I guess I have to disagree with that part of their definitions. BUJ: A non-random (bounded) jitter that does not relate directly to the data signal. Examples include crosstalk, power supply break-through. In this document the term will be used to refer to the stress component used to emulate the effects of such signals. This is achieved by modulation of the data transitions in the time dimension, usually using some form of PRBS signal. Wavecrest paper differentiates between BUJ and all other DJ: http://www.wavecrest.com/technical/VISI_6_Getting_Started_Guides/6unders tanding.pdf Mike Li, if you're on this reflector, can you explain why - by the definitions below - PJ is not part of BUJ? DCD is, I think, clearly correlated to the data pattern. It is even more clearly bounded. So why would it not be BUJ? Bounded-Uncorrelated Jitter Definition "Deterministic jitter that is caused by other than the data on the signal under test."6 Periodic Jitter (PJ) Definition PJ can quantify crosstalk effects from EMI sources. (adjacent lines, power supply noise, etc.) The last 3 papers seem to follow the lead of the lab equipment manufacturers above in separating periodic jitter from BUJ. Guide Tech paper differentiates between BUJ and PJ: http://www.guidetech.com/pdf/wp/GTW2004_BUJ_Characterization.pdf Guide Tech paper differentiates between BUJ and and all other DJ: http://www.jitter.com/pdf/wp/itc2004_JitterModelsforSerialInterconnects. pdf EE Times Asia article differentiates between BUJ and other DJ, including PJ: http://www.eetasia.com/ARTICLES/2003MAR/B/2003MAR16_HBM_NTES_TA.PDF?SOUR CES=DOWNLOAD