Talk:Signal integrity

Comments
I am not sure who is writing this page, but there are several typographical and factual errors. The content seems to be geared towards promoting "select" textbooks and does not list the fundamental work done by several experts. The most notable exclusion is of a text by Dally and Poulton. It seems written too hastily.

Hi! I'm Lou Scheffer. There was no entry at all for Signal Integrity, so I made an initial stab at one. I'm an IC guy, and in this field the first and best known authors are Vinod Kariat, Ken Shepard, and others from their group at IBM, with their initial paper on Harmony. (These folks, after IBM, founded CadMOS, a SI startup that was bought by Cadence). I got permission to take part of a chapter he wrote and use that as the article.

Of course, there are also lots of SI issues on PC boards, which were not addressed in the initial article. Looking at the edit log, it appears that "Howard Johnson" added most of this part. Then lots of others made smaller changes.

If you find factual errors and/or typos, please just go ahead and fix them. That's the beauty of Wikipedia!. Adding more reference texts is also welcome.

LouScheffer 18:43, 27 December 2006 (UTC)

It makes me sick to the bottom of my stomach knowing how people misuse this space for self promotion.

Perhaps instead of complaining, you should contribute? You say there are errors - go ahead and fix them. There is fundamental work by several experts that is not listed? Add the work, and the references. You say it refers to only a few "select" textbooks? Then add the textbooks that should be listed. You say there is too much self-promotion? Then re-write the article and you can remove the references you think are inappropriate. All of these are possible, just as easy as adding complaints here. LouScheffer 05:39, 28 December 2006 (UTC)

Where should the the ref "Noise Considerations in Digital ICs, by Vinod Kariat." point to?
Maybe it should go with the phrase "In CMOS..."? Woz2 (talk) 23:08, 3 February 2008 (UTC)


 * It was the basis for the initial text of the article (with permission). Of course a lot has changed since then, in particular the PCB section has been added.  It goes along with most of the CMOS stuff, but not the PCB stuff. LouScheffer (talk) 01:09, 4 February 2008 (UTC)

Ok I'll put the ref close to the CMOS section Woz2 (talk) 01:21, 4 February 2008 (UTC)

Thanks, Lou. Done! Woz2 (talk) 01:26, 4 February 2008 (UTC)

Hi! I think at least the two overview type articles are best left as general (not numbered) references. I'd guess there are 50-60 statements that can be tracked back to one of these two sources. Numbered references are great for a specific point, but they tend to imply that all the rest is general knowledge, which is not the case here. Most of the unreferenced statements can be confirmed by one of these two articles, one concentrating on ICs and one on PCBs.

Also, I re-wrote the intro to try to make it clearer to the non-specialist. I tried to follow the normal order of technical argument - what is the problem, what are the consequences, what can be done about it (divided into the two main cases - IC and PCB/package). LouScheffer (talk) 18:34, 4 February 2008 (UTC)

This page is poorly written, and attempts to improve it have been undone
I attempted to rewrite the introduction to remove vague, colloquial words like "flaky" and replace them with specific, well defined communication engineering terms like "impairment" but these edits were undone with no justification. The reference section is mixture of bullets and numbers which doesn't adhere to the style guide. Again I tried to fix this but it was undone, again for no real reason. Just because the original author saw the key word had no page, doesn't mean that that author "owns" the page, and it doesn't mean that the author's book has to be self-promoted by being place first reference. This is wikipedia, not a gold mine claim on the Wild West frontier. The Johnson and Bogatin books are much more populaar (Check the ranking on Amazon for example), are more comprehensive, and have more than one chapter on the topic. What do other folks think?

Woz2 (talk) 21:23, 9 February 2008 (UTC)


 * Hi! I appreciate that you are trying to improve the article.  However, I think editors can reasonably differ over what level of rigor is appropriate in an introduction.  The main goal of an introduction is to explain a topic to someone who knows little of it.  I think (but others may differ) that a term like 'various effects can degrade a signal to the point where errors can occur' is shorter and simpler than 'deleterious physical effects ('impairments') can degrade a digital signal to the point where decision errors reach unacceptable levels.'  I too would be interested in what others think is right for an introductory sentences.  Also, I think it makes sense to first mention the problems that SI can cause, before getting to the details of fixes.
 * On many other pages, it has taken many iterations to get the introduction to where all editors are happy with it. See the history of 'Simplified method of operation' on the GPS page, for example.  Often convergence is achieved a sentence at a time - if you do not like the word flaky, you can take it out and replace it with 'unreliable', and hence work towards a common understanding.
 * The references are an interesting question. I certainly agree that being first to write an article gives no priority.  However, the references are supposed to be the sources used to write the article, which they are.  This is independent of popularity - it's a matter of recognizing sources.  The question is how you separate a 'this reference forms the basis of large portions of the text' from 'this is the reference to consult to back up a specific claim'.   If you make them all numbered references, there is no way to tell this. I am certainly open to other resolutions of this (in academic papers it typically says something like 'the following overview was derived from []', which might work).  Another possibility is to re-write the article entirely, so that the general references are no longer needed. LouScheffer (talk) 01:00, 10 February 2008 (UTC)


 * The reason why the article relies so heavily on quotes from your book is that you keep deleting any additions from other authors. The page smacks of blantant advertisting for your book which a) Has a sales rank of 624,452 on Amazon, compared to Eric's book which is 75,144 b) Isn't about signal integrity apart from one chapter which touches on ICs (but not, judging by your latest mis-edits, on preemphasis filters, DFE, or CDR circuitry) c) doesn't cover the important PCB and backplane aspects of SI. This page has the highest PageRank on a Google search for "signal integrity". But it's nowhere near the best article about SI on the web. Caveat emptor. Woz2 (talk) 02:46, 11 February 2008 (UTC)


 * Perhaps we should split the page into 'Signal Integrity for ICs' and 'Signal Integrity for PCBs and systems'?  I personally feel they have many aspects in common, and are best discussed together.  However, references tend to discuss one or the other.  For example, the books by Johnson and Bogatin do not cover any of the on-chip signal integrity issues at all.  Also, the fixes used, the ability to use measurements, and the typical problems encountered are very different.  By the way, I did not write any of the article contents concerning PCB signal integrity - it was all added by the editor who added the references to the 'Johnson' books. LouScheffer (talk) 03:21, 11 February 2008 (UTC)


 * Thank you. I'm not in favor of splitting the page. The challenge of chip-to-chip SI comes from the interaction of IC and its environment and therefore needs to be solved wholistically, by co-operation up and down the supply chain between IC and system engineers. (IBIS for example was set up to facilitate the dialog between people and their EDA tools). I agree that the intra-chip issues relate only to the chip. However the chip-to-chip issues relate to both the chip (or at least its I/O tranceiver) and the PCB/backplane. For example the equalizers, DFEs, and CDRs, are implemented on the IC. Vendors like Altera offer these features on their chips to help their customers design the chip into thier board, and to open up the eye by setting various register values (eg tap weights). I think the solution is to rewrite the page with with a general perspective. With that structure, sections for on-chip, chip-to-chip, rack-to-rack techniques can be added underneath with info that is specific to the domain. Woz2 (talk) 12:28, 11 February 2008 (UTC)


 * I really prefer specific engineering language. Saying a 0 or 1 must be "pure" is vague and wouldn't make it past a journal editor. It wouldn't even be acceptable in a general audience mag like Scientfic American. The correct goal is to achieve some specific bit error rate, not to make the bits "pure" . "Pure" suggests the use of moral guidance as mitigation :-) Bit error rate is a wikiword, so it's simple to define the term for those not familiar with it.Woz2 (talk) 11:36, 11 February 2008 (UTC)

Rewrite: How about this intro and structure?
Something like:

In digital electronics, a stream of binary values is represented by a voltage (or current) waveform. Over short distances and at low bit rates, a simple conductor can transmit this with sufficient fidelity. However, at high bit rates and over longer distances, various effects, usually called impairments, can degrade the signal to the point where errors occur, and the device fails. Signal integrity or SI engineering is the task of analyzing and mitigating these impairments. Signal integrity engineering is an important activity at all levels of electronics packaging, from internal connections of an integrated circuit (IC), through the package, the printed circuit board (PCB), the backplane, and inter-system connections. While there are some common themes at these various levels, there are also practical considerations that cause substantial differences in the approach to signal integrity for on-chip connections versus chip-to-chip connections. Therefore, this article has separate sections for each.


 * This looks good. My only change I would suggest would be to change the phrase "usually called impairments" to "often called impairments".  Searching scholar.google.com, which contains all the IEEE and ACM journals, shows 15K references to "signal integrity", but only 283 contain the words impairment or impairments (and only one of these talks about signal integity for signals purely on-chip).  This matches my personal experience - as far as I know the term is not commonly used for issues with SI entirely on chip.
 * Also it might make sense to put the PCB/system part first. There are surely more PCB designers than IC designers, and the field itself is older (back to the days of telegraphs). LouScheffer (talk) 17:49, 11 February 2008 (UTC)


 * OK, I started the division. Feel free to improve, LouScheffer (talk) 03:33, 12 February 2008 (UTC)

History
(history section)

Challenges and Solutions Common To on-chip and chip-to-chip Signal Integrity
(general observations about crosstalk, bandwidth, risetime.

On-chip Signal Integrity
(IC internal connection section)

Need a section on off-chip SI, e.g. loss tangent and other related losses and distortions. PCB dielectric loss particularly, e.g. needs work. --74.107.74.39 (talk) 00:26, 27 March 2011 (UTC)

Chip-to-chip Signal Integrity
(Backplanes and IC I/O drivers)

Edit request: Change Agilent Technologies to Keysight Technologies
I suggested to change Agilent Technologies to Keysight Technologies of application notes's title in Notes section because Agilent's Electronic Measurement business is now Keysight Technologies. Keysight Technologies was spun off from Agilent Technologies in 2014. Keysight compromises of Agilent’s test and measurement products and software. The two applicate notes are: 7. "Eight Hints for Debugging and Validating High-Speed Buses," Application Note 1382-10, Agilent Technologies 14. "Using Clock Jitter Analysis to Reduce BER in Serial Data Applications", Application Note, literature number 5989-5718EN, Agilent Technologies Please let me know if you have any question. SOPHIASONGPANDA (talk) 08:16, 15 April 2021 (UTC)

"Submit an edit request"