Talk:Scattering parameters

Merge s-parameters into scattering parameters
The page titled "s parameters" should probably be merged into this one.

I have just added quite a bit more from some of the classic references especially Pozar and Gonzalez and the infamous Agilent application notes. I hope this is a good start, the problem being avoiding rambling on for too long. Definitely should be merged with s parameters because this is just an abbreviation of the correct full descriptionChrisAngove 21:58, 11 April 2006 (UTC)

I have just added a few corrections and improvements. In a few days I plan to merge s-parameters into this one. Nobody has objected and I consider s-parameters to be more shall we say blog style than encyclopedic style.ChrisAngove 22:40, 19 August 2006 (UTC)

Why I have merged s-parameters into scattering parameters
They are of course the same subject. People less familiar with RF engineering often ask what the S stands for so the better title is the full name with redirection from s-parameters. I have read S-parameters and added a few things I had overlooked in scattering parameters. Now please can we have some more comments from the huge numbers of electrical/RF engineers who must surely read this. ChrisAngove 20:39, 19 September 2006 (UTC)


 * The merge was a good idea. Needed to be done. --Thesilmarilion 18:38, 26 October 2006 (UTC)

Renormalized/General
Nice article! - but I believe that what you describe as The General S-Parameter Matrix is usually called the Renormalized S-Matrix or Normalized-to-50Ω S-Matrix. The Generalized S-Matrix is for a matched load (as opposed to 50Ω) on each port (so that in $$A = S B$$, $$B$$ is the amplitude of the in-going waves, and $$A$$ is the amplitude of the scattered-out waves). I expect plain S-matrix could mean either normalized or not depending on whether the context was circuits or microwave devices. --catslash 09:50, 21 September 2006 (UTC)


 * I'm not following what you mean by I expect plain S-matrix could mean either normalized or not since S-parameters are always normalised to the port impedance. It's just that the port impedance is not always 50Ω.
 * It's also worth pointing out that each port impedance could be different such as is the case for a matching network. As a result referring to system impedance is only relevant for systems where all the port impedances are the same. --Thesilmarilion 18:12, 26 October 2006 (UTC)

Yes, it was the reference to system impedance that I was pointing out. Considering (say) a 50Ω to 25Ω quarter-wave transformer in microstrip, I would define the s-matrix with matched ports, so that when it was working properly, I would have s11 = s22 = 0 and |s12| = |s21| = 1. However (awkward and annoying) people tell me that they must have the s-parameters normalized to a single system impedance (i.e. the scattering between a source and load of the same impedance (e.g. 50Ω) on each port). The reasons given are (1) they want the s-parameters as input to a circuit simulator, which requires this, and (2) they want Touchstone format, which requires this. Hence the reference to system impedance made me suppose that the article was limited to renormalized s-parameters. --catslash 22:33, 26 October 2006 (UTC)


 * Thanks for the clarification. I think it would be worthwhile then to point out in the article that S-parameters may be normalised to any arbitrary impedance at any port with zero and infinity being the only exclusions but that there are some simulators out there that insist on requiring the specification of a system impedance where all ports are normalised to the same impedance. --Thesilmarilion 15:53, 27 October 2006 (UTC)

Spatial/temporal phase
I think it would be useful to explain the difference between 'spatial' and 'temporal' phases in the General S-Parameter Matrix section. As the text stands at present the terms are used without explanation or reference. --Thesilmarilion 18:13, 26 October 2006 (UTC)

Yes I agree with that. It seems like magic that you can just forget about the $$e^{j \omega t}$$ so easily.ChrisAngove 20:23, 1 December 2006 (UTC)

I think what is missing is to note that the *relative* difference in phase between two ports remains constant over time. Or, we could equivalently say that the phase change was measured simultaneously at both ports. It may be misleading to differentiate between spatial and temporal phase - after all, a single frequency has only one phase. A phasor expression for a wave propagating in the z direction is $$e^{j( \omega t - \beta z) }$$, where $$\beta$$ is the phase constant, and this shows how the phase varies in both space and time. How about rewording the sentence to something like the following?

The phase part of the s-parameter tells the amount of phase change on one port relative to another for a given frequency. -Peter 129.237.165.40 06:00, 11 January 2007 (UTC)

The present sentence is meaningless as far as I can tell - I shall delete it. --catslash (talk) 00:22, 17 January 2009 (UTC)

math improvements, shortening
I have forced png rendering in the math tags using the \, characters to make the equations consistent. Also started to shorten sentences as the article is apparently somewhat long. May be possible to split it to say a definition and a measurement article.ChrisAngove 16:28, 1 January 2007 (UTC)

I agree that isolating a topic such as "S-parameter measurements" from "S-parameter definitions" would be useful. Thesilmarilion 13:54, 3 January 2007 (UTC)

Lossy/Lossless networks
In the text $$S\cdot S^* \leq I$$ was used, this is valid in most cases, since for a lossless reciprocal network $$S_{i,j} = S_{j,i}$$. However to be correct, the following should be used $$S^H\cdot S \leq I$$, where $$\{\}^H$$ indicates the hermetian matrix (complex conjugate transpose). This way it will also be valid for non-reciprocal network, like circulators. Btw: The hermetian should be the first term. Erik, 212.187.90.11 20:36, 19 February 2007 (UTC)

Agilent link
I added the Agilent link at the bottom of the page as a hyper-ref in good faith. It's a commercial site, but the application note is general and well written. I think I'm missing something about Wikipedia's linking policy, but why was it removed as spam? Surely a link just saves a Google search? Thanks, Arthurtech 14:22, 5 May 2007 (UTC) (aka 86.21.13.156).

4-Port S-Parameters
Added a short section on 4-port S-parameters and differential s-parameters. Correct and expand on as necessary.

DrillDriver (talk) 23:04, 13 February 2009 (UTC)

Scattering transfer parameters or T-parameters
Usually, they are defined different, that is, [a1 b1] = T * [b2 a2], for example, in MATLAB or Network scattering parameters

Someone should explain this point to avoid confussions.

Suggestion: I believe that this section should be better placed in section 3 or 4, since it deals only with 2-port systems.

X-parameters
Regarding the recent deletion of the X-parameters link (not me by the way), are these notable? or is the X-parameters page just an Agilent advert (I realize that this question should be addressed on that page first)? I'm happy for there to be a Touchstone File page, since that is a commonly used standard, but X-parameters seems to be something Agilent just made up last year, and has yet to be commonly adopted. Short version: I agree with the deletion. --catslash (talk) 20:59, 11 November 2009 (UTC)


 * You might be right, I restored it mostly on the grounds that no reason was given to explain the deletion. On the other hand, Agilent may have invented the concept, but it has to be admitted they have some influence here and there seems to be enough academic papers out there to claim notability.  I would agree though, that describing S-parameters as a sub-set of Agilent's X-parameters does not feel right for this article, at least not in the lede.  Sp in ni  ng  Spark  23:29, 11 November 2009 (UTC)


 * Yes, I think it is the fact that S-parameters are far more notable than X-parameters that makes it feel wrong. I have re-removed the sentence.
 * Are there academic papers out there on X-parameters? The references on that page do include a conference paper and a couple of magazine articles (no papers in IEEE Trans. MTT-S), but they all seem to be self-promotion. If X-parameters are a new idea, and only the inventor has written about them, are they notable (whether or not they are brilliant idea)? I had half a mind to challenge that article under WP:Notability, WP:NOT etc. Could you go back and take a critical look at it? - or suggest somebody qualified to do so? Thanks --catslash (talk) 00:17, 12 November 2009 (UTC)

Voltage and power parameters
I suspect there is some confusion between the power S11 parameter and the voltage s11 parameter. I rederived the whole thing. If rho is the voltage transmission coefficient, then the scaled reflected power goes as rho^2, and the scaled transmitted power goes as (1-rho^2). This would be if a short square-wave pulse for instance impacts the discontinuity. See, where the field derivation is made (which assumes continuity of the E field, like voltage, and H-field, like current at the boundary) 82.166.221.34 (talk) 12:18, 15 March 2011 (UTC)


 * It comes as a surprise to me that there are two kinds of S parameter. I thought there was only one kind as measured by network analysers.  Sp in ni  ng  Spark  19:14, 15 March 2011 (UTC)

Hi spinning spark. s-parameters are unitless. You get the same value no matter if you ratio the power waves b/a or the voltage waves (b*2*sqrt(Zref))/(a*2*sqrt(Zref)). Woz2 (talk) 01:31, 31 May 2012 (UTC)
 * That only works if Zref is the same at both ports, which is why that is not the definition of S-parameters.  Spinning Spark  09:16, 31 May 2012 (UTC)


 * This does need to be clarified -- I'm pretty sure most modern network analyzers (Aligent, etc) and textbooks (Pozar, etc) use the voltage definition of scattering parameters. The way the article currently stands it introduces them as power ratios, but most of the formulas are for voltages which is probably confusing a lot of people.  a13ean (talk) 18:24, 18 October 2012 (UTC)


 * Not just clarified. It is completely wrong to say 'You get the same value no matter if you ratio the power waves b/a or the voltage waves', regardless of Zref. b/a is different to b^2/a^2. In addition, the article defines S-parameters as 'power ratios' but all the math shown assumes they are voltage ratios (which is correct). — Preceding unsigned comment added by 75.140.140.189 (talk) 18:48, 22 March 2018 (UTC)


 * The article does not define the S-parameters as 'power ratios', but as the ratio of [the amplitude of] 'power waves', i.e. as the ratios of square-roots-of-powers. The article fails to make this clear - possibly out of concern for handling the arcane case where the port impedances are not taken to be real.  S-parameters are voltage ratios only between ports of equal impedance.  For ports of different impedance, there will be a different constant of proportionality between the voltage and the square root of the power at each one.  Consequently the s-parameter linking two ports of different impedance is not equal to the voltage transmission coefficient between them. --catslash (talk) 00:47, 23 March 2018 (UTC)


 * In my experience voltage scattering parameters are often used with transmission line (as opposed to waveguide) networks. When the port impedances are all the same, these are identical with "true" scattering parameters, but when the impedances differ, the matrix becomes asymmetric. Conversion between the two is simply a matter of dividing the columns and multiplying the rows (or vice-versa), by the square root of the impedance of the corresponding port. However, last time I looked, I could not find any mention of these voltage scattering parameters in any textbook, so they probably shouldn't appear in WP either. --catslash (talk) 21:51, 19 October 2012 (UTC)
 * What it is perhaps, is that we calculate a set of voltage reflection and transmission coefficients, and then informally in an abuse of terminology refer to these as scattering parameters or voltage scattering parameters for want of a more suitable term. --catslash (talk) 23:22, 19 October 2012 (UTC)

Why so much detail about a Vector network analyzer?
You can measure S-parameters without a VNA, but I would agree that a VNA is usually the most appropite instrument for the job. But I fail to see the need to go into length about a VNA. A sentence or two should be sufficient. Instead we have needless information about VNA error correction, and systematic errors. That should be (and is), in the article about VNAs. The block diagram has been copied from that too.

Surely the purpose of this should be to teach someone about S-parameters, not a VNA. For a VNA, there's a separate entry in Wikipedia. I don't see the need to copy a significant fraction of that.

Much more useful would be a discussion of other ways of measuring S-parameters. Drkirkby (talk) 21:46, 10 May 2017 (UTC)

Elaboration on measurement unit and six-port network analyzers
The article mentions the following:

Contrary to popular belief, the quantities are not measured in terms of power (except in now-obsolete six-port network analyzers).

Can someone elaborate on this, as I think it is a bit confusing without context? What kind of six-port network analyzer measures S-parameters as power, and why are six-port analyzers now obsolete? — Preceding unsigned comment added by 185.34.132.200 (talk) 08:16, 3 June 2022 (UTC)