Recent changes to feature-requestshttps://sourceforge.net/p/qucs/feature-requests/2017-02-10T15:00:36Z#63 Enable negative-length transmission line2017-02-10T15:00:36Z2017-02-10T15:00:36ZClaudio Girardihttps://sourceforge.net/u/in3otd/https://sourceforge.net2deef0f459bb9c7891cd8de931e0d20b9b367449<div class="markdown_content"><p>Thanks for the pre-simplified equations Gert, :)<br/>
Not sure of what can be done if S is singular, maybe playing with its Singular Value Decomposition and trying cutting the smallest singular values, since the de-embedding matrix looks very similar to the S matrix inverse. The Qucs code has routines for the SVD, but I don't think these are exported and available in the equations.<br/>
Or try to fit a rational function (vector fitting) and use that as "real" S-parameters. This will remove the noise but you could still have a singular matrix somewhere...</p></div>#63 Enable negative-length transmission line2017-02-10T12:25:24Z2017-02-10T12:25:24ZGerthttps://sourceforge.net/u/cuypersg/https://sourceforge.net431e057000a915d1bb8a68f91907801ca4759d63<div class="markdown_content"><p>Claudio,<br/>
exactly that method is one I have also used myself. More specifically this (S->T, Tinv=T^-1, Tinv->S) translates to:<br/>
D11= S11./(S11.<em>S22 - S12.</em>S21);<br/>
D12=-S21./(S11.<em>S22 - S12.</em>S21);<br/>
D21=-S12./(S11.<em>S22 - S12.</em>S21);<br/>
D22= S22./(S11.<em>S22 - S12.</em>S21);<br/>
So it is sensitive to a singular S-matrix. I wonder if there is a more robust method to do this, because usually the data is noisy. Something like the difference between a zero-forcing solution and a minimum-mean squared solution (if you are familiar with these). That could be a very nice addition to Qucs. But maybe too far-fetched or too experimental. :-)</p></div>#63 Enable negative-length transmission line2017-02-10T11:11:40Z2017-02-10T11:11:40ZClaudio Girardihttps://sourceforge.net/u/in3otd/https://sourceforge.nete369342ef1cd03e8d615e37b4972de6e4d6ff339<div class="markdown_content"><p>Thanks Frans,<br/>
seems a nice tool. I'll take a closer look :)</p>
<p>While we are on the subject, I wanted to add a simple 2-port de-embedding in Qucs since some time. I usually do all the preprocessing in Octave (S->T, Tinv=T^-1, Tinv->S) to prepare the de-embedding matrix but will be easier to adapt the S-parameter file component to do this directly also. Not as flexible as your tool but it can be sometimes useful in Qucs.</p>
<p>And also the versatile <a class="" href="https://github.com/scikit-rf/scikit-rf" rel="nofollow">scikit-rf</a> is often handy.</p></div>#63 Enable negative-length transmission line2017-02-10T10:56:46Z2017-02-10T10:56:46ZFranshttps://sourceforge.net/u/fransschreuder/https://sourceforge.netaedd6b0e23141b4e42c378754bef4a225b453afb<div class="markdown_content"><p>Dear Claudio, a little off-topic, but you may also be interested in a deembedding tool that I wrote recently:<br/>
<a href="https://github.com/fransschreuder/deembed/wiki" rel="nofollow">https://github.com/fransschreuder/deembed/wiki</a></p></div>#63 Enable negative-length transmission line2017-02-10T10:13:13Z2017-02-10T10:13:13ZClaudio Girardihttps://sourceforge.net/u/in3otd/https://sourceforge.nete17ea2078f5bbbce73c63971506ca5abe8a6a1d0<div class="markdown_content"><ul>
<li><strong>status</strong>: open --> closed</li>
</ul></div>#63 Enable negative-length transmission line2017-02-10T08:57:21Z2017-02-10T08:57:21ZGerthttps://sourceforge.net/u/cuypersg/https://sourceforge.netfc94303f162462cb46540b7af906e36132c19a48<div class="markdown_content"><p>Sorry, my bad. I have checked and just realised the problem was somewhere else, the transmission line works as desired.</p></div>#63 Enable negative-length transmission line2017-02-09T19:19:49Z2017-02-09T19:19:49ZClaudio Girardihttps://sourceforge.net/u/in3otd/https://sourceforge.net99ca1fc6747383f5ad57fa0050f264c0c756cada<div class="markdown_content"><p>As you said, this can certainly be useful for de-embedding.<br/>
A quick check of the code shoes that most of the ideal lines already allow negative lenghts, while microstrips require a positive length. Is there some model in particular that is not working with negative lenghts?</p></div>Enable negative-length transmission line2017-02-09T17:46:47Z2017-02-09T17:46:47ZGerthttps://sourceforge.net/u/cuypersg/https://sourceforge.net5acc3b4283934a58d1fe04d78b29ab8ce7c75a2d<div class="markdown_content"><p>It would be interesting to be able to use a negative-length transmission line (iirc this was also possible in MDS/ADS, which Qucs is more or less a clone of).</p>
<p>The purpose of this is to be able to allow a simple track de-embedding for (measured) S-parameter blocks. Limiting to the theoretical transmission line would be good enough, microstrip or so is not really needed. I imagine that no real programming is needed to enable this, just not check for a positive length. The mathematics don't care for negative length.</p></div>Enable negative-length transmission line2017-02-09T17:46:47Z2017-02-09T17:46:47ZGerthttps://sourceforge.net/u/cuypersg/https://sourceforge.net01c68a15035922e96af923d276424b3a3d8109f3<div class="markdown_content"><p>Ticket 63 has been modified: Enable negative-length transmission line<br/>
Edited By: Claudio Girardi (in3otd)<br/>
Status updated: u'open' => u'closed'</p></div>#62 Delay(Signal.v, ms) function requested2016-08-20T15:10:52Z2016-08-20T15:10:52ZClaudio Girardihttps://sourceforge.net/u/in3otd/https://sourceforge.neta0af48f4bfb020dae1fd52a8c655d98e0a38c9eb<div class="markdown_content"><ul>
<li><strong>status</strong>: open --> closed</li>
</ul></div>