I got the following question regarding the impedance of a grounded coplanar waveguide(gcw) and microstrip as shown in the image attached: If I increase the spacing w in the gcw to large values I am expecting getting a similar impedance results for the microstrip if the other parameters remain the same since both geometries equalise. At least that both result converge. Why is this not the case?

Thanks for any advice enter image description here

  • 1
    \$\begingroup\$ The coplanar calculator doesn't allow track thickness to be entered as a parameter - maybe you should find one that does. \$\endgroup\$ – Andy aka Aug 5 '16 at 11:36
  • \$\begingroup\$ Could you reference the calculators you are using? It would improve my answer. \$\endgroup\$ – N.G. near Aug 5 '16 at 12:39
  • \$\begingroup\$ Interestingly, CGI-Wcalc agrees with all 4 of your results, to within a couple 0.1's of ohms. Anybody have access to Polar? \$\endgroup\$ – The Photon Aug 5 '16 at 16:19
  • \$\begingroup\$ I think perhaps our man has got s and w transposed - this would explain a lot, hence the request to reference the calculator used. Secondary effects you mention will be negligible at the 200 Mhz \$\endgroup\$ – N.G. near Aug 5 '16 at 18:17

The field patterns on microstip and CPW are very different, and it is the field distribution that determine characteristic impedance. What you have shown is actually GCPW or grounded coplanar waveguide. CPW can be and often is used without a backside ground.

E field patterns on MS and CPW lines *Note for the pedants - the diagram bottom right is slightly incorrect as the E fields should be normal where they meet the conductors (enforced boundary condition)*

Consider this; if the separation s, shown as w in your diagram (I will use s, as w is generally used for the width of the track) were to be infinite - you would infact have a microstrip, so you are correct to assume this.

I refer you to Wadell for models of transmission lines https://books.google.co.uk/books/about/Transmission_Line_Design_Handbook.html?id=MyxTAAAAMAAJ

Be very aware when using online calculators however, as they are often just coded up, naive application of models. All models have an accuracy and range of applicability over which they are valid.

(Note to ADS users - ADS does not always issue a warning when models are used outside area of validity)

For GCPW, the separation s need to be quite small before the impedance is influenced. If s > w the effect is minimal and the microstrip formulas are adequate for most purposes.

I have often heard other engineers somewhat erroneously refer to a shielded microstrip as co-planar, as it looks to be of a CPW topology.

Track thickness does play a part as commented, but it is usually not major, as thickness t is usually very much smaller than the height h and separation s. Don't get hung up on very small inaccuracies for general interconnecting transmission lines anything between 45 and 55 ohms is perfectly acceptable. for example the mismatch between 50 and 45 ohms yields a return loss of better than 25dB or a reflection coefficient of 0.05, which instrument grade connector quality.

Comments welcome

  • \$\begingroup\$ I get nearly the exact same results (for all 4 geometries, Z0 matches within 0.2 ohms) when estimating with CGI-Wcalc. CGI-Wcalc includes terms for surface roughness, copper resistivity, and dielectric loss tangent. \$\endgroup\$ – The Photon Aug 5 '16 at 16:23
  • \$\begingroup\$ thanks for these answers so far. even with other calculators I get the same result like in my first post: MS: eeweb.com/toolbox/microstrip-impedance GCPW: wcalc.sourceforge.net/cgi-wcalc.html And I didn't mix trace width and spacing. Just using different variables respectively for MS and GCPW in the image. So am I right that theoretically I should get same results in MS and GCPW for large spacings in GCPW? Just that the online calculators are wrong for such case. \$\endgroup\$ – r.walther Aug 8 '16 at 6:00
  • \$\begingroup\$ GCPW with large spacings is effectively microstrip, yes. For most lines of c. 50 ohms, the w/h ratio will be around unity - if it's less than 0.5 or more than 2 - take a closer look. \$\endgroup\$ – N.G. near Aug 10 '16 at 7:59
  • \$\begingroup\$ "GCPW with large spacings is effectively microstrip, yes." So why do I get different impedance results in both simulators? I think I still don't get the point. Is the GCPW simply not made to simulate large spacings so that it becomes a microstrip model? \$\endgroup\$ – r.walther Aug 10 '16 at 9:34
  • 1
    \$\begingroup\$ I discovered wcalc through this post, but have found that its calculation for coupled microstrips is broken. It turns out there was a bug filed against it (sourceforge.net/p/wcalc/bugs/2), but I don't currently see a fix for it, and wanted to caution anyone else getting wcalc to sanity check it with other calculators. \$\endgroup\$ – Toybuilder Jul 29 '17 at 18:13

GCPW with large spacings is effectively microstrip, yes.... So why do I get different impedance results in both simulators? I think I still don't get the point. Is the GCPW simply not made to simulate large spacings so that it becomes a microstrip model?

Correct. The limit as gap spacing goes to infinity is microstrip. However, these calculators are using curve-fitting models tracing back to the Wheeler and Getsinger era (60's 70's etc). If the parameters exceed the allowable ranges, the curve-fit models blow up. The CPW models underestimate the influence of the ground plane underneath the T-line as noted by the Zo being much too high as the gap spacings are made very large.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.