CPW Media impedance does not account for strip thickness

[mhuser]

When writing #665 , we observed that Media.CPW strip thickness t is not used to correct the characteristic impedance.
Currently, it only affect the conductor loss.
Qucs project provide equations and references to correct the impedance and effective permittivity to account for strip thickness.

[mhuser]

After looking on Qucs CPW page and Qucsator CPW source, there is a big discrepancy between them.

Qusator use equations that account from the substrate height h and provide a boolean to compute the case were the coplanar waveguide is backed by metal under the substrate (also known as grounded coplanar waveguide coplanar with lower ground plane or CPWG) that is often seen on printed circuit boards 😄

In ADS simulator there is even two distinct components : CPW and CPWG

Both Qucs and ADS correct impedance and effective permittivity for strip thickness.

[mhuser]

More comparison between Qucs and current Media.CPW

• red CPW with zero thickness, agreement of return loss phase and magnitude (thus of characteristic impedance), over-optimistic insertion loss but with proper phase : this is maybe because the current model does not compute dielectric loss, not investigated yet
• green CPW with strip thickness and limited substrate height (not implemented yet, but it shows these parameters have significant influence)
• blue CPW with strip thickness, limited substrate height and metal backing (not implemented yet)

[Vinc0110]

If I understand #665 (comment) correctly, this issue is not yet resolved.

[jhillairet]

indeed. Was closed automatically.

[mhuser]

Promising results, but still some works to achieve characteristic impedance agreement:

[mhuser]

Got it !

[mhuser]

I think #676 is ready, but there is still the question if we shall do a comparison vs ADS CPW and CPWG.
It would not be a surprise if the results deviate a bit from Qucs because the method to account for thickness seems to make a less general agreement than for microstrip models.

[mhuser]

@JAnderson419 would you be interested in this quest ?
I could also bear if we do not make the comparison 😉
CPW w = 3mm, s = 0.3mm, t = 35um, h = 1.55mm, t = 35um, ep_r = 4.5 (frequency invariant), tand = 0.018, rho = 1.7e-8 ohm/m, 0.2GHz - 20 GHz, 100pts
This one could be run once more with t = 0 if we want to compare without thickness correction

CPWG w = 1.6mm, s = 0.3mm, t = 35um, h = 1.55mm, t = 35um, ep_r = 4.5 (frequency invariant), tand = 0.018, rho = 1.7e-8 ohm/m, 0.2GHz - 20 GHz, 100pts

[mhuser]

@JAnderson419 would you be interested in this quest ?

@JAnderson419 are you in for this or shall we leave things as is, hoping for the best ?

[JAnderson419]

My apologies @mhuser - I was out sick last week and missed this when catching up on emails. I'm happy to help. Thank you for all the work you are putting into these comparisons between tools!

CPW_ADS.zip

[mhuser]

@JAnderson419 First of all, I hope you are getting better or that you will get well soon!
Thank you so much for these precious simulation data !

[mhuser]

I also had to remove frequency dispersion of impedance and permittivity, which are not included into ads model (based on doc)

The agreement with ADS is quite poor unless thickness = 0. For this case :

• S11 magnitude within 0.5 dB : not bad
• S11 phase within 6°, not that good but acceptable
• S21 magnitude within 0.1 dB : validate the loss computation.
• S21 phase is within 0.1°, which validate permittivity computation

The computation to account for non-zero thickness must be different.

[mhuser]

@JAnderson419 Could you run a last simulation with CPWG and all parameters kept equals but metal thickness = 0 ?
This will help to validate the base computation with metal backing.
Thank you a lot for the already generated precious data !

[JAnderson419]

Here you are.

CPWG_ads_t0um.txt

[mhuser]

Here you are.

Thank you a lot @JAnderson419 !

[mhuser]

Above is a comparison ADS / QUCS for coplanar waveguide with/without ground backing and zero thickness. The solid curves are s-parameters generated by these simulators while dots are simulated with skrf.media CPW in ads resp. qucs mode.

The computation of static impedance and effective relative permittivity are the same for the two simulators for zero thickness.

The difference between these simulators are:

• ADS does not compute frequency dispersion of characteristic impedance and effective permittivity, QUCS does.
• QUCS keep relative permittivity of the substrate real and constant while ADS use a complex value with tand.

skrf generated curves (dots) shows a good agreement when these differences are applied (compatibility_mode set to 'ads' or 'qucs' for comparison sake)

[mhuser]

For skrf default behaviour, my proposition is to pick QUCS behaviour with dispersion with the addition of the complex effective permittivity.

[mhuser]

For coplanar waveguide with/without conductor backing, skrf has good agreement with QUCS

The agreement with ADS is poor for insertion loss magnitude, which mean the computed static impedance is slightly different.

QUCS is using an empirical formula from Gupta, Garg, Bahl, Bhartia, Microstrip Lines and Slotlines, 2nd ed.Artech House, Inc., 1996. ADS claim to use a correction from Cohn, Thickness Corrections for Capacitive obstacles and Strip Conductors, IRE Trans. on Microwave Theory and Techniques, Vol. MTT-8, November 1960, pp. 638-644.

However, by reading last paper, there is no ready-to use thickness correction inside. After stepping back, reverse-engineering ADS thickness correction would need simulations with various thickness and... is probably not useful because it does not sounds like a standard and widely used solution.

By default, my proposition is to just keep the same thickness correction as QUCS.

[mhuser]

There is a good agreement on insertion loss of skrf media model and a CPWG mesurement I did in the past. I will put that in an example notebook for CPW media.

The 180° jumps in the phase seems caused by deembedding algorithm but I did not investigate more yet.

[mhuser]

Fixed by #676 that add thickness to CPW and more, and #684 for phase jumps in 2xthru deembedding.