s21>0 for passive device!!

[Decesicum]

For an empty TEM cell I get s21<0 (in CST Studio) but, when loading it (I put a glass inside) s21 is sometimes slightly greater than zero!! Why is it so as long as there no active device there? I thought that it might be some oscillations due to improper spatial resolution (meshing?) and time-step. Might it be this? Any help is greatly appreciated!
Thanks!

 

[volker_muehlhaus]

Time domains solvers calculate the S-parameters from the time domain signal using an FFT. To get accurate results, the time signal must be sampled over a long enough period so that energy in the analysis volume is down to zero. There is some setting in the software that determines the acceptable error by setting a limit on the residual energy. The lower that residual energy is, the more accurate the FFT results are. The higher the residual energy in the analysis volume is, the more ripple you will see in S-parameters.

Compared to an empty TEM line, there will be more resonances with DUT inside, and the time signal will take longer to decay. I suspect that you should check the stop criteria for your simulation (time period too short and/or residual energy limit too high).

 

[Decesicum]

Re: s21>0 dB for passive device!!

Thank you very much for your answer! I increased the required accuracy (at the price of increasing the simulation time; it takes though a bit too long, almost 2 hours...) Anyway, I still don't get what I was supposed to for the absorbed power. I used a VNA to measure the S-parameters for the TEM cell (loaded with a dish only) Did the same in CST. Got similar curves for the measured and computed S and, in consequence, I thought that I should have similar absorbed power values. I computed the absorbed power by the formula P_abs=P_incident x (1-|s11|^2-|s21|^2) for both, measured and computed S. The one resulted from measurements gets 100 times greater than the one from simulation, although S are quite simillar. I don't know where I am wrong, I have checked few times the values, formulas etc...

 

[volker_muehlhaus]

Your equation is fine and makes perfect sense. Can you share the results of this power loss for simulation vs. measured? Is the difference some dB or some 0.0x dB?

The one possible problem that comes to my mind is radiation from the TEM cell. This should be seen when you calculate the power loss for an empty TEM cell. Does radiation increase when you distort the field by placing some object in the TEM cell? If that's the case, it shoudl also be seen in simulation with absorbing boundary conditions.

 

[Decesicum]

Re: s21&gt;0 for passive device!!

the power loss in the dish obtained from measured S-paremeters is 25.9mW while for computed S, I get a power of 11.27mW. What's more confusing is that CST says is that the absorbed power is 0.26mW! This is what I get out from SAR info and what I get from "Loss and Q calculation" I check out now to see if I get different power losses in the TEM's case (which is made of copper) with and without the dish.
Well, the power dissipated in copper increases a bit when the dish is placed in but I think it's negligible (from 540.74 uW to 546.90 uW) I suppose it owes to reflections from the dish.

 

[FvM]

I don't know in which units you are measuring s21, without additional information, I would expect a complex number. The only limitation for passive devices is |s21| <= 1, I think, so what's the actual problem?

 

[Decesicum]

Re: s21&gt;0 for passive device!!

I don't know in which units you are measuring s21, without additional information, I would expect a complex number. The only limitation for passive devices is |s21| <= 1, I think, so what's the actual problem?

Yes, it is a complex number, but what I use is the magnitude (dimensionless). It is the linear magnitude actually, not the one expressed in dB. The problem was that I have a passive device and s21 was greater than 0 dB! This is fixed now. What I don't understand is why the power dissipated in the simulated structure is much much smaller than the one obtained from the above formula. They should be identical in my opinion.

- - - Updated - - -

Or maybe I was wrong when considering that the absorbed power is given by the formula P_abs=P_incident x (1 - |s11|^2 - |s21|^2) ?? (|s| given also by CST!) With this formula I get a value, while CST outputs another value for dissipated power...

 

[volker_muehlhaus]

Re: s21&gt;0 for passive device!!

the power loss in the dish obtained from measured S-paremeters is 25.9mW while for computed S, I get a power of 11.27mW.

I can't interpret these numbers because I don't know what the power levels are.

Can you give the results for 1-(|S11|^2 + |S21|^2) in db?

 

[Decesicum]

Re: s21&gt;0 for passive device!!

View attachment loss.txt so, the stimulated input power (in CST Studio) is 1W. At a certain frequency (200 MHz), |s11|=0.03091351 and |s21|=0.9978061 (of course, these are not in dB)
It results that P_absorbed[W]= P_input[W] x (1-(|S11|^2 + |S21|^2)) which is P_absorbed= 1W x 0.003427= 3.42 mW. This is what I get using the previous formula. In the attached file is what CST outputs for power losses. When requiring SAR information in the whole volume (which mean copper and dish) it outputs 0.27391 mW as absorbed power [I suppose it's in the dish] Why is that value (so small)?

 

[FvM]

This sounds like stretching a simulation result beyond reasonable accuracy limits.

 

[volker_muehlhaus]

I agree with FvM that your S-parameter results might not be accurate enough to extract such a small absorbtion accurately. As you had seen with the initial results, there is always some error from truncation of the time domain data and FFT. You want to extract a small difference from two large numbers (almost all input power is reflected or transmitted, only very little is absorbed) and that's sensitive to precision of the input data.

Not sure about the SAR calculation. That might actually be a smart way to overcome the S-parameter accuracy issue for simulation. Time to talk to CST support, I guess.

However, for measurement you still have the same problem with precision of the underlying S-parameters. How accurate is VNA calibration?

 

[Decesicum]

I see. So I cannot trust the power I get from computed S-parameters... As regards the VNA, the calibration was performed for 12dBm input power, 1.5 kHz IF BW and 20000 samples between .1 and 1.2 GHZ. It seemed resonable to me.

 

[volker_muehlhaus]

I see. So I cannot trust the power I get from computed S-parameters...

You can trust it if and only if the S-parameters are accurate.

As regards the VNA, the calibration was performed for 12dBm input power, 1.5 kHz IF BW and 20000 samples between .1 and 1.2 GHZ. It seemed resonable to me.

These parameters don't tell us much about the accuracy of calibration. The S-parameter that you are measuring are S11 at -30dB and -0.02dB S21. To measure these accurately, you would need VNA accuracy after calibration that is an order of magnitude below these values - not trivial at all.

 

[Decesicum]

You can trust it if and only if the S-parameters are accurate.



These parameters don't tell us much about the accuracy of calibration. The S-parameter that you are measuring are S11 at -30dB and -0.02dB S21. To measure these accurately, you would need VNA accuracy after calibration that is an order of magnitude below these values - not trivial at all.

Ok, I understand now what you're saying. I don't know how to set the accuracy for VNA, I have to look for it.
As regards the power I think about introducing something in that dish, a liquid or smtg with known parameters which would absorb some greater power than the dish itself.

 

[volker_muehlhaus]

I don't know how to set the accuracy for VNA, I have to look for it.

There is no easy way to set the accuracy - it is a result of careful VNA calibration, high quality measurement cables that keep their phase when bent, and several other influences.

In general, it is best to use a measurement setup where your device under test has a large influence on results. The setup with obstacles in a TEM cell, that change the S-parameters only slightly, is sensitive to measurement error.