lumped or distributed ADS simulation

[J85]

Hi,

I'm creating two circuits using ADS for the frequency range 1-100MHz. The circuits will be separated by a 10m coaxial cable. Is it appropriate to model this cable using distributed parameters? I am concerned about the frequency response of the lumped component circuits either side of the cable for the lower frequencies in this range where the wavelength is greater than the length of the cable. Wavelengths are between 300-3m. I'm particularly concerned about any resonances within this frequency range between the circuits and the cable (if behaving as lumped parameter).

When including the cable as a pi model the results are poor. Can anyone shed light on this? Do I need to assess the circuit with both lumped and distributed representation of the cable? I am aware of the various rules of thumb for choosing lumped or distributed components, I would like to know what others do when they have a frequency range that meet both criteria?

Many thanks

Jon

 

[pancho_hideboo]

Less than λ/10, lumped approximation.

freq=1MHz, λ=300m, λ/10=30m
freq=100MHz, λ=3m, λ/10=30cm

So if your concerns are 1MHz-100MHz, you can approximate by lumped components model for size less than 30cm.

 

[volker@muehlhaus]

When including the cable as a pi model the results are poor.

You have ADS, so the easiest solution is to use a transmission line model (line impedance and loss). Then ADS takes care of modelling and there is no need to worry about lumped model segment lenth. This model is then accurate wideband, from DC to RF.

If for learning purposes you want to model your cable as distributed pi model (RLCG), you need to use cascaded segments that are no longer than ~ 1/10 wavelength each.

What is NOT valid: You must not create one single PI model that lumps all the L and C of 10m length into one single pi segment. That model will NOT be valid for wavelength < 100m (frequencies > 3MHz). For high frequencies, you must cascade multiple pi models for shorter segments, to properly model the distributed nature of the line's L and C.

 

[J85]

thanks for the replies. In all honesty I did make a single pi model to compare, even though I knew the requirement for cascaded sections, lazy mistake. So the transmission line model in ADS will account for full interaction between itself and external lumped components and not just calculate reflection/transmission coefficients and attenuation? I have capacitors either side of the transmission line and I'm concerned that when constructed, the cable could shift their self-resonance to within the 1-100MHz bandwidth, which is not shown in ADS with the tline model?

 

[volker@muehlhaus]

No need to worry. The transmission line model will be an accurate model for the line from DC to GHz and include the effect of source/load impedances that you connect. They are valid for all combinations of voltage & current at any frequency. Or in other words: it will be more accurate than any segmented pi model for the line.

Smallprint:
The one assumption that is made by these models is that all current that goes into the (+) terminal of a port leaves at the (-) terminal. Or in other words: there must be no additional propagation path where current can flow that goes only partially through one of the line conductors. Example: a coax that has a common mode RF voltage against a 3rd conductor, that's would be 3 conductors in total and require a more sophisticated model.

I have capacitors either side of the transmission line and I'm concerned that when constructed, the cable could shift their self-resonance to within the 1-100MHz bandwidth, which is not shown in ADS with the tline model?

If there is such an effect, you will see it in simulation. Take care with modelling realistic impedances for the ports in ADS if your hardware environment is not 50 Ohm source/load.

 

[J85]

thank you for the reassurance, I will keep in mind the realisitic source/load impedances.