ADS and relative dielectric constant variations, schematics simulation vs layout simulation

Thread Starter

simozz

Joined Jul 23, 2017
125
Hi all,

I am designing a bandpass stripline filter @ 17.2 GHz using parallel coupled lines with Keysight ADS, and I am experiencing problems with schematics simulation vs layout simulation results.

The schematics simulation plots the expected results based on design requirements and equations

Screenshot_2020-11-07_16-36-56.png

while layout simulation results with a frequency response shifted approx. 1.3 GHz below.

Screenshot_2020-11-07_16-39-03.png

It results that I always have to manually tune the substrate dielectric constant quite a lot in EM substrate settings.
For example, for DiClad 880 εr= 2.2, I have to tune it down to εr = 1.875.

I know that εr is frequency dependent, but such variation of -0.147 is too much from what I can see from the Roger's Diclad 880 DS (https://rogerscorp.com/-/media/proj...lish/data-sheets/diclad-series-data-sheet.pdf)

Screenshot_2020-11-07_16-29-16.png

Is it normal such a variation of εr in ADS ?
Is it possible that I am doing something wrong in layout simulation ?

Thanks,
s
 
Last edited:

Papabravo

Joined Feb 24, 2006
21,159
Seems like a disconnect in geometries between the circuit simulation and the layout simulation. I would check my models and measurements. At this frequency a wavelength is 17.44 mm or around 3/4". Small dimensional errors can certainly add up.
 

Thread Starter

simozz

Joined Jul 23, 2017
125
Hello @Papabravo

I would check my models and measurements
What do you mean exactely ? The transmission lines dimensions are calculated with ADS Line-Calc tool. Even and odd mode impedances are correctly determined based on parallel coupled analysis (schematics simulation is quiet correct).

s.
 

Papabravo

Joined Feb 24, 2006
21,159
Hello @Papabravo



What do you mean exactely ? The transmission lines dimensions are calculated with ADS Line-Calc tool. Even and odd mode impedances are correctly determined based on parallel coupled analysis (schematics simulation is quiet correct).

s.
At this distance I have no idea why your schematic simulation and your layout simulations are producing different results. Your assumption that they should be the same is apparently not borne out in practice. You are in the best position to find the answer. Based on my experience with simulations and bench results you need to recheck all of your assumptions to find out where the difference are. Have you actually fabricated this filter to see which simulation, if either, produces the closest results?

Are you in possession of a trivial test case that does produce identical results from schematic to layout? If not, why?
 

Thread Starter

simozz

Joined Jul 23, 2017
125
Have you actually fabricated this filter to see which simulation, if either, produces the closest results?

Are you in possession of a trivial test case that does produce identical results from schematic to layout? If not, why?
There is no practical design. It's all about simulation that, of course, tries to replicate a real situation as best as possible.
I have tried to implement the same type of filter at lower frequency, for example at 5 GHz, and I did not issued such a variation for εr in layout simulation.

s.
 

Papabravo

Joined Feb 24, 2006
21,159
There is no practical design. It's all about simulation that, of course, tries to replicate a real situation as best as possible.
I have tried to implement the same type of filter at lower frequency, for example at 5 GHz, and I did not issued such a variation for εr in layout simulation.

s.
I would expect that for at least two reasons:
  1. As we go up in frequency things become less and less intuitive.
  2. The behavior of components and structures changes as their dimensions become a significant fraction of a wavelength.
Nature and science are mechanistic, and there is an explanation for the things we observe. This is equivalent to saying that we don't believe in magic. Just because we don't understand and cannot explain our observations does not mean that it is a permanent state of affairs. Perhaps a conversation with the tech support folks who make the simulator might be in order.
 

Papabravo

Joined Feb 24, 2006
21,159
Yes, this is probably the best things to do... Unfortunately the ADS forum doesn't seem to have a lot of participation... :confused:
I sympathize with your difficulty. It has happened to me numerous times with language compilers and Integrated Development Environments. Limited user base and useless or unknowledgeable tech support.
 
Top