Agilent ADS slot/strip planes difference

Hi

In agilent ADS momentum there are two ways to model a ground plane: slot_planes and strip_planes.
In slot_plane method the program only meshes the voids, like antipads and splits, while in strip_plane method it meshes the copper areas. The strip_plane method results in a lot more mesh cells in a simulation.

I would think that using strip planes would be more accurate to simulate the effect of an imperfect ground plane (like antipad fields) and return currents, since it meshes the plane itself.
Am I right or wrong?

What is the exact accuracy difference between the 2 methods? I know how the 2 methods are different in building the models (electric currents vs magnetic currents), but how do they correlate to measurements (the 2 types separately) in case of heaviliy antipad-ded ground planes?
Basically how does the slot_plane method model the return currents, when the path of those return currents is not part of the model?

If you build your model on a slot layer, then ground areas will be infinite (consider a GCPW case) which is not the actual case.
Also only perfect conductors could be defined on slot layers which is also an ideal case.
Therefore basically it's not as accurate as strip layer. But it reduces simulation time if you can tolerate a bit of reduced accuracy.

i have tried both for a structure with 4 diffpairs.
The strip plane method seems to be 77x slower, i have simulated less than 1 frequency point overnight on a dual core 2.0GHz machine, while the slot plane version went through with 15 freq points in 4 hours.

What i am really interested in typically, is if a diffpair goes thorough an antipad field with 20 antipads. The effect of this is what i am after. Can both the slot and the strip method accurately simulate this?

buenos said:

i have tried both for a structure with 4 diffpairs.

What i am really interested in typically, is if a diffpair goes thorough an antipad field with 20 antipads.

Click to expand...

Sorry, I don't get your drift. what is antipad?

on a PCB (printed circuit board), we normally have these differential pairs coming out of BGA packages. In the first half inch of routing the tracks go next to these antipads. The BGA components have hundreds of pins/signals go all over the board, they are connected to innter layers through VIAs. Where a signal via goes to an inner signal routing layer, it needs a circular cutout in the ground planes around the VIA. This cutout is the antipad. Under BGAs these antipads cut out so much of the copepr from the ground planes, that they normally halve the amount of copepr in some areas. In other areas the antipads are so close to each other that there will not be any copper left between them (touching), so any return currents will have to go around a group of antipads like this. This has signal integrity degradation which I want to analyze.

buenos said:

Hi

In agilent ADS momentum there are two ways to model a ground plane: slot_planes and strip_planes.
In slot_plane method the program only meshes the voids, like antipads and splits, while in strip_plane method it meshes the copper areas. The strip_plane method results in a lot more mesh cells in a simulation.

I would think that using strip planes would be more accurate to simulate the effect of an imperfect ground plane (like antipad fields) and return currents, since it meshes the plane itself.
Am I right or wrong?

What is the exact accuracy difference between the 2 methods? I know how the 2 methods are different in building the models (electric currents vs magnetic currents), but how do they correlate to measurements (the 2 types separately) in case of heaviliy antipad-ded ground planes?
Basically how does the slot_plane method model the return currents, when the path of those return currents is not part of the model?

Click to expand...

If you are modeling microstrip/stripline/etc. structures, then you need to use a strip_plane ground interpretation. This way, ADS knows to determine the currents on the metal conductors, rather than focusing on determining the fields inside the gaps (slots). If you are developing your PCB to use slot-style waveguide structures, then using a slot_plane ground interpretation would make sense, since you care little about the currents on the rest of the structure (typically flooded planes with only a few openings), and want the solver to focus on the fields developed inside the waveguide slots.

To speed up your simulation, check your highest frequency of interest and what frequency your mesher is using (might be going too high, and making too many cells). Also, turn on mesh reduction and turn off thick-metal models for your metal layers to speed up your initial simulation. Also, try reducing the amount of "stuff" on your design... if you are trying to model the loss of a differential line through a pin-field, then you don't need to include the rest of the artwork on the PCB.

thanks for the advices, but i have already considered these.
the only way to avoid a 2weeks simulation on my dual-core PC is to use slotplanes.

the slotplane method models magnetic currents inside the antipads. does this effect the trace s-parameters in the same way as modeling the plane-copper instead? i mean the antipad is the discontinuity, the rest of the copper does not hurt the signal.