pml with cst

Hello,
I simulate microstrip configurations, normally using metal boundary.
For open structures the boundary has to be sufficiently far from the strip.
This results in large calculation rooms, with many senseless mesh cells.
I think about it, if I can reduce the calc. room using the open boundary (perfectly matched layer PML).
My big metal box has, in some cases, only small waveguide ports (due to stabilty and higher order modes problems); so the incident wave can only leave the domain through this small ports. The rest of the wave will be reflected.
Using PML, every wave, which contacts the boundary will be absorbed (not reflected).
I already saw many antenna models using PML, but never a strip configuration.
Any ideas ?
Thanks

You cant use metalic boundary for an antenna simulation! You MUST use PML boundary...... all you have to do is correctly define your ports on the microstrip antenna and then put PML on all boundaries.

@aaron412
Right. But I simulate microstrip waveguide applications. No antenna problems.
The question was: Is it possible to reduce the calculation room with PML, when doing interconnection simulations.

I think you forgot to give the dimension of your PML vaccum box, so people cannot help much!
Just put your PML vaccum box at distance lambda/2 (minimum) from the closest microstrip.
It's a small box.

@ COSTOX
I take 5 times the distance strip to ground plane as distance to the copper boundary, when simulating transmission lines. To make sure the metal boundary doesn't serve as gnd plane for the strip.
The idea was to place PML boundaries closer to the strip.
Why do you say lambda / 2 ?

Here's what I've read from everyone:
1) Everyone know the VACCUM box has to be at least lambda/2 away from the microstrip.
2) only a VACCUM box can be PML (perfect match) because VACCUM doesnot reflect. METAL BOX will always reflect, at least a little bit, so cannot be set to PML.

@costox
Now, I understand. Of course there is a dielectricum, respectively vacuum or air around my strip.
Normally I used copper boundary for this dielectricum.
But I would be pleased, if PML boundary gives accurate results. If so, I could reduce the size of the air box to 2 times the strip-to-gnd-distance.
This is even more than lambda/2.

I think:
Actually you don't need any metal boundary. Just put an vacuum box, then select it, then assign boundary as PML (which is the same as OPEN). That's it.
So you design will have: air box, substrate, microstrip.
Thanks

@costox
Ok. So far, so clear.
But will this PML-model give the same accurate results as the metal boundary one ?
This would be perfect, because the PML model does not need this large calculation room, because there is simply no disturbing metal wall.

PML (also called open) is much more accurate because there's no reflection.