Momentum bugs

Updated July 11, 2009

Update July 2009! The bug is still there. When will Agilent get around to fixing this? Or even explaining it?

New for December 2008! Momentum is the electromagnetic solver that is offered by Agilent. It works great, does some amazing things, but once in a while it gives bad results. Hopefully you'll recognize bad data when you see it!

Have you ever seen this warning when you are running a Momentum simulation?

WARNING :

The geometry contains parallel ports which will be simulated as uncoupled ports. This may yield incorrect results if the ports are too close together. To simulate them as coupled ports modify the layout so that the ports coincide with one common reference line.

Every day you probably get a hundred garbage messages from your computer, warnings like this are like the boy who cried wolf. This message might not be anything, and it is certainly not something that is understandable, so you ignore it.

We uncovered a bug that is serious which triggers this warning, and it will provide you with very bad data below about 10 GHz. Or as Agilent says, "unexpected response". Below is a simple two-port circuit with both ports facing the same direction.

Here's Momentum's predicted response. Port 1 looks like fifty ohms at low frequency as it should.

Here we've modified the circuit, so the two ports are on the same side but at different latitudes:

This generated the warning you see above. Now look at the response. What the heck?

The circuit response above about five GHz appears normal. By the way, the problem goes away if you enable "RF mode" which will reduce accuracy at higher frequencies. It also goes away if you never have ports facing the same direction. So when you mesh up a four-port network for analysis, be sure to cut your circuit so that the ports face North, South, East and West.

Agilent doesn't have an application note that covers this problem. We'll leave this page right here until they either do that or get rid of the bug. Why should we have to move the ports to they are in "one common reference line" like the warning suggests? How about instead of a warning, you fix the the math that causes this annoying and unnecessary bug? Sonnet never had this problem...