I use ADS to simulate a simple microstrip line. but the result seems big different between model and EM mode.
does anyone know the reason ? and which one I should trust ?
Dielectric is FR4, the height of the substrate is 0.9012 mm, the width of the line is 20 mil.
There can be many mistakes in measurement (i.e. connectors, VNA port calibration)) and simulation.
From your S11 plot, it seems that electrical length is rather different. To avoid trivial mistakes in VNA port calibration, measure one coax connector which you short directly at the PCB side. Make sure that this appears as a short in Smith chart, with no additional phase offset.
There can be many mistakes in measurement (i.e. connectors, VNA port calibration)) and simulation.
From your S11 plot, it seems that electrical length is rather different. To avoid trivial mistakes in VNA port calibration, measure one coax connector which you short directly at the PCB side. Make sure that this appears as a short in Smith chart, with no additional phase offset.
I see ... I was confused because you didn't follow the usual workflow for Momentum model results as an emModel view (link), but this doesn't explain your differences.
For sure, this difference is not expected, so I recommend that you start all over again. My best guess is that you have compared to some outdated Momentum results obtained with other geometry parameters (width w is a tuned variable). If you use the emModel workflow from my link above, you don't need intermediate S2P files, and reduce the risk of comparing apples to oranges.