Hi, sas11!
My few thoughts here:
-
substrate seems quite thick
look at radiation patter at
fundamental mode ~400MHz (not the 900MHz, it's second mode!!) , it's slightly assymetric (maximum not in the antenna's normal direction).
I guess, pattern degradation is thus caused by unwanted surface waves excitation.
added: ground plane is also very thick, why?
-
simulation frequency
for TEM/QTEM feeders (with no cut-off frequency), simulation from f=0 speeds up calculation time-due to the shape of time excitation signal which becames gaussian pulse. for TE/TM you should set fmin>fcut-off, then it's excited by "sinus*gaussian" taking always more time to pass. It comes from the fourier properties of band-limited signals.
added: for high Q structures like this, increase accuracy (energy stop limit) in solver setting to say -50dB
-
boundaries
at X, Y boun. there should be rather "open add space" due to accounting intensive fringing fields. It will also put S11 closer to measurement, it approaches real-life problem more accurately. At Zmin you can put open only because of p atch radiation principle. With open at Zmin you'll simulate real life antenna with finite ground. Of course you need to add groundplane PEC brick there.
-
symmetry planes
for calculating only fundamental mode is OK but some higher modes will (thanks to this symmetry) not be accounted!!
Hope it helps..
eirp