Circularly Polarized Patch Antenna A 77 GHz - Axial Ratio Bandwidth

Hi all,

I'm currently trying to design a circularly polarized patch antenna in CST for a project at university. As mentioned in the title, the antenna has to operate in the W-band with a center frequency of 77 GHz

Other given properties are e.g. the substrate (Rogers RO3003, e_r = 3, h = 0.127 mm) and a minimum line width of 50 μm. Moreover, the patch has to be fed with microstrip transmission lines (based on Z_0 = 50O).

Regarding this requirements, I decided to use a square patch with dual feeds like shown here

Since space for the feeding network is kind of unlimited, I decided to use a 2-stage quarter wave transformer in every feed branch to match the input impedance of the patch (approx. Z_in = (350+j0)Ω at 77 GHz) to the 100Ω transmission lines (50Ω -> 2x100Ω) at the T-junction power divider. One of the 100Ω lines is ΔL = λ/4 longer then the other one.

This works quite well with an axial ratio of < 3db for a beamwidth of θ = +/-10° BUT unfortunately just in a quite narrow band (< 1 GHz)

I guess poor axial ratio bandwidth is neither due to the frequency-dependent quarter wave transformers (since they are identical for both feed branches), nor result of the frequency-dependent 90°-phase shift in the 100Ω lines (I tested the feeding network with 2 discrete ports instead of the patch and the deviation from 90° should be about +/- 4° between 76 and 79 GHz).

For this reason I'm afraid that it's due to Z_in of the patch (complex for f≠fr) which is finally transformed differently by every path. Which means that two quite different impedances can be "seen" at the outputs of the T-junction and power is splitted in a way that finally one of the modes is excited more then the orthogonal one.

I am sorry for this extensive description
but I was wondering if my assumption is right so far and if there might be a simple possibility two enhance the impedance bandwidth and thus axial ratio bandwidth? (without changing the substrate properties but maybe by using another patch geometry or parasitic microstrip elements)

regards, Tom