Hybrid power divider vs ratrace power divider vs Wilkinson power divider.

Equal power divider can be realized in many ways. Simplest is T-junction microstrip power divider. Next three types provide isolation: Wilkinson power divider, provides isolation between two output ports. Hybrid power divider (with one port terminated with 50 Ohms). And also ratrace coupler can be used as power divider (with one port terminated with 50 Ohms).
I have seen only T-junction, Wilkinson and hybrid-coupler power dividers.

Why some designs use hybrid coupler power dividing (phase shift is not required). One part of signal is goes to TX, another part to RX path. Why not using simple Wilkinson?

Anyone used rat-race as a power divider? As am mixer it has wider bandwidth than hybrid couplers, i guess as a power divider it performs better too?

I am sure all those cases could be compared in simulation, but i can't guarantee i can design each one at it's best performance, so comparison would be wrong.

Any of those mentioned (T-junction, Wilkinson, Rat-Race, Hybrid) can be used as a 3dB power divider.
What is making the difference, is the isolation between the divided ports.
Wilkinson is the best, because can get more than 30dB isolation, when T-junction has only 3dB isolation.

I can't understand why hybrid coupler used for power dividing sometimes. Not transceiver, just simple power dividing. Why not Wilkinson power divider?

It can be fully printed with no resistor? Just a guess.

Maybe. But usually unused port is terminated with 50 ohm resistor, but maybe it is not necessary, maybe it can be matched open circuit at narrowband.

I tried to do simple S-parameter simulation of three configurations:
Wilkinson, ratrace and hybrid power dividers.

Ratrace - best isolation >-40dB , bandwidth is somewhat less than Wilkinson
Wilkinson - best pass bandwidth, but isolation only about -30dB
Hybrid - narrowest bandwidth, isolation is between ratrace and Wilkinson.

I only used simplest configuration (50 and 74 ohm lines. Hybrid is made in 74Ohm system)

If you do an EM simulation using standard FR4 you can see that Wilkinson still keep good isolation when Rat-Race isolation drops a lot.
On the other hand, on Wilkinson you have 3 ports where have to keep good 50 ohms termination and 2 TL with constant impedance vs frequency, when in Rat-Race there are 4 ports and 4 TL where have to do the same.

I did FR4 simulation with Er = 4.2 and tangent loss of 0.020, i do not see bad isolation of Rat-Race. Tried different ring lengths for frequencies from 5 to 10GHz. Unused port is terminated with 50 Ohm ideal resistor. If resistor is removed, then isolation is very bad (tried open stubs). I only see that Wilkinson divider is not sensitive for Er fluctuations. S21 and S31 of Wilkinson is always pretty good (Er=3.8, Er=4.5, etc.), and isolation is ok. With rat-race i see very good isolation, but center frequency of best isolation can move a lot with Er tolerances.
Thank for your note on FR4, i am going to try both dividers, make prototypes and compare them.

50 Ohm terminations are very important for coupler performance. I am currently trying to achieve best termination using smd resistor. The best S11 i achieved was around -8dB! It seems that it is very difficult to achieve using vias. Now trying narrowband solution using quarterwave stub for RF ground.