Wilkinson Power Divider
I have this simple question, I know that a wilkinson power divider is to be implemented with lambda/4 transformers of sqrt(2)Z0 each and a resistor of 2Z0, but my doubt is whether this same divider can be implemented with 3*lambda/4 and 5*lambda/4 or for that matter any odd multiple of lambda/4 but of unequal lengths, I understand that this is theoritically possible, but is that practically good. Has anyone tried it. WHat did u observe?
Waiting for reps,
Thanx in advance.
The problem is routing the transmission lines around to connect to the resistors. One problem with the multiple quarter wave length lines of constant impedance is the narrow bandwidth. You are better off cascading different impedance lines to widen the bandwidth.
Hi,
There are several drawbacks of having long(er than needed) lines.
1. Losses will increase
2. Exact phasing of the divider network is difficult in practice. As Flatulent poined, your bandwidth will narrow even in theory when you add extra multiples of quarter wavelength. However, in practice things are going to be even worse as exact electrical length is critical for optimum matching and isolation between outputs. Sections should be exactly 90 deg and resistor have no electrical length (0 deg). Real resistors have electrical "length and width" so real situation is not as in the drawing or simulator where sections and resistor are connected in a dot having zero electrical length. The more extra electrical length you add, the more dispersive this non-ideality will be.
However, there is another point considering length. As frequency increases, geometrical length of sections is shrinking. It is important to shrink down the size of a resistor as well and to narrow the microstrip line width as well to keep this connecting point and resistor as "lumped" as possible comparing to the length of the section (quarter wavelength). Typically, geometrical length of a resistor should be 1/10 of section length, this is from an egineering practice thing.
flyhigh