New for September 2025. On this page we will show you the advantages of a non-planar approach for a three-way Wilkinson divider. And we will show you preliminary, non-optimized microstrip layout. This page came about because we saw a photo of such a part on a company booth banner at the 2025 IMS exhibition. We will leave the company name anonymous because they did not show any interest in further sharing what was public information.
Ideal three-way using a star resistor
The star resistor approach is the ideal way to add isolation resistors. We used Microwaves101 transformer free download to calculate impedance values for a 150-to-50 ohm transformer for a three stage Wilkinson that spans 1.5 to 4.5 GHz. 150 ohms is needed because when you put three of them in parallel you are back to fifty ohms. Then we let the MWO optimizer come up with isolation resistor values to provide good spilt-port return loss (S22 in this case) and isolation (S31). In case you can't read the graphic, the line on the left is 40 ohms, the next stage is 87 ohms and the final stage is 62 ohms. We don't like the idea of using 87 ohms (it will be a very skinny microstrip line) , but this is not a real design so we will leave that alone for now. The resistor values are 43 and 95 ohms.
We see 20 dB return loss and isolations across 1.5 to 4.5 GHz, or 100% bandwidth. You would buy one of these if it was available, right?
Three-way planar Wilkinson
Let's take a step backwards and look at how much performance is lost in a planar approach with four resistors instead of six. The schematic below uses the same line impedances, but the four resistors have been re-optimized to 75 and 201 ohms.
Now compare the response. S11 of course is exactly the same. But the isolations have been compromised and only hit 20 dB from 2 to 4 GHz, or 67% bandwidth.
Three way non-planar layout
The heart of the design is what we call the "tombstone trio" of resistors. For this design we would use 0603 resistors. The center one is mounted vertically, while the other ones are tipped into it. Below is the schematic. The top node is not connected to anything but the resistors, in microwave parlance we say it "floats" above ground.
Here is how the resistors are mounted. The gray material is solder. You will want to get the lid on the module as soon as possible, it is not pretty!
Here is a schematic that uses physical elements so that it can generate a microstrip layout. We used a material with ER 3.0, and height 254 microns.
We used optimization to get the performance in the right ballpark, but this is far from a finished design. We are leaving a good bit of the possible bandwidth on the table.
Below is the layout generated by Microwave Office. The dotted area is an error layer, we used it to size up the product. It measured 44 x 31 mm. A good amount of the length is because we insisted on the three split arms being phase matched. The important thing about the layout is that the arms of each stage are bent up to have the same physical length, and they are brought together in close proximity so that the "tombstone trio" resistors can be mounted. Center-to-center they are ~1.3mm. The resistors measure 1.5mm by 0.76mm, so it seems like they will fit. However, the resistors are wider than the lines they will solder to. We need to redo this circuit on thicker material to increase the linewidths so that the resistors don't add capacitance to ground. Dagnabbit!
If anyone is interested we could continue this lesson and tweak it up.