Broadband Four-way and Eight-way Wilkinson Example

Click here to go to our main page on Wilkinson power dividers

Click here to go to our page on multistage Wilkinson power dividers

This design was contribured by Alexander, it's a great example of giving back to the community.  This design cascades multi-stage Wilkison dividers (for wide band performance) in four-way and eight-way configurations, built up as circuit card assemblies. You can find the multi-stage Wilkinson spreadsheet that he used on our download page. and it is further described here.  From here down, we'll let Alexander talk to his design...(The Editors)

First of all THANKS A LOT for maintaining this great portal - being seasoned engineer I somewhat lacked RF specific knowledge and sure enough learned nice trade tricks and simple to the point explanations and guides. My desire was to build specific to dual-band WiFi (2.4 - 2.5 GHz and 5.0 - 6.0 GHz) splitter/combiner to complement my wireless security monitoring appliance design. I used your multi-stage Wilkinson calculator in order to get some sense in what is desired as well as followed your examples of PCB layout. While there are plenty of articles and examples of Wilkinson layout on a web, I somewhat didn't like them much for multiple reasons. By the end I think I came up with almost perfect design concept that I feel like deserves sharing with community (unless you think otherwise).  So, here you go.

Initial design goals:

1) Frequency coverage 2.4-2.5 GHz and 5.0-6.0 GHz

2) FR4 PCB material with Err=4.6 for cost purposes

3) Compact layout that would not have unnecessary traces for compensate losses presented by FR4 on high frequencies (based on PCB manufacturers reports losses at 6 GHz are approximately 0.4 dB per inch)

4) 1:4 and 1:8 designs that will allow easy transition to dual 1:4 and dual 1:8 assembly (rationale is that almost all modern WiFi radios are dual-chain and as such require feed off x2 antennas)

Preliminary calculations are based on a fact that I desired to have approximately equal losses in both bands (2.4-2.5 GHz has approximately 0.2 dB/inch and 5.0 - 6.0 GHz approximately 0.4 dB/inch losses). So here are my initial calculations based on a standard resistors values - multiplying resistor value x2 for each consecutive stage worked just fine.

After calculating all required traces  width for two layer 0.8mm FR4 PCB (I will explain later why 0.8mm was a choice), time came to design actual layout. After some experimenting and try/fail attempts I came up with layout that uses strictly 90 degree mitered bands, and presents a minimum amount of them. See below:

Notice a minimum number of bands to accomplish 3 stages plus a fact that traces do "run away" from isolation resistors very nicely. Distance between traces were calculated to be exactly "end-to-end" size of 0603 packaging resistors; I didn't like these resistors pads that are "sticking out" from main RF path. It is probably not a big deal, but I wanted it to be just perfect. So this is now what my 1:4 and 1:8 designs look like:

Note that this design is done using ONLY mitered bends without any unnecessary additional traces.

You can notice these additional pads and symmetrical layout - a reason why 0.8mm PCB was chosen is because I can mate two boards together back-to-back and "join" them using a regular 1.6mm SMA connector, which in turn creates extremely compact and efficient dual splitter assemblies that one will not be able to achieve using a legacy approach. Below are my assemblies - 1:4 (60mmx60mm), 1:8 (120mmx120mm), dual 1:4 (same 60mmx60mm) and dual 1:8 (same 120mmx120mm).