wilkinson combiner

I am new to PCB design. I am making a 2-section broadband Wikinson power combiner on PCB centered at around 2.5GHz. I am running into some problems, and I wonder if anyone is willing to help.

This combiner needs to be able to handle about 1W, and I rate the two resistors to be 500mW. I am using Agilent's ADS 2008 to do my simulations (as shown in the attachment, the impedances of CPWs are 60 Ohm and 85 Ohm). When I replace ideal resistors in my simulation with BCR1/2237F and BCR1/297 SMT resistors from the ADS library, the gain is then sharply reduced at high frequency (more than 6dB loss at 2.5GHz). I guess, this is the result of the modeling of parasitic to-substrate capacitance of 0.5W resistors (2010 package). If my guess is correct, is the modeling of parasitic capacitance of resistors in ADS fairly accurate or over-pessimistic?

Any suggestions to get reasonably less loss?

Thank you very much

I don't get you. 1 Watt goes in, 0.45 watts go out one port, 0.45 watts go out the other port. How do you get 0.5 Watts into either of the isolation resistors? Are you using cold fusion or something.

Analyze the worst case dissipation, including if one of the outputs ever gets open circuited.

To biff44,
It is a combiner. 0.5Watt goes in one port, another 0.5Watt in the other port, 1 Watt out if in phase and 0 watt if completely out of phase. So I rated the resistors to be 1/2W. Thank you

combiner, splitter...they are the same thing. >5 watt in one port, .5 watt in the other port, .95 watt out the combined port. How can you get 0.5 watts dissipated in both isolation resistors? You would have to be putting in 1 watt each into both inputs, and getting only 1 watt out, to do that.

To biff44,

.5 watt in one port, .5 watt in the other port, you can still get almost 0 watt at the combining port if the two signals come 180 degree out of phase (odd mode). Of course you should get about 0.95 watt if two signals come in phase (even mode). In my specific application, two signals at two input ports are not necessary always in phase (in fact, occasionally two inputs would come in 180 degree out of phase), so I need to rate the resistors' power high at 0.5 watt. Hope this makes some sense now.

If you use it as the splitter, it is guaranteed to be in phase (even mode), and resistors always dissipate almost no power. In my case, they are not.


Anyway, 0.5 watt resistors need a package as big as 2010... And ADS simulation tells me that's bad in terms of parasitics. Any suggestions? Also if you spot any other mistakes I made in the schematics attached, please let me know. Please help. Thank you very much!

Don't worry zhipeng, Biff44 is gruff, but lovable.

When combining uncorrelated signals, half of the power apprears at the input and a total of half the power dissipates in the resistors. 1/4 each resistor.

If combining 2 signals 180° out of phase, no power appears at the input, all the power is dissiapted in the resistors. 1/2 each resistor.

Of course, the resistor is an open circuit to 0° correlated signals.
If combining 0° correlated signals @ -3dBm, 0dBm appears at the "input".

You ^&&%S&er!

You are right. If you are combining 180 deg out of phase signals, then you are dumping power into those resistors. What misled me was that you were using a wilkensen structure. Most people either use a 4 port magic tee or hybrid coupler for such a combiner--that way you can have a nice fat 50 ohm resistor load on the fourth port, mounted to the housing so that it can disipate power all day long.

If you like the wilkensen structure, I would suggest using ALN resistors, which can handle .5W and still be small. Your only problem is getting the heat out of the resistor chip and into the housing. Soft boards do not conduct heat so well.

Thank you very much biff44.

You are right. For a hybrid coupler (rat-race) where the isolation resistor is grounded, the gain could not be as vulnerable to the parasitic capacitance-from-pad-to-substrate of the resistors.

But it is not as immediate to me how to make the rat-race broadband. It is easy to do 2-section Wilkinson.

Any recommendation on specific AlN resistor vendor?

Thank you again!