Using a slotted line (poor man's VNA) to build a matching network...

[Tj256]

I think you'll find this is an easy question if you are well versed in taking impedance measurements with a VNA (or even a slotted line).

I believe my setup is flawed in some subtle way. I am able to plot changes on a smith chart as I add/remove components to my PCB using the technique I am about to describe, but sometimes things don't work quite as I'd expect. The parts I'm using are thin film, designed for 5.8 ghz.

I'm building a PA based on the HMC408 and I have to match a 4 ohm output (with maybe a slight imaginary component) with a 50 ohm connector. Mind you, I'm at 5.8 ghz so there is probably little room for slop. I thought I'd use my slotted line to help me measure the impedance to make sure it will work.

My test setup looks like this:

Signal generator---> Slotted line ---> Output connector on the PA board (what I'm working on)

I replaced the MMIC with a 50 ohm resistor on the PCB, and took some measurements with the slotted line. Don't get too hung up on this though. I am easily able to measure SWR (to draw the circle on the smith chart) and see the phase shift from my changes relative to a short I calibrate with. Things move around on the smith chart MOSTLY as I'd expect.


The signal path on the board looks like this :
1)SMA connector
2)50 ohm TX line
3)Shunt capacitor (1.3pf)
4)Inductor (.4nh)
5)50 ohm TX line
6)PA chip (50 ohm resistor at this point)

I have a program called "smith tool" that lets you drive around the smith chart by inserting components into a matching network. If I "drive around" from the center of the chart starting at 50 ohms and part #1 above, I can get to my desired impedance with the above list. The final inductor and TX line rotates clockwise around the chart to hit the desired impedance.

I'm finding that in real life, that final inductor and even the txline has NO EFFECT on the final result. Almost as if it doesn't count. This has been verified as the behavior by not only my slotted line measurements, but by a software program I use called "ansoft designer", if I try to plot S11 on the smith chart on the same network. I can boost the inductor to 1 or 2 nh and still no change in the measured impedance.

In other words, that final txline should be adding a phase delay, and rotating the measured impedance. It doesn't.

Clearly I am missing something here. Is this a valid way to measure my matching network to make sure it's "tuned", and why are these inductors/transmission lines having no effect on the measured S11?

I'm hopeful I can avoid buying a VNA. And again, I don't think this is the problem because the simulator shows the same deal. Try it yourself: Port 1 goes to a shunt 3pf cap. Then through an inductor. Then through a txline. Then to port2, a 50 ohm port. You'll notice that S11 doesn't change as you'd expect if you play with the inductor/txline.

Maybe I am not doing this right...

 

[Tj256]

READ FIRST:

I may have figured it out!

Ideally, I want to be able to plug my matching network into my slotted line and measure the 4 ohm impedance on my smith chart.

If I use a smith tool such as "supersmith", I notice that if I set the load at 4 ohms and the source at 50 ohms, I get the match using a shunt capacitor(1.9pf) and series inductor(.37nh) just fine.

Now if I then change the 4 ohms to 50, so both the source and load are 50 ohms (as in my test) suddenly the smith chart shows an impedance that matches the crazy value I was seeing with the slotted line!

So it appears that if you use the slotted line, and you terminate the other side with 50 ohms, you are going to wind up trying to match to a different location on the smith chart.

Maybe the answer is to terminate with 4 ohms? Someone out there has to know the procedure for this.

The attached image shows it best... Check it out and it will be obvious what I'm seeing.

Thanks so much for any help.

 

[Tj256]

And finally....

I believe what is going wrong is that the order you build up the network, and the side you inject the signal, and the termination have to be all correct, or you get bad results. Duh.

What seems to work in the simulator is to inject the signal at the PA side (opposite of what I was doing), terminate the output with 50 ohms, and build up the network from the output on backwards. This lets you incrementally move on the smith chart from the center (50 ohm output) towards the desired 4 ohms (PA impedance) step by step.

You just have to keep resetting the reference plane along the way, so all the measurements are from that component forward. This is easily done with a short.

For my case, I'd put the reference plane just before the capacitor, zero out (measure) the phase, then replace with the shunt capacitor. Then move backwards towards the PA and repeat for the inductor, and so on. In this way I'll build up the matching network and move on the smith chart from the center to the 4 ohms in an intuitive, step by step way.

I hope this works when I try it tonight.

 

[biff44]

Some comments:

Adding a series length of line does NOT change the standing wave that a slotted line sees. It only changes the phase (or physical location) of where a SWR minimum is seen.

Trying to match to a 4 ohm impedance is....hard. There is not much difference between a 4 ohm impedance and a zero ohm impedance, as far as a slotted line goes! But there is a really big difference as far as a matching network goes. So expect to have a lot of trial and error in your matching network attempts.

Maybe, to get the hang of it, you try to make a matching network for a 25 ohm load. That should be a little easier, and you can learn the process before diving in for a 4 ohm load!

http://www.MaguffinMicrowave.com

Maguffin Microwave wireless design consulting

 

[Tj256]

Thanks! Well the more I play with the slotted line, the more I realize that I really need a way to move the test point around the board instead of the short. I think my method is measuring the end result of the network. As you move the 4 ohm termination back toward the PA, adding in more components, you are adding them to the beginning of the network.

The end result is that it's easy to goof up and not be able to figure out what you need to do to adjust the network.

It seems better to move the test point where the slotted line measures from slowly down the chain. But no idea how to do that reliably at 5.8 ghz.

I wonder what the pros do? I mean, when they're building up a matching network, can or do they go through step-by-step moving a termination or test point like I'm describing? And if so, what's the procedure?

I think my conclusion is that you really have to have your network already designed pretty close, then when you terminate the PA with it's 4 ohms (resistor) and measure the other side, you can only really measure the impedance it's presenting. You can only really make sure that's 50 ohms. If not, you'll have to play with the simulated network on the PC to see what you need to tweak.

Maybe it's a dream to be able to do it step-by-step like I wanted.