Network Analyzer question

[mdwebster]

I'm implementing a TI reference design for their CC2500 radio chip using a chip antenna. From the reading that I've done, I surmise that I will need to do final matching of the antenna to the balun after setting the board into its final plastic enclosure. I have a rough idea of how to do this using a network analyzer and swapping out components on a final Pi matching network.

However, my dilemma, hence my question, is how do you hook a network analyzer up to a device? Most analyzers I've looked at use N-type or SMA connectors. I don't really want to spin a second board with SMA connector pads coming off the antenna feed trace. Even if I did do that, wouldn't they change the line impedance to some degree? Is there some sort of RF probe you can use to check the traces directly without using some matching connector?

Oh, one last detail, the frequency range of interest is 2.4-2.5GHz.

Thanks in advance for any help or pointers to help...
Mike

 

[Mikebits]

In the past, I have used SMA to pcb connectors as in image. I solder the connector directly to the termination of the trace and ground. You have to calibrate the Network Analyzer setup with the same type of conn you use to solder to board. by that I mean you make a SMA conn with a short and the other open. So cal out the connector then when you solder one on the DUT(Device under test) the Z change is tolerable.

 

[mdwebster]

Yeah, I figured it was going to be something like that. Oh well, I guess I can always order a few extra boards and cut one of them with a paper cutter or a Dremel tool or something. Maybe just scoring it with an exacto knife using a metal straight-edge a few dozen times will be enough to snap it in two.

Thanks for the calibration tip too. I had a vague idea about needing to do some sort of de-embedding thing for the connector, but explicit direction is always appreciated.

 

[RadioRon]

My favorite approach, which I've used up to 2.2GHz successfully, works like this. Start with a blank or scrap version of your board. Choose a point along the feedline where you can cut the line or omit a component and solder a short pigtail coax to that point using extremely short exposed center conductor (like 2mm or less). Reflow solder on both sides of the coax braid to surface mount the coax to the board. Choose your connection point so that the matching components can be placed along the feedline (like perhaps choose to attach the coax in place of the balun?). So, now you have a coax feedline coming off the board, and the board has the antenna and its matching parts to get to 50 ohms. Your pigtail should be kept short, like less than 30mm as its length will add some error. The other end of the pigtail is an SMA connector.

Do a full 1-port calibration (open/short/load) at the end of your analyzer test cable, which should be an SMA. Your reference plane for the measurement of S11 is now at that connector. Now, when you plug your pigtail antenna board into this connector, the display is not accounting for the additionial coax length. You will have a phase and amplitude error due to the pigtail. We can remove the phase error quite easily and ignore the amplitude error. We remove the phase error by using Port Extension, like this.

Display S11 in Smith Chart form. With no connection of your test cable to your pigtail, the trace will appear to show an open, moving a short distance around the perimeter over on the right hand side. This is to be expected as you calibrated to roughly this point and it is presently open. Now, attach your pigtail. Now, depending on what is soldered down to your board, the trace will appear to rotate around the smith chart and is generally a bit of a mess. What we will do now is put a short circuit across the end of the pigtail, the end that is soldered to the board. We will do this using a tool that has a nonconductive handle and a tiny piece of metal in the tip. I use a tuning tool meant for use on tiny potentiometers but you may have to make something. It is important that the metal piece be small and have a straight edge. My metal end is about 2x2mm.

Before we actually apply the short, let's set up the analyzer. I will refer to the operation of the Agilent 8753E but others are similar. Go the CAL menu and select MORE and you will see the PORT EXTENSIONS selection. Push that and you face several choices of which port. Choose EXT PORT 1 if that is your test port. Also turn PORT EXTENSIONS to ON if there is an ON/OFF function here at the top of the list. Now note that the analyzer is displaying a value in picoseconds, currently at zero. Now is the time to touch your shorting probe between the tip of the coax pigtail as it touches the pcb, and the nearby ground of the coax, which should be very very close. Note that once you make this short circuit, the analyzer trace simplifies to a bumpy line roughly following the outside edge of the smith chart. This represents a short as seen through a random length of coax. Now, holding the short on the board, rotate the analyzer knob to increase the value of pSec displayed. As you do this, note that the trace rotates and slowly becomes shorter and shorter. As you approach the correct amount of delay in pSec, the trace will gather around the zero ohm point of the smith chart over on the left side. Stop rotating when you feel that most of the trace has gathered at that point. It will be a bit rough since your short circuit isn't perfect. At this point you might have a value of around 200 pSec dialed in. As long as you leave PORT EXTENSIONS switched ON, the analyzer is now accounting for the additional length of the pigtail, at least as far as phase is concerned. Hopefully, the amplitude loss of your pigtail is so low that the loss can be ignored.

By the way, I should have mentioned that the pigtails I use are made of an SMA connector mounted on a short piece of 1.8 mm diameter coax. Belden 83265 is the number on it. Fatter coax isn't very practical at these frequencies.

It is also wise to slide a ferrite bead or two over your pigtail to help isolate antenna currents from flowing down the outside of the coax. Choose a ferrite mix suited to the highest frequencies available.

Once you have done this port extension procedure, your reference plane is at the end of the pigtail coax. If the microstrip on your board is an accurate 50 ohms it is possible to calibrate at various spots down your line in the same way. If the line between your reference plane and your first matching component is very short, you can infer directly on the smith chart what value is needed there. For components further along, you will mentally have to adjust what you see be rotating your locus a little bit to account for the distance from the reference plane to that component.

 

[mdwebster]

Wow! Thanks, Ron. That was very detailed and informative. I, and I'm sure others far off in the future, appreciate your effort here.

 

[asdf123]

Thanks man