Antenna Tuning Equipment

[hebnern]

I am designing a board that includes a WiFi module (TI CC3000) and chip antenna (Johanson 2500AT44M0400). My board is a 4 layer-board based on TI's CC3000EM reference design. This design includes a basic Pi filter matching network. I have selected similar components to what TI used, however, I realize that I will likely end up needing to do a bit of tuning to match the antenna perfectly to my board.

In the past few days, I have been doing lots of reading on the subject and have learned a ton about the required measurements and tuning methodologies. After lots of reading and lots more head scratching, I think that I am starting to understand the underlying concepts. Here is my current understanding of the antenna tuning process:

1. Measure the complex reflection coefficient S11.
2. Plot S11 on a Smith chart (and admittance chart).
3. Select parallel or series components that will move the plotted point to the center of the chart.
    a. Series inductors move clockwise along circles of constant resistance
    b. Series capacitors move counter-clockwise along circles of constant resistance
    c. Parallel inductors move counter-clockwise along circles of constant conductance
    d. Parallel capacitors move clockwise along circles of constant conductance

My big question at this point is how to accomplish number 1. I know that the normal way to accomplish this is to use a VNA, but I don't have access to one or the funds to buy one. Copper Mountain Technologies' Planar R54 Vector Reflectometer that looks like it would do the job for around $3k. I have also seen mention of using a dual directional coupler to measure return loss, but does this just give you the magnitude? What is the minimum equipment required to perform this type of measurement?

[madires]

Do you really need a perfectly matching impedance? If the antenna is built for the band of the WiFi module that should be sufficient as long as the impedances match roughly. And since the output should be limited to 100mW an antenna tuner wouldn't make much sense anyway.

[hebnern]

Do you really need a perfectly matching impedance? If the antenna is built for the band of the WiFi module that should be sufficient as long as the impedances match roughly. And since the output should be limited to 100mW an antenna tuner wouldn't make much sense anyway.

I guess I just assumed that some amount of tuning would be required. Everything I read seemed to indicate that its not as simple as just slapping an antenna on a board and being done with it. I had a very tough time getting another 900MHz board that I designed to work reliably, so I figured I would likely have similar troubles with this board. That said, I do not know how important it will be in the end. It is certainly possible that I am just over-complicating it.

My current plan is to build up the board, compare the signal strength indications to the reference design board that I have, and if its good enough, be done with it. If its lacking, then I will start looking more seriously at tuning. I just wanted to learn about what is required while I am waiting for the boards to come in.

[madires]

Most antenna tuners are based on a directional coupler for measuring the VSWR. You could check some RF antenna tuners for HAMs ( manual and automatic ones) to get the idea, for example http://www.mfjenterprises.com/Categories.php?sec=220.

[dr.diesel]

You won't find any Ham stuff that will reach 2.4Ghz.  Array Solutions makes a reasonably priced unit but it hits a max of 1Ghz, I have the 4170C version, but only good to 180Mhz.

Anything capable of accurate measurements in that range will be cost prohibitive, go with your calculated values and you'll be ok for the power levels involved.

[nctnico]

AFAIK a spectrum analyser with tracking generator or external generator and a directional coupler should be sufficient. First measure the baseline with the directional coupler terminated with 50 Ohm and then measure again with the antenna. The cable between the directional coupler and the antenna should be really short (a few cm).

[Rory]

Take a look at Prof. Dellsperger's page, particularly the downloads section. He has a great Smith chart application that you may get a lot of use from.

http://www.fritz.dellsperger.net/

[larry42]

How are you planning to actually do the measurement (i.e. to go from a multilayer board to the coax of your network analyzer?

I've just had look at the CC3000 EV board manual, the datasheet and the antenna datasheet.

The antenna datasheet gives a rather different antenna tuning suggestion (and a much wider BW than the CC3000 EVAL board).

To be honest I would follow the EVAL board layout, layering and layer thickness recommendations, buy the CC3000 EVAL board recommended matching components, buy the Johanson datasheet matching components and mount to see which performs better. NB digikey has johanson low loss cap and inductor kits for reasonable prices - I bought these and I like the Johanson components - way cheaper for High-Q caps than ATC!

Note that even a return loss of 6dB (3:1 VSWR) will hardly make a real difference to the signal strength, we're talking about a 1-2dB loss.

Regarding the first question - yes, you could calculate L & C values, once you've figured out your reference plane, and if you have accurate models for your "lumped" components. If I had to really match this board then I would make a TRL cal kit, cut up some boards and put them in an Anritsu 3680K test jig, hooked up to a VNA.

[hebnern]

AFAIK a spectrum analyser with tracking generator or external generator and a directional coupler should be sufficient. First measure the baseline with the directional coupler terminated with 50 Ohm and then measure again with the antenna. The cable between the directional coupler and the antenna should be really short (a few cm).

From what I have gathered, a spectrum analyzer + tracking generator is essentially equivalent to a Scalar Network Analyzer. This allows you to measure the magnitude of the return loss, but not the phase. Thus you can tell how well your antenna matches your characteristic impedance, but you can't really tell what components you should use to correct it. For that you need a Vector Network Analyzer or Vector Reflectometer. Magnitude is better than nothing though and can be used to guide a guess and check approach, which might be the best option based on the cost of a real vector analyzer.

[hebnern]

How are you planning to actually do the measurement (i.e. to go from a multilayer board to the coax of your network analyzer?

The antenna feed line is on the top layer (grounded coplanar wave-guide), so I figured I could just solder an SMA connector to the transceiver pad on an unpopulated board and use that to connect to an analyzer.

I've just had look at the CC3000 EV board manual, the datasheet and the antenna datasheet.

The antenna datasheet gives a rather different antenna tuning suggestion (and a much wider BW than the CC3000 EVAL board).

To be honest I would follow the EVAL board layout, layering and layer thickness recommendations, buy the CC3000 EVAL board recommended matching components, buy the Johanson datasheet matching components and mount to see which performs better. NB digikey has johanson low loss cap and inductor kits for reasonable prices - I bought these and I like the Johanson components - way cheaper for High-Q caps than ATC!

Note that even a return loss of 6dB (3:1 VSWR) will hardly make a real difference to the signal strength, we're talking about a 1-2dB loss.

Ok, so it is sounding like I should just use a guess and check approach. Try some different component values and see what works best. I will order a couple of those capacitor and inductor kits.

Regarding the first question - yes, you could calculate L & C values, once you've figured out your reference plane, and if you have accurate models for your "lumped" components. If I had to really match this board then I would make a TRL cal kit, cut up some boards and put them in an Anritsu 3680K test jig, hooked up to a VNA.

This approach is mainly what I was wondering about, but it sounds like its probably out of my reach and that it wouldn't really make much of a difference for this application. Really appreciate the tips.

[larry42]

How are you planning to actually do the measurement (i.e. to go from a multilayer board to the coax of your network analyzer?

The antenna feed line is on the top layer (grounded coplanar wave-guide), so I figured I could just solder an SMA connector to the transceiver pad on an unpopulated board and use that to connect to an analyzer.

At those kind of frequencies it's not really possible to "just" solder a SMA to a pad, at least not when you want to find matching inductors in the ~nH range. BTW, I think that their feed-lines are microstrip for all practical purposes - the top ground is too far away to call it GCPW.

[hebnern]

At those kind of frequencies it's not really possible to "just" solder a SMA to a pad, at least not when you want to find matching inductors in the ~nH range.

I meant that I would use a semi-rigid coax cable with an SMA connector. I have seen a whole bunch of articles that recommend hooking up to the antenna in this manner. Why don't you think that this would work?

BTW, I think that their feed-lines are microstrip for all practical purposes - the top ground is too far away to call it GCPW.

This is another topic that I spent a lot of time researching. TI isn't very consistent in their descriptions. In some places they say that the feeds are microstrips, in others they say that they are GCPW. Their calculations for feed-line geometry from the CC3000EM manual seem closer to GCPW, although I could not get much closer than around 60 ohms with the GCPW calculators that are available but microstrip calculators are even further off. From what I have read, typical GCPW calculators break down when the feed-line width approaches the L1/L2 separation distance as is the case with TI's example where the feed-line width is 14.3 mils and the L1/L2 separation is 12 mils. In this case an RF simulator like Sonnet with its thick metal model is required. I was able to get a bit closer to 50 ohms using some free RF simulators. In the end, I just went with the geometry from the CC3000EM gerbers (which is different from what they describe in the manual).

[vk6zgo]

Ultimately,the object of all "antenna tuning" is to maximise the radiated signal.

You could make up a field strength meter & adjust for a maximum,or just use the signal strength indication on the associated receiver.

[KJDS]

I'd start using a network analyzer, even if it meant hiring one, however once I was close then the approach stated below is the final step.

Ultimately,the object of all "antenna tuning" is to maximise the radiated signal.

You could make up a field strength meter & adjust for a maximum,or just use the signal strength indication on the associated receiver.

[leafi]

How did your antenna tuning go?

We need to do the same thing at work for a blue tooth antenna. I have my personal spectrum analyzer with a tracking gen which can go up to 3GHz but I do not have a directional coupler and cables for 2.4Gz. (Nor the calibration kit). When I add the 50 ohm coax without the 50 ohm terminator to the in port of the directional coupler like it should be hooked up I can see the Return loss increase to the same amount that it does when I add the 50 ohm terminator to that 50 ohm cable. I need better cables and terminators and a directional coupler which can do 2.4Ghz. Mine is only rated to 2Ghz and mini circuits said that the VSWR starts to get really bad beyond 2 GHz.

On our board we will not have room for a SMA connector so I may make a SMA pigtail and solder it to the board. Then make a separate SMA pigtail with the connector so I can calibrate the sweep. I'm not sure how much this will affect the results and I'm not sure how much having the pigtails will affect the results.

I think I will have them just rent a vector network analyzer to do it correctly. Additionally it is calibrated and we will rent the calibration kit and cables and / or buy what we need.