If you are getting started with antenna matching, most of the information in books and on the Internet is either highly theoretical or overly simplified. This post will describe a practical procedure of matching an on-board antenna.
What Is an On-Board Antenna?
In the 80s, expectations for an antenna built into a circuit board were low. Some 46/49MHz cordless phones used them. People didn’t expect the same performance out of them that they’d get with a telescoping whip or rubber duck.
Most mobile phones now have built-in antennas, and users expect good performance. Three years ago poor performance of an iPhone antenna became a scandal.
Chip antennas use the least amount of board real estate. You can get a few dB improvement in performance over a chip antenna just using a trace that meanders away from the ground plane and all circuit components.
In either case, you need to match the antenna. Using a chip antenna does not obviate the need for a matching network between the transceiver and antenna.
Solder a Connector to the Board
Solder a small connector to the board near where the 50-ohm transceiver will be connected.
Depopulate parts such that only the matching network and the antenna beyond it can be “seen” from this connector.
I use connector of the U.FL or UMC style. These connectors are easy to solder down to the board and robust enough to mate to coax. One thing to be aware of is U.FLs have exposed center conductor under the part. Try to elevate it a little on blobs of solder to so that the bottom of the connector is not pushed right up against ground or other copper on the board.
Vector Network Analyzer
You need a vector network analyzer (VNA) to match an antenna. It must cover the frequency of interest and have a “test set” that allows the instrument to generate a signal and look at the reflected signal coming from that same port. Older equipment requires a separate “test set”. Modern VNAs have this built in.
The least-expensive VNAs available are from AEA. I recommend the most expensive one, which goes up to 2.5GHz and has built-in spectrum analyzer function.
If you need something better, you typically pay over $10,000. VNAs are widely available for rental, and in the low-interest-rate environment in 2013, monthly rent runs only 1/20th of equipment purchase price.
Calibration or Cable Nulling
It’s important to cal out the effects of the cable. VNAs have a calibration routine that instructs the user to present an open, short, and 50 ohm load to the end of the cable. 
A good cable won’t attenuate the signal much. If the wavelength is not much longer than the cable length, the wave’s phase will be different at the ends of the cable. If you fail to calibrate out the cable, you will get reasonably good information about how much power than antenna is reflecting back to you, but will not be able to work out changes to the matching network to improve it.
If the frequency is over 1GHz, you will benefit from doing the calibration right on the board, soldering an open, short, and 50 ohm right next to the connector. For VHF and lower, you can use an off-the-shelf cal kit or homemade connectors with 50 ohms, short, and open soldered onto them.
As the frequency exceed 2GHz, the unwanted inductance of a 0-ohm resistor begins to become significant. Consider using a capacitor at its series resonant frequency (SRF) in place of a “short”. This will not be a perfect short, but it’s better than a 0-ohm resistor.
Once this is done right, the 50-ohm load will result in the S11 trace staying close to the center across the spectrum. As a sanity check, you can test the cal by connecting a short, open, and 50-ohm load to the end of the coax and verifying the S11 trace appears near the left, right, and center respectively for these cases. If any of these are radically far from the expected locations, repeat the cal to make sure there was not an unwanted short or open at some point in the calibration.
Once you have it right, the VNA will show you the exact impedance seen at a point on the board right where the connector is mounted. You are ready to start experimenting with matching networks.
Next Week - Matching Procedure
Next week I will offer a procedure to use VNA readings to get to a good match across the band of interest.
