(All images are at end of article)
Connecting RF test and measurement equipment to an Antenna Under Test (AUT) usually involves tradeoffs among measurement accuracy, electrical considerations, cost, and mechanical ruggedness. In this article, we describe some aspects of this connection with practical advice for such test and measurement scenarios. Some of the items used are shown in Figure 1.
Testing via an RF Test Connector: U.FL vs Coaxial switch
When provided, an antenna test port will usually be at a 50 ohm point. If the design includes a coax switch for inspection and final test purposes, an antenna test sample can be produced or reworked with the switch rotated 180°. This reverses the in and out terminals so connecting a test cable interrupts the feedline and the cable faces the antenna instead of the device.
A test port and test cable of the appropriate type reduce the need for strain relief as the cable can be connected just before measuring. A right-angle connector at the device end may further reduce stress and allow easier routing through the device housing. Commercially prepared cable assemblies are available from Amphenol, Crystek, Emerson Network Power Connectivity/Johnson, Hirose Electric Co. Ltd, KSM Electronics, Molex, Phoenix, Pomona, RF Solutions, Samtec, Taoglas Limited, and TE Connectivity. The website www.digikey.com is an excellent source for these. Note that most ultra-miniature connectors have a mating life on the order of a few dozen cycles, but failure will become self-evident in the measurements.
The example PCB in Figure 2, left has a U.FL connector on a very short stub. One can isolate the WiFi device by omitting the 100pf cap and installing the U.FL to obtain coaxial access to the antenna. Alternatively, and as shown in Figure 2, right, an RF switch can be installed, allowing for testing either the WiFi module or the antenna.
Custom Cal Standards
Lumped element tuning for PCB antennas requires that your VNA calibration is in the proximity of the next series or shunt element to be added. You have options:
Use a coaxial calibration standard, usually to an SMA interface, then extend the reference plane across the length of your board connection. This extension may compensate for an SMA-U.FL adaptor, or maybe a short cable. Add reference plane time until the “Open” is at the right edge of the Smith chart.
Create a home-brew set of calibration standards, as shown in Figures 4-7. Take three sparsely populated boards, add 50 Ohm shunt, 100pf shunt (short) and one left open.
If the reference plane isn’t extended by either custom calibration standards or by manually extending the reference plane, then matching efforts will be extremely confused because the Smith chart will be rotated.
In Figure 3, this test board omitted the WiFi IC, installing a U.FL connector instead. We made four of these boards, one each configured for Load, Open, Short and Thru to the unmatched antenna.
Now we could calibrate to the entry point of the antenna with the short, open and 50 ohm load. In production, there are no extra components or space required.
Alternatively, if we had used coaxial calibration standards (N or SMA), we would subsequently “extend” the reference plane by an adjustment in the VNA, usually called Port Extension. How far? Use the Open and Short configurations (shown on the right) and extend the reference plane until these align with the right and left edges of the Smith chart.
If your instrument and cabling mass is substantial compared to the counterpoise (ground plane) of the board, add a DC-DC connector prior to calibration. When you start to match, move the reference plane (Port Extension) backwards, bringing the reference plane back with you as you add tuning elements.
Tuning a Loop
The keyfob design in Figure 9 was based on a radio chip with a differential output. The round shape of the board and differential drive suggested a loop antenna, which is less susceptible to detuning than a monopole, dipole or folded F. Dielectric detuning happens when the keyfob is held by different hands or in different orientations or sits on different surfaces.
On the right (in Figure 9) is the production board with differential RF feed at the top of the image. On the left (in Figure 9) is a test board with the same antenna and same ground plane. Notice tuning elements both at the entry of the differential drive and at the midway point. L5 enables the loop to be extended, in case simulations (or back-of-the-hand guesses) are off when judging loop size/characteristics.
Testing Without a Coaxial Connector
Sometimes there just isn’t an opportunity to have a U.Fl or coaxial switch. Maybe it was overlooked in design, or maybe the product requirements were just too tight. Here are some coaxial cable options if you need to forge a reliable transmission line connection.
First, you will most likely want to identify a position on your board where the impedance is, or is expected to be, 50 ohms. This isn’t always required, but simplifies things. Remember that you still face two calibration/reference plane options — calibrating with normal standards at the “connector” end of your coaxial cable, then extending the reference plane, or building multiple boards and implementing a custom “calibration set.”
The latter option can be a problem in this case, however, because it requires that your coaxial cable have the same electrical length on each of 3-4 boards, or that you have to regularly solder your Open/Short/Load values in and out of place.
50 ohm matches are increasingly typical at the output of mixed signal RF SOC ICs, but if not, review your options at your duplexer, diplexer, balun or within your RF matching circuit to the antenna. The ideal test point has close by access to the antenna’s ground plane, preferably one well reinforced with vias.
Whether your coaxial cable attached with “horizontal” or “vertical” launch, the goal is to maintain the TEM modes that exist between the center conductor and shield of the test cable the feedline microstrip and its ground plane through the transition between them. An excellent discussion of transition issues and some of the alternatives can be found at http://www.artechhouse.com/uploads/public/documents/chapters/holzman_941_CH04.pdf
A 50 ohm microstrip feed line may be as little as 0.03 to 0.1 mm (.015” to .025”) wide if over an embedded ground plane of a thin circuit board assembly. Protecting such a thin trace from being pulled up by the test cable requires a mechanically sound connection between the cable sheath or shield and the ground plane, preferably in an area where the ground plane is reinforced by vias.
For a horizontal launch, to the extent possible the test cable should access the broken microstrip feedline trace directly from the ground plane and in the direction of the feedline. Keep the feedline and center conductor and their separate ground return currents close by, in proximity, parallel to each other, and minimally different in length. Avoid 90° connections and bends until both currents are safely within the test cable. Eisenhart’s “edge-launch design” (described at the Holzman link above) can be approximated even if the test point is not on the edge of the circuit board.
For a vertical launch, it may be practical to break the feedline trace and reflow solder a U.FL or similar surface mount connector to the feedline at the break and the ground plane on both sides. This gives the ruggedness advantage of the connector, albeit at the presence of some reflection. If this isn’t practical, the device doesn’t need to remain functional, and ground plane is accessible on the other side of the circuit board assembly beneath the feedline, the approach below may work. A clearance hole is drilled for the test cable dielectric, which is trimmed to the board thickness, and the emerging center conductor soldered closely to the feed trace at the break. The shield should be tightly flared so as to be soldered closely to the ground plane beneath.
Using a Balun
A further refinement might help exclude the test cable from the measurement if the DUT has insufficient ground plane and common mode currents on the cable result. This can be done by forming a 1-turn loop around the minimum bending radius for a semi-flexible or hand-formable test cable, or with a ferrite bead or two placed over the test cable where it departs from the ground plane. While ferrite beads are not characterized above 500 MHz and all of them lose performance as frequency increases, they are inexpensive and unlikely to impair the test results. Type 61 ferrites are usually recommended above 300 MHz but Type 43 is available in a greater range of dimensions and may work.
A bead can be selected for fit and evaluated by placing a sample at the shorted end of a coax stub and extending the R54 port to that point. As a useful example, the Kemet B-20L-44 (Digikey 399-10822-ND) in an RG-178 stub demonstrated |Z| in excess of 58 ohms out to 3 GHz, which includes the GSM, UMTS, and LTE to Band 7. It was in excess of 30 ohms out to 5GHz. Any common mode current from the DUT can be arbitrarily reduced by stacking multiple beads.
Cable Length and Type
Regarding the choice of cable, here are some points to consider:
Test Cable: The test cable should be long enough to place the test instrument out of the near field of the AUT and preferably a few wavelengths away. This reduces reflections off the test instrument itself so the antenna measurement will better approach the limits of the test environment.
Test Cable Type: The thinner and more flexible the cable, the less strain relief required for ruggedness under test; on the other hand, the thinner the cable, the greater the attenuation loss, which limits the measurement accuracy. Table 1 is a table of cable types typically available in commercial cable assemblies.
If the required test cable length introduces significant loss, it may be possible to include the coaxial adaptors and test cable in the instrument calibration using a tightly constructed Short, Open, and a 50 ohm chip resistor on a brassboard of similar construction to the DUT.
The typical integral antenna in the 2.4 GHz ISM bands and up is a compromise to begin with because of multiple band requirements and size and space constraints; accordingly, the typical SWR may be far more significant than reflection or attenuation from a less than ideal test cable interface. Every device will be different, but keeping the TEM modes in mind in making the test connection won’t hurt and will make for the most accurate test.
If you have questions about the potential test approaches for your application, please contact us at firstname.lastname@example.org and we will be glad to help you.