Connectors made into test fixtures let you get consistent measurement results when characterizing SMT capacitors for signal and power integrity designs.
In my previous post, Solder-wick trick characterizes bypass caps, you saw a very simple home-made fixture . It takes only two coaxial cables and some solder wick. The fixture is convenient for quick tests and for situations when a little error due to the contact resistance and due to the variable shape of the solder wick contacts is acceptable. If you want more accurate and consistent data, you need to use fixtures with fixed geometry and solder the component to it.
A fixed-geometry fixture produces consistent and more repeatable results. For instance, you can make fixed-geometry fixtures out of certain connectors. PCB-mounted SMA connectors are convenient for mounting typical bypass capacitor sizes. These connectors have two distinct sides: the SMA side interfacing with the SMA connector on your cables and the PCB side that attaches to the PCB. The SMA side comes with the choice of regular male or female, but there are also reverse-polarity SMA male and female connectors, so you need to pay attention when you buy these parts.
Common cables come with regular male SMA connectors. Thus, you can use regular female SMA connector for your fixture. The PCB side also has a large variety of different geometries, dependent on how you want to mount the connector to the PCB. If the mounting is perpendicular to the board, then solder the four posts at the corners to the board either as surface-mount or through-hole connections. If you want to use the connector as edge-mount, the spacing between the posts let the connector straddle the board with its specified thickness. Some of these varieties are illustrated in Figure 1. From left to right you see SMA-female-to-surface-mount, SMA-male-to-narrow-base-vertical-through-hole-mount, SMA-female-to-wide-base-vertical-through-hole-mount, SMA-female-to-narrow-base-vertical-through-hole-mount, SMA-female-to-wide-base-edge-mount, and SMA-female-to-narrow-base-edge-mount connectors.
You can build a fixture to measure bypass capacitors using SMA-female edge-mount connectors, which are the last two connectors on the right in Fig. 1. You must, though, consider the capacitor’s base size. The narrow-base connectors in Fig. 1 have posts on a 5 mm grid and these connectors are best suited for smaller-size components, such as 0805, 0603 and 0402 capacitors. You can also attach 1206 and 1210 size components to these fixtures. The wide-base edge-mount connector has posts on an 8 mm horizontal grid. You need them for bigger components, such as D-size (7.3 mm×4.3 mm) packages.
To create a fixture, solder two of these connectors back-to-back. You can connect both ends to SMA male cable connectors and you attach the bypass capacitor you want to measure to the center pin and ground frame in the middle of the fixture. Figure 2 shows wide-base (l) and narrow-base empty fixtures.
Figure 3 shows fixtures with 1210-size ceramic capacitors attached.
Before making measurements, you must calibrate the system. Up to about 30 MHz, a simple Response-Through calibration is usually sufficient. For the response-through calibration, you can simply connect an empty fixture (no capacitor) between your vector network analyzer (VNA)’s Port 1 and Port 2. After the calibration, you will solder the DUT between the center pin and ground frame of the fixture. You can also create multiple fixtures with identical geometry, keeping one just for response through calibration and reserving the others for measuring DUTs. You measure the S21 transfer parameter with the network analyzer and from that you can calculate the complex impedance of the capacitor with this simple formula:
Where Z0 is the port impedance of the VNA, usually 50 Ω. The network analyzer used for these measurements has an option available that does this transformation inside the instrument . If you are interested in the derivation of the formula above, you can find it on page 132 of .
When you select the measurement’s frequency range for the measurement, you face a the tradeoff between the start and stop frequencies and the nature of components you want to measure. If you want to start the sweep anywhere below 30 kHz and want to measure components with low impedance at low frequencies, such as low-ESR high-capacitance parts, you may run up against the cable-braid loop error, described in Section 7.1.1 of . Depending on how you reduce the cable-braid loop error, the chosen solution may come with its own limitations at high frequencies. For this article, I used a home-made common-mode choke to reduce the effect of the cable braid resistance, with an upper bandwidth of approximately 50 MHz. Data was collected from 300 Hz to 30 MHz.
Figure 4 shows the setup of this measurement without the common-mode toroid. The common-mode toroid and its use will be described in later blogs.
Figure 5 shows the measurement results for a 47 µF 1210-size X5R capacitor. The excitation was set to −10 dBm RF source power level and 0 V DC bias. The VNA uses the Impedance Analysis Option . The screen is set up with four simultaneous traces: impedance magnitude (upper left), equivalent series resistance, Rs (upper right), equivalent series capacitance, Cs (lower left), and equivalent series inductance, Ls (lower right). The logarithmic horizontal scale starts at 300 Hz and ends at 30 MHz. The 70 Hz IFBW setting provides a good compromise between sweep speed and noise floor.
While these fixtures are still very simple to make and they provide fixed and repeatable geometry, they lack a planar structure with power and ground planes. They don’t represent our typical PCB applications. For this reason, you probably need to ignore the inductance in the measured data. For applications where you are looking only for the capacitance information and possibly also for the equivalent series resistance (ESR) data, this simple fixture is an acceptable solution. If you really need the inductance to also represent that of your application, the best you can do is to create fixtures with the same or similar stack-up as the final application and connect the DUT with the escape patterns that you plan on using on your board. These fixtures are tailored to our PCB geometry and therefore have more limited applications, but will best represent the DUT’s performance in real application over a wide frequency range. For such fixtures, see Figure 7.13 on page 206 of Reference 3.
Istvan Novak, PhD is Senior Principle Engineer at Oracle.