Verify capacitor performance with a simple “homemade” test fixture.
Bypass capacitors are used in large numbers in power distribution networks. Most vendors today supply not only typical characteristics, but also various simulation models. Nevertheless, doing our own characterization of these components is still useful and often necessary. In this short article, I’ll show you how to create simple home-made fixtures for these measurements.
For bypass capacitors, measuring impedance over a reasonably wide frequency range is the way to go. This gives us the small-signal equivalent behavior of the part. By post-processing the complex impedance, we can obtain the capacitance, effective series inductance (ESL), and effective series resistance (ESR) as functions of frequency. If needed, DC- and AC-bias-voltage dependence and temperature may be added to the mix of input parameters. All these have been described in earlier publications . Similarly, the instrumentation and measurement setup for this purpose is well established . To measure the impedance of bypass capacitors with high capacitance and low ESR, a good choice is a suitable vector network analyzer  in the two-port shunt-through connection.
Once we have our instrumentation ready, the next challenge is to decide how we connect our sample to the instrument. For quick and simple measurements, home-made fixtures work just fine.
This is a simple and rudimentary fixture, yet it works amazingly well at low frequencies. We start with two thin and flexible coaxial cables, with connectors at one end and pigtails at the other. You could use an RG-178 SMA jumper and cut it in the middle. This gives us two cables of identical length. The type of connector does not matter much at low frequencies; SMA connectors are relatively small, low cost, and readily available. After cutting, the cables should be long enough to bridge the distance between the network analyzer and our fixture, placed on the bench in front of it (if we have a choice, we should always opt for the shortest cable that can make the connection).
Strip the plastic jacket at the open ends, untangle the braids for about ¼”, and create pigtail connections. Next, cut two pieces of 1” solder wick, slide on short heat-shrink tubes, solder the two coax cables in a series fashion, slide the heat-shrink tubes in place, and finally activate the heat-shrink tubes with a heat gun. We get a flexible fixture shown on the top of Figure 1. We need to mark the strips; which is the one connecting the center wires, and which is connecting the braids of the coax cables. We can use colored heat-shrink tubes, or, as in Figure 1, the somewhat longer heat-shrink tubes mark the return (braid side) of the coax.
Figure 1 Solder-wick fixture (top) and a D-size capacitor in the fixture (bottom).
The capacitor can be placed between the two exposed solder-wick conductors; this will create a two-port shunt-through measurement scheme. On the bottom of Figure 1, a D-size (7.3 × 4.3 mm) capacitor is shown in the fixture. This fixture is best suited for pressure-mount connections, not for soldering. The benefit is that by using pressure contacts, we avoid heat stress to the component and can swap out samples quickly. The pressure-mount connection can be implemented by a spring-loaded plastic or wood clip, or we can just squeeze the fixture between our fingers with the component in place. If we are worried about the error caused by our body impedance, we can grab the solder-wick electrodes with our fingers but without a capacitor in place and take the impedance reading. As long as it is much higher than the impedance of the capacitor, we can ignore this error.
If we use short cables and limit the frequency to below 10 MHz, a simple response-through calibration on the VNA is enough. Note that the calibration and measurements are done without changing/disconnecting the cables, which improves the consistency of the measurements.
We should also make additional reference measurements. We have to measure the fixture with no DUT in place (open) and with a short, similar in size to the DUT we want to measure later. Figure 2 shows the impedance readings for these reference cases. The photo of our shorting device is shown in Figure 3. An extra capacitor sample was taken and the terminals on the bottom of the part were shorted with a strip of solder wick.
Figure 2 Impedance magnitude of Open and Short
Note that the impedance of the short reference piece is not exactly zero; it has finite resistance and inductance. For measuring really low impedances, we would need to characterize our shorting device, and do a more complex calibration.
Figure 3 D-size Short reference
In this very rudimentary fixture there is direct connection between the two VNA ports through the braid, which creates a “sneaky path”. As a result, what we get for the short reading (and for all other readings) is a mix of, a) the actual impedance of the DUT, b) some contact resistance, and c) the residual error created by the sneaky path. The open and short readings give us an impedance range where we can trust our data with this fixture with just a response-through calibration: the DUT impedance should be at least 3-5× (preferably 10× or more) away from these limits. The traces corresponding to open and short run around 20 kΩ and 2 mΩ at low frequencies, respectively. The rising tail of the short impedance trace is due to its inductance. Note also that the lower measurement limit set by the instrument noise was below 0.1 mΩ.
Figure 4 shows the impedance magnitude of ten DUTs measured with this fixture. The data was collected by pressing the solder wick against the capacitor terminals by hand.
Figure 4 Impedance magnitudes of ten capacitor samples
Readings are close at 1 kHz indicating close tolerances on the measured 470 µF capacitors. The lines begin to deviate beyond 10 kHz and hit maximum spread around the series-resonance frequency of 1 MHz. Considering that for bulk capacitors the data sheet guarantees only a maximum value of ESR (but no typical or minimum), the spread we see here is considered to be typical.
Of course, with this simple fixture, we also need to consider the spread and consistency of contact resistance. Above the series-resonance frequency, the spread continues. The upslope between one and ten megahertz represents inductance. Inductance is related to current-path geometry. As the body size and shape of these capacitors is very consistent, the likely source of inductance spread is the variation of the loop size of the solder-wick connections as we press the fixture to the DUT.
We can always re-measure the outlier samples. This will tell us whether the data needs to be updated or in fact some parts were outliers. We need to remember to periodically clean the surface of the solder-wick strips to remove contamination and we also have to make sure that the component terminals are clean.
In summary, this simple fixture may have some downsides: the sneaky path raises the error floor, the flexible solder-wick connections make the discontinuity between the cables more uncertain, and the contact resistance can vary. But, for measurements below 10 MHz and above a few milliohms in values, the simple construction, ease of connection, and simplicity of calibration far outweigh the drawbacks.
—Istvan Novak is Senior Principal Engineer at Oracle, with over thirty years of experience in high speed digital, RF, and analog circuit and system design. He is a Fellow of the IEEE, author of two books on power integrity, teaches signal and power integrity courses, and maintains an SI/PI website.
 “DC and AC Bias Dependence of Capacitors Including Temperature Dependence,” DesignCon East 2011, Boston, MA, September 27, 2011
 “Accuracy Improvements of PDN Impedance Measurements in the Low to Middle Frequency Range,” DesignCon2010, Santa Clara, CA, February 1-4, 2010