Vector network analyzers (VNAs) aren’t for everyone, but if you need to characterize components at high frequencies, a VNA may be the best tool for the job. It is one of the few types of measurement instruments that achieves high accuracy and repeatability at microwave frequencies. The test results it provides can be invaluable to designers governed by power, cost, time scale, and size constraints.
A related high-frequency instrument—the spectrum analyzer—is a scalar measurement tool that determines the amplitude of an unknown signal. A VNA measures the phase and magnitude effects that a DUT has on the VNA’s closely controlled source signal and determines the characteristics of the DUT. Spectrum analyzers measure signals; VNAs characterize components.
Figure 1 is a block diagram of a two-port VNA with transmission/reflection capability. The source output is split and directed to port 1 and reference receiver R. A directional coupler samples the reflected signal at port 1 and drives receiver A. The DUT output signal is connected to receiver B.
This type of VNA is adequate for DUTs with good input/output isolation. It measures the magnitude and phase of the input-transmitted signal, reflected signal, and the DUT output signal. It is lower cost than the fully reversible VNA with S-parameter capability shown in Figure 2.
In addition to making port 1 transmission/reflection measurements, the VNA in Figure 2 drives port 2 and measures reverse (port 2-to-port 1) parameters. By suitably combining the forward/reverse and transmitted/reflected measurements, a complete set of S-parameters can be derived to describe the DUT’s electrical characteristics.
S-parameters always are measured with source and load impedances equal to the characteristic impedance of the system, typically 50 W or 75 W. Shorts and opens are not required during characterization of the DUT as, for example, with Y and Z parameters. This is especially important at high frequencies because shorts and opens can cause oscillation or destruction of DUTs. It also is difficult to produce an accurate open standard because of fringing effects.
Power can be measured, but because the VNA normally measures ratioed voltages, a special calibration is needed for accurate power measurements. A broadband power meter is used for this purpose. To make calibration more convenient, calibration constants should be sent from the meter to the VNA via a local RS-232 or IEEE 488 bus. Transferred data will include calibration factors at several frequencies throughout the band of interest for both the meter and the sensor.1
Sources of Error
Conceptually, a VNA is very simple, consisting of a source and up to four couplers and receivers. Unfortunately, the real components corresponding to the block-diagram elements are not ideal, and all contribute to measurement errors. In addition, the implementation of each element requires VNA design engineers to make compromises. For example, is it better to use a coupler with low loss but an inherently poor low-frequency response or a higher loss bridge with a better low-frequency response?
Whether to use narrowband or broadband detectors is another important decision. VNAs generally use tuned receivers to detect signals because they provide much lower noise and greater dynamic range than a broadband detector and are necessary for phase measurement. Of particular benefit, tuned receivers eliminate source harmonics. Broadband detectors often are found in microwave scalar network analyzers because they cost less than tuned receivers.
Figure 3 shows the sources of errors that calibration can remove. The port 1 directional coupler measurement output will contain a crosstalk contribution from the input signal. The ratio of the directional coupling factor to the crosstalk level is termed directivity.
If the characteristics of the port 1 reflected signal receiver A differ from those of reference receiver R, a reflection tracking error exists. Similarly, if the behaviors of DUT output receiver B and reference receiver R are not identical, a transmission tracking error exists.
There is a mismatch associated with each of the four impedances: source port, DUT input, load port, and DUT output. Their values are all complex (magnitude and phase) functions of frequency and will differ in some way from Z0, the characteristic impedance of the test system, usually 50 W or 75 W. The fundamental definition of mismatch is given in Equation 1 for the reflection coefficient G:
G = Vreflected/Vincident = (ZL – Z0) / (ZL + Z0) (1)
where ZL is the complex load impedance.
Mismatch is expressed as
Return Loss = -20 log (r) (2)
where r = ê G ê. Large values of return loss are desirable because they indicate near-ideal impedance matching.
Finally, a degree of crosstalk will exist between ports 1 and 2. All together, six error terms have been defined. In each case, the signal was applied to the DUT input port 1 connection; that is, in the forward direction. Six more error terms can be defined when the signal is applied to the DUT output via port 2 in the reverse direction for a total of 12 terms.
A reflection/transmission VNA can only correct errors caused by directivity, source match, and reflection tracking (Figure 1). In contrast, the S-parameter type of VNA can measure and correct the effects of all 12 error terms (Figure 2).
Each element of a VNA must individually perform as well as possible for calibration to provide the best results. For example, unless the source has little sensitivity to temperature changes, calibration could be required quite frequently. And, unless the connectors and adapters fit correctly, test results won’t be repeatable.
There are other VNA configurations besides those shown in Figures 1 and 2. Some models have the capability for a third, optional port to simplify testing devices such as duplexers, combiners, and couplers. Without the third port, an external switch must be used to multiplex the two VNA ports to the three DUT ports. Alternatively, the connections can be changed manually.
Also, it may be possible to add a second, independent source for intermodulation distortion (IMD) work. Because the two sources are totally independent, the spacing between their output frequencies can be adjusted to suit two-tone IMD tests.
Calibration
Calibration is the key to the very high accuracy achievable with a VNA. Both absolute and relative accuracy are guaranteed by using National Institute of Standards and Technology (NIST)-traceable standard terminations. In addition, meticulous attention to instrument construction details ensures repeatability.
To illustrate this point, design and construction of the Rohde & Schwarz Models ZVR and ZVRL VNAs include matching the electrical length of the measurement and reference channels. This precaution eliminates phase variations due to temperature drift, for example. Also, the standing wave ratio (SWR) bridges and a corresponding reference-channel compensation line are constructed from the same batch of cable to ensure tracking.
As an example of the stability that can be achieved for ZVR and ZVRL VNAs, “a calibration is typically valid for months.. After normalization, switching off and completely cooling down the analyzer, and switching on the cold unit, within a few minutes typical values for analyzer stability are <0.02 dB and <0.5° for transmission measurements." After being switched off overnight and switched on the next day, "after one hour the typical differences are <0.005 dB and <0.1°."3
If you are making very critical measurements, a calibration must be performed immediately prior to each measurement, and the instrument should be operated in a temperature-controlled room. A few hours of warm-up ensure that any slow, turn-on-related drifts have been completed.
The primary technique used for calibration is known as Short, Open, Load, Thru or SOLT. Standards kits are available that include these terminations for specific VNA models. The characteristics of the four physical standards have been accurately determined and are defined in a calibration-kit definition file.
The VNA uses this file to determine correction factors that will cause measured results to coincide closely with theory. Applying the standards in prescribed combinations produces 12 equations that the VNA solves for the 12 error-correction terms. The correction terms are all complex functions of frequency.
David Ballo, a component test applications engineer at Agilent Technologies, said, “Phase is required for a number of measurements including group delay and deviation from linear phase, but it also is necessary for error correction. The only way to separate a VNA’s systematic error terms is to measure both magnitude and phase. That is part of the process that happens when you measure known calibration standards.
“With either a scalar network analyzer or a tracking generator with a spectrum analyzer, you don’t have the phase information so you can’t do sophisticated vector error correction,” he continued. “A key feature of the standards themselves is NIST-traceability. That’s how we ensure traceable absolute accuracy.” Consequently, even though a combination of a spectrum analyzer and a tracking generator may have a similar dynamic range as a VNA, the VNA’s accuracy will be significantly better.
In using the SOLT calibration-kit terminations, you implicitly establish the measurement plane. The terminations and thru standard screw directly onto the test ports. If the DUT being tested does not have coaxial connectors, a separate test fixture will be required, and the measurement plane will be altered.
A set of calibration standards can be produced for use with the fixture in place of the DUT. This approach allows the effects of the test fixture and the altered measurement plane to be corrected by the calibration process. The standards may be fabricated on PCB material for design/development work but must be more rugged and made with an appropriate contact system for production applications.
Thru-reflect-line (TRL), an alternative calibration method, has been developed for microwave noncoaxial environments such as waveguide, wafer probing, and test fixtures. TRL provides the same 12 error correction terms as SOLT but uses a different set of calibration standards. At microwave frequencies, noncoaxial TRL standards generally are easier to make than SOLT standards.
A fourth receiver, another splitter, and an attenuator must be added to the block diagram in Figure 2 to perform true TRL calibration. With these changes, the difference between the forward and reverse matching of both ports can be included in the error-correction terms.
References
- “Compression Measurements With the 37300 Series Vector Network Analyzers,” Anritsu, 1999, p. 3.
- members.tripod.com/~zoam/lordkelvin.html/kelvin quotes.
- Ostwald, O., “Frequently Asked Questions About Vector Network Analyzer ZVR,” Application Note 1EZ38_3E, Rohde & Schwarz, 1998.
Information for this article was taken from “Network Analyzer Basics” by David Ballo of Agilent Technologies and can be downloaded at www.agilent.com/find/basics.
For more information on vector network analyzers,
see “Loss Improves Matching.”
Published by EE-Evaluation Engineering
All contents © 2000 Nelson Publishing Inc.
No reprint, distribution, or reuse in any medium is permitted
without the express written consent of the publisher.
February 2000