Vector network analyzers grade microwave components

-April 01, 2001

Vector network analysis is critical to ensuring the quality of microwave components manufactured for wireless communications systems. If you are designing or manufacturing microwave components, you need instruments that can generate microwave stimulus signals and monitor the phase and magnitude of DUT response signals. The most effective systems for measuring microwave vector parameters are vector network analyzers (VNAs). They embody the necessary sources and receivers within one box, they combine the computational power and software needed to make sense of measurement results, and they come with test-set options that simplify the measurement of DUT impedance, signal magnitude, and signal phase.

Vector measurements (simultaneous measurements of a signal’s magnitude and phase) are critical for characterizing device performance. They extract S-parameters (Ref. 1) and also measure distortion-related parameters such as group delay (the derivative of a signal’s phase with respect to noise) and intermodulation distortion (IMD), which arises in a nonlinear component subjected to multifrequency inputs.

Scalar vs. vector

Even scalar measurements such as noise figure (a DUT’s output SNR divided by its input SNR) and reflection coefficient can benefit from the use of a VNA. A VNA can provide the vector error correction that compensates for impedance mismatches between itself and the DUT, and it also can compensate for error sources that can degrade scalar measurement accuracy.

In addition, VNAs can provide better noise performance than scalar network analyzers can. VNAs employ tuned receivers while scalar analyzers tend to rely on diode-detection schemes, which are inherently more noisy. That said, vendors of scalar network analyzers are beginning to address the need for low noise in scalar measurements. In February, IFR (Wichita, KS) introduced a scalar analyzer whose noise performance rivals that of VNAs. Such an instrument can be a cost-effective choice if you need good noise performance but can get along without S-parameter measurement capabilities.

If a VNA is right for your application, you’ll want to consider several factors before making your selection (see Table 1 for a list of VNA vendors). Choosing an instrument that meets your frequency requirements is most critical. The closer you match a VNA’s frequency range to that of your application, the more economical your choice will be. Anritsu, with its Lightning series, provides a wide choice in the 65-GHz and under range. If you can get by with the firm’s 50-GHz model, which starts at $99,500, you can save more than $20,000 over the cost of a 65-GHz version.

Beyond frequency range and cost, you should evaluate noise performance, measurement speed, test-set and calibration-set availability, built-in processing power, onboard software, the GUI, and compatibility with external computers, particularly those used to simulate microwave designs. A look at how a VNA operates can help you make effective choices in these areas.

VNA operations

A VNA continually measures a DUT’s response to a synthesized DUT input signal that sweeps over a programmed frequency or power range. A frequency sweep proves useful for characterizing passive or active filters, while the power sweep can describe nonlinear performance of a power amplifier. “Sweep” is a slight misnomer for what the signal-source portion of a VNA does, because modern VNAs employ digitally controlled synthesizers that provide discrete steps rather than a continuous sweep. A “sweep” typically consists of anywhere from 201 to 1601 discrete points.

The VNA receiver employs a programmable intermediate-frequency (IF) bandwidth, which you can adjust to optimize sweep speed (and thus total measurement time) vs. noise and dynamic-range performance. Generally, a narrow bandwidth implies lower noise and, hence, greater dynamic range—for a fixed maximum output level, your dynamic range depends on the level at which low-level signals become lost in the noise floor.

Consequently, narrow bandwidths afford more accurate characterization of your DUT—but at the cost of increased test times. For example, the venerable Agilent Technologies 8753E (introduced by Hewlett-Packard in 1986), using a 6-kHz IF bandwidth, could perform seven and a half 201-point sweeps per second with a trace noise of 0.04-dB peak-to-peak. Reducing the IF bandwidth to 3 kHz improved trace-noise levels down to 0.02-dB peak-to-peak, but reduced the sweep rate to 4.75 sweeps per second. In general, a tenfold reduction in IF bandwidth affords a 10-dB reduction in a receiver’s noise floor (Ref. 2).

TMW01_04f1fig1.gif (36780 bytes)
Figure 1. Targeting wideband CDMA applications, Credence Systems’ MI 4115A modulated vector network analyzer features an acquisition bandwidth as wide as 15 MHz.

Not all VNA vendors agree that a narrow bandwidth is a good thing. Credence Systems, a newcomer to the VNA arena with its MI 4115A MVNA (modulated vector network analyzer), is betting that requirements for testing wideband CDMA and Bluetooth devices are best served by a wideband VNA that applies real-world modulated signals—not the pure sine waves of traditional VNAs—to DUTs. The goal of such an approach is to expose DUT nonlinearities that could occur in real-world operation but that could remain hidden during narrow-band swept tests. The MI 4115A is a benchtop system with vector error correction that extracts S-parameter data by means of 15-MHz wideband and 6-kHz narrowband acquisition bandwidths (Figure 1) . Wideband dynamic range is 80 dB; in wideband mode, the instrument can acquire 60 Msamples/s. Its wideband capabilities do come at a price premium: $125,000 for the 2.4-GHz MI 4115A vs. about $40,000 for a traditional narrowband instrument in the same approximate frequency range.

VNA-based tests typically consist of several steps: An operator connects the DUT to the test setup, takes a VNA measurement, tunes the DUT, and repeats the measurement, and then repeats the entire process until the DUT meets spec. The operator then moves to the next DUT. Time the operator spends waiting for a VNA to perform a sweep and acquire, process, and display measurements is time wasted.

Boosting measurement speed

TMW01_04f1fig2.gif (30089 bytes)
Figure 2. Available in 3-, 6-, and 9-GHz models, the Agilent Technologies PNA series instruments achieve dynamic range to 143 dB in direct-receiver-access mode.

Not surprisingly, traditional VNA vendors, including Agilent Technologies, Anritsu, and Rohde & Schwarz, have been working to increase measurement speeds while maintaining good noise performance. Agilent’s PNA models (Figure 2), introduced last year, provide at least a sixfold improvement in measurement speed at a given noise level. The company reports that a sweep requiring 43 s to provide a 120-dB dynamic range on the 8753E takes just 1.2 s on a PNA model.

But take care when evaluating sweep speeds. You’ll see specs such as “35 ms per point,” but what’s really important is the total time required to complete a sweep of (typically) 51, 101, 201, 401, 801, or 1601 points. For that, you’ll have to look beyond the product-brochure blurbs and delve into an instrument’s data sheet (such as Ref. 3, which tabulates the Anritsu Scorpion’s sweep speed as a function of IF bandwidth and frequency range).

Total test times depend on more than just instrument sweep and processing speeds. An instrument that’s difficult to set up and whose results are hard to interpret can drastically impede test times as the operator struggles with the instrument. The ability to store test setups for later recall is critical for reducing operator errors. Features such as color displays and markers can help operators correctly and quickly interpret test results.

Beyond test speed, you’ll want to evaluate a VNA’s ability to work within your company’s network without overloading the network with measurement data. VNAs such as Anritsu’s Scorpion provide onboard nonvolatile memory and disk storage, so constant data transfers over external communications channels aren’t necessary. (The Scorpion stores 1,601-point S-parameter data files in 102.8 kbytes.) When you are ready to transfer Scorpion data externally, you can use the instrument’s IEEE 488, RS-232, SCSI-2, or Ethernet ports. The Scorpion also includes connectors for an external VGA monitor, parallel printer, and keyboard.

Agilent’s PNA goes a step further with respect to connectivity, integrating the entire Windows 2000 operating system. The PNA enables design and test engineers to use a variety of tools, including COM/DCOM technology; Windows-compatible programming languages like Agilent VEE, Microsoft Visual Basic and Visual C++, or National Instruments’ LabView; and office applications for processing measurement data and generating test reports. The PNA’s Windows interface enables you to run an instrument electronic-calibration wizard and also access online, context-sensitive help, built-in tutorials, and manuals. You can set up four stimulus settings and display 16 traces simultaneously.

Several types of error can plague microwave measurements. Environmental factors such as temperature affect many measurements, not just those of microwave networks. These errors—called drift errors—can be controlled by controlling the environmental factors that cause them. In other words, if temperature drift is impeding your measurement accuracy, you should upgrade your air conditioning, or else develop a compensation factor that adjusts your measurement results to real-world environmental conditions. Drift errors can also occur as test-setup hardware, such as connectors, degrades over time. You can alleviate these errors by replacing mechanical test components before they begin to degrade measurement results.

Random errors, too, can plague measurements. They can range from unpredictable inconsistencies in an operator’s connection of a DUT to the test setup to random fluctuations in instrument noise level. Solutions to such random errors can range from improving operator training to increasing instrument power (to improve SNRs) or reducing receiver IF bandwidth to lower the noise floor. Data averaging, too, can minimize the effects of random errors, and most VNAs provide some form of data-averaging capabilities. Anritsu’s Scorpion models, for example, can average as many as 4096 data points, performing the averaging function at each data point during a sweep. A front-panel control toggles the averaging function on and off.

Other errors, however, are specific to the mechanics of your test setup and, hence, are predictable. A VNA can mathematically compensate for these “systematic errors,” if it has the necessary smarts. If you are used to low-frequency measurements, these systematic errors can be analogous to the resistance of your DMM leads as you make measurements on low-resistance components. At microwave frequencies, the possible errors become much more complex. For a two-port network, you’ll find these sources of systematic errors for a stimulus applied at port 1 with its response measured at port 2:

• Crosstalk. Absent a DUT, an input signal at port 1 will result in a crosstalk signal at port 2.

• Directivity. A synthesizer generates a signal to be applied to a DUT at port 1. In a VNA, that signal’s value is determined by a reference receiver, ideally obtained by a coupler, whose response varies with the direction of energy flow.

• Source mismatch. Ideal VNA performance depends on an ideal impedance match between the DUT and the test system, an ideal that’s never met.

• Load mismatch. Ideally, a VNA receiver connected to port 2 will match DUT output impedance, an ideal that, as above, never occurs.

• DUT-input to reference-receiver signal tracking (reflection tracking).

• DUT-input to DUT-receiver signal tracking (transmission tracking).

These six errors repeat for a stimulus applied at port 2 and measured at port 1, resulting in 12 error sources, necessitating “12-term vector error correction.” Ref. 4 explains vector error correction in more detail, and Ref. 5 describes its derivation in terms of S-parameters.

VNAs come equipped with the software and computing power to do these calculations for you. All you need to do is hook up calibration standards—available from your VNA vendor or third parties such as Flann (, Focus Microwaves (, Maury Microwave (, Oleson Microwave Labs (, or Storm Products (

Time-domain analysis

TMW01_04f1fig3.gif (51667 bytes)
Figure 3. Providing quad-display capability, the ZVR/ZVC VNAs from Rhode & Schwarz and Tektronix can show Smith charts as well as magnitude-vs.-frequency plots. They also provide time-domain analysis.

VNAs essentially operate in the frequency domain—their main job is to directly measure S-parameters, which are analogous to complex impedances, over a frequency range of interest. VNAs such as Rohde & Schwarz’s ZVR (Figure 3), however, can perform inverse Fourier transforms on frequency data to yield a result much like that of a time-domain reflectometer (Ref. 6).

Time-domain displays can help localize cable faults, and time-domain display of impulse and step responses can provide component designers with insights that wouldn’t be readily apparent in the frequency domain. The ZVR includes a gating function that can screen out unwanted time-domain signals from measurement results. The gated time signal can then be transformed back to the frequency domain to provide more accurate S-parameter representations of a DUT without spurious data from cables external to the DUT.

You can expect to see changes in VNA technology as manufacturers adapt their instruments to changing applications and as VNAs contend with ever more capable scalar analyzers and as specialized instruments such as impedance analyzers intrude on traditional VNA turf.

An article on a competitor’s Web publication goes so far as to say “network analyzers provide an inadequate solution” for impedance measurements “because of their unreliable test results and the complexity of removing fixture errors from the measurement data” (Ref. 7). That’s a rash claim, but there is no doubt that instruments such as impedance analyzers are economical choices for impedance-only applications. But evolutionary improvements will keep VNAs at the forefront of microwave-component characterization. T&MW


1. Nelson, Rick, “What are S-parameters, anyway?” Test & Measurement World, February 2001. p. 23.

2. “Improving Throughput in Network Analyzer Applications,” Application Note AN 1287-5, Agilent Technologies, Palo Alto, CA, 2000. Editor's Note 10/24/03: This page has moved:

3. “MS4622 A/B, MS4623 A/B Vector Network Measurement Systems,” Anritsu, Morgan Hill, CA. p. 17. Editor's Note 10/24/03: This page has moved:

4. “Applying Error Correction to Network Analyzer Measurements,” Application Note AN 1287-3, Agilent Technologies, Palo Alto, CA, 2000.  

5. “AutoKal Automatic Calibration of Vector Network analyzer ZVR,” Application Note 1EZ30_2E, Rohde & Schwarz, Munich, Germany, 1998.

6. “Time Domain Measurements using Vector Network Analyzer ZVR,” Application Note 1EZ44_0E, Rohde & Schwarz, Munich, Germany, 1998.

7. “Impedance measurements and the role of the impedance analyzer,” by Agilent Technologies, Test and, January 3, 2001,

Rick Nelson received a BSEE degree from Penn State University. He has six years experience designing electronic industrial-control systems. A member of the IEEE, he has served as the managing editor of EDN, and he became a senior technical editor at T&MW in 1998. E-mail:

Table 1Microwave vector network analyzers




Agilent Technologies
Palo Alto, CA






300 kHz to 1.3 GHz


100 kHz to 1.8 GHz

PNA Series


300 kHz to 3 GHz

300 kHz to 6 GHz

300 kHz to 9 GHz


50 MHz to 13.5 GHz


50 MHz to 20 GHz


50 MHz to 40 GHz

8510C Series

45 MHz to 110 GHz

Morgan Hill, CA

Scorpion Series

10 MHz to 3 GHz
10 MHz to 6 GHz

Lightning 37000C Series

22.5 MHz to 8.6 GHz
40 MHz to 13.5 GHz

40 MHz to 20 GHz

40 MHz to 40 GHz

40 MHz to 50 GHz

40 MHz to 65 GHz

Credence Systems
Fremont, CA

MI 4115A

800 MHz to 2.4 GHz

Rohde & Schwarz*
Munich, Germany



9 kHz to 4 GHz


20 kHz to 8 GHz


10 MHz to 20 GHz


10 MHz to 40 GHz

*In North America, Tektronix (Beaverton, OR, 800-833-9200, distributes the models ZVR and ZVC.

Loading comments...

Write a Comment

To comment please Log In

Currently no items