It may be the single most important measurement tool in the RF/microwave industry. It is used for research and on production lines. It gathers critical data from active and passive devices and components that can in turn be transformed into software models in computer-aided-engineering (CAE) programs. This majestic instrument, of course, is the vector network analyzer (VNA). Among test equipment, it is unique to this industry, in many ways a descendent of the common voltmeter, but so much more complex. Although VNAs are offered by only a handful of suppliers worldwide, a surprisingly varied selection of instruments is available for RF and microwave applications, ranging from economy models to full-featured systems extending well into millimeter-wave frequencies.

Selecting an RF/microwave VNA from available choices is not an easy matter, given the complexity of these instruments. If a spectrum analyzer can be thought of as a sensitive receiver (see p. 38), a VNA can be considered as a complete communications system, with transmitter and receiver electronics. To qualify as a measurement tool, VNAs maintain extremely tight tolerances among components while minimizing noise and distortion in the signal path. Unlike a spectrum analyzer, which looks at amplitude as a function of frequency, a VNA examines amplitude and phase and can thus measure the complex impedance of a device under test (DUT). Armed with amplitude and phase information, a computer-based VNA can also perform signal processing, such as an inverse Fourier transform, to generate time-domain displays of discontinuities as a function of distance (from the point of incident power).

By measuring scattering parameters (Sparameters), a VNA can perform a detailed characterization of an active or passive DUT. The S-parameters, which essentially describe how a DUT transmits and reflects incident energy as a function of its port impedances, are typically shown on a polar Smith chart display on a VNA's display screen, although the VNA offers a wide array of displays in both frequency and time domains (depending upon instrument options). S-parameters are essential for building a model of a DUT, which can then be used for analysis and optimization in a computer-aided-engineering (CAE) circuit simulation software program. But rather than delve into a longer discussion on how a VNA measures S-parameters, application note AN 1287-1 from Agilent Technologies, "Understanding the Fundamental Principles of Vector Network Analysis," provides a good primer and is available for free download in PDF file format from the Agilent web site.

As with a spectrum analyzer, a choice of VNA is usually initially guided by required frequency range. VNA suppliers typically offer several choices of frequency range, with similar levels of performance across the different frequency ranges in a product family. For extremely broadband coverage, and as the front cover of this issue shows, Anritsu has unveiled the VectorStar line of VNAs, with the broadest frequency range available within a single VNA at 70 kHz to 70 GHz (see p. 89). While the frequency coverage is impressive, it may be overkill for some test applications, such as for components used in wireless communications systems such as WCDMA cellular and WiMAX systems. Of course, a VNA's bandwidth should also provide adequate range to capture spurious and harmonic signals for a DUT of interest.

VNA suppliers such as Agilent Technologies, provide different product families depending not only on frequency range but on performance needs. Agilent's ENA series of instruments, for example, is touted as an economy line of instruments when absolute accuracy and highest performance are not required for an application. The ENA series from Agilent, for example, includes models covering 9 kHz to 4.5 GHz, 100 kHz to 4.5 GHz (with bias tees for active devices), 9 kHz to 8.5 GHz, 100 kHz to 8.5 GHz (with bias tees), and 300 kHz to 20 GHz (with bias tees). When more bandwidth, dynamic range, and other performance enhancements are needed, the company also offers its PNA series of VNAs, covering 10 MHz to 20 GHz (model E8362C), 10 MHz to 40 GHz (model E8363C), 10 MHz to 60 GHz (model E8364C), and 10 MHz to 67 GHz (model E8361C).

What sets an economy VNA apart from a higher-performance VNA other than frequency range? Obviously, the frequency range is important, especially for those performing broadband microwave or millimeter-wave measurements. But other factors to consider when comparing VNA models are the key performance parameters that differentiate VNAs, such as the number of ports, level and frequency accuracy, amplitude measurement range, dynamic range, measurement speed, phase noise, system noise levels, and display capabilities. As with a spectrum analyzer, comparing numbers between different VNA models is not always straightforward since the conditions of a measurement must be normalized. For example, the noise floor of an analyzer will be lower when measured with a narrower intermediate-frequency (IF) filter in the VNA, and any comparison of noise levels should be made using the same width filters.

As an example, the test-port noise floor for the PNA analyzers is typically -97 dBm to 10 GHz and -94 dBm to 50 GHz, both measured with a 1-kHz IF bandwidth. But if a 10-Hz IF bandwidth is used, the test-port noise floor for these VNAs is -117 dBm to 10 GHz and -114 dBm to 50 GHz. For the other end of the dynamic range, the typical third-order intercept (IP3) for the PNA instruments is +26 dBm to 20 GHz and +19 dBm to 50 GHz. These analyzers offer wide dynamic ranges which, for VNAs using error correction, are typically specified by both system uncorrected and corrected dynamic ranges. For the PNAs, the standard uncorrected dynamic range is 123 dB to 20 GHz, 110 dB to 40 GHz, and 104 dB to 50 GHz.

For measurements where phase noise is critical, such as for oscillators and frequency synthesizers, the PNA instruments feature excellent phase noise. But again, phase noise depends upon measurement conditions, notably the offset from the carrier, and these conditions must be normalized when comparing the performance of different VNAs. For the PNAs, the phase noise is -70 dBc/ Hz to 10 GHz and -55 dBc/Hz to 50 GHz, when measured at a 10-kHz offset frequency. But when the offset frequency is pushed to 1 MHz, the phase noise improved to -103 dBc/Hz to 10 GHz and -90 dBc/Hz to 50 GHz.

In a VNA, available test-port power will decrease with increasing frequency, as with the PNA instruments that provide test power of -87 to +3 dBm to 10 GHz but -87 to -17 dBm at 50 GHz, with the power level controlled with 0.01-dB resolution. The amplitude response of a VNA is judged by several different parameters, including level linearity and level accuracy. For a PNA analyzer, the level linearity is 1 dB to 50 GHz while the level accuracy is 4 dB to 50 GHz. Other factors to consider in a VNA are the directivity (corrected and uncorrected), the source and load match (a measure of the tight tolerance of the VNA port complex impedance), reflection and transmission tracking, and measurement speed.

Comparing measurement speed, for example, also depends on certain test conditions, as shown in performance specifications for the R&S ZVA family of VNAs from Rohde & Schwarz. The family includes two-port and four-port models covering different frequency ranges, such as the R&S ZVA8 (300 kHz to 8 GHz), R&S ZVA24 (10 MHz to 24 GHz), R&S ZVA40 (10 MHz to 40 GHz), and the highest-frequency VNA in the line (Fig. 1), the R&S ZVA50 (10 MHz to 50 GHz). Each features 1-Hz frequency resolution with outstanding accuracy and dynamic range. For the ZVA40, for example, the dynamic range is typically 135 dB to 20 GHz and 115 dB to 40 GHz.

The sweep speed of these VNAs is affected by the number of measurement points, the measurement bandwidth, and the start and stop frequencies, as well as by the type of instrument calibration used. When comparing sweep times among different VNAs, they should include the times for display retrace and internal band switching. For the R&S ZVA8, for example, the measurement time per point is less than 3.5 ms in CW mode when using a 1-MHz IF bandwidth. The switching time between channels is less than 1 ms for 2001 points or less. So, factoring in different times for different sweep ranges and center frequencies, the sweep times for the R&S ZVA8 are less than 6 ms for 2001 measurements at a center frequency of 1.1 GHz and less than 4.5 ms for 201 measurements at a center frequency of 5.1 GHz. Although these differences in speed are slight, they must be normalized when comparing VNAs from different suppliers.

For the R&S ZVA instruments as well as other VNAs, accuracy depends on the frequency range and amplitude range being measured, as well as the ambient temperature. For the best accuracy, the temperature of a VNA system should be maintained as close to room temperature (+25C) as possible. For the ZVA40, for example, transmission measurements are accurate within 0.1 dB for measurements from 2 to 24 GHz and levels of -55 to +5 dBm and within 0.2 dB for the same frequency range and levels within -70 to -55 dBm. At higher frequencies, the accuracy is expected to degrade somewhat. For the ZVA40 at 32 to 40 GHz, however, the accuracy of a transmission measurement is still a respectable 0.2 dB for levels from -55 to +5 dBm and and 0.4 dB for levels within -70 to -55 dBm.

Continue on Page 2

Page Title

In addition to its new VectorStar line of premium performance VNAs, Anritsu supplies numerous more mature VNA products, including its Lightning family of VNAs. The Lightning 37000D VNAs are available in versions for 40 MHz to 20 GHz (model 37247D), 40 MHz to 40 GHz (model 37269D), and 40 MHz to 65 or 67 GHz (model 37397D). The 37300D series instruments (Fig. 2) are two-port instruments designed for active device testing, with step attenuators, internal bias tees, and wide power ranges, while the 37200D series instruments are twoport instruments geared for passive device testing. The cost-effective analyzers, with 1-Hz frequency resolution and 0.01-dB power resolution, provide four independent display channels for showing all four S-parameters simultaneously.

Also, Anritsu offers several variants of a portable VNA line known as the VNA Master, for true two-port on-site VNA measurements. The instrument provides vector network measurements from 2 MHz to 6 GHz and can be equipped for spectrum analysis from 9 kHz to 7.1 GHz. The VNA Master instruments can provide one-port measurements of return loss, cable loss, and distance to fault in the field, as well as two-port measurements of cable loss, phase, and group delay. For example, the model MS2036A VNA Master combines a VNA and spectrum analyzer for total frequency range of 9 kHz to 7.1 GHz. It features a 65-dB typical dynamic range across its full 2-MHz-to-6-GHz VNA range. It has 10-Hz frequency resolution and 25 PPM frequency accuracy. Its error-corrected directivity is 42 dB from 2 MHz to 6 GHz.

As a spectrum analyzer, model MS2036A offers 1-Hz tuning resolution from 9 kHz to 7.1 GHz and is capable of a DANL as good as -163 dBm when using a preamplifier and 1-Hz resolutionbandwidth (RBW) filter. The spectrum analyzer exhibits phase noise of -100 dBc/ Hz offset 10 kHz from the carrier.

Along with the "Big Three" of VNA suppliers, one of the lesser-known microwave VNA suppliers, Gaotek (www. gaotek.com), offers the 9-GHz 3629A series VNAs as well as the 40-GHz 3629 series analyzers. Two-port instruments in the 3629A series operate from 300 kHz to 9 GHz with 1-Hz frequency resolution and test source power of -85 to +5 dBm. The VNAs feature 38-dB directivity to 9 GHz and 110-dB dynamic range to 9 GHz. Two-port VNAs in the 3629 series operate from 45 MHz to 40 GHz

1-Hz frequency resolution and test source power of -85 to -5 dBm. They deliver 34-dB directivity and 110-dB dynamic range to 40 GHz.

Given this wide choice of VNAs, the choice of instrument is not a simple matter. But any comparison should consider normalized specifications for as much of a frequency range as required for an application, including harmonics, and with enough bandwidth to include future projects.