A unique instrument
A network analyzer is a powerful instrument that, when used correctly, can achieve extremely high accuracy. Its applications are wide-ranging and indispensable in many industries, especially in measuring the linear characteristics of wireless radio frequency (RF) components and devices. Modern network analyzers can also be applied to more specific situations, such as signal integrity and material measurements. With the introduction of the industry's first PXI network analyzer—the NI PXIe-5630—you can completely break free from the high cost and large footprint of traditional network analyzers and easily apply network analyzers to design verification and production line testing.
The development of network analyzers
You can use the NI PXIe-5630 vector network analyzer shown in Figure 1 to measure the amplitude, phase, and impedance of a device. Because a network analyzer is a closed excitation-response system, you can achieve excellent accuracy when measuring RF characteristics. Of course, a thorough understanding of the basic principles of a network analyzer is crucial to maximizing its benefits.
Figure 1. NI PXle-5630 Vector Network Analyzer
Over the past decade, vector network analyzers have surpassed scalar network analyzers in popularity due to their lower cost and efficient manufacturing technology. While network analysis theory has existed for decades, the first modern stand-alone benchtop analyzer wasn't developed until the early 1980s. Prior to this, network analyzers were bulky and complex, consisting of numerous instruments and external components, and their functionality was limited. The introduction of the NI PXIe-5630 marks another milestone in the development of network analyzers, successfully bringing vector network analysis capabilities to a flexible, software-defined PXI modular instrument platform.
Typically, extensive measurement practice is required to achieve accurate amplitude and phase parameter measurements and avoid major errors. Due to the inherent uncertainties in RF instrument measurements, small errors can easily be overlooked. However, network analyzers, as sophisticated instruments, can detect even extremely small errors.
Network Analysis Theory
Network is a frequently used term with many modern definitions. In network analysis, a network refers to a set of internally interconnected electronic components. One of the functions of a network analyzer is to quantify impedance mismatches between two radio frequency (RF) components, maximizing power efficiency and signal integrity. Whenever an RF signal enters from one component to another, a portion of the signal is always reflected, while another portion is transmitted, similar to the diagram in Figure 2.
This is analogous to light emitted from a light source striking an optical device, such as a lens. The lens is like an electronic network. Depending on the lens's properties, some light is reflected back to the light source, while the rest is transmitted. According to the law of conservation of energy, the sum of the energy of the reflected and transmitted signals equals the energy of the original or incident signal. In this example, heat loss is usually negligible and can be ignored.
Figure 2. Using light as an analogy to a fundamental principle of network analysis
We can define the reflection coefficient (G), a vector containing amplitude and phase, representing the proportion of reflected light to the total (incident) light. Similarly, we define the transmission coefficient (T) to represent the vector ratio of transmitted light to incident light. Figure 3 illustrates these two parameters.
Figure 3. Transmission coefficient (T) and reflection coefficient (G)
By analyzing the reflection and transmission coefficients, you can gain a deeper understanding of the performance of the device under test (DUT). Recalling the analogy of light, if the DUT is a mirror, you would want a high reflection coefficient. If the DUT is a lens, you would want a high transmission coefficient. Sunglasses, on the other hand, may possess both reflective and transmissive properties.
The measurement methods for electronic networks are similar to those for optical devices. The network analyzer generates a sinusoidal signal, typically a swept-frequency signal. The DUT responds by transmitting and reflecting the incident signal. The strengths of the transmitted and reflected signals generally vary with the frequency of the incident signal.
The response of a distributed underpinning device (DUT) to an incident signal characterizes both the DUT's performance and the system's characteristic impedance discontinuity. For example, a bandpass filter has a high reflection coefficient outside the band and a high transmission coefficient inside the band. If the DUT deviates slightly from its characteristic impedance, it will cause impedance mismatch, generating additional unwanted response signals. Our goal is to establish an accurate measurement method for the DUT response while minimizing or eliminating uncertainties.
Network Analyzer Measurement Methods
The reflection coefficient (G) and transmission coefficient (T) correspond to the proportions of reflected and transmitted signals in the incident signal, respectively. Figure 3 illustrates these two vectors. Modern network analysis extends this idea based on scattering parameters or S-parameters.
S-parameters are complex vectors that represent the ratio of two RF signals. They include amplitude and phase, represented as real and imaginary in Cartesian form. S-parameters are represented in an S-coordinate system, where X represents the output of the DUT being measured, and Y represents the input of the DUT excited by the incident RF signal. Figure 4 illustrates a simple two-port device, which can be characterized as an RF filter, attenuator, or amplifier.
Figure 4. S-parameter representation of a simple two-port device
S11 is defined as the proportion of energy reflected from port 1 to the incident signal at port 1, and S21 is defined as the proportion of energy transmitted to port 2 of the DUT to the incident signal at port 1. Parameters S11 and S21 are forward S-parameters because the incident signal originates from the RF source at port 1. For a signal incident from port 2, S22 is the proportion of energy reflected from port 2 to the incident signal at port 2, and S12 is the proportion of energy transmitted to port 1 of the DUT to the incident signal at port 2. These are inverse S-parameters.
You can extend this concept based on multi-port or N-port S-parameters. For example, RF circulators, power dividers, and couplers are all three-port devices. You can use similar analysis methods to two-port devices to measure and calculate S-parameters such as S13, S32, and S33. S-parameters with consistent subscripts, such as S11, S22, and S33, represent reflected signals, while S-parameters with inconsistent subscripts, such as S12, S32, S21, and S13, represent transmitted signals. Furthermore, the total number of S-parameters equals the square of the number of ports on the device; this is necessary to fully describe the RF characteristics of a device.
S-parameters characterizing transmission, such as S21, are similar to other common terms like gain, insertion loss, and attenuation. S-parameters characterizing reflection, such as S11, correspond to voltage standing wave ratio (VSWR), return loss, or reflection coefficient. S-parameters also offer other advantages. They are widely recognized and used in modern RF measurements. You can easily convert S-parameters to H, Z, or other parameters. You can also cascade S-parameters for multiple devices to characterize the RF characteristics of composite systems. More importantly, S-parameters are expressed as ratios. Therefore, you don't need to set the incident source power to an exact value. The DUT's response will reflect any minute differences in the incident signal, but these differences are eliminated when the transmitted or reflected signal is characterized as a ratio relative to the incident signal.
Network Analyzer Structure
Network analyzers can be divided into two types: scalar (containing only amplitude information) and vector (containing both amplitude and phase information). Scalar analyzers were once widely used due to their simple structure and low cost. Vector analyzers can provide better error correction and more complex measurement capabilities. With technological advancements, increased integration and computational efficiency, and reduced costs, the use of vector network analyzers is becoming increasingly widespread.
The network analyzer has four basic functional modules, as shown in Figure 5.
Figure 5. Basic functional modules of a modern network analyzer
The signal source generates the incident signal, supporting both continuous and discrete frequency sweeps, and its power is adjustable. The signal source is fed into the DUT input via a signal separation module, which can be considered a test device. Here, the reflected and transmitted signals are separated and measured in different components. For each frequency, the processor measures the signal and calculates parameter values (e.g., S21 or VSWR). User calibration primarily provides error correction for the data and will be discussed in detail later. Finally, when interacting with the network analyzer, you can view the parameters and corrected values on the display and use other user functions, such as zooming in and out of waveforms.
Depending on the performance and cost of the network analyzer, there are several ways to implement the four modules in the architecture. The test setup can be designed as a transmission/reflection (T/R) setup or a full S-parameter setup. Among them, the T/R test setup is the most basic implementation, and its structure is shown in Figure 6.
Figure 6. Structure of the network analyzer T/R test setup
The T/R configuration includes a stable signal source capable of providing a sinusoidal signal of a specified frequency and power; and a reference receiver R, connected to a power divider or directional coupler, used to measure the amplitude and phase of the incident signal. The incident signal is emitted from port 1 of the network analyzer and fed into the input of the DUT. The directional coupler receiver A measures any signal reflected back to port 1 (including amplitude and phase). The directional coupler functions similarly to a resistor bridge, both used to separate signals; you can choose based on performance, frequency range, and cost requirements. The signal is transmitted through the DUT to port 2 of the network analyzer, where receiver B measures the amplitude and phase of the signal.
Receivers have different structures to meet different characteristic requirements and can be viewed as narrowband receivers with downconverters, intermediate frequency filters, and vector detectors, similar to vector signal analyzers. They can extract the real and imaginary parts of the signal for calculating amplitude and phase information. Furthermore, all receivers use the same phase reference as the signal source, allowing you to calculate the phase relationship between the received and incident signals under the same phase reference.
The T/R structure offers high cost-effectiveness, simple structure, and good performance. However, it only supports forward parameter measurements, such as S11 and S21. To measure reverse parameters, the DUT needs to be disconnected and reversed, or controlled by an external switch. Since the source (incident signal) cannot be switched to port 2, port 2 has limited error correction capabilities. If the T/R structure design meets your project requirements, this structure is a high-precision and cost-effective option.
The full S-parameter structure is shown in Figure 7, with a switch embedded in the signal path after the reference receiver coupler.
Figure 7. Full S-parameter network analyzer
When the switch is connected to port 1, the analyzer measures the forward parameters. When the switch is connected to port 2, you can measure the reverse parameters without resetting the external connections of the DUT. Targeted receiver B at port 2 measures the forward transmission parameters and the reverse reflection parameters. Receiver A measures the forward reflection parameters and the reverse transmission parameters.
Because the switch is positioned on the network analyzer's measurement path, user calibration must account for switch uncertainties. Even so, slight differences may still exist between the two switch positions. Furthermore, over time, switch contacts wear down, requiring more frequent user calibrations. To address this, the switch can be moved to the source output, and two reference receivers, R1 and R2, can be used, corresponding to the forward and reverse directions respectively, as shown in Figure 8. This higher-performance architecture comes with increased cost and complexity.
Figure 8. Full S-parameter network analyzer with dual reference receivers
The basic structure of a network analyzer is largely implemented in the test setup. Once the analyzer measures the amplitude and phase of the incident signal (R reference receiver) and the transmitted signal, or the amplitude and phase of the reflected signal (A and B receivers), it can calculate the four S-parameter values, as shown in Figure 9.
Figure 9. Four S-parameters of a full two-port network
You can choose the appropriate network analyzer architecture by considering factors such as application, performance, accuracy, and cost.
Error and uncertainty
Understanding the sources of uncertainty in a vector network analyzer helps you adopt effective user calibration methods. For the complete two-port network analyzer architecture shown in Figure 10, we start the analysis from the front.
Figure 10. Uncertainty of a complete two-port network analyzer source.
First, the initial uncertainty lies in signal loss due to frequency misalignment or divergent paths of the transmitted and reflected signals. Second, there's the difference between the DUT's input impedance and the network analyzer or system impedance. Similarly, similar issues exist at the DUT outputs, specifically source and load mismatches.
The efficiency of directional couplers used for signal separation also needs to be considered. An ideal directional coupler produces an output signal in the coupling arm that is proportional to the standard signal in one direction of the main arm, while signals in the opposite direction produce no output signal. The difference between the coupler output (coupling arm) and the standard input signal (straight-through arm) is the coupling coefficient. The coupling coefficient is typically between 10 dB and 30 dB, meaning that when the input signal passes through the straight-through arm in the appropriate direction, the output RF power level is 10 to 30 dB lower than it is.
A directional coupler should not produce an output for signals traveling in the opposite direction. However, this is practically difficult to achieve. Even small signals traveling in the opposite direction can still produce an unwanted response at the output of a real coupler. This unwanted signal is called coupler leakage. The difference between the coupling coefficient and the coupling leakage is called the directionality of the coupler.
Finally, there's isolation. The receiver at port 2 detects a small amount of radiated or conducted signal from port 1; in modern network analyzers, this unwanted leakage is typically minimal. Generally, it doesn't affect measurements unless the DUT has very high losses. Although recommended in many modern vector network analyzers, isolation is only an optional operation during calibration.
A complete network analyzer's forward uncertainty sources include: transmission and reflection tracking; load and source matching; directionality and isolation; these, combined with six backward error terms, result in a total of 12 error terms. User calibration needs to fully account for these 12 errors to obtain appropriate correction factors for use in measurement data. This correction is the main reason for the significant accuracy of vector network analyzers.
calibration
Calibration of RF equipment often requires periodically sending the instrument to an accredited instrument calibration laboratory to ensure that the instrument operates within the manufacturer's specifications. The laboratory also typically adjusts the instrument's performance to a standard, such as the standard specified by the National Institute of Standards and Technology (NIST).
Network analyzers are no exception. They require so much periodic calibration that they sometimes fall short of high accuracy, and user calibration is also frequently necessary. Network analyzer calibration is typically accomplished using a set of calibration standards included in a network analyzer kit, or by user-defined standards. A series of correction parameters are generated by comparing known data stored in the network analyzer with measurement data produced according to the calibration standards. These are then used in the calibration test data to compensate for the error sources discussed in previous chapters.
Many factors determine how often user calibration needs to be performed. Factors to consider include the required test accuracy, environmental factors, and the repeatability of the DUT connections. Typically, network analyzers require user calibration every few hours or every few days. You should determine the calibration frequency based on verified standards, identification of sources of test instability, and personal experience. It's important to note that this discussion uses the term "periodic calibration" to describe user calibration and should not be confused with the recommended annual certified factory calibration.
Three series of calibrations are frequently used in the calibration of network analyzers:
1. Short-circuited, open-circuited, loaded, direct (SOLT)
2. Direct, reflective, linear (TRL)
3. Automated calibration using an external automated calibration model
Because each calibration series has many different requirements, the method used needs to be determined based on the DUT, test system, and test requirements. Since SOLT is widely used, we will use it to illustrate the variations within a calibration series.
SOLT requires the use of short-circuit, open-circuit, load-type, and straight-through standards in the system (and DUT) and impedance. Precise standard data, determined by their mechanical characteristics, is loaded into the network analyzer before calibration. The location where you connect the calibration standard (network analyzer port, cable end, or inside the test fixture) is where the test begins and ends. This is the reference platform or test platform.
To further clarify, you must create a pass-through connection using a pluggable connector. For example, a male-to-female connection, or any other connection that does not require external devices or adapters, will complete the pass-through during SOLT testing. Inserting any device during calibration, or using a device that is not used in the calibration measurement, will result in measurement errors.
If you cannot make a straight-through connection, it will be called non-pluggable. There are several ways to handle non-pluggable situations. The simplest is to use a set of phase-matched adapters (included in most calibration kits) and one for each type of short, open, and load. Use one adapter to make the straight-through connection during calibration, and then use a suitable adapter to switch with the DUT during calibration testing.
Other calibrations in the SOLT series include response-type calibration. It's relatively fast, but not as precise as removing bandwidth loss at frequency. It only considers both forward and reverse cases in a 12-error model. You can perform a one-port calibration by placing short circuits, open circuits, and loads on port one. This can save time if you only need to perform a single-port measurement, such as the return loss of an antenna. An enhanced one-port calibration is equivalent to a full one-port calibration and uses a straight-through connection to measure port two, which is common in T/R structures where port two has no source. Finally, there is a full two-port SOLT calibration that, according to the calibration specifications, can be performed with short circuits, open circuits, and loads on both ports. Figure 11 summarizes these common SOLT series calibrations.
Figure 11. Typical SOLT calibration
SOLT and TRL calibrations have many variations. You can use TRL calibration in applications where actual terminals are absent, such as probe nodes, or if the DUT is in a test fixture. Because TRL does not require a load, it can be implemented well in these cases.
Automated calibration is a relatively new approach that has quickly gained popularity due to its speed, repeatability, and ease of use. Furthermore, it eliminates most human intervention, significantly reducing the probability of errors during calibration. These units typically include an electronic component, such as a diode, terminal, or other marker, along with encoded, detailed electronic description information stored in an EEPROM. When connected to a network analyzer, automated calibration is set to different states. During calibration, these states are measured and compared with the corresponding states stored in the EEPROM to achieve the correct correction values.
Regardless of the calibration method used, random sources of error should be avoided. Reducing IF bandwidth, using averages to decrease noise, and providing better results are all beneficial. When calibrating a network analyzer, high-quality components, solid measurement practices, and a comprehensive understanding of calibration procedures and the instrument are equally important.
Process requirements
When performing precise measurements with a network analyzer, it is essential to understand and correctly execute each step to obtain optimal results. This requires the use of high-performance components and comprehensive measurement practices. Consider the RF connection between a well-calibrated network analyzer with correction parameters and a high-performance DUT requiring precise measurements:
• Are there cables, adapters, and other high-performance components?
• Did you clean them properly?
• Was the appropriate torque used?
Even the best network analyzer is useless if the performance of the RF connected to the DUT does not match the specified system accuracy.
When using a network analyzer, procedures are very useful. Procedures can enhance operations and improve results. Below is an example architecture using a network analyzer.
Prepare
• Prepare network analyzer and DUT
• Clean, inspect, and measure all connectors.
• If using SOLT calibration, select a method for handling non-inserted connections.
• Connect the analyzer cable and adapter to the analyzer.
operate
• Pre-tuned network analyzer
• Configure source parameters, including frequency, power, speed coefficient, and IF bandwidth.
• Connect the DUT, verify the installation, cables, adapters, and operation.
• Select S-parameter measurement and display format
• If possible, set specific measurement targets, such as the expansion of the reference plane.
• Observe the response
• Remove DUT
calibration
• Select an appropriate calibration toolkit or define an input calibration standard.
• Set the IF bandwidth and average it to minimize noise during calibration.
• Manual calibration or automatic calibration
• Verify calibration quality using well-known verification standards.
• Save instrument status and calibration
implement
• Connecting the DUT
• Obtain appropriate calibration parameters from the calibration procedure.
• Measure and save DUT parameters
One instrument, multiple applications
When used correctly, a network analyzer is one of the most accurate RF instruments, typically with an accuracy of ±0.1 dB and ±0.1 degrees. It can perform accurate, repeatable RF measurements. Modern network analyzers offer configurations and measurement capabilities as extensively as their applications. Choosing the right instrumentation, calibration, features, and employing reliable RF measurement methods can optimize your network analyzer's results.
