12-Port VNA Characterizes Crosstalk On High-Speed Lines

This 12-port vector network analyzer and novel multiport calibration algorithm combine for signal-integrity analysis on transmission lines and devices at multi-Gigabit speeds.

Harry Momjian
Marketing Manager
Anritsu Co., 490 Jarvis Dr., Morgan Hill, CA 95037-2809; (408) 778-2000, FAX: (408) 776-1744, E-mail: [email protected], Internet: www.us.anritsu.com.

Differential circuitry can effectively remove common- mode noise in highfrequency, high-speed designs. Differential devices and transmission lines are commonly used in highspeed digital buses, but are also working their way into many RF and microwave products, including cellular telephones. Testing such devices and lines requires additional test ports compared to measurements on traditional single-ended components. Fortunately, Anritsu Company (www.us.anritsu.com) has developed a practical 12-port vectornetwork- analyzer (VNA) test system capable of Sparameter measurements on single-ended, mixed-mode, and differential devices and components from 40 MHz to 65 GHz. The test solution is ideal for signal integrity measurements on high-speed devices and systems.

The Anritsu 12-port, 65-GHz system consists of a two-port model 37397D VNA system with an external test set and easily positioned port modules. The mobile test ports can be positioned close to almost any form of device under test (DUT), which makes the measurement system especially powerful for use with wafer probe systems. Calibration for the system, which is scalable for 4-, 8-, and 12-port applications, is handled with the help of flexible calibration and measurement software developed by PAF Company (www.pafmicro.com). This powerful software helps operators navigate through the most difficult multiport test scenarios, including with on-wafer measurements. For complex measurements with 144 S-parameters, the 12-port VNA system and software support an accurate 78-term error model with as few as 17 connections.

VNAs have traditionally been used for scatteringparameter (S-parameter) measurements on singleended, 50-O components and devices. As digital communications systems and buses increase in speed (and frequency), however, multiport, mixed-mode Sparameters represent an effective tool for characterizing the signal integrity (SI) of signal lines, buses, and components operating at high digital speeds. For example, a VNA can directly measure crosstalk on high-speed channels. Although the channels of a high-speed backplane are designed to be independent of each other, they often suffer from crosstalk between channels with high-speed/ high-frequency signals. For digital communications standards such as USB 3.0 or PCI-Express Gen 3, which can readily exceed speeds of 10 Gb/s, availability of a 12-port, 65-GHz VNA test system can provide meaningful SI measurements under full-speed conditions. A 12-port VNA system provides the needed measurement ports for simultaneous measurements to be made across multiple channels.

In addition to highspeed backplanes, an increasing number of wireless components and devices rely on differential (balanced) architectures for reduced susceptibility to electromagnetic interference (EMI). While a fourport VNA system can accommodate measurements on a single differential channel or device, more complex devices or components require a greater number of measurement ports. In fact, single-ended measurements for highspeed transmission lines can provide misleading results of loss performance since those lines are designed for differential operation.

As with high-speed backplanes, crosstalk between adjacent differential channels can degrade performance. In a pair of differential channels, crosstalk will be caused by energy from one channel, referred to as the aggressor line, coupling to an adjacent channel, referred to as the victim line.¹ In order to characterize the crosstalk of two differential channels with a VNA system, four test ports would be needed for the aggressor line and four test ports for the victim line. Of course, in a multichannel communications system or set of differential lines, pairs of lines cannot realistically be considered as isolated from surrounding lines. It is generally more practical to characterize the crosstalk from the two adjacent aggressor lines on a victim line, which requires four test ports for each line, or a total of 12 test ports.

Several architecture choices are available for assembling a 12-port VNA system. Two choices can be described by starting with a basic two-port VNA system that employs a pair of samplers and usually a dual directional coupler on each port to measure incident and reflected signals at those ports. In the first design, the number of ports is increased by adding a pair of samplers and its associated high-frequency hardware for every additional test port required. While straightforward, this "bruteforce" approach adds a great deal to the complexity of the test system as well as to its cost. In the second approach to creating a higher-port VNA system from a two-port starting point, a switch matrix is added to the twoport VNA in order to route signals to the original test ports. While this approach places high performance expectations on the switch matrix, it is considerably less complex and expensive than the first approach. Studies at Intel Corp. (www.intel.com) have also shown that, when used with the proper calibration techniques, this type of system architecture can provide comparable accuracy to multiplesampler test system approaches.

The Anritsu 12-port, 65-GHz system (Fig. 1) is based on a cost-effective extension of a four-port measurement engine, using a millimeter-wavefrequency switch matrix to route signals from the remote port modules to the four test ports, thereby increasing the effective number of measurement ports. The switch matrix, designed and produced at Anritsu, relies on low-loss, broadband 65-GHz switches (Fig. 2).

Choices for switch components included electromechanical switches and solid-state (PIN diode) switches. Electromechanical switches can achieve low insertion loss and high isolation over broad bandwidths, but their moving parts tend to result in relatively short mean time before failure (MTBF) and low performance repeatability compared to solid-state switches. PIN diode switches offer outstanding reliability across millions of switching operations, but tend to lack the outstanding RF/microwave performance of mechanical switches. To overcome the electrical shortcomings of solidstate switches, especially at millimeter- wave frequencies, Anritsu's engineers developed a novel design that combines the electrical performance of mechanical switches with the proven MTBF reliability of solid-state switches. The switch matrix achieves better than 95-dB isolation with low insertion loss of less than 6.0 dB at 60 GHz. This low loss contributes to the unit's high dynamic range. Placement of high-performance couplers within the mobile test ports and close to the DUT contributes to optimum raw directivity. Both factors lead to improved measurement stability and repeatability for 12-port measurements.

The use of the switch-matrix multiport VNA architecture also supports full calibration flexibility. The 12-port, 65-GHz VNA system includes a model 37000D VNA, a test set with the switch matrix with multiple model SM6272 test port modules, and calibration and measurement software. The model SM6272 external test port modules (Fig. 3) incorporate solid-state switches and highperformance directional couplers for high accuracy and repeatability through 65 GHz. They are compact enough (only 4.5 x 5.0 x 7.0 in.) to facilitate placement close to a DUT, even for on-wafer measurements. The modules can be added to a basic system to create a test solution with an increased number of ports (upgrading a four-port system to an eight-port system, for example).

Page Title

High measurement accuracy with a VNA system, whether it is a simple two-port system or a more complex 12-port VNA, requires a well-chosen and well-executed calibration procedure. Calibration is designed to de-embed or subtract the electrical contributions of test fixtures, cables, and other components used to transmit and receive signals from the device under test (DUT). By minimizing the lengths of test cables, as is done in the Anritsu 12-port VNA system, and employing high-performance test components, good raw directivity can be achieved. An effective calibration process uses precisely known electrical standards to characterize the test system's performance so that, ideally, measurements of a DUT reveal only the electrical performance of the DUT, and not the test system, fixtures, probes, or other test accessories.

An effective calibration procedure must gather enough data on a test fixture, probe, and other hardware used to route signals into and out of a DUT so that the effects of internal delays, mismatches, connections, and switching losses can be determined and subtracted from the measurement values for the DUT. A wide range of calibration procedures are currently used with VNA test systems, including short-open-load-through (SOLT) and through-reflect-match (TRM) methods. The choice of calibration technique is usually a function of the type of DUT and the availability of practical standards. Each calibration approach attempts to remove error terms from the raw measurement data taken on a DUT. As the number of ports for a VNA system increases, so too does the number of error terms.

These established VNA calibration approaches have been developed for traditional multiple-sampler/ directional-coupler VNA architectures rather than multiport VNA test systems employing a switch matrix such as the Anritsu 12-port, 65-GHz system. Anritsu worked with PAF (Turin, Italy) to create practical measurement and calibration software for the 12-port system. PAF, a firm known for its work on multiport VNA calibration algorithms, offered commercial software products for multiport VNA measurements and calibration through 20 GHz; Anritsu required accurate extension of such tools through 65 GHz.

The software reduces the time required for calibration and test and offers great flexibility, since the 12-port VNA measurement system would be used for coaxial measurements, with fixtured devices, and for on-wafer measurements. The system's split hardware architecture, with port modules that could be placed close to a DUT test fixture or right on a wafer probe stations, required calibration flexibility in support of multiple test applications. Such flexibility is critical for 12-port on-wafer test applications, in particular, where the number of available connections and calibration standards may be limited.

The calibration procedure that they developed for the Anritsu 12-port, 65-GHz VNA is based on an algorithm that dynamically computes the sequence of standards connections based on a minimum number of connections while still achieving the highest possible accuracy.² The resulting MMS-NT calibration and test software from PAF (Fig. 4) provides support for a wide range of calibration procedures, including SOLT, TRL, line-reflect-match (LRM), line-reflectline (LRL), short-open-load-reciprocal (SOLR), quick SOLT (QSOLT), and even half-leaky calibrations.

Because of the number of test ports in the new system, and the number of possible connections for a calibration, such as a 12-port SOLT, working two ports at time, minimizing calibration and test time was an important issue for the new calibration software. The MMS-NT software makes it possible to calibrate the 12-port VNA system with a single 12-term calibration between two opposite ports, which takes six connections for a TRL calibration, then measure sets of five throughs between each one of the original ports and five ports on the opposite side, adding only 11 more connections. The process can be readily implemented in a semi-automatic on-wafer system. Even in a coaxial test environment, a 17-connection calibration can still be achieved in less than 20 minutes.

In essence, the MMS-NT Calibration software adapts to unique test port configurations and provides the best accuracy with the minimum number of connections. The MMS-NT software also provides generous measurement and display capabilities, supporting 12-port differential parameters, time-domainreflectometer (TDR) measurements, eye-diagram displays, and data exports to Excel spread sheets for report generation and in Touchstone file formats for use with software simulation tools.

The calibration software contains internal algorithms and data structures that automatically compute and manage the error coefficients of the multiport measurement system. Once the multiport connection plan is defined, the software determines the optimum, most-efficient connection scheme for calibration of all defined ports. The software automatically defines the system ports and connector organizations, describes the proper set of calibration standards, optimizes the sequence of calibration standard connections, minimizes the number of required connections, computes the error coefficients, and de-embeds the measurement data.

The 12-port VNA system's use of compact SM6272 test port modules makes it suitable for on-wafer testing by positioning the test ports as close as possible to the DUT. The switches and all RF signal-routing cables are located inside each module and behind the test port directional couplers for minimum insertion loss and optimum measurement stability. Each test port module contains two test ports with ruggedized male V connectors. The MMS-NT software automatically converts the measurement connection plan into the best multiport calibration scenario including compensation for external test fixtures. Anritsu Co., 490 Jarvis Dr., Morgan Hill, CA 95037-2809; (408) 778-2000, Internet: www.us.anritsu.com.

REFERENCES

1. Thomas Ruttan, Brett Grossman, and Evan Fledell, "Comparison of Multi-Port VNA ArchitecturesMeasured Results," 68th Automatic RF Techniques Group (ARFTG) Conference Digest, 2006.
2. Andrea Ferrero and Valeria Tepatti, "Development and Implementation of New Multi-port Calibration Algorithm," 70th Automatic RF Techniques Group (ARFTG) Conference Digest, 2007.