Introduction

The last ten years have witnessed an increasing interest in multiport S-parameter measurements, i.e. S-parameter measurements of devices with more than two ports, for two main reasons: the first one is the increasing complexity of modern microwave devices and circuits and the use of more complex MMICs.

But the main reason is definitely the shift toward microwave frequencies of the personal computer's processors speed, which implies that such digital applications must now face typical microwave challenges. These topics have recently been addressed in [1]. Preserving the signal integrity of a microwave signal through the packages, sockets, connectors, and PCB traces, commonly found in today's computer systems, is one of the main issues. System architectures with hundreds of parallel channels, operating at higher and higher data rates, involve microwave multiport measurements for the characterization, design, and analysis of the structures and their effects on the signals. Microwave designers and engineers are thus facing new challenges in multiport measurement hardware and calibrations.

The first challenge comes from the typical media of digital interconnections: the PCB. It can include both planar and three-dimensional (3-D) DUTs, as found, for example, in memory modules. So, on one hand many data lines must be connected and measured simultaneously, and they do not necessarily lie on a single plane. On the other hand, these connections from the boards to the typically coaxial test ports of the VNA must have good performances at microwaves, i.e. be “transparent” for the measurements.

In the last few years, the field of microwave testing has been evolving rapidly with the development and introduction of digital techniques and microprocessor based instruments, and reaching higher and higher frequencies. Nevertheless, the basic underlying concepts, such as frequency synthesis, network analysis and calibration, and spectrum analysis, still constrain even the more modern equipment.

In recent years, microwave instrumentation has had to meet new testing requirements, from 3G and now LTE wireless networks, for millimeter wave and THz applications. Thus instrumentation and measurement techniques have evolved from traditional instruments, such as vector network analyzers (VNAs), to increasingly more complex multifunction platforms, managing time and frequency domains in a unified, extensive approach. We can identify two main directions of evolution:

• linear measurements, essentially S-parameter techniques;

• nonlinear measurements, for high power and nonlinear device characterization.

S-parameter measurements have been moving towards the multiport and millimeterwave fields. The first to characterize multi-channel transmission structures such as digital buses, and the latter for space or short-range radio communication or security scanner applications. New calibrations and instrument architectures have been introduced to improve accuracy, versatility and speed.

Nonlinear applications have also evolved. Traditional high power transistor characterization by load-pull techniques now also typically includes time domain waveform measurements under nonlinear conditions. These techniques can nowadays also handle the broadband signals used in most communication links, or pulsed signals. Moreover, even nonlinear measurements had to evolve to multiport, with differential and common mode impedance tuning, due to the spreading of amplifiers and devices exploiting differential configuration.

Introduction

Although VNAs are probably the most advanced microwave systems, with broadband sources, high-speed and high-dynamic-range receivers, the intrinsic property of distribute components makes a calibration procedure mandatory to obtain reasonable results due to an enormous systematic error. To stress this fundamental problem, imagine weighing 300 g of ham with a one-ton plate scale! This is more or less the same influence as systematic phase error introduced by a 1 meter teflon cable in front of a VNA port at 10 GHz if we are trying to measure 1 degree of phase-shifting on a DUT S11 parameter. Not only the phase, but also the magnitude as well is affected, due to different attenuation paths in various system sections. Clearly without a proper correction the measurement quality would be unacceptable.

In the early development of VNA, hardware compensation with line stretchers and variable gain amplifiers was attempted, but it's only with the introduction of computer controlled digital VNAs, that specific signal processing techniques allow a real-time correction of the most important errors. During the last forty years several algorithms have been proposed especially for one- or two-port VNAs; some of them like TRL, SOLT, LRM, and SOLR, became a de-facto standard in all modern VNA firmware; however, many others have been proposed to solve particular problems [1–4]. This chapter presents a review of the error models and the main VNA calibrations, by focusing the attention on their commonalities and by pointing out their different fields of application. The development follows the system approach born in the early 1990s, rather than the traditional one based on the analysis of all possible sources of error [3–5].

Introduction

Chapter 12 has explained how S-parameter measurements in small-signal conditions are not adequate to characterize an active device for a relevant number of applications. In large-signal conditions the performance of the active device under test (e.g. output power, gain, and efficiency) does not only depend on the chosen bias point and the excitation frequency, but also on the input power level and the loading conditions at the input and output ports.

The “Rieke diagrams” were already used in the early 1940s to show how the performances of microwave tubes and oscillators vary as a function of load impedance [1]. The oscillating frequency and output power used to be plotted on the Smith chart as contours at constant level. The modern term of “load-pull” comes from these early applications, where a change of the load impedance would have “pulled” the output oscillator frequency.

More precisely, the term “source and load-pull systems” refers nowadays to the set of instrumentation (at microwaves and at lower frequencies) needed to address two main issues:

• setting and monitoring the DUT loading conditions,

• measuring the DUT performance of interest.

Chapter 12 has already described in detail the most modern large-signal measurement techniques to address the second point. However, it will be clear in this chapter that the two problems are strictly correlated, as the solutions chosen for the former have a considerable impact on the latter.