The modern age of Vector Network Analyzers (VNA) started with the introduction of the HP 8510, in 1984. This was the first, fully integrated, fully automated VNA, and was the gold-standard for measurement instruments. It may come as a surprise that new generation of hardware and software improvements has led to new capabilities in VNAs, taking them out of the realm of linear S-parameters and moving into much more complex measurement classes, including frequency offset (mixer) measurements, power measurements, pulsed, compression and non-linear measurements, and most recently, noise figure measurements.
To start, let’s consider the block diagram of traditional network-analyzer architectures, such as the HP 8510 or the HP 8753. In the 8510, the source consisted of an external swept-frequency signal generator, such as the HP 8340, 8350 or 8360 series.
These were controlled by a GP-IB interface, and which limited the speed and flexibility sweep types. In the HP-8753 and 8720 series, the source was integrated with the analyzer, but the frequency was controlled by phase-locking it to a low frequency synthesizer, through the reference receiver down conversion. In both systems, the receiver was typically sampler-based, with a low-frequency synthesized Local-Oscillator (LO) driving a high-harmonic sampling down-converter. This resulted in a high noise figure (typically 40 dB), made it very susceptible to spurious responses, and required a low-loss path through the reference channel to allow phase-locking the source. The final IF path consisted of either a synchronous detector follow by DC digitizers (8510), or a low frequency digital IF (8753, 8720). The main processors were custom built and used a proprietary operating systems.
Recently, advances in RF and microwave hardware, including custom RF-ICs dedicated to network analysis applications, allowed a revolutionary leap in VNA architectures which completely changes the role that these new systems play in test and measurements applications. That, coupled with utilizing the best of new commercial processors provided the foundation for a new class of instruments: multi-function capable analyzers such as the PNA-X. A block diagram of the new architecture is shown in Figure 1.
Figure 1: Block diagram of an advanced VNA architecture, including dual reflectometers on each port, switchable access loops at the rear panel, dual sources with internal pulse modulators and generators, and a switchable reference path.
The first major architectural change was to provide separate synthesizers for the source and the LO, removing the requirement of using the reference channel for phase-locking the source. This, coupled with the introduction of network analyzers with two internal sources (another key architectural change), provided the hardware for a new class of measurement applications: Frequency Converter Measurements. As an example, the PNA series of network analyzers from Agilent provides a Frequency Converter Application that completely characterizes mixers and frequency converters with a very simple setup and calibration. Further, for the first time, a single instrument can perform magnitude, phase and group delay measurements, fully corrected, utilizing the Vector Mixer Calibration (VMC) method developed by this author [1], eliminating the need to synchronize between external sources and the VNA, always a point of difficulty. The mixer phase measurements required the additional architecture change of adding a bypass switch in the reference path, which allowed a reference mixer to be automatically added into the path during the conversion measurements, but removed from the path during the match measurements.
In addition, new software phase-locking techniques allow measurements on frequency converters with an embedded LO, such as found in satellite systems, and removes the requirements for access to the LO or even a 10 MHz reference[2]. This provides the best accuracy for a phase or group delay measurement of a mixer. For ease-of-use and best amplitude accuracy, the Scalar Mixer Calibration (SMC) method combines the time-honored techniques of two-port vector error correction along with a precision power calibration to provide what is arguably the best measurement of mixer conversion-gain available today. For example, during the power meter measurement, the match of the power meter is also measured, and mismatch between the source and the power meter is compensated for. Similarly, when the source is connected to the receiver to provide a receiver calibration, the mismatch between the source and receiver is removed.
Other non-linear measurements have also been updated using new calibration and measurement methods. The just introduced Gain Compression Application of the PNA-X also utilizes the precision power level calibration and vector match correction, along with a new error correction method that deals with non-linear gain responses to improve accuracy of 1 dB gain compression measurements. New two-dimensional data taking sweeps provide at least two orders of magnitude speed improvement in swept frequency gain compression, and adaptive sweep algorithms prevent any possibility of overdriving the amplifier-under-test by stopping the power sweep when the compression point is reached.
In the past, Intermodulation-Distortion (IMD) measurements required multiple sources (or sources with multi-tone capability) and a spectrum analyzer. However, two new architectural innovations in the PNA-X allow the dual-source model to provide IMD measurements at unprecedented speeds. The first is switched, internal combiner that allows the dual sources to be combined at a single port, providing two-tone capability. However, typical network analyzer sources have harmonics of only about -30 dBc, so the quality of source is not sufficient for two-tone measurements. But the sources in the PNA-X include an internal switched-filter that provide more than 60 dB (typ.) harmonics, and the combiner has more than 25 dB of isolation, meaning the internally generated IMD tones in the source are below -85 dBc.
When the combiner switch is not used, each of the dual sources provides an independent output, and the dual sources have individual phase control capability, so that they can be used together to provide a true-differential signal to a balanced device. New error correction techniques have been developed that not only correct for mismatch errors (after the measurement) but modify the drive signal, in the presence of the DUT and its mismatches, to ensure that the drive signal at the DUT reference plane is truly differential with exactly 180 degrees phase offset, and 0 dB amplitude offset. This correction is done real time, as each new DUT may present a different mismatch, causing a different error in the balanced signal [3].
Many active devices cannot tolerate continuous power, but require pulsed RF inputs. In the past, this required complicated test setups with additional equipment such as pulse generators and pulse modulated, along with complicated triggering schemes. But with the new class of custom RF-ICs, the pulse modulators can be integrated into the source block diagram, and can be controlled by the same DSP that is used for data taking. In the PNA-X, a high speed DSP allows 17 nsec pulse width steps, and pulses as narrow as 35 nsecs. Using the same DSP to control the source pulses and the receiver ADC readings allows the PNA-X to use sophisticated digital gating techniques that dramatically improve the system noise floor in pulsed mode. The basic idea is that the DSP can ignore ADC readings when the pulse is off, eliminating one source of broadband noise during the measurement.
In many cases, higher power, or additional measurement equipment may be required to be inserted into the signal path. In the past, special switching test sets or multiplexers were required between the VNA and the DUT. But in the architecture of figure 1, we can see that each port also has a switchable “RF-loop” that allows access to the signal path, behind the directional couplers. This is an ideal place to put a booster amplifier, or to inject an additional signal (perhaps from a complex modulated source), or to sample an output signal (perhaps to a vector spectrum analyzer), without degrading in any way the VNA performance.
Finally, the last addition in the block diagram is a new noise receiver path, figure 2. The noise receiver requirements are quite different from the VNA receiver requirements, but they can both share the synthesized LO. The mechanical signal switching ensures that in the noise figure measurement mode, the lowest loss path from the noise receiver to the DUT interface is maintained. Adding another directional coupler ensures that full S-parameter measurements can be made even when in the noise mode, without switching the mechanical switch. This is critical in high production volume applications. In addition, another bypass switch is added to the port 1 path to allow an electronic noise tuner to be added to the input path. This noise tuner is used to pull the match on the DUT over several impedances, from which the exact 50 ohm noise figure is determined. The switch is set to the internal mode when Intermodulation measurements are needed, avoiding distortion caused by electronic tuners.
Figure 2: Switchable paths for adding the noise tuner (at port 1) and a low loss path to the noise receiver (port 2).
With this final addition, the new VNA architecture becomes ideally suited to multi-function testing in both R&D and production environments. In almost every application, the effects of instrumentation and cable loss or mismatch can be removed by full vector calibration. The highly flexible switch paths provide access for almost any conceivable additional test equipment, making this an ideal architectural base for even more complicated test stands. Finally, the very high speed data taking capability (up to 20000 points in less than two seconds) means multi-function, multi-dimension device characterizations are practical in minutes now, which would otherwise take hours utilizing more conventional methods.
References:
[1] Vector mixer characterization for image mixers, Dunsmore, J.; Hubert, S.; Williams, D.;Microwave Symposium Digest, 2004 IEEE MTT-S International, Volume 3, 6-11 June 2004 Page(s):1743-1746 Vol. 3
[2]New methods for testing frequency converters with embedded local oscillators Dunsmore, Joel P.; European Microwave Conference, 2007, 9-12 Oct. 2007 Page(s):234-237
[3] Complete Pure-Mode Balanced Measurement System,Dunsmore, Joel; Anderson, Keith; Blackham, David; IEEE/MTT-S International Microwave Symposium, 2007., 3-8 June 2007 Page(s):1485-1488