Improved UHF Nonlinear Measurements Using Active Load Pull

With the move to digital television and the subsequent reallocation of UHF band frequencies in the US and other countries, some very attractive spectrum is being opened up that creates opportunities for commercial and military wireless technologies and applications in the 200 to 800 MHz frequency band. This, in turn, is leading to a growing need for high efficiency UHF power amplifiers and a renewed development in advanced amplifier topologies that can operate over a wide fractional bandwidth. In this article, a new active load pull technology is presented, that not only simplifies UHF testing, but also removes several of the limitations and inaccuracies common to passive load pull measurements.

Of course, the demand for more efficient and lower cost design spans all wireless communications at all frequency ranges. Since this trend is directly tied to power amplifiers, an improvement in power-added efficiency (PAE) across a wider fractional band of operation can lead to the need for fewer components in multi-band designs. Traditional testing approaches are starting to reach their limits, leading to such problems as multiple design cycles due to poor correlation between expected and actual results. For complex wireless radio formats, including those technologies proposed for implementation at UHF frequencies, efficiency can be improved by operating devices in their nonlinear regions, where the natural device distortion products can be represented by harmonic components.

The Cardiff Centre for High Frequency Engineering, at Cardiff University in Wales, has been researching time domain techniques and load pull measurement technologies in close cooperation with industry leaders in the mobile wireless and aerospace defense area for over a decade. The need for a practical design methodology for optimizing power amplifier designs has led to the development of the open-loop, active harmonic source and load pull techniques.

Evolution of Nonlinear Measurement Techniques

Network analyzers and S-parameter theory are typically used for testing both passive and active components. Across any frequency, as long as active components are kept in the linear region, S-parameter measurements are the accepted standard for CW testing. Network analyzers can be used at frequencies as low as 9 kHz and beyond 100 GHz. However, as active components move into nonlinear regions, the theory and measurements become more complex. Typical network analyzers are swept frequency instruments and the need for accurate phase correlation across the harmonics has been a challenge. While recent advances in calibration techniques make it possible to measure an accurate harmonic response for a repetitive signal, device and PA designers still need load pull capability to test under a variety of impedances.

Passive load pull solutions have been a valuable tool for device and (PA) power amplifier designers for some time. A power sensor, VNA or sampling scope can be used for nonlinear measurements, as shown in Figure 1. Impedance values of the harmonics at the device under test (DUT) input and output can be adjusted using tuners to provide impedance values for designing matching circuits and power levels. A typical configuration consists of one tuner at the input, to reduce the mismatch between the input source and the device under test, and one tuner at the output to generate the required loading condition. Tuner positioning is highly accurate and repeatable and this allows them to be calibrated. During the calibration, an accurate relationship is established between the position of the slug and the sliding short. Based on this calibration stage, the power inserted into the DUT can be determined from the power level set by the input source, while the output power can then be calculated from the power sensor reading. Tuners can be concatenated to allow for the additional control of harmonic impedances and to look at multiple harmonics. Passive tuners are also available that can handle multiple harmonics in a single tuner.

The biggest disadvantage of these systems is that they generate impedances over a large frequency range and not just the harmonic impedances; they can only control impedance at a single frequency. Impedance control is achieved by positioning the slug, which physically affects all remaining tuner operating frequencies. Consequently, harmonic impedances are not only uncontrollable, but also change their value with every new position, resulting in measurement artifacts that are not representative of real circuits. This can lead to significant performance variations between load pull measurements and realized power amplifier performance.

The same disadvantage is true for harmonic tuners (passive tuners with multiple slugs and sliding shorts) that allow a limited control of harmonic impedances as the higher harmonics (above 3^rd harmonic) are not controlled and still have large variance from real circuits. For instance, a small current coming from the DUT can be transformed, due to ohm’s law, into a very large voltage with harmonic impedances (above 3^rd harmonic) that are easily generated by the tuners. The uncontrolled load variations make it impossible to achieve clean waveforms as required for waveform engineering. Further, coupling artifacts can create significant capacitive and inductive loading, making the time-domain waveforms highly distorted.

The position of the passive tuner between the DUT and the measuring receiver makes it difficult to distinguish artifacts from the tuner and the DUT itself. This effect has potential impact on the input or output matching network in a PA design. Also, due to the losses between the DUT and the tuners, only part of the Smith Chart can be covered. At lower frequencies, this effect is mitigated by the relatively low losses. However, it can increase significantly for higher frequencies, such as the harmonics of a signal.

For UHF applications, passive tuners represent a physical challenge as their size is proportional to wavelength, as shown in Table 1. The control of the reflection coefficients at lower frequencies is limited by the prohibitive length requirements for the coaxial line for the respective wavelength. Passive tuners that operate in the several hundreds of megahertz can be three feet long or more. Adding these tuners to either side of the DUT creates a number of challenges. For on wafer applications, the tuner size and weight can add to the cost and complexity of the probe station. In addition, there are vibration concerns as impedances on these large tuners are varied. The move to increase PA performance at UHF frequencies clearly creates a problem for passive load pull.

Open-Loop, Active Load Pull

A new approach to nonlinear measurements, that addresses many of the problems just described, is called open-loop, active load pull or active load pull for short. This technique uses a separate coherent signal source to stimulate either the source or load side of the DUT, thus removing any uncontrolled interaction between the DUT and load pull system. This configuration is shown in Figure 2. The open-loop system absorbs the signal that is generated by the device under test and injects a signal that is generated by an independent source back into the device. The amplifier bandwidth is large enough to cover all harmonic frequencies where impedance control is required.

For UHF applications, the need for passive tuners—and their physical size challenges—disappears. With active load pull, impedance variations are created electronically. Eliminating the need for passive tuners removes the physical constraints for performing load pull measurements at low frequencies. Measurements at 2 GHz, 200 MHz, or 200 kHz become a function of source, not wavelength.

Within the open-loop architecture, all in-band and out-of-band impedances are tightly controlled. All the frequencies at which the active load pull is operating are controlled by the load source, while all other frequencies are terminated into 50 Ω. When no signal is output by the arbitrary waveform generator (AWG), the active source/load pull architecture presents a broadband 50 Ω impedance environment and, therefore, a reflection coefficient that is nearly zero over the entire bandwidth of the system. The 50 Ω environment is changed only at the frequencies that the AWG produces. As a consequence, the open-loop active load pull architecture does not face problems with coupling artifacts, as previously noted with passive tuner techniques.

With an AWG capable of generating any signal within a more than 5 GHz bandwidth on each channel, it is possible to control and modulate all frequency components in phase and magnitude irrespective of whether it is a CW or a complex multi-tone signal. Interestingly, the open-loop architectures can even be safely used to generate reflection coefficients larger than unity. This allows for unique investigations of the interaction between a driver and main PA stage. Due to the stability of the open-loop architecture, it can be easily used in measurement systems.

Another usage is related to load pull measurements at baseband frequencies. Electrical memory effects are a resultant hysteresis of a rapidly changing modulation envelope signal due to the parasitic capacitance across the surface of the transistor. The phenomena and effect are represented in the baseband measurement of the modulated signal. The impedance control at baseband (below 50 MHz for most modulated signals) is an important part of being able to accurately predict the memory behavior.

With DC to more than 5 GHz of bandwidth, the open-loop architecture can be used at baseband, fundamental and higher harmonic frequencies. The use of a sampling scope as the nonlinear receiver opens this up for lower frequencies. At higher power levels, it is often more cost efficient to implement the amplification over the required number of harmonics through use of a narrowband PA with center frequencies located around the fundamental and harmonic frequencies. The separation of the harmonic frequencies can be obtained by means of a multiplexer. The same multiplexer can then be used re-combine the harmonic signals.

The fact that the active load pull system is positioned further away than an impedance network within a real circuit design can be readily compensated by controlling the phase and magnitude of each frequency component within the signal generated by the AWG. As the active load pull system is placed outside the calibrated path (comprised of couplers and their connection to the sampling scope) the load pull can be reconfigured without the need to re-calibrate the measurement system.

Accurate, Fast Large-Signal Simulations

Designers have been challenged over the years from the discrepancies observed between simulations and measured performance. This can result in the need for multiple design cycles that increase development costs and often add months in development cycle time. Traditionally, characterizing nonlinear device behavior has involved the use of measurements and modeling to achieve optimum results. On the measurement side, existing products have been expanded with application software and hardware in an attempt to address this market and assist with the creation of behavioral models. However, these PA measurement techniques often lack a coherent integration with a harmonic source/load pull system resulting in devices and amplifiers being characterized at impedances that are different from their final application. This makes it challenging to translate the measured device performance into a PA design or achieve the potential performance available from the device or the employed PA architecture.

Active load pull-based systems allow for accurate and fast large-signal simulations, including both high power and non-50 Ω applications and data sets beyond the Smith Chart for complex multi-stage amplifier design optimization. Because the system can measure basic voltage and current waveform data, it provides the information needed for harmonic balance and envelope simulations. The time domain data collected by the sampling scope can be transformed into the frequency domain and stored in a format that can be imported into EDA tools. As shown in Figure 3, this technique provides excellent replication of measurement data within nonlinear simulations. Once the data is imported, it can be used for spot-analysis and the authentic recreation of the device behavior with the look and feel of a behavioral model. This allows for complete characterization of nonlinear devices and amplifiers and their use in common nonlinear simulation engines. These behavioral models can be used by designers in complex component or system-level simulations.

Conclusion

New applications and technologies are being developed at UHF frequency bands for commercial, aerospace and defense uses. When used to help optimize power amplifier designs, traditional passive load pull solutions have both limitations as well as challenges at these low frequencies. Designers need next generation tools to optimize their RF device and PA designs. Active load pull technology not only simplifies UHF testing, but also removes several of the limitations common to passive load pull measurements. Active load pull and waveform engineering enable the designer to more accurately understand the higher order source/load harmonics and thus achieve near theoretical performance. Improving the correlation between design tools and measured results has the potential to reduce design cycles and enable more efficient designs. n

Darren McCarthy holds a BSEE from Northwestern University in Evanston, IL. He is the RF technical marketing manager for Tektronix. He has worked extensively in various test and measurement positions for more than 25 years, including R&D engineer and project manager, product planning, business and market development. He has worked, designed or developed solutions for a wide variety of industries, including radio communication and satellite payload test, radar and direction finding, and EMC and surveillance systems.