Power levels are increasing in RF/microwave transistors, presenting new opportunities for amplifier designers, and new challenges for those who must characterize these devices. Newer wide-bandgap (WBG) semiconductor devices are achieving impressive levels of power density, but generating large amounts of heat in the process. The heat poses problems when making measurements on these high-output transistors, since elevated temperatures can modify device behavior. Fortunately, pulsed DC and load-pull measurement systems from FOCUS Microwaves provide the capabilities to accurately characterize these WBG transistors with controlled, repeatable measurements to 26.5 GHz and beyond.

The improving performance levels of WBG semiconductors are providing amplifier designers with the means of reaching unprecedented output power in solid-state designs. Compared to traditional GaAs MESFET devices, WBG transistors based on silicon carbide (SiC) and gallium-nitride (GaN) substrates are reaching new ground in terms of solid-state power at high frequencies. But the high power densities of these devices, coupled with less-than-optimum efficiency, results in excessive generated heat. Before amplifiers can be designed around these devices, the semiconductors must be fully characterized, but the amounts of heat produced by these devices can render test data worthless at high output levels because of temperature-dependent device behavior. For example, self-heating phenomena can reduce carrier velocity through the device, reducing the effective cutoff frequency and output power. In addition to thermal issues, these new WBG devices suffer from material, such as trapping effects, that can limit performance and must also be identified.

The best way to understand the limitations of newer WBG transistors is through careful device characterization, made possible by measurement systems offered by FOCUS Microwaves and AMCAD Engineering. The two test systems allow characterizing WBG devices using pulsed measurements: pulsed load-pull (PLP) and pulsed current- voltage (PIV) and S-parameters. The PLP system is based on the firm’s patented impedance tuners integrated with commercial test equipment, such as oscilloscopes, power meters, and microwave vector network analyzers (VNAs). The PIV system is based on high power pulse generators, synchronized with (VNAs) working in pulse mode as well.

A PIV test system helps extract quasi- isothermal device current-voltage (I-V) and RF characteristics as well as S-parameters for use in creating new device models or for validating existing models. Enhanced measurement accuracy is possible by performing these measurements under pulsed signal conditions, to minimize thermal effects. In contrast, when characterization is performed in continuous DC mode under large-signal conditions, thermal effects will degrade device performance, especially under high-power (large-signal) conditions. The PIV system can be used for evaluating parasitic device phenomena, such as trapping effects, when developing a device model.

Several guidelines should be followed to ensure accuracy when using the PIV system. The pulse duration should be large enough to achieve steady-state operating conditions. This will depend on the size of the transistor and the measurement setup.

The pulse duration must also be short enough to avoid temperature swings to the DUT during testing, which could lead to invalid data. For RF measurements, the pulse duty cycle should be large enough to avoid strong pulse desensitization. Pulsed S-parameter measurements made with a microwave VNA in narrowband mode will employ triggering and timing based on the pulse duty cycle in order to effectively gate the information acquired during the measurement. Finally, the pulse duty cycle should be sufficiently low to avoid device self-heating effects. By balancing these guidelines, the pulse duration for test signals can be determined within an acceptable range.

PIV systems can be assembled for test-fixture or on-wafer measurements. As an example of the latter, Fig. 1 shows a PIV system that makes pulsed I-V and load pull measurements and can be synchronized with a VNA for S-parameter measurements for nonlinear device characterization and electro thermal modeling. Pulsed RF test signals are synchronized and combined with DC bias by means of wideband bias tees. Two different types of bias tees can be used with this system, supporting bandwidths as wide as 0.5 to 40.0 GHz, one based on LC networks and one based on wideband 3dB couplers. In pulsed mode, these bias tees can feed peak DC power levels to +50 dBm at duty cycles to 3 percent. PIV systems can be assembled as new systems or designed as upgrades to existing load-pull test systems.

For new PIV systems, the pulsed gate voltage can be swept from -40 to +40 V while the Quiescent Point (QP) drain voltage can reach as high as 120 V with pulsed levels to 250 V. Maximum current is 10 A for an associated maximum mean power level of 50 W. The maximum peak power is as high as 2 kW. Depending upon the applied power level, typical drain pulse rise times vary from 50 to 200 ns. The DC measurement capabilities of this system are embodied in the power heads. For 15-b measurement accuracy, the settling time is about 300 ns.

The complement of test instruments in a PIV system depends on a customer’s requirements. For example, for pulsed RF measurements from 10 MHz to 26.5 GHz, the VNA is a model PNA-X from Agilent Technologies equipped with internal RF modulators. To extend pulsed measurements to 40 GHz, an external pulse modulator is used with a PNA series VNA.

Pulsed I-V and S-parameter measurements were made on a commercial 30-W packaged GaN transistor to study the device’s static characteristics. Figure 2 shows plots for three different I-V characteristics, in DC mode (blue), pulsed mode with a cold QP at 0 mA and 0 V drain-source current (Ids0) and voltage (Vds0), respectively (green), and pulsed mode with a hot QP at 200 mA and 50 V drain-source current and voltage, respectively (red). The hot QP corresponds to self-heating of +30°C or a global temperature of +50°C. Measurements were made at room temperature, with a thermal resistance (Rth) of 3°C/W. A 400-ns pulse width test signal was used at a 10-percent duty cycle. The system records transient I-V shapes and S-parameters for each I-V measurement, providing model developers with the data needed to evaluate device performance and extract an electrothermal nonlinear (large-signal) model.

The pulsed load-pull (PLP) system is ideal for further validation of a nonlinear electro thermal device model, as well as for development of impedance-matching networks for pulsed amplifier applications. Pulsed load-pull measurements help determine input and output device impedance conditions for optimum operation under pulsed conditions, by precisely tuning the source and load impedances shown to a device under test (DUT) under different operating conditions. In lieu of a complete device model (Fig. 3), pulsed load-pull measurements can provide useful data for power amplifier designers.

Existing load-pull or harmonic load-pull systems can be upgraded to pulsed I-V or load-pull systems by adding a modular PIV (MPIV) along with a pulsed RF signal source, peak power meter and digital oscilloscope. The MPIV system consists of a master drain pulser module (DPM) and a slave gate pulser module (GPM). The drain pulser uses an external high current/voltage power supply while the gate pulser includes its own high precision voltage supply. The drain pulser module can be controlled from a PC program using an ActiveX library through a TCP/IP interface, simplifying integration into existing test programs.

This modular system is ideal for integration into systems, such as on-wafer measurement stations, where space is limited. The MPIV system is suitable for measuring pulsed I-V curves and for biasing the transistor in synchronism with an RF test pulse. The system can be triggered from a personal computer (PC) or synchronized with external trigger. It features capability to 200 V and 18 A with pulse widths as short as 300 ns and duty cycles as short as 0.15 percent.

Continue to page 2