Picosecond Pseudomorphic AlGaAs/InGaAs MODFET Large-Signal Switching Measured by Electro-Optic Sampling

Abstract

We present a study of the large-signal switching characteristics on a picosecond time scale of an AlGaAs/InGaAs modulation-doped field-effect transistor (MODFET) using electrooptic sampling. The MODFET has been of great interest for high-speed digital and analog applications since its demonstration. Recent progress in the fabrication of such devices has resulted in devices with extrinsic cutoff frequencies exceeding 250GHz, corresponding to response times on the order of a picosecond. The rapid increase in the speed of MODFET devices has outstripped the ability to measure their performance with all-electronic equipment. Devices are most often characterized at lower frequencies and then the results are extrapolated to much higher frequency. These difficulties can be avoided using very high bandwidth optoelectronic techniques based on ultrafast lasers, where it is routine to generate and detect electrical signals with subpicosecond rise and fall times. These optoelectronic techniques have been applied to small-signal characterization of passive coplanar striplines to 1THz and MOD-FET devices to 100GHz. In addition to the very high bandwidth of optoelectronic methods, the ability to generate large electrical signal amplitudes allows the study of switching of active devices. Large-signal characterization of active devices is important because of the wide range of high-speed large-signal device applications, including analog sampling circuits, oscillators, and all digital circuits. It is also important in understanding device operation since large-signal operation on very short time scales involves rapid changes of carrier distributions throughout the device. As the time scale of the signals at the device input and output approaches that required for re-equilibration of carrier distributions, it will become necessary to include these non quasi-static effects in nonlinear device models. Large-signal measurements on a very short time scale will allow the development and verification of nonlinear circuit models, and their comparison with existing models based on bias-dependent small-signal measurements.