Amplifier reuse is a common way for mobile communications designers to reduce manufacturing material costs to a minimum as they develop new networks. For example, the same power amplifier (PA) developed for GSM and GPRS is also be used for EGPRS applications. Unlike GSM and GPRS, however, EGPRS is sensitive to PA nonlinearities, which can significantly and negatively affect the performance of an EGPRS handset. To use a GSM PA for EGPRS, developers must add calibration factors to compensate for the amplifier's nonlinearities. These calibration factors are based on information about phase and amplitude with respect to time, and they are required in manufacturing test to ensure that the EGPRS phone transmits properly.

Mobile manufacturers can introduce the necessary phase and amplitude versus time (PAvT) measurement into manufacturing test through either of two calibration methods: external calibration using test equipment or internal self-calibration using an additional component on the amplifier's printed-circuit board (PCB). The method a manufacturer chooses involves weighing the initial cost of purchasing a test solution (or upgrading existing equipment) versus the ongoing material costs and PCB space requirements associated with the additional internal component. For manufacturers who place a premium on board space and already use test equipment on the production line, external calibration offers an efficient and cost-effective solution.

To make the PAvT measurement externally, manufacturers need a test set with speed, accuracy, and—perhaps most important—flexibility to optimize PA designs as their handsets evolve. They also need the handset chip-set supplier's proprietary software, which is provided under the terms of a licensing agreement with the chip-set supplier. The proprietary software is used to generate the appropriate test signal from the mobile telephone and calculate the resulting calibration factors for the PA.

The European Telecommunication Standards Institute (ETSI) adopted an eight-state phase-shift-keying (8PSK) modulation format for EDGE-based systems that enables high data rate communication. The specifications for modulation accuracy in EDGE present challenges in the design of vector PAs. Such amplifiers provide variable shift and gain; however, their internal parametric nonlinearities can produce significant error-vector magnitude (EVM) in the output signal.

To understand the cause of the error, consider an ideal 8PSK signal (Fig. 1). The eight magnitude/phase states of the output signal occur at precisely defined locations on a constellation diagram. These locations lie along a circular contour and are evenly spaced at π/4 intervals (Fig. 2). Furthermore, each magnitude/phase state occurs at a defined offset from the constellation's defined phase origin. In an EDGE system, these locations would occur typically at integer multiples of 3π/8 intervals from the constellation's coordinate reference angles.

Now consider the effects of passing this ideal signal through the PA stage with its inherent nonlinear behavior. A phase shift is introduced by the amplifier at varying power levels. This is the power gain. Changes in the amplifier's power gain may prevent the desired output power from being achieved. These effects are apparent in an 8PSK constellation whose phase trajectory is not circular and whose phase states are not located precisely at the expected points—a phenomenon known as polar distortion-(Fig. 3).

The PAvT measurement provides mobile manufacturers with a means of characterizing the magnitude and phase variations introduced by the nonlinear behavior of power amplifiers operating over a wide dynamic range. Variations in power magnitude, phase and frequency are measured as a function of the transmitting device's varying power output levels. From the measurement results, manufacturers can derive tables of control data, which are used to predistort the magnitude and phase-control signals that modulate the final power-amplification stage. This technique gives a more accurate phase trajectory of the modulated output signal.

Characterizing the magnitude, phase, and frequency variations of an RF transmitter as a function of the output level requires only a CW signal, whose amplitude can be controlled with simple pulse amplitude modulation. One advantage of using a CW signal is that as the capabilities of devices evolve, mobile manufacturers can characterize the device hardware using the same technique, without having to worry about providing unique test signals for each design iteration. The first step is to generate a customized test signal that exhibits the desired power versus time profile. The next step is to measure, at prescribed time intervals, the power, relative phase, and relative frequency of the output signal of the device. These measurements can then be used to create control data correction tables in whatever proprietary format suits the design of the device under test.

A class of test signal suitable for this calibration procedure has a power vs. time profile consisting of a number of discrete power steps (Fig. 4). The profile shows a piecewise monotonic power versus time pattern, which can be customized by the mobile manufacturer to suit specific requirements. Note that monotonicity is not necessarily a requirement.

The design of this profile includes elements to control what portions of the waveform are measured. These elements include a defined physical event to trigger the acquisition of waveform samples by the measurement system, a power-level reference, and user-specified locations and time intervals for each settled, stable, and measurable portion of each power step.