Envelope tracking (ET) brings many advantages to power-amplifier (PA) design, ranging from higher efficiency to increased output power, improved operation into mismatched loads, and insensitivity to temperature variations.

But while the criteria and metrics for designing traditional fixed-supply PAs are well defined and known, ET PAs have added complexity requiring more sophisticated characterization techniques.

ET Essentials

ET is used to improve the efficiency of PAs carrying signals with high peak to average power (PAPR). High-PAPR signals such as Long-Term Evolution (LTE) transmissions are needed to achieve high data throughput within limited spectrum resources.

While traditional fixed-supply PAs are highly inefficient under these conditions, efficiency is vastly improved by varying the PA supply voltage in sync with the envelope of the RF signal. This is envelope tracking.

When the PA is operating in ET mode, though, its fundamental output characteristics (power, efficiency, gain, phase) depend on the complex interplay of two “control” inputs: RF input power and supply voltage.

The first step in such a design is to build a simple “quasi static” behavioural model of the PA: one that ignores memory effects. The model can be constructed from AM/AM and AM/PM characteristics.

Along with other key PA metrics such as power and efficiency, the mapping between the instantaneous RF envelope and applied supply voltage profoundly influences these characteristics. In an ET system, the contents of a “shaping table” in the envelope path determine this mapping (Fig. 1).

The PA’s fundamental characteristics (output power, efficiency, gain, and phase) over the full range of supply voltage and input power must be measured to define the shaping table. These values could be measured using a continuous-wave network analyzer and a variable dc supply.

However, this technique generally provides unsatisfactory results for ET PAs. Thermal effects, ranging errors, and drift in phase measurements make it impossible to capture accurate enough data. It also is far too slow to allow load-pull techniques to be used.

Instead of continuous-wave techniques, a pulse characterization solution could be implemented using Automatic Test Equipment (ATE)-controlled standard test equipment. Then, there wouldn’t be any need for a high-bandwidth, low-impedance supply. This technique also is sufficiently fast to make load pull viable. However, it’s difficult to accurately measure phase in the pulse testing approach.

This leaves no alternative but to use real waveforms, varying the shaping table to allow all combinations of input power and supply voltage to be measured. This introduces the cost and complexity of a power supply modulator, but it’s very fast and enables accurate phase information to be gathered. It also can be used to characterize memory effects.

A practical setup emulates transceiver signals by generating synchronised RF and envelope waveforms. An ET modulator is used to supply the PA under test and a high-precision current probe is used to capture instantaneous voltage and current, while the PA supply voltage is being dynamically modulated using real ET waveforms.

Fast RF power meters are connected to the input and output of the PA, allowing the system to compute instantaneous PA efficiency (Fig. 2).

“Basic” ET PA characterization can be used to create a quasi-static (i.e., memory-less) data model of the PA with output power, phase, and efficiency as outputs and input power and supply voltage as inputs. Application software running in MATLAB allows the shaping table to be defined. The model then can be used to predict PA system performance parameters such as adjacent channel power ratio (ACPR), error vector magnitude (EVM), and instantaneous efficiency for standard test waveforms.

In addition to being used for PA device level characterization, the same hardware can be used to directly verify PA system performance using the defined shaping table (Fig. 3). Such measurements also can be used to identify sources of nonlinearity.

For example, with higher-bandwidth waveforms, PA memory effects can become significant. This is because PA output parameters such as efficiency, AM/AM, and AM/PM distortion now depend on signal history, so time becomes a factor as well as instantaneous input power and supply voltage.

Memory effects show up in the PA characterization as a “broadening” of the AM/AM and AM/PM characteristics and can result from electrical time constants in input or output bias circuits and physical interconnects, thermal time constants associated with local die heating, or technology-specific “charge storage” effects.

Efficiency Optimization

To fully optimize the efficiency of an ET PA, the device characterization can be extended to include sweeping the load impedance (fundamental or harmonic load-pull) in addition to input power and supply voltage.

Analysis of the large dataset produced by such a characterization can be automated (e.g., using MATLAB) to predict the average PA efficiency when operating with a specific set of ET parameters.

For example, using this characterization methodology it’s possible to predict how a PA’s average efficiency varies with shaping function and output voltage swing range back off from maximum power and waveform statistics when operated in ET mode (Fig. 4).

Self-Linearization With ET

ET PA performance over temperature is commonly expected to vary more than with fixed-supply counterparts. Characterisation using a modulated power supply shows that the reverse is true: unlike a fixed-supply PA, an ET PA’s performance is much less sensitive to changes in the gain of the RF chain driving the PA than it is to changes in the supply voltage. As the characteristics of the supply voltage can be much better controlled over temperature than RF gain, little variation in PA linearity is observed for extreme temperature variations (Fig. 5).

Another counterintuitive characteristic of ET PAs is apparent when the test bench emulates a real-world handset environment, where the load impedance presented to the PA is poorly controlled owing to reflections from nearby objects. This can result in the PA having to work into load mismatches as high as 3:1 voltage standing-wave ratio (VSWR).

The ET PA’s “self linearization” principle observed with temperature variations can also apply under high VSWR conditions, resulting in significantly improved EVM and ACPR performance compared with the same PA operated in fixed supply mode (Fig. 6).

In ways like this, the use of a test environment that sweeps supply voltage as well as input power can reliably predict the system performance of ET PAs. Collecting substantially more data than fixed-supply measurements demonstrates that, in contrast with fixed-supply PAs, the performance of an ET PA is not “self contained.”

Using an appropriate “system characterization” bench not only allows designers to optimise the system efficiency benefits of operating PAs in ET mode, but also quantify other useful system benefits, such as increased output power, improved operation into mismatched loads, and insensitivity to temperature variations.