Digital pre-distortion provides substantial improvements to the performance of modern cellular and Wi-Fi power amplifiers.
We rely heavily on mobile devices for a wide range of our needs, from serving as our primary method of communication to being a flashlight or alarm clock. The last thing any user wants is a dead battery. To extend battery life and improve efficiency, designers are focused on the power efficiency of the radio, particularly for modern communication standards such as LTE and 802.11ac. The transmitter’s final-stage power amplifier (PA) often consumes the most power. This, combined with the increasing complexity of modern communications signals, requires engineers to make difficult tradeoffs between efficiency and linearity, and has led to the introduction of a wide range of new PA design and test techniques.
One of the most fundamental evolutions of the handset PA is the use of digital pre-distortion (DPD). Although DPD does not produce dramatic improvements in PA efficiency, it does improve the quality of a signal that a PA will produce when operating at its point of peak efficiency. Given the widespread adoption of DPD techniques in the radios of modern mobile devices, it is worth explaining both the theory and challenges of testing DPD-enabled PA’s.
Fundamentals of PA Efficiency
To best understand how DPD can be used to improve PA performance, it’s important to review the theoretical behavior of a PA, as illustrated in Figure 1.
The PA achieves its point of peak efficiency at an output level that approaches thesaturation region of the amplifier. In an ideal PA, every 1 dB increase in input power would result in a 1 dB increase in output power, yielding the 1:1 relationship described by the “ideal pin vs. pout” trace shown in Figure 1. However, as the PA approaches its saturation region (where efficiency is highest), the output power is no longer a linear function of input power. In fact, the resulting nonlinear behavior of the PA significantly distorts the output signal.
At first glance, nonlinearity might not seem particularly troubling. After all, the difference between 19 dB of gain and 20 dB of gain seems like a small difference. However, when amplifying modulated signals, the ability to produce a constant gain over a wide range of power levels is extremely important. For example, consider a modulated signal such as a 64-QAM waveform. In this modulation scheme the power of a modulated transmission will vary, often substantially, from one symbol to the next.
As we observe the constellation plots shown in Figure 2, nonlinear gain results in a constellation where the symbols closer to the origin experience a larger gain than those farther away. The resulting distortion degrades the modulation quality of a transmission (typically measured as an error vector magnitude metric), and can increase the resulting bit error rate (BER) of the receiver intended to demodulate the transmission. Even worse, distortion introduces spectral regrowth in the frequency domain, which results in unwanted emissions in bands adjacent to the modulated signal.
As one might guess from Figure 2, the more complicated the constellation plot, the more nonlinear distortion affects the modulation quality. In fact, a general assumption is that signals with high peak to average power (PAPR) characteristics require higher linearity requirements than those with a lower PAPR.
Historically, communications systems, such as global systems for mobile communications (GSM), used specialized modulation schemes, such as gaussian minimum shift keying (GMSK) to minimize the PAPR of the signal and allow the use of nonlinear PA’s. However, the demand for higher data rates (and hence more bits per symbol) has pushed the PAPR characteristics of modern communications systems even higher. For example, a typical UMTS (3G) transmission using quadrature phase shift keying has a PAPR of approximately 5.5 dB. By contrast, signals that employ the orthogonal frequency division multiplexing multi-carrier techniques, such as the LTE downlink and 802.11ac, experience a PAPR typically ranging from 8 to 13 dB.
Digital Pre-distortion
Although the effects of distortion on modulation quality can be severe, the spectral regrowth introduced by nonlinear distortion is often more problematic. In fact, when driving a PA deep into compression, the resulting spectral regrowth in bands adjacent to the modulated signal can potentially interfere with wireless devices on those channels. These challenges and the desire to operate a PA with maximum efficiency create the motivation to improve PA performance in its saturation region through DPD.
Engineers historically reserved most DPD implementations for extremely high power base station PA’s that were required to amplify multi-carrier downlink signals with a high degree of linearity. However, the demands of modern communications signals, such as LTE and 802.11ac, combined with the improved signal processing capabilities of the mobile device, have made DPD commonplace in a large number devices.
Testing DPD-Enabled PAs
Today, DPD provides substantial improvements to the performance of modern cellular and Wi-Fi PA’s. However, it also introduces significant test challenges. For example, when utilizing more sophisticated DPD techniques, such as a memory polynomial model, engineers are increasingly required to perform more advanced stimulus-response measurements that enable them to correlate modulated input and output signals. In general, these stimulus-response measurements require tight synchronization between RF signal generation and acquisition, as well as extremely wide bandwidth.
For a sophisticated DPD model, it is often useful to generate and acquire 3x to 5x the bandwidth of the modulated signal. So, it’s essential to use instruments capable of meeting both synchronization and bandwidth challenges.
DPD is a great example of the constant innovation we see from engineers in the wireless industry. This technology, combined with techniques such as load pull and envelope tracking serves.
Filed Under: Aerospace + defense, M2M (machine to machine)