Effective use of digital predistortion (DPD) in power amplifiers for wireless-communications systems requires the right measurement tools.
Power amplifiers for wireless applications are expected to provide exceptional linearity and efficiency in order to handle the complex waveforms used in modern communications systems. Rather than building RF power amplifiers with more pure performance, which adds cost, lowers efficiency and creates reliability issues, designers today can also opt for adding digital processing power with the use of digital-predistortion (DPD) techniques that can help maximize power-amplifier (PA) efficiency, increase reliability, and reduce operating costs.
Digital techniques offer many advantages in cost, power consumption and reliability when compared to analog methods. Because of these advantages, older narrow band, single-carrier, triple conversion systems are being replaced with wide band, multi-carrier transmitters enabled by digital signal processing (DSP) and DACs that produce direct IF, or even direct RF outputs to the RF amplifier.
Wireless systems are providing users with a host of services and benefits. Unfortunately, the benefits of advanced wireless technologies often come at the expense of increased power consumption and cost of operations. Modern cellular and wireless technologies—especially digital RF communications networks—transmit and receive more data, more video, and more voice than ever before. New standards such as HSDPA, HSUPA, 1xEVDO, WiMAX, and Long-Term Evolution (LTE) necessitate greater power usage, create more and larger RF waveform peaks, and allow for larger data bursts. Consequently, modern wireless devices are producing RF signals with unprecedented peak-to-average ratios (PARs) and the potential for distortion within an already crowded RF spectrum.
With power consumption and modern PARs at all-time highs, power amplifiers are being strained like never before—and causing transients and cost inefficiencies as a result. Bigger amplifiers that can accommodate more power inflate capital expenses in the short-term and operating expenses in the long-term. Larger, more costly batteries are required for the same amount of backup capability. And greater power consumption and production intensifies thermal and electrical conditions, which can create reliability problems.
When working with power amplifiers that support advanced wireless technologies, designers and network operators can choose one of two paths: They can add more brawn, or they can add more brains. Whereas the former effectively adds to the aforementioned cost and reliability concerns, the latter is prompting new strategies for digitally predistorting waveforms prior to power amplification for maximum efficiency and tight spectrum control. With the right test instruments, digital-predistortion (DPD) techniques can be honed to allow for smaller, more efficient power amplifiers—reducing development and operating costs while improving network and device reliability.
Whether it is a high-power satellite ground station, a multi-carrier cellular base station or even a low-power mobile system, modern transmitters employ a variety of predistortion techniques to reduce out-of-channel interference and optimize operating efficiency. One of the most popular and effective distortion reduction methods is Adaptive DPD.
This approach uses a sample of the transmitter's output to calculate error vectors and generate correction coefficients, which are then used to predistort the incoming signal. To reduce analog circuitry distortions, the signal in the chain is kept in digital format for as long as possible.
Figure 1 shows how a portion of an amplifier's output signal is tapped, then down-converted and digitized. The digitized signal is used to feed the DSP circuitry, which performs analysis of the non-linearities present in the signal and creates non-linear correction coefficients. These non-linear coefficients are used to alter the incoming in-phase (I) and quadrature (Q) signals in the transmit chain. The signal, now predistorted and with PAR reduction applied, is fed to the amplifier after being converted back to the analog realm by the DAC, as seen in the transmit chain. The resulting output signal exhibits reduced spectral distortion and improved adjacent-channel-leakage-ratio (ACLR) performance than the signal without predistortion techniques.
Digitally predistorted amplifiers offer improved spectral efficiency with much higher power-added efficiency (PAE) than previous feed-forward architectures, dramatically reducing heat concerns, improving reliability, and lowering operating costs. This approach has branched beyond cellular base stations, and we are now seeing feedback linearization for cellular handsets, satellites and even adaptive phased-array radars.
This scenario creates a wide variety of troubleshooting challenges not seen in traditional analog systems, however. Digital artifacts may be introduced into the transmit chain by the ADC and DAC, or by any DSP performed on the signal prior to analog conversion in the transmit path. These artifacts are frequently transient in nature and are difficult or impossible to capture using conventional spectrum analyzers. They may only occur rarely and can cause frequency-domain effects in the adjacent and alternate channels. Effective troubleshooting of transient frequency-domain signals requires not only the detection of the problem, but also the ability to trigger on it and capture a record for analysis.
Characterizing these systems presents new challenges as well. In the development stage, a variety of predistortion and PAR reduction methods may be tested and optimized prior to the availability of the entire transmit chain. The signals in the feedback path must often be captured using test equipment, and calculation of the new non-linear distortion coefficients is performed in offline software prior to the availability of completed hardware (ASICs or FPGAs). Correction algorithms using these coefficients are then applied to the initial I and Q signals and the result is loaded into arbitrary waveform generators (AWGs) to test their performance.
The rate of signal and power changes is also problematic. Since many wireless signals employ a burst format (such as 1xEV, HSxPA, and WiMAX signals), use pulsed waveforms (such as radar, RFID/NFC, and Zigbee signals), or rely on adaptive techniques (with changes in coding or modulation rate), the RF power level changes quickly. Often, these changes occur faster than the feedback loop can respond. Unlike previous linearization architectures, such as feed-forward amplifiers, the amplifier is blind to fast temporal changes while the feedback loop is sensing and adapting to these changes. This can lead to unintended signal performance that can be damaging to network reliability and operation.
Implementing, testing, and streamlining DPD is not new. Traditional swept-frequency spectrum analyzers and vector signal analyzers (VSAs) enable a certain level of DPD testing—even enough to pass most standards requirements. But transients and other unseen effects inevitably persist because these legacy instruments are only capable of showing sweeps or snapshots of the RF spectrum.
To truly see the entire RF environment, evaluate how digitally pre-distorted waveforms are operating, and identify what transients are leaking into the spectrum, designers must use an advanced real-time spectrum analyzer (RTSA). This unique and flexible test instrument has quickly become and essential tool for effective DPD troubleshooting and characterization.
The RTSA is a multidomain instrument capable of displaying time, frequency, and modulation measurements. This allows the RTSA to replace several traditional instruments with a single unit. Since each measurement domain is derived from the same seamless time record, measurements are precisely time-correlated. The user can place a marker on a spectral or modulation anomaly and correlate it with the exact signal that produced it. Time-correlated displays greatly enhance diagnostic insight and characterization accuracy by providing a critical causality component. Traditionally, such precise time-correlated displays were difficult to obtain, requiring synchronization of multiple instruments.
Unlike swept-frequency spectrum analyzers and VSAs, an RTSA can process time-domain data into the frequency domain in real-time. This allows the RTSA to continuously analyze its input spectrum by taking real-time Digital Fourier Transforms (DFTs) of the input signal prior to triggering and capturing data. Conversely, other analyzers randomly capture the data or trigger on time-domain levels and then analyze what has been captured offline, leaving large portions of the signal unanalyzed between sweeps or captures. The RTSA's ability to continuously analyze the signal in the frequency domain and trigger only on events of interest is ideal for many RF applications.
To simulate RF conditions when testing DPD and power-amplifier performance, AWGs are often used in conjunction with RTSAs. Modern AWGs can approximate real-world RF signals and simulate modern signal conditions. They enable users to create and directly insert complex, modulated I/Q and RF signals into wireless systems and networks, which is invaluable for power-amplifier linearity testing.
Figure 2 shows a DPD development system. An AWG is used in place of the I and Q signals and the DAC, and the RTSA is used in placed of the correction loop downconverter and the ADC. The I and Q vectors from the RTSA are then sent to an offline processor where DPD and PAR reduction techniques are applied.
The AWG in this system must have sufficient resolution, bandwidth, and memory depth to replace the digital system used in the transmit chain. The RTSA must have sufficient frequency range, capture bandwidth, capture fidelity, and memory depth for the application. Moreover, the capture bandwidth must be a minimum of three times the transmit bandwidth to assure that third-order distortion products are digitized. In fact, many systems now use fifth-order distortion products in the calculation of the pre-distortion coefficients, requiring a capture bandwidth at least five times that of the transmitted signal.
For example, 3GPP WCDMA multicarrier-power-amplifier (MCPA) applications using a four-carrier test configuration produce distortion products that require significant measurement bandwidth. Viewing the fifth-order products of the 20-MHz four-carrier bandwidth requires 100 MHz of acquisition bandwidth. Fortunately, RTSAs are available with 110 MHz of capture bandwidth and dynamic range that is sufficient to acquire the intermodulation products for many modern systems.
The signals captured during DPD development may contain very long sequences of specialized data, which are intended to exercise the limits of the amplifier by creating the worst-case operating scenario. These sequences may be one second or more in length, depending upon the design requirements. The best RTSAs have the ability to capture up to 1.7 s of I and Q data at their maximum capture bandwidth of 110 MHz. Longer capture times are possible at reduced capture bandwidth. Capturing long record lengths allows the user to examine the performance of devices in response to real-world signals. The ability to capture many packets of data is very useful, especially as it relates to changes in PAR, including changes in modulation type, the number of active code channels, and adaptive power levels.
In addition to exceptional capture bandwidth and deep memory, the best modern RTSAs feature unique capabilities, such as digital phosphor displays and frequency mask triggers (FMTs). Digital phosphor technology, which has traditionally been used in high-speed digital storage oscilloscopes, has recently been applied to the RF domain for unmatched insight into RF signal behavior. The technology enables RTSA users to view "live RF" signals for the first time.
Using a parallel processing architecture, digital phosphor live-signal display technology produces nearly three orders of magnitude improvement in the spectrum capture rate compared with swept spectrum analyzers and VSAs. RTSAs with digital phosphor technology deliver a remarkable 48,000 spectrums per second. By continuously converting time-domain signals into the frequency domain at that rate, digital-phosphor technology provides a means of displaying both frequent and infrequent events, distilling real-time FFT frame rates far above what is perceivable by the human eye into an intuitive, full-motion display. With variable color-graded persistence that displays transients at rates beyond the human eye's response time, RTSAs with digital phosphor technology can reveal elusive glitches, anomalies, and other transient events. The frequency of occurrence is color-graded to accurately represent the rate of occurrence that energy is found at each display pixel. This gives the digital-phosphor display a transparent-like quality that enables viewing of spectral information below the peak amplitude of the spectral envelope as shown in Fig. 3.
Figure 4 shows a DPD power amplifier using both traditional spectrum analyzer displays as well as advanced digital phosphor displays. Prior to correction, "shoulder" transients can be seen in both displays. However, after correction, the digital phosphor display reveals fast transients that go undetected by the spectrum analyzer. The level of insight delivered by digital-phosphor technology allows for faster, more thorough testing and correcting of DPD power amplifiers and leads to more reliable devices and networks.
RTSAs also feature a unique FMT, which allows a user to trigger an acquisition based on specific events in the frequency domain. The flexible FMT is a powerful tool for detecting and analyzing dynamic and digitally pre-distorted RF signals, capturing low-level transient events that occur in the presence of more powerful RF signals and detecting intermittent signals and specific frequencies within a crowded spectrum. Complex frequency masks can be stretched around signals of no interest with a few mouse clicks, and the mask can be set just above the noise floor to avoid false triggers. Once the frequency mask is set up, any spectral event of interest that falls outside the mask will trigger a signal capture. The FMT is essential for finding short duration or time varying signals while troubleshooting RF circuits. It can detect sporadic signals, the presence of transient intermodulation products and spectrum containment violations.
Due to its wide capture bandwidth, deep memory and inherently correlated measurements, the RTSA is an ideal tool for the analysis and troubleshooting of DPD techniques and RF power amplifiers. Today's leading RTSAs allow spectrum and vector measurements to be performed over bandwidths to 110 MHz with high dynamic range and low residual EVM. In addition, measurement correlation across multiple domains, Digital Phosphor displays and FMTs greatly improve troubleshooting efforts. With these capabilities, the acquisition, measurement, and characterization of digitally modulated and pre-distorted RF signals are quick, efficient, and accurate. Figure 5.