5G testing: Modulation distortion method speeds power amplifier characterization

Article By : Jessy Cavazos

This article explores different ways to perform 5G power amplifier characterization and provides several measurement examples.

Linearity and power efficiency in power amplifiers are critical to the signal quality and battery life of 5G designs. Traditional methods for characterization testing of power amplifiers (PA), however, face increasing challenges as systems move to millimeter-wave frequencies. The new modulation distortion method for performing characterization test promises simpler, quicker, and more accurate results.

Modern communication systems use the orthogonal frequency division multiplexing (OFDM) waveform for digital signal demodulation. The lack of linearity in an OFDM waveform, however, generates errors in the demodulation process, causing signal-quality issues. Poor power efficiency in components making up a device, such as PA, also reduces battery life. Increasing PA linearity while maintaining a high level of efficiency in a design is challenging, though. This challenge is even greater in the context of 5G due to the move to millimeter-wave (mmWave) frequencies and wide signal bandwidths, both of which complicate the characterization testing designers need to perform in order to optimize their designs.

Common industry practice

The common practice for performing PA characterization consists of two stations. The first station uses a vector network analyzer (VNA) to make basic characterization measurements such as S-parameter, gain compression, intercept point third order (IP3), and sometimes noise figure. The second station features a signal generator and a signal analyzer and is used to generate error vector magnitude (EVM) and adjacent channel power ratio (ACPR) measurements, the figures of merit (FOM) for measuring PA non-linearity (Figure 1). The device is first tested with the VNA and then brought to the other station.

power amplifier characterizationFigure 1 The process traditionally used for PA characterization requires two steps, first with a VNA and then with a signal generator and analyzer.

Because it uses the higher frequencies of the mmWave spectrum, referred to as frequency range 2 (FR2), and the wide bandwidth of OFDM signals, 5G makes EVM measurements for PAs a lot more difficult than in the past. Measuring EVM for a 5G device with the traditional method, for instance, requires you to first modulate the signal with a signal generator that has a specific scheme for 5G new radio (NR) that includes a preamble, pilot, and data. You then need to capture the waveform, demodulate it with the specific scheme, draw the constellation diagram, and measure the error between the ideal constellation and the measured one to determine the EVM.

But residual EVM, meaning the EVM of the test system itself, is very close to the EVM of the device in a 5G FR2 scenario because of the wide carrier frequency. The captured wideband signal includes wideband noise. The signal-to-noise ratio (SNR) degrades as bandwidth increases. Cable loss and high frequency response also contribute to reducing signal quality, and the high SNR makes test automation difficult.

Modulation distortion setup

A new way to perform PA characterization has emerged recently that addresses the shortcomings of the traditional method: modulation distortion. The modulation distortion setup provides all the VNA measurements as well as ACPR and EVM in a single station using a VNA and a signal source (Figure 2).

modulation distortion setupFigure 2 The modulation distortion setup for PA characterization provides all characterization measurements in a single station.

The first step in the modulation distortion setup is to generate a stimulus signal called the compact test signal. The VNA firmware selects a slice of the original waveform that presents the statistical characteristics of that waveform, then removes spectral leakage using a brick-wall filter. Although it only uses a slice of the waveform, the compact test signal’s frequency signature is the same as that of the parent signal. The compact test signal’s complementary cumulative distribution function (CCDF) can differ slightly from that of the parent signal but using a longer test signal reduces the difference between the two, with a slight impact on measurement speed.

After stimulating the device with a compact test signal, you can measure the device’s nonlinearity using frequency domain analysis. By measuring the input and output simultaneously, the measurement is coherent. Also, vector correction helps minimize errors from a mismatch in the measurement system.

When measuring a wideband signal, bandwidth limitations of the VNA digitizer prevent you from measuring the whole band at once. To address this challenge, the VNA measures every 30 MHz of bandwidth and moves its local frequency to capture all the frequency spectrum in the band of interest.

A technique called spectral correlation decomposes the output signal spectrum into the linear and distortion parts, which makes computing FOMs like EVM and ACPR possible. ACPR is computed from the channel power of the in band and adjacent band. EVM is computed from integrating in-band distortion spectrum from the measurement result. Computing EVM from the frequency domain is mathematically the same as computing it from the time domain, which can be explained from Parseval’s theorem.

In a traditional setup, the signal generator and analyzer stimulate the device, capture the signal in the time domain, and plot the constellation to compute the EVM. In contrast, the modulation distortion setup compacts the waveform, repeats the compact test signal, stimulates the device, captures input and output spectrum in the frequency domain, and then decomposes the output spectrum into the linear and distortion parts to compute the EVM.

This setup can be much simpler and make it easier to accurately characterize distortion contribution of PAs, especially in wideband applications. The wide system dynamic range generates low residual EVM and the VNA calibration technique enables high signal fidelity at the device-under-test (DUT) input. Modular distortion delivers consistent measurement results while increasing measurement speed.

Measurement examples with both methods

Let’s have a look at some concrete measurement examples (Figure 3).

power sweep measurement resultsFigure 3 Comparing the traditional (orange) and modulation distortion (blue) methods’ power sweep measurement results for the same DUT shows the latter’s improved accuracy.

In this measurement, we use the traditional method (orange) and the modulation distortion method (blue) to characterize the same DUT. Careful characterization of power out (Pout) [dBm] allows for an apple-to-apple comparison.

While both methods provide the same results for the 100 MHz QPSK waveform, the results differ slightly for the 100 MHz 64 QAM waveform in the high-power region when the PA is highly distorted. This difference is the result of a common error due to symbol skipping that only occurs while performing demodulation with the traditional method.

For dense constellations like 64 QAM, an error larger than the gap of the QAM’s constellation leads to an underestimated EVM with the traditional setup. The 400 MHz QPSK results show the same issue in the high, non-linear region.

In the lower power region, the modulation distortion method also provides better EVM results, the reason for this being the low noise floor of the VNA. The measurement results for the 400 MHz 64QAM waveform are similar.

Figure 4 shows other measurement examples with the modulation distortion setup for a 5G FR2 100 MHz 4CC signal. The setup computes the EVM for each of the carriers and ACPR for the entire band.

5G FR2 4CC signal measurementFigure 4 The 5G FR2 4CC signal measurement using the modulation distortion method shows EVM for each carrier and ACPR for the entire band.

Figure 5 shows a pulsed measurement example with the modulation distortion method. With the pulsed compact test signal as the stimulus, you can synchronize the measurement with the pulsed compact test signal and trigger the SMU to synchronize for bias voltage. Performing pulsed measurements with the modulation distortion setup is easy.

Pulsed measurement for 5G NR 400 MHz signalsFigure 5 Pulsed measurement for 5G NR 400 MHz signals is easy with the modulation distortion method.

5G is a complex technology, but making PA EVM measurements does not have to be. A modulation distortion setup is a much simpler and easier-to-use alternative to the traditional method for performing PA EVM and ACPR measurements, with the added benefit of higher accuracy.

You can get more information on the method and see additional measurement results by viewing the Keysight on-demand webinar Modulation Distortion: How to Make High-Accuracy EVM Measurements for 5G​.

Jessy Cavazos is part of Keysight’s Industry Solutions Marketing team.

Related articles:

Leave a comment