While some 5G products – radios, power amplifiers, and phased arrays – have begun to appear and many more are in development and prototyping, research continues at universities around the world. After all, once you get a technology to work, you then must optimize it for performance, cost, and manufacturability. A technical session held at the 2018 International Microwave Symposium on Wednesday, June 13 called "5G Sub-Systems: From Predistortion to Complete Link" highlighted research in phased array antennas, MIMO, and mmWave technologies.

The session opened with "Single-Input Single-Output Digital Predistortion of Power Amplifier Arrays in Millimeter Wave RF Beamforming Transmitters" by Eric Ng, Yehia Beltagy, Patrick Mitran, Slim Boumaiza from the University of Waterloo (Canada), who modeled and tested digital predistortion (DPD) in a power amplifier (PA) at frequencies below 6 GHz. The team's intent was to calculate a DPD model and apply it to an array of four single-input, single-output (SISO) PAs, which should linearize the PAs and boost efficiency. The model also accounted for antenna crosstalk and impedance mismatches. Figure 1 shows the thought process.

Figure 1 A team of student engineers at the University of Waterloo have developed a process for simulating nonlinearities in a 64-antenna array and its power amplifier and applied a digital predistortion feedback loop to linearize the PA.

The team applied the DPD model to a 64-antenna array manufactured by Anokiwave operating at 28 GHz driven by a 320 MHz digitally modulated signal. The drive signal for the PA was generated by a Keysight Technologies AWG. Figure 2 shows the test setup.

Figure 2 The test setup consists of a Keysight Technologies arbitrary waveform generator, an upconverter, and an Anokiwave radio head that includes an RF beamformer, PAs, and a 64-element antenna array.

Test were performed by rotating the radio head to angles at 0°, 20°, and 40°. Using beam steering, the researchers kept the beam pointed at a receiving antenna, which was connected to a Keysight PXA signal analyzer (Figure 3).

Figure 3 Test results show improved normalized mean-square error (NSME) in the left plot and adjacent channel power ratio (ACPR) with and without DPD versus motor angle.

Reducing feedback loops on transmitters
Continuing with the DPD theme, Ampleon's Andre Prata described a technique to improve efficiency, integration, and cost for a sub-6 GHz massive MIMO (m-MIMO) transmitter. The block diagram from the slide in Figure 4 shows a dedicated analog front-end module (AEFM) for each antenna in a m-MIMO system. Each antenna also uses DPD to compensate for nonlinearities in the PA.

Figure 4 Using a dedicated AEFM per m-MIMO antenna can be costly and inefficient.

Prata then described a method of using a series of multi-tone sine waves to approximate the signals of each transmitter, where the tones representing the signals from each transmitter are separated in such a way that they don’t overlap in frequency, maintaining the independence of each transmitter's signal. In their experiment, the team of engineers used two transmitters to demonstrate the feasibility of the concept. An FFT of the transmitter signals allows for placement of the multi-tone frequencies. An inverse FFT then creates a time-domain digital representation of each signal path, which allows for calculation of the DPD coefficients. The technique should enable a reduction in the number of feedback loops on a set of transmitters.

Figure 5 shows a slide containing a diagram of the validation setup. In the diagram, you can see a single combiner and RF ADC in a feedback loop for two PAs.

Figure 5 The proposed optimization technique for reducing feedback components in an m-MIMO transmitters shows a single feedback path for two transmit paths.

The research team used 50 tones to mimic a 5 MHz TLTE signal, with 100 kHz between tones. In their paper "Optimized DPD Feedback Loop for m-MIMO sub-6GHz Systems," the engineers reported an ACPR improvement of 15 dB using this method to generate DPD coefficients versus no DPD. When mimicking a 20 MHz LTE signal (200 tones), the engineers reported an ACPR improvement of 10 dB (Figure 6).

ACPR improvement
Figure 6 Using 200 tones to mimic a 20-MHz LTE signal resulted in and ACPR improvement of 11 dB.

No calibration at 28 GHz
Multiple input/multiple output (MIMO) antennas and beam steering continue to be a subject of much research. Although some commercial products have appeared, there's still plenty of room for development, as there always is early in a technology's lifecycle. A team at the University of California San Diego headed by Prof. Gabriel Rebeiz continues development of phased-array transceivers operating at 28 GHz. The team presented a paper called "A Scalable 64-Element 28 GHz Phased-Array Transceiver with 50 dBm EIRP and 8–12 Gbps 5G Link at 300 Meters without any Calibration". Thus far, the team as demonstrated data rates of 8-12 GHz at distances of 300 m with a 64-element array. What's most interesting about the design is that it doesn't require calibration.

The design consists of sixteen 2×2 antenna arrays mounted on a 12-layer PCB (Figure 7). The diagram on the left of the slide shows the PCB stack.

UCSD pahsed array 28 GHz
Figure 7 A 64-element phased array developed at UCSD uses sixteen 2×2 antenna arrays and requires no calibration.

Figure 8 shows the integrated array along with measurements on gain and phase for each element.

Figure 8 Measurements on the UCSD 64-element phased array show a gain span of ±2 dB with normalized phase of ±20°.

Figure 9 shows the array as demonstrated in December 2017 at the Keysight 5G Tech Connect Summit.

UCSD 64-element phased array
Figure 9 This 64-element phased-array developed at UCSD showed results showed high performance and repeatability, even without calibration. Image courtesy of Keysight Technologies.

For testing the array, the UCSD team set up a 300 m link and sent data over 16QAM and 64QAM signals generated with a Keysight M8195A waveform generator. The modulated signal was then upconverted to 28 GHz. The array is capable of steering the beam in azimuth to ±50° and elevate the beam to ±25°. Figure 10 shows the equivalent isotropically radiated power (EIRP) measurements taken with a Keysight DSO-S804A real-time oscilloscope running VSA 89600 software.

Figure 10 A 300 m link at 28 GHz shows promise for 5G mmWave communications over a wide scan angle.

These and several other papers presented at IMS 2018 show that research into 5G technology goes on. Which of these technologies – developed in university and industry labs – will find their way into commercial products is yet to be seen.

Martin Rowe covers test and measurement for EDN and EE Times. Contact him at martin.rowe@AspenCore.com

Related articles: