Application of 5G Ultra-Wideband Millimeter Wave Test Bench

While fourth-generation (4G) cellular systems—LTE and LTE-Advanced—are still being widely deployed, research and development work on fifth-generation or 5G systems is already underway. Key attributes of 5G systems are likely to include highly integrated, dense networks consisting of small cells that support data rates on the order of 10Gbps with round-trip latency of 1 ms or less. In most studies, researchers assume that multiple air interfaces operate at microwave or millimeter-wave frequencies, known as new radio (NR). There are three basic use cases for 5G/NR: enhanced mobile broadband (eMBB), massive machine-type communications (mMTC), and ultra-reliable low-latency communications (URLLC). However, in 3GPP's TR 22.891, 74 more basic use cases are defined, many of which are directly or indirectly derived from next-generation mobile networks (NGMNs).

To realize this vision, evolution and change are required at the technology, business model and policy levels. For example, in terms of policy, the Federal Communications Commission (FCC) recently announced a series of new rules to quickly develop and deploy next-generation 5G technologies and services. These rules create new microwave flexible application services in the 28 GHz, 37 GHz and 39 GHz bands and create a new public band at 64-71 GHz. Together with the existing 57-64 GHz public band, this new band expands the range of public spectrum to 14 GHz (57-71 GHz).

The expansion of public spectrum creates new opportunities for 5G applications that require high data throughput using wideband digital modulation; however, to benefit from the high frequency bandwidth within this public spectrum (such as 5 GHz occupied bandwidth), new technologies are needed to achieve the highest system performance.

Testing these new designs will also require advances in measurement technology. For example, today’s vector signal generators offer modulation bandwidths up to 2 GHz. We will need to generate signals differently to meet the wideband requirements allowed by new spectrum allocations.

Digital pre-distortion (DPD) is a technique introduced in 3G and 4G that modifies the input signal to allow the power amplifier (PA) to operate at peak signal levels while performing AM/AM and AM/PM conversions. By measuring the AM/AM and AM/PM conversions of the amplifier, the inverse of these functions can be applied to the input waveform to produce the ideal waveform at the amplifier output.

In contrast, AM/AM and AM/PM plots are only a preliminary analysis of the characteristics of a power amplifier, and memory effects are an important consideration. Common circuit models used to design and simulate power amplifiers cannot predict memory effects. The only practical approach is to test the power amplifier and then capture the modulated signal in the time domain after it passes through the power amplifier.

Established techniques typically require generating and measuring input signals at three to five times the signal bandwidth. Existing equipment capable of testing 4G power amplifiers can easily handle such sampling rates, even for the widest 20 MHz LTE signals. For 5G and 802.11ad signals with bandwidths up to 2 GHz, most current vector signal generators and vector signal analyzers based on 4G technology are powerless.

This white paper will discuss a new broadband mmWave test platform that uses digital technology and compact mmWave converters to generate and analyze very high bandwidth mmWave test signals (> 2 GHz to 8 GHz). While this approach provides sufficient capabilities for some applications that require large bandwidth, it still cannot completely replace traditional vector signal generators and vector signal analyzers.

The test bench will be used to generate and analyze V-Band (50-75 GHz) and E-Band (60-90 GHz) test signals with up to 8 GHz of occupied bandwidth. The test bench solution software will then be used to perform digital pre-distortion (DPD) on a Skyworks power amplifier operating in V-Band. Using the test bench’s simulated pre-distortion algorithm, improvements in adjacent channel power (ACP) and error vector magnitude (EVM) will be demonstrated. In this DPD application, up to 7.5 GHz of signal generation and analysis bandwidth is utilized. Specifically, this is 5 times the QPSK and 64QAM waveforms used for PA DPD, which have a symbol rate of only 1.5 GHz.

New test bench approach for ultra-wideband signals in V- and E-band

The traditional method of generating high-frequency broadband signals is to first create analog I and Q waveforms, modulate them onto an intermediate frequency (IF) or radio frequency (RF) carrier, and then use an external upconverter to convert the frequency to millimeter waves.

When analyzing signals, an external mmWave downconverter is often used to shift the input signal down to IF or RF for characterization using an RF signal analyzer or digital oscilloscope. While this approach works relatively well for moderately wide modulation bandwidths (e.g., 2 GHz), it can be a bit cumbersome for very wide modulation bandwidths (e.g., > 2 GHz) because of impairments such as I/Q gain imbalance and amplitude or phase variations with respect to frequency.

The wideband mmWave testbed shown in Figure 1 is designed to address the large bandwidth issues at mmWave frequencies. The testbed uses high-performance digital techniques to generate and analyze wideband IF signals in order to overcome the shortcomings of analog RF.

它还使用了一组 VDI 公司（Virginia Diodes, Inc.）的紧凑型上变频器和下变频器。这样便可从任意波形发生器（AWG）直接驱动中频输入，并直接用数字示波器进行分析。紧凑 型 VDI 上变频器和下变频器对本振（LO）频率采用有效的 2 倍频系数，为上变频信号带来改进的信噪比（S/N）和显著降低的转换损耗，相对而言，传统系统中采用的是 6 倍频系数。此外，2 倍频系数能够使用高质量的本振源，确保较高频率下的低相位噪声。

Note: An E-band converter is shown here; a V-band converter can also be used.

Figure 1. This recommended test platform includes a variety of powerful hardware and software.

The various elements of the test bench provide the following key functions:

• The multi-channel 8-bit Keysight M8195A 65 GSa/s arbitrary waveform generator is used to generate wideband modulated IF signals.

• 
  

• Infiniium S-Series DSOS804A high-resolution oscilloscope (8 GHz, 20 GSa/s, 10-bit resolution) is used in conjunction with Keysight 89600 VSA software to digitize and analyze wideband IF signals. An oscilloscope with higher bandwidth is also available.

• 
  

• VDI E-Band compact upconverters (N9029ACST-U12) and downconverters (N9029CST-D12), both of which can be driven directly by an AWG and connected directly to a high-resolution oscilloscope. Alternatively, VDI's compact V-Band converters can be used.

• 
  

• The compact N5183B MXG X-Series microwave analog signal generator provides high-quality local oscillator signals for compact VDI converters.

• 
  

• 89600 Vector Signal Analyzer software, providing advanced demodulation and analysis for 5G candidate waveforms.

• 
  

• Keysight SystemVue software is used to create the waveforms and process and extract the DPD data. Some key considerations of this test approach compared to traditional methods are discussed below.

• 
  

• Wideband signal generation

There are two different approaches to generating wideband signals beyond the analog bandwidth of an AWG: analog I/Q modulation and digital up-conversion. Figure 2 shows a simplified block diagram of both approaches. Each approach has its advantages and disadvantages, which are discussed below.

Figure 2. Both approaches use arbitrary waveform generators to generate wideband signals.

Note the blue and black arrows in both diagrams: the blue ones correspond to mathematical formulas or digital signals, and the black arrows represent real analog signals and voltages.

The figure above is analog I/Q modulation. In this case, an arbitrary waveform generator generates the I and Q signals, which are fed into an I/Q modulator that is typically built into a vector signal generator. One advantage of this approach is that the analog bandwidth required at the AWG output is only half the modulation bandwidth achievable by the RF signal at the output. Therefore, using a 500 MHz AWG, this approach can generate a signal with a 1 GHz modulation bandwidth. The disadvantage is that the analog I/Q modulator generates unwanted distortion components (such as image and carrier feedthrough), and these distortion components can only be partially offset. In addition, the relationship between I/Q gain imbalance and frequency can become a major source of EVM error, especially as the modulation bandwidth increases.

In digital upconversion (lower half of Figure 2), I/Q modulation is performed mathematically—either in real time using a digital signal processor or in advance using software. The result of this calculation is fed into a digital-to-analog converter (DAC) and then upconverted using a mixer. This effectively separates the signal from the high frequencies of the local oscillator (LO), so image and LO feedthrough can be filtered out or rejected. Because this technique uses digital or software I/Q modulation, I/Q gain imbalance is less of an issue than with analog I/Q upconversion. The modulated IF output from the AWG can be converted directly to mmWave frequencies using an external upconverter; in this case, an analog I/Q upconverter is not required. The downside is that this improved high-frequency signal generation technique requires a higher AWG bandwidth.

One type of impairment affects both methods: non-ideal amplitude and phase variations in frequency. These variations typically occur in the RF output path from the AWG to the upconverter, and in the signal path from the RF input of the downconverter to the digitizing oscilloscope. Sources of variation include I/Q modulators, frequency converter mixer or multiplier stages, various passband filter or amplifier variations, and the flatness of the oscilloscope input channel.

Most of these frequency response errors are linear and can be characterized using the adaptive equalizer in the VSA software. Once equalization is complete, the frequency response of the adaptive equalizer can be used to pre-correct the waveform to compensate for the linear frequency response in the signal path (Figure 3).