EVM in 5G NR: addressing measurement challenges

EVM quantifies the difference between the ideal and actual received signal points in a constellation diagram, and lower EVM values are crucial for minimising bit errors. In this article, Rohde & Schwarz explore the importance of EVM in 5G NR, its calculation, and the challenges posed by residual EVM, which is introduced by the measurement instruments themselves.

This article originally appeared in the September'24 magazine issue of Electronic Specifier Design – see ES's Magazine Archives for more featured publications.

The evolution of cellular technology has consistently focused on increasing end-user throughput. Each new generation, including 5G NR, has achieved this by combining wider channel bandwidths with higher-order modulation schemes. Even with equal channel bandwidths, 5G NR surpasses LTE in throughput, largely due to its use of higher modulation orders.

In both LTE and 5G NR, subcarriers are modulated using quadrature amplitude modulation (QAM). QAM conveys information by altering the amplitude and phase of the carrier signal between different states, known as symbols. A symbol represents a unique combination of amplitude and phase, and a greater number of symbols allows for more bits to be transmitted per symbol. For example, a 16-symbol modulation scheme can encode 4 bits per symbol, while a 256-symbol scheme can transmit 8 bits per symbol.

The importance of EVM in modulation accuracy

Symbols are often visualised in a constellation diagram, where each symbol is the endpoint of a vector with a specific magnitude and phase. The modulation order refers to the number of possible symbols, such as the 16 symbols in a 16QAM constellation, as shown is Figure 1. However, in practice, received signals rarely match these idealised points exactly due to various distortions, leading to errors in magnitude and phase. This discrepancy is represented by an error vector, which has both magnitude and direction, but typically, only the magnitude of the error is considered significant.

Figure 1. 16QAM constellation

Modulation accuracy is thus quantified by Error Vector Magnitude (EVM), with larger EVM values indicating a greater distance between measured and ideal symbol points, and consequently, a higher likelihood of bit errors (see Fig 2).

Figure 2. Error vector magnitude

EVM is calculated at each symbol interval and is reported as a normalised quantity, either relative to the maximum power or the RMS power in the received signal constellation, with RMS power being the more common reference. EVM is typically expressed as a percentage or in decibels (dB), with lower values being more desirable. For 5G NR networks, typical EVM values range from -40 to -50dB, or in low single-digit percentages.

The importance of minimising EVM grows with the modulation order. Higher-order modulation schemes, like those used in 5G NR (64QAM and 256QAM), feature symbols that are closely spaced in the constellation diagram. This proximity increases the likelihood of errors due to deviations in magnitude or phase. For instance, 16QAM in 5G NR allows a maximum EVM of 12.5%, while 256QAM requires an EVM of 3.5% or less. Figure 3 shows constellation diagrams for 16 QAM, 64 QAM, and 256 QAM.

Figure 3. 16 QAM, 64 QAM, and 256 QAM

Measuring EVM in 5G NR networks

Given that EVM is a crucial metric for assessing modulation quality in 5G NR networks, it is essential to accurately and consistently measure EVM. These measurements are usually performed using a spectrum or signal analyser capable of decoding the received 5G NR signal and calculating EVM for each decoded symbol. In some cases, a vector signal generator may also be used to create a modulated 5G NR signal for testing a device under test, such as a power amplifier.

Residual EVM

When measuring EVM with a spectrum analyser (see Fig 4), it’s important to recognize that the measured EVM reflects both the EVM of the device under test and the residual EVM introduced by the analyser itself. Residual EVM arises from the analyser’s imperfections and can be influenced by factors such as phase noise, frequency response, non-linearities, and wideband noise. Traditionally, the measuring instrument’s EVM should be at least 10dB better than that of the device under test, but achieving this margin can be challenging in 5G NR.

Figure 4. EVM measurement setup with vector signal generator and spectrum analyser

The residual EVM of an analyser has four primary sources: phase noise, frequency response, non-linearities, and wideband noise. The first three sources are relatively manageable. Modern high-performance spectrum analysers can limit phase noise through high-quality local oscillators. Frequency response variations can be calibrated or compensated for, and attenuation can address non-linearities by limiting the amplitude of received signals to prevent compression within the analyser.

Wideband noise, however, presents a more significant challenge in EVM measurements. This noise includes both thermal noise and noise from individual components and scales with bandwidth, making it a greater issue for the wideband signals used in 5G NR. Accurate EVM measurements for 5G NR devices, therefore, require methods to reduce or mitigate the impact of wideband noise on residual EVM.

IQ noise cancellation

Figure 5. Improvement in residual EVM using IQ noise cancellation (IQNC) as a function of attenuation and number of captures

Several methodologies can be employed to reduce the noise added by the analyzer, but one of the most promising is IQ noise cancellation. IQ noise cancellation is a multi-step procedure that involves measuring the total noise contribution (internal and external), isolating the analyser’s internal noise by terminating its input, and performing multiple captures to estimate a noise-free signal (see Fig 5). This process operates on raw IQ data - the digital representation of the received RF signals - improving EVM measurement accuracy by reducing wideband noise without requiring hardware modifications. IQ noise cancellation is effective across different modulation types and orders and doesn’t eliminate noise from the signal generator or the device under test. Depending on the input attenuation, IQ noise cancellation can improve EVM measurement performance by approximately 5dB, a significant enhancement for 5G NR testing.

Conclusion

In summary, Error Vector Magnitude (EVM) remains a vital metric for evaluating modulation quality in wireless networks, particularly in 5G NR, where higher modulation orders and wider bandwidths present unique challenges. Although EVM measurement is a well-established process, the increased demands of 5G NR require advanced techniques like IQ noise cancellation to ensure accurate and reliable measurements. These advancements enable the verification of compliance with stringent 5G NR modulation quality standards, ensuring optimal performance in next-generation networks. Rohde & Schwarz offers test and measurement solutions at the forefront of innovation which help engineers meet the most demanding EVM requirements.