Error vector magnitude (EVM) is the most important figure of merit for signal quality in 5G NR. A new method improves measurement accuracy by reducing noise.
Each new generation of cellular technology increases end-user throughput or bit rate over previous generations. Each new generation accomplished higher data rates through a combination of both wider channel bandwidths and higher-order modulation. Even when we compare LTE and 5G New Radio (NR) with equal channel bandwidth, 5G NR delivers higher throughput through higher-order modulation. Unfortunately, higher-order modulation makes receivers more sensitive to noise and thus bit errors. The good news is you can compensate for the noise in your test equipment.
Both LTE and 5G NR modulate their signal carriers (or rather, subcarriers) using quadrature amplitude modulation (QAM). QAM conveys information by changing the carrier’s amplitude and phase between different states or “symbols.” A symbol is a unique combination of amplitude and phase. The number of symbols means the number of bits transmitted with each symbol. For example, a modulation scheme that has 16 symbols can encode four bits per symbol, whereas a system using 256 symbols can convey eight bits per symbol.
Figure 1. A 16 QAM constellation diagram shows sixteen combinations of amplitude and phase. Each point represents four bits, and the distance from the center to each point is the vector.
A constellation diagram shows the modulation, where each symbol is the endpoint of a vector having a given magnitude and phase. Modulation order is simply the number of possible symbols. The 16QAM constellation shown in Figure 1 has 16 symbols or vector endpoints.
In practice, a signal’s amplitudes and phase shifts don’t precisely fall on the defined symbol endpoints. The error may be due to magnitude error, phase error, or, most commonly, a combination of both. Should an amplitude and phase combination deviate too far from the ideal point, the receiver could incorrectly decode it, which leads to bit errors.
EVM defined
You can find the difference between ideal and measured symbols by connecting these two points with an error vector (Figure 2). Like all vectors, this vector has a magnitude and a direction. In most cases, the error magnitude, rather than its direction, matters. Therefore, modulation accuracy is quantified as the error-vector magnitude (EVM). Larger values of EVM mean greater distance between the measured and reference points and thus a higher probability of bit errors.
Figure 2. The difference between an amplitude and phase point and its ideal location represents an error vector.
A signal analyzer calculates EVM at each symbol time and reports it as a normalized quantity, either relative to the maximum power or to the RMS power in the received signal constellation. Most standards use RMS, but you must verify the method when comparing EVM values. EVM uses units of percent or dB, usually as statistical values (mean, max, min, etc.) over some period. Analyzers may also plot EVM for successive symbols to see whether EVM remains constant during a transmission. Lower values of EVM, that is, smaller percentage values or lower (more negative) dB values, are always more desirable than greater values. Typical EVM values for 5G NR networks typically run -40 dB to -50 dB or single-digit percent values.
The importance of good EVM increases as the modulation order increases. In higher-order modulation, such as OFDMA, where the symbols or constellation points are close together, errors in the received signal’s magnitude and/or phase are more likely to lead to incorrectly decoded symbols because the symbols are close together. Figure 3 shows constellation diagrams for 16 QAM, 64 QAM, and 256 QAM.
Figure 3. As order modulations (16 QAM, 64 QAM, and 256 QAM) increase, the vector points on a constellation diagram get closer.
5G NR achieves higher throughput in part by using higher-order QAM modulation, in particular 64 QAM and 256 QAM. These modulations do, however, require both better performance at the transmitter and receiver, as well as a “cleaner” RF environment than 16QAM. Like many other wireless standards, 5G NR places limits on the maximum permissible EVM, which decreases as modulation order increases. In 5G NR, 16 QAM requires an EVM of no greater than 12.5%, while 256 QAM requires an EVM of 3.5% or less.
Figure 4. This EVM measurement setup uses a vector-signal generator and spectrum analyzer to measure EVM in a device under test.
Because EVM is the primary “figure of merit” for modulation quality in 5G NR networks, you must accurately and repeatably measure a device’s or system’s EVM. Use a spectrum analyzer or signal analyzer to make EVM measurements. These instruments can decode the received 5G NR signal and calculate its EVM. In some test scenarios, you need a vector-signal generator to create a modulated 5G NR signal, which serves as the input to a device under test, such as a power amplifier, as shown in Figure 4.
Get the actual EVM
When measuring EVM, remember that the measured EVM is a combination of both the EVM of the device under test (and potentially the channel) and the EVM created by or within the measuring instrument. The contribution of the analyzer to overall EVM is sometimes referred to as residual EVM. Traditionally, the requirement for accurate EVM measurements was that the measuring instrument should have an EVM that was at least 10 dB better than the DUT’s EVM. Unfortunately, even with high-performance instruments, this margin can be difficult to obtain. The fact that you must make some 5G NR measurements over-the-air rather than in a conducted test environment further enhances the need for good analyzer EVM performance, especially with low signal levels due to free-space path loss or other factors.
An analyzer’s residual EVM has four primary sources:
- phase noise,
- frequency response,
- nonlinearities, and
- wideband noise.
The first three are relatively easy to address. Using high-quality local oscillators, high-performance spectrum analyzers can limit the contribution of phase noise to residual EVM. You can calibrate out or compensate for the effects of frequency response, the variation in received-signal characteristics as a function of frequency. Attenuation addresses nonlinearities such as harmonics and intermodulation products by limiting a received signal’s amplitude, which avoids compression within the analyzer.
Wideband noise is, however, a more challenging issue in EVM measurements. This noise is normally characterized by using traditional noise-figure measurements. It includes both thermal noise and noise contributions from individual components. Furthermore, this noise scales with bandwidth, meaning that wideband noise is an even greater issue given the wider bandwidth signals commonly used in 5G NR. Because 5G NR often requires the measurement of signals having a wide bandwidth, accurate EVM measurements for 5G NR devices require some method of reducing or mitigating the impact of wideband noise on residual EVM.
Noise reduction methods
Various methods can remove or reduce analyzer-added noise. IQ noise cancellation has emerged as the most promising method. Depending on the amount of input attenuation, IQ noise cancellation can improve EVM measurement performance by approximately 5 dB, a significant improvement when measuring EVM in 5G NR networks.
Figure 5. Improvement in residual EVM using IQ noise cancellation (IQNC) occurs as a function of attenuation and number of captures.
Performing an IQ noise cancellation procedure requires several measurements. Figure 5 shows the additive effect from
noise sources.
- Make a measurement containing all noise contributions, both internal and external.
- Make a measurement with the analyzer input terminated to find the noise contribution of the analyzer alone.
- Make multiple captures on the signal to estimate an ideal, noise-free capture.
Make the measurements on raw IQ data, that is, on the digital representation of the received signals. This method reduces the contribution of wideband noise to residual EVM better than other methods for several reasons.
- IQ noise cancellation requires a single measurement path in the measuring instrument. It can be performed entirely in software and implemented without requiring a hardware change.
- IQ noise cancellation is also independent of the modulation type or order.
- IQ noise cancellation won’t cancel out noise from the signal generator noise or from the DUT.
Conclusion
EVM continues to be the most important figure of merit for modulation quality in wireless networks. It quantifies the distance between the ideal constellation or symbol points and the actual received or measured symbol points. Lower values of EVM (expressed in units of percent or dB) are always desirable. Many standards, including 5G NR standards, specify a required maximum EVM level for a given modulation order.
EVM is normally measured with a signal analyzer or spectrum analyzer. Although this measurement is well understood and has been used in previous generations of cellular technology, EVM measurements are particularly important in 5G NR due to higher modulation orders and wider bandwidths. Accurate EVM measurements of 5G NR signals are, however, challenging. Fortunately, recent developments such as IQ noise cancellation let existing instruments make reliable and repeatable EVM measurements and thus verify compliance with relevant 5G NR modulation quality requirements.
Tell Us What You Think!