What impairments contribute to EVM?

EVM is a composite metric that captures all signal impairments. Phase noise: the LO's random frequency fluctuations smear constellation points tangentially (along circles of constant amplitude). Integrated phase noise directly adds to EVM: if integrated PN is minus 35 dBc, it contributes approximately 1.8 percent EVM. IQ imbalance: gain mismatch between I and Q paths distorts the constellation into an ellipse. Phase mismatch between I and Q rotates and skews it. 1 degree phase error contributes approximately 1.7 percent EVM. PA nonlinearity: AM-AM distortion compresses outer constellation points inward. AM-PM distortion rotates them. Together they are often the dominant contributor, especially at high output power. LO leakage (DC offset): creates a fixed offset of the entire constellation from the origin, visible as an energy spike at the carrier frequency. ADC quantization: each bit of resolution contributes approximately 0.5 percent RMS EVM. A 10-bit ADC floor is approximately 0.03 percent.

How is EVM related to SNR and BER?

For an AWGN-limited system, EVM and SNR are directly related: EVM(percent) = 100 / sqrt(SNR_linear), or equivalently EVM(dB) = minus SNR(dB). This means 1 percent EVM = 40 dB SNR, 3.5 percent EVM = 29 dB SNR, 10 percent EVM = 20 dB SNR. The relationship between EVM and BER depends on the modulation scheme. For M-QAM: BER approximately equals (2(1 minus 1/sqrt(M))) / log2(M) times erfc(sqrt(3 log2(M) / (2(M minus 1) times EVM squared))). In practice, the standard specifies EVM limits that guarantee a target BER with margin. 5G NR 256QAM requires 3.5 percent EVM, which corresponds to approximately 29 dB SNR and a raw BER below 10 to the minus 3 before FEC. The FEC (LDPC in 5G) then reduces this to the target BLER of 10 to the minus 1 for the initial transmission.

EVM measurement requires a vector signal analyzer (VSA) or equivalent software on a spectrum analyzer with IQ demodulation capability. The VSA digitizes the signal, demodulates it to recover the constellation, and compares each received symbol to its ideal position. The measurement process includes: synchronization (find the start of the data payload), channel estimation and equalization (remove the known channel response using reference signals or pilots), and then computation of the error vector for each data symbol. EVM is reported as the RMS over all symbols in the measurement period. Per-subcarrier EVM analysis in OFDM reveals frequency-dependent impairments: high EVM on edge subcarriers indicates filter roll-off or IQ skew. EVM versus power analysis (sweeping output power) reveals the PA compression point and the optimal operating point. The measurement must use the correct standard-specific equalizer and reference signal configuration; using the wrong settings gives misleading results.

Signal Quality

EVM

/ee-vee-em/ — Error Vector Magnitude

|error|/|reference| × 100%. Captures all impairments: phase noise, IQ imbalance, PA nonlinearity, LO leakage, ADC quantization. EVM = 1/√SNR for AWGN. 5G NR: QPSK 17.5%, 16QAM 12.5%, 64QAM 8%, 256QAM 3.5%. 1%=40dB SNR. PA compression: dominant EVM contributor at high power. VSA measurement with standard-specific equalizer.

256QAM: 3.5%

64QAM: 8%

QPSK: 17.5%

Understanding EVM

EVM is the single most important metric for evaluating the quality of a digital transmitter or receiver. Unlike older metrics that measure individual impairments separately, EVM captures everything at once: it is a direct measure of how close the actual constellation is to the ideal. If EVM is within spec, the system works; if not, something is wrong.

The beauty of EVM is its diagnostic power. By analyzing the pattern of constellation distortion, an engineer can identify the specific impairment: phase noise creates circular smearing, PA compression pushes outer points inward, IQ imbalance makes circles into ellipses, and DC offset shifts everything. EVM is both a pass/fail metric and a debugging tool.

EVM Equations

EVM definition:
EVM(%) = √(P_error/P_ref)×100
EVM(dB) = 20log(EVM/100)

SNR relationship:
SNR ≈ −EVM(dB)

From IQ errors:
EVM² = ΔI² + ΔQ² + PN + nonlinearity

EVM Requirements by Standard

Modulation	EVM req	EVM (dB)	SNR equiv	Standard
BPSK	<35%	−9 dB	9 dB	802.11
QPSK	<17.5%	−15 dB	15 dB	LTE
16QAM	<12.5%	−18 dB	18 dB	LTE
64QAM	<8%	−22 dB	22 dB	WiFi 6
256QAM	<3.5%	−29 dB	29 dB	WiFi 6
1024QAM	<1.8%	−35 dB	35 dB	WiFi 6

Common Questions

Frequently Asked Questions

Impairments?

Phase noise: tangential smearing (integrated PN −35 dBc ≈ 1.8%). IQ imbalance: elliptical (1° ≈ 1.7%). PA nonlinearity: compression inward + rotation (AM-AM/PM, dominant at high power). LO leakage: constellation offset. ADC: ~0.5%/bit. All RSS combine to total EVM.

SNR/BER relation?

EVM(%) = 100/√SNR. 3.5% = 29 dB SNR. Maps to BER via modulation-specific equations. 256QAM @3.5% EVM: raw BER < 10³. LDPC FEC reduces to BLER 10¹. EVM limits guarantee target BER with margin. Lower EVM = better BER = higher throughput.

Measurement?

VSA: digitize, demodulate, equalize, compare to ideal. RMS over all symbols. Per-subcarrier: reveals freq-dependent issues (filter rolloff, IQ skew). EVM vs power: find PA compression/optimal point. Must use correct standard equalizer/reference signals. Wrong settings = wrong results.

Related Terms

← Even Mode EVM Budget →

Signal Quality

Request a Quote

Need EVM analysis, transmitter characterization, or DPD optimization? Contact our team.

Get in Touch