How do I measure the EVM of a transmitter using a vector signal analyzer?
EVM Measurement and Analysis
EVM is the most comprehensive single metric for transmitter signal quality because it captures all impairments: noise, distortion, phase noise, I/Q errors, and frequency errors in one number.
| Parameter | SOLT Cal | TRL Cal | eCal |
|---|---|---|---|
| Accuracy | Good | Excellent | Good-very good |
| Standards Needed | 4 (S,O,L,T) | 3 (T,R,L) | 1 (module) |
| Bandwidth | Broadband | Band-limited | Broadband |
| Setup Time | 5-10 min | 10-20 min | 1-2 min |
| Best For | Coaxial, general | On-wafer, waveguide | Production, speed |
Calibration Procedure
The total EVM is the RSS of individual impairment contributions: (1) Phase noise contribution: EVM_PN ≈ sqrt(2 × integral(f_min to f_max) L(f) df) (the integrated phase noise over the signal bandwidth). For a 20 MHz LTE signal with LO phase noise of -100 dBc/Hz at 1 kHz: EVM_PN ≈ 1-2% (-40 to -34 dB). Phase noise primarily affects higher-order modulations (256-QAM, 1024-QAM). (2) PA nonlinearity: compression and AM-to-AM/AM-to-PM distortion cause constellation point displacement. For a PA operating at 3 dB below P1dB: EVM_PA ≈ 3-5% (-30 to -26 dB). With DPD: EVM_PA reduces to 1-2%. (3) I/Q imbalance: gain and phase mismatch between the I and Q modulator branches: gain imbalance of 0.5 dB: EVM contribution ≈ 3% (-30 dB). Phase imbalance of 3°: EVM contribution ≈ 3% (-30 dB). (4) LO leakage (carrier feedthrough): DC offset in the I/Q modulator creates a carrier component at the center of the spectrum. EVM contribution depends on the leakage level relative to the signal: -30 dBc LO leakage: EVM ≈ 3% (-30 dB). -40 dBc: EVM ≈ 1% (-40 dB). (5) Quantization noise: DAC resolution limits the EVM floor. For a 12-bit DAC: EVM floor ≈ 0.02% (-74 dB). For 10 bits: 0.1% (-60 dB). For 8 bits: 0.4% (-48 dB). (6) Thermal noise: the SNR at the VSA input determines the noise contribution. EVM_noise = 1/sqrt(SNR_linear) × 100%. For SNR = 40 dB: EVM_noise = 1%. Total: EVM_total ≈ sqrt(EVM_PN^2 + EVM_PA^2 + EVM_IQ^2 + EVM_LO^2 + EVM_noise^2).
Error Sources
(1) VSA input level: the signal must be within the VSA linear range. Too high: the VSA compresses, adding its own distortion to the measurement. Too low: the VSA noise dominates, increasing the measured EVM beyond the DUT actual EVM. Optimal: set the VSA input level 15-25 dB below the VSA maximum input (typically 0 to -20 dBm at the VSA input). (2) VSA frequency accuracy: the VSA reference oscillator must be more stable than the DUT. If the VSA and DUT use different references: the carrier frequency offset (CFO) must be within the VSA synchronization range. For most VSAs: CFO < ±10 ppm is sufficient. For very narrow-bandwidth signals: use a common 10 MHz reference between the DUT and VSA. (3) Triggering: for burst signals (Wi-Fi, Bluetooth): the VSA must trigger on the burst preamble. Use the VSA trigger function (level trigger or IF trigger). For continuous signals (LTE, 5G): no trigger needed (the VSA uses the sync signals in the standard for frame synchronization). (4) Averaging: the EVM is computed over a specified number of symbols or frames. More symbols/frames: more stable, averaged result. Minimum: typically 10-20 frames for a stable result. Standards may specify the minimum measurement interval (e.g., 3GPP specifies 10 subframes = 10 ms for LTE EVM measurement).
- Performance verification: confirm specifications against the application requirements before finalizing the design
- Environmental factors: temperature range, humidity, and vibration affect long-term reliability and parameter drift
- Cost vs. performance: evaluate whether the application demands premium components or standard commercial grades
- Interface compatibility: verify impedance, connector type, and mechanical form factor match the system architecture
Fixture Considerations
(1) EVM per subcarrier: for OFDM signals, the VSA can report EVM for each subcarrier individually. This reveals frequency-dependent impairments: a dip in EVM at specific subcarriers indicates passband ripple or filter notch at that frequency. Higher EVM at band edges: indicates filter roll-off or group delay distortion. (2) EVM vs time: for time-varying signals (TDMA, pulsed), EVM can be plotted vs time within the burst. Degradation at the start of the burst: indicates PLL settling or PA transient. Degradation at the end: indicates PA thermal effects. (3) EVM vs power: measure EVM at multiple output power levels. At low power: EVM is constant (limited by noise and I/Q impairments). As power increases toward P1dB: EVM degrades due to PA compression. The maximum output power for a given EVM specification is the usable linear power of the transmitter.
Frequently Asked Questions
What EVM does my transmitter need for 256-QAM?
3GPP LTE/NR specification: EVM < 3.5% (-29.1 dB) for 256-QAM. In practice: design the transmitter for EVM < 2.5% (-32 dB) to provide manufacturing margin. This requires: PA operating at least 6-8 dB below P1dB (or with DPD), LO phase noise < -100 dBc/Hz at 10 kHz offset, I/Q imbalance < 0.3 dB gain and < 2° phase, LO leakage < -35 dBc, and DAC resolution ≥ 12 bits. For Wi-Fi 6 1024-QAM: EVM < 3.2% (-29.9 dB), requiring even tighter impairment control.
Can I measure EVM with a spectrum analyzer instead of a VSA?
A standard swept spectrum analyzer cannot measure EVM because it does not demodulate the signal (no phase information, no symbol recovery). A signal and spectrum analyzer (SSA) with vector signal analysis software (e.g., Keysight 89600 VSA software running on an X-series SA, or R&S FSW with K70 option) can measure EVM. The SSA digitizes the IF signal and the VSA software performs the demodulation and EVM calculation. This approach combines the hardware of an SA with the analysis capability of a VSA. Alternatively: use a dedicated VSA instrument (Keysight M9393A, R&S FSW-K, or NI PXI-based VSA).
Does EVM include the effect of the measurement system?
Yes. The measured EVM is the combined EVM of the DUT and the measurement system: EVM_measured ≈ sqrt(EVM_DUT^2 + EVM_VSA^2). To determine the DUT EVM: subtract the VSA residual EVM in quadrature: EVM_DUT = sqrt(EVM_measured^2 - EVM_VSA^2). The VSA residual EVM (with a clean signal generator as the source) is specified by the manufacturer: typical -40 to -50 dB for mid-range VSAs. If the DUT EVM is -30 dB and the VSA residual is -45 dB: EVM_measured = sqrt(10^(-3) + 10^(-4.5)) ≈ sqrt(1.032e-3) = -29.9 dB (0.1 dB error from the VSA). If the DUT EVM is -38 dB and the VSA residual is -40 dB: EVM_measured ≈ -36 dB (2 dB error). For accurate measurement: the VSA residual EVM should be at least 10 dB better than the DUT EVM.