What it measures
Each digital symbol has an ideal position in the phase and amplitude plane. In practice the real signal falls near that point, but not exactly on it. The error vector is the difference between the measured position and the ideal one. Error vector magnitude sums up those differences in a single number: the square root of the mean error power divided by the mean reference signal power, expressed as a percentage or in decibels.
How it is measured
- A vector signal analyzer captures the signal with a conversion to baseband.
- It synchronizes, recovers the carrier and the symbol clock, and demodulates.
- It reconstructs the ideal signal from the decided symbols and compares it with the real signal, symbol by symbol.
- It calculates the mean of the errors and displays the value along with the constellation and the error breakdowns.
The measurement parameters, such as the pulse-shaping filter, the symbol rate and the modulation type, must match those of the transmitter. With a wrong configuration, the value means nothing.
What degrades it
- Noise: it scatters the points into a circular cloud.
- Phase noise: it rotates the symbols, so they stretch in the tangential direction.
- Amplitude and phase imbalance between I and Q: it deforms the constellation.
- Local oscillator leakage: it shifts the entire constellation.
- Amplifier compression and non-linearity: they compress the outer points.
- Synchronization or filtering errors: they produce intersymbol interference.
How to interpret the result
The overall value indicates quality, but the useful part is the shape of the error. A round cloud points to noise; a rotation points to phase noise; a squashed constellation, to compression. Denser schemes demand a lower EVM, because the points are closer together and any error crosses the decision boundary sooner. Acceptable limits are set by the standard or the system specification, and should be consulted in their current edition.
The instrument’s own EVM limits the measurement: measuring a very good transmitter requires an analyzer with a clearly lower residual EVM.