EVM Measurement and Correction for Digitally Modulated Signals

We proposed an error vector magnitude (EVM) measurement method for 16-quadrature-amplitude-modulated (16QAM) waveform. A 16QAM signal with 20GS/s sample rate, 1Gbaud/s symbol rate and 4GHz carrier frequency is simulated. By changing orders of the root-raised-cosine (RRC) filters, we set the EVM of the simulated signals to 0.001% (increasing the order of the filter reduces the EVM but can never reach zero) and 1% respectively. We then upload these two signals to the arbitrary waveform generator (AWG), and measure the outputs of AWG using a real-time digital oscilloscope (RTDO). After a series of processing like carrier-recovery, time-recovery, frequency and phase shift estimation and correction, the EVM of the measured signals are 6.0% and 7.0%. The nonlinearity of the AWG is the main reason of the bad results, so we use an algorithm to correct it. After several iterations, the EVM of these signals reduce to 2.2% and 2.9% respectively.