I usually use

EVM = sqrt[ 1/10^(SNRdB/10) ] x 100

It is possible to calculate EVM by measuring the output SNR of the PA/transmitter. If you have code to measure SNR (MATLAB) then you can use that, but it isn't necessary to have a full receiver to measure symbol SNR or BER. Of course if you do have a receiver (VSA) then the measurement is straight forward.

All that is really necessary is that you have a way to find what portion of the output signal is uncorrelated with the input signal (over the signal bandwidth). You can do this several different ways

1. Crosscorrelation: If you can easily obtain a complex envelope representation of the output signal (envelope simulation) then you could filter the output to the actual signal bandwidth (to remove off channel noise/distortion) then calculate the crosscorrelation between the filtered output and the input (or ideal input) signal. The crosscorrelation tells you what percentage of the output signal is correlated to the input signal. The remainder is distortion/noise that is uncorrelated, so from this you have SNR.

2. Feedforward: You can also measure the uncorrelated distortion/noise by using feedforward cancellation. You can take the output signal and subtract the input signal by attenuating the output signal by the gain of the circuit and then changing the phase until the linear/desired output signal is fully cancelled. You will reach a floor in the cancellation that should be the uncorrelated distortion that exists across the signal bandwidth (looking a the power spectrum). The ratio of the original output signal power to the remaining inband distortion (after cancellation) is the effective SNR.

I, and others, have used and compared both techniques and they are equivalent. Crosscorrelation is the easier method. The feedforward technique is interesting for a simple lab measurement since it can be performed from two measurements using a spectrum analyzer; although, it requires attenuators and a phase shifter to carryout the feedforward cancellation.

Ok, so this was probably more information than you were looking for. Hopefully some of it is helpful.