The source impedance of a signal generator, for use under terminated conditions, has a direct effect on the accuracy of the load voltage. Signal-generator specifications rarely give detailed information about the variation of this parameter with frequency. Consequently, the user is generally forced to assume that it is constant and equal to the nominal value. The view is maintained that it is impracticable to apply corrections for this parameter even if its value is known.Three methods are analysed for calibrating the output level of signal generators with a view to determining that method which minimizes the error in the load voltage due to source impedance variation. The methods are each compared in terms of the voltage ratio produced across an arbitrary load impedance by an imperfect generator to a perfect generator. The voltage ratios are expressed in terms of the reflection coefficients of the generator and load impedances. The minimum and maximum of the ratios are given in terms of voltage standing-wave ratios. The results are equally applicable to coaxial and waveguide systems.The recommended method employs the concept of equivalent source e.m.f. or equivalent available power. The requirements for a suitable instrument to measure these quantities are briefly discussed.Recommendations are made regarding the manner of specifying the tolerances on signal-generator parameters which affect the accuracy of the load voltage, namely the calibration level, the attenuation and the source impedance.