The field in the conventional sphere gap is appreciably influenced by the surrounding earthed surfaces, and the expected accuracy of the sphere-gap voltmeter may be increased by modifying the configuration of the earthed electrode so as to make the field less dependent on the distance to these surfaces. Accordingly it is proposed that the conventional earthed sphere and supporting shank be replaced by a hemisphere placed on the ground plane, and that the upper sphere be made movable to vary gap length. Alternatively, the distance of the hemisphere above the ground can be varied if this is strongly preferred; for example, if difficulties arise in providing a motor driven high-voltage electrode. Use of the proposed electrode arrangements allows a considerable reduction in cost, especially when requirements of laboratory space are taken into account, and renders voltmeters using sphere diameters of 1 m and above very much simpler to use.Calculations of field factor and measurements of flashover voltage confirm that a sphere-hemisphere gap is affected very much less than a sphere gap by variations in the clearance to walls and large pieces of equipment near the spark gap. It is therefore possible to work with smaller clearances than are required by BS 358 and IEC52 without increasing the dependence of the voltmeter reading on the value of the clearance. Tests show that the sphere-hemisphere gap has a flashover voltage about 5% lower than the sphere gap when the gap length is equal to the sphere radius, and the difference is smaller as the gap length is reduced. A correction factor k is proposed to convert sphere-gap calibrations to sphere-hemisphere gap calibrations, where k = (1.015 − 0.06g), g ≥ 0.25, and g is the ratio of gap length to sphere radius.Two quite different means of supporting the top sphere of the sphere-hemisphere gap have been proposed, and both give entirely satisfactory performance when used with alternating, direct or impulse voltages.