Although impulse-voltage acceptance tests are, by assumption, nondestructive when successful, it is not easy in practice to establish working criteria of insulation failure and to co-ordinate these criteria with observations. A tentative definition of impulse failure is used as a basis for considering the significance of indications given by electrical methods of fault detection applied to transformer windings. Exploratory tests on high-voltage motors are described, in which it was found that fault indications were likely to be obscured by disturbances arising from non-cyclic magnetization of the iron and from other causes not determined. Acceptance impulse testing for highvoltage capacitors and cables is briefly discussed. It is concluded that the techniques for acceptance testing of transformer windings are broadly satisfactory, but that, for motor windings, existing methods are inadequate.LIST OF SYMBOLS V -Impulse voltage test level. v = Impulse breakdown voltage of one turn. N = Number of turns in winding, .s = Factor depending on distribution of impulse voltage.(1) INTRODUCTION Ideally, any test made on a completed article which is afterward put into service should be entirely non-destructive. It is permissible for an article that has failed in the test to be damaged by the test, provided that all failures are detected with certainty. In the practice of the impulse-voltage testing of power apparatus, it cannot readily be proved that a successful withstand test has been entirely harmless. Moreover, the diagnosis of insulation failure, particularly in windings of transformers and machines, can present great difficulties.For economic reasons, the discharge inception level of the insulation of power apparatus is generally below the specified impulse test voltage. The problem of discharges in the insulation of transformer windings under impulse test 1 has been considered at some length, and it has been claimed that discharges have been detected. 2 Estimates have been given of the cumulative effect, on the electric strength of insulation, of repeated impulse voltage applications. 3 ' 4> 5 A large number of papers have been published describing methods of fault detection in windings and the performance of these methods. References to a few of these papers are given, 1 -2>6~13 but many others can be found in Reference 14.The problems which need to be solved can be summarized as follows:To fix a criterion by which to decide whether the insulation has failed.To co-ordinate this criterion with observations. To show that impulse withstand testing is non-destructive or. if it is not, to determine the degree of destructiveness.To devise detection methods capable of giving satisfactory fault indications.Mr. White is with the Electrical Research Association.