High voltage equipment will be subjected to several types of electrical stress during operation. A battery of factory tests is defined to ensure that the equipment will perform satisfactorily in service. One of the crucial tests is to apply a simulated lighting impulse as standardised to a double-exponential impulse with at front time of 1.2 µs (± 30 %) and a time to half value of 50 µs (± 20 %). Although this wave-shape only approximates natural lightning, there is a solid body of experience within industry, proving that reliability of equipment in service is adequately proven by the standard waveform. It is however crucial for consistency of results that the both voltage level and wave-shape are correctly measured. This paper discusses the requirements and performance of the recording instruments used, leaving the properties of high voltage impulse dividers outside the discussion. The requirements for the recording instrument – transient recorder – are given in IEC 61083-1. The standard provides requirements for, and/or tests to verify, that the recorder has moderately fast response, fast settling time, high resolution, linearity under dynamic conditions, high accuracy and reasonably low internal noise. This is partly in contrast to major trends in transient recorder development, where fast sampling and fast step response are prioritized ahead of high accuracy and fast settling without creeping response. We have therefore evaluated several commercially available recorders in order to find one with respectively flat and reasonably fast step response. In this campaign, a proprietary step generator based on the use of a mercury reed relay has been used. Evaluation of this device is submitted to ISH 2017. It has been found that the measured flatness of the step response directly after the step is a good first indicator of the performance of the transient recorder. This is identified in IEC 61083-1 clauses 1.5.2 and 1.5.3, as a requirement on stability of the recorded step from 0.5 T1min to T2max. For lightning impulse this means from 0.42 µs to 60 µs. For approved transient recorders the requirement is to be within 1 %. For reference transient recorders, a limit of not more than 0.5 % should be applied. Further proof of the accuracy of the transient recorder can be achieved by convolution of an ideal waveform with the recorded step response and analysing the resulting curve with lightning impulse parameter software. A third possibility is to make direct calibration of the transient recorder, using a calculable impulse calibrator. Several state-of-art transient recorders have been evaluated and the results show that only a few are suited for measurement of lightning impulse. Also, the variation of the performance between the ranges and channels of one instruments are significantly large. Both direct assessment of step response as well as result of convolution with a theoretical 0.84/50 µs impulse will be reported. The agreement with results obtained with a calculable impulse calibrator will be illustrated.