From a technical point of view, electroretinography is one of the most difficult things in biomedical electronics. We have a very small potential, sometimes some hundred microvolts, but often only some ten or even less. The potential contains a broad spectrum of frequencies, sharp a-waves and wavelets, slower b-waves and perhaps c-waves with a duration of one second and more. This potential is disturbed by many other bioelectrical and technical components, often ten or hundred times greater in amplitude than the desired signal. For instance by the noise of the necessarily high input impedance of the aJmplifiers, by skin potentials, electrodepolarisation, eye blinking, line hum, and so on. Moreover, all these artefacts appear in the same frequency band as tile retinogram, and therefore, ordinary filtering is a doubtful procedure, because the signal is influenced by it too.Nevertheless, we find a lot of successful work and important results in the past. But nowadays, with the enormous development of electronics and measurement technics, we have to think over, what can be done to utilise these and to accelerate the progress and usefulness of retinography. We feel, that progress in clinical retinography will depend on two main points. Firstly we have to replace the more or less rough qualitative description of a ERG by quantitative and comparable measurements, and secondly to change from single ERG recordings to a more complex functional analysis of the behavior of the b-wave to various stimulating conditions. With the common stripchart-records, one gets a registration of ten or fifteen millimeters and cannot read it with more accuracy than -k 1 ram. So normally, there is already a measuring error of' at least twenty per cent. In addition, to get a good signal-to-noise ratio, only large signals can be handled, evoked by very strong stimuli. Therefore, the examinations very often are performed near by or sometimes beyond the upper limit of
