As a nonmember reader of the Institute's Journal I have had occasion to refer to your more recent articles on the diode-rectification method of detection of radio-frequency signals. The complete theoretical treatment and experimental verification available in articles over the past few years, the latest of which to date are the excellent discussions by Messrs. F. E. Terman and J. R. Nelson,* have been very helpful.However, I noticed that to date this method of detection has one shortcoming, specifically that it cannot help but distort at 100 per cent modulation and thereabouts. Dr. Terman expressed it in his December discussion in a statement to the effect that for distortionless detection the ratio X/R must theoretically equal infinity, where X is the grid condenser reactance and R the leak resistance. I have used a circuit with a slight change from Dr. Terman's in which there is no distortion theoretically or practically up to and including 100 per cent modulation, and yet in which X/R theoretically does not need to be quite as high as would ordinarily be used. I submit an explanation below with the hope that the modification which to my knowledge is new, will be of interest to Messrs. Terman, Nelson and other members.Theoretically, the ideal grid leak for a detector would be a constant-current device from zero applied voltage to the highest. This could cause a constant discharge rate for the condenser-leak combination, which for distortionless detection would be made to equal the maximum slope of carrier envelope to be encountered at all percentages of modulation.Of course, there is no such constant-current leak practical, but if the ordinary leak be disconnected at the cathode of the rectifier and reconnected to a point on the direct-current supply positive with respect to the cathode by several times the peak of the rectified envelope, the leak current may be made very nearly constant as the bias fluctuations are then only a small part of the total voltage across the leak. Of course the resistance of the leak should then be increased to bring its average direct current back to what it was before.As an example, suppose a diode detector is to be designed for usual broadcast carrier frequencies to detect without distortion all audio frequencies up to 5000 cycles and up to 100 per cent modulation, but with maximum instantaneous envelope slope not to exceed 20,000 volts per second. Assume a value of C of 0.00025 microfarad, a maximum negative peak for C of five volts, and a positive tap for the leak of 22-2 volts above the cathode potential. Then the discharge rate of C by R must be equal to 20,000 volts per second, so we have 20,000 a 22.5 + 5/2 20,000 0.00025 X 10-6 X R or, R = 5 X 101 ohms. There appears to be no disadvantage in connecting the leak from bias to positive instead of to ground unless a positive potential on the detector bias during starting before the detector cathode has become warm is so considered. The current in such case, as the cathode heated, would be hundreds of times less than the plate ...