At our Murray Hill, New Jersey, laboratories we worked on a different approach. We exposed transistors like those used in the decoders to large doses of radiation. We also exposed entire spare decoder units to accelerated radiation to find out where their weakest points were (_see illustration_). And then we built and tested decoders using radiation-resistant transistors to see if they worked better. After a week of intensive laboratory work, we had some pretty good evidence. The tests of individual transistors definitely showed that heavy radiation would cause them to deteriorate. Testing of the complete decoders also led to some failures, and, when we analyzed them, they turned out to be the kind that would be caused by faulty transistors. We also discovered that the most sensitive part of a decoder circuit was the zero gate, which recognizes the zeros in the one-and-zero code that commands the satellite. [Illustration: _A typical command signal sent to Telstar I. It consists of seven pulses: a three-unit “start” pulse and a binary code made up of three two-unit “one” pulses and three one-unit “zero” pulses._] Fooling the Decoder Now we thought we knew the guilty component, but the hardest job still lay ahead of us. We had to do something to the commands so that they would bypass this troublesome zero gate. Each of the fifteen satellite commands is a binary code of seven pulses, as illustrated in _the diagram above_. The first—the _start pulse_—is three units wide. Then follow six more pulses of which three are two units wide (_one pulses_) and three are one unit wide (_zero pulses_). The arrangement of this group of six ones and zeros determines the particular command. Each time a one pulse arrives at the decoder, a _one gate_ counts the pulse and stores a one in its memory. A _zero gate_ counts the zero pulses, but does not store anything. So, if the zero gate is blocked, the decoder will not count the zeros in any of the coded commands and thus cannot decode them properly. [Illustration: _The special “notched one” pulse that was invented to fool Telstar I command decoder._] What could be done about this? The answer seemed to be to devise a new type of pulse—a pulse that would be enough like a one so that it would pass through the one gate and advance the counter, but, at the same time, be enough unlike a one so that the one gate would not store it in its memory. This led to the invention of the special long pulse with a dip or notch in it that is shown in _the diagram above_. When we tested it in the laboratory on one of the duplicate decoders we had exposed to radiation, this new notched pulse worked as we hoped it would. It passed through the one gate and advanced the counter, but was not stored as a one in the one gate’s memory. Thus it fooled the decoder by doing just what a zero is supposed to do, even though it had gone through the one gate rather than the zero gate.