Bernard Widrow

"my experience with computing started with work on memory. I always had a perspective on computers which is sort of a memory's eye view. I look for the memory and see what you have to connect around it. In my work for a master's degree I wanted to improve the signal-to-noise ratio of the sensing signal coming out of core memory. In those days, the materials for making magnetic cores were very poor compared to what they finally evolved into."

"What was needed was a very square hysteresis loop, and they couldn't get that exactly. The signals coming from the selected core in the memory plane containing the bit that the computer was trying to read became corrupted with noise, and sometimes the signal could be noisy enough to cause an error. I had the idea of driving the Cartesian x and y axis grid lines of the core with currents of two different frequencies, and I chose 10 mhz and 10.5 mhz."

"The phrase "signal processing" was not used. But in fact, I was dealing with signals and determining what was happening with nonlinearity and mixing in the frequency domain. All these concepts were well understood."

"What was new was the way of putting this together to get information from a magnetic core in a memory plane. I was taking a graduate course called Sample Data Systems from Professor William Linvill. Today the same course might have two names."