The MIT Geophysical Analysis Group (GAG) from inception to 1954

Abstract

The beginning of digital signal processing took place in the years 1950 to 1954. Using an econometric model, E. A. Robinson in 1951 came up with the method of deconvolution, which he tested on 32 seismic traces. Norbert Wiener, George Wadsworth, Paul Samuelson, and Robert Solow were his advisors. On the basis of this work, the MIT president's office in 1952 set up and sponsored the Geophysical Analysis Group (GAG) in the Department of Geology and Geophysics. GAG was made up of graduate students doing research in digital signal processing. In 1953, a consortium of oil and geophysical companies took over the sponsorship. At first, GAG used the MIT Whirlwind digital computer. In order to do the larger amount of computing required by the consortium, the Computer Service Section of Raytheon Manufacturing Company was enlisted in 1953. The Raytheon people who played key roles were Richard Clippinger, Bernard Dimsdale, and Joseph H. Levin, all of whom had worked on ENIAC, the world's first electronic digital computer. As originally built, ENIAC did not use programs stored in memory as does a modern computer; instead, the programming was done by rewiring the physical components for each new problem. In 1948, Clippinger was responsible for converting ENIAC into the world's first operational stored-program computer. ENIAC had 20 accumulators but no other random access memory (RAM). The programs were stored in the function tables, which acted as programmable read-only memory(PROM). For GAG work in 1953, Raytheon used the British Ferranti Mark 1 computer (which was the commercial version of the Manchester Mark 1 computer, for which Alan Turing played a key role). This computer was installed at the University of Toronto to help in the design of the St. Lawrence Seaway. Raytheon was plagued by frequent breakdowns of the computer but still produced several hundred seismic deconvolutions for the summer GAG meeting in 1953. The consortium was pleased with the geophysical results but was disheartened by the unreliability of the current state of digital technology. As a result, GAG was directed to find analog ways to do deconvolution. Instead, GAG found that all of the analog methods, and in particular, electric frequency filtering, could be done by digital signal processing. In fact, the digital way provided greater accuracy than the analog way. At the spring meeting in 1954, GAG proposed that all analog processing be thrown out and replaced by digital signal processing. Raytheon was at the meeting and offered to obtain or build all the elements required for digital signal processing, from input to output. The conversion to digital was not done at the time. However, that step did happen in the early 1960s, and exploration geophysics has the distinction of being the first science to experience a total digital revolution. Digital processing today provides seismic images of the interior of the Earth so startling that they compare to images of the stars made by the Hubble telescope. (In fact, the digital method of deconvolution first developed in geophysics made possible the digital correction of the lens of the Hubble telescope.)