722 THE BELL SYSTEM TECHNICAL JOURNAL, MAY 1956 



to give access to the emergency drum memory, or to a "foreign area" 

 memory where such extra memory capacity is necessary. 



Let us now return to the discussion of Fig. 4 and consider the assign- 

 ment of the translation information to the drum surface where it is stored. 

 Recall that the drum surface is effectively divided into a grid by the co- 

 ordinates of tracks, each passing under an individual write-read magnetic 

 head, and "slots," each defined by the appearance of a timing pulse in a 

 rhythmic train synchronized from the drum itself, and that the "cells," 

 at the coordinate intersections, each accommodate one bit of code infor- 

 mation. 



Since each card in the card translator accommodates 38 bits of input | 

 code and 116 bits of output, about 160 cells, divided in the ratio of one 

 cell for input to every three cells for output, must be assigned to each i 

 translation item. One simple and direct assignment would be to place 

 the entire translation item in a single slot composed of 160 cells. With i 

 this layout the slot containing the desired translation would be identi- 

 fied by reading, or "matching" the input code, and during this same in- 1 

 terval the output information in the same slot would be gated-out to the I 

 translation delivery circuits. A 1 ,000-translation drum would then be' 

 long and narrow, and far too many reading amplifiers would be required. ' 

 Another evident arrangement would be to assign the entire input code: 

 to the first of each group of four slots proceeding under the heads, with 

 the output code following in the next three slots. Such an allocation 

 would require only 40 reading amplifiers but the drum necessary for the 

 desired capacity, with the cell-spacing chosen, would have been larger' 

 in diameter than the mechanical designers cared to undertake in their 

 first trial. A logical choice, therefore, was to place each translation item 

 in a pair of adjacent slots, and this was done, although it was later recog- 

 nized that other, more sophisticated, arrangements might offer eertainj 

 advantages. 



In Fig. 4, the apparent location of one translation item is sketched inj 

 relation to the drum surface. This sketch is not drawn to scale, since thef 

 slot width is actually only 0.020 inch, and the track width is comparable. 

 It is also geographically inaccurate; actually the cells of any one slot arei 

 positioned in four quadrants on the drum, the associated heads being! 

 positioned in four stacks for mechanical reasons. However, all of the 

 cells in a time slot pass under all of the heads at the same instant and the| 

 presentation of Fig. 4 was adopted for the sake of clarity. 



Note, then, that the input code and one-third of the output code arel 

 recorded in the first or a slot of a slot-pair passing undcn- the reading^ 

 heads, and that the remaining two-thirds of the output code occupies 



