738 THE BELL SYSTEM TECHNICAL JOURNAL, JULY 1952 



from the digrammer will be more peaked toward zero amplitude than 

 that of the monogrammer. This is illustrated by the signals in the figures. 

 At the receiver the same type of device, but with an inverse mask can be 

 used to convert the signal back to its original form. 



The digrammer can, with a little assistance, supply all the data re- 

 quired to prepare the encoding mask. If typical signals from the message 

 source are applied to the cathode ray tube (without mask) for a long 

 time, and a time exposure is made of the face of the tube, a lattice of 

 spots will be obtained on the film. These spots will be dense where the 

 high probabihty combinations occur and less dense elsewhere. The order 

 of decreasing density in each column is noted, and the filter transmis- 

 sions are arranged in the same order. 



It is, of course, not necessary to use a phosphor, optical filters, and a 

 photocell. An array of targets each of which connects to the appropriate 

 tap on a load resistor might be simpler and more efficient. The cathode 

 ray tube itself can be replaced with an appropriate diode switching net- 

 work. Relay networks could be used for low-speed operation. 



At the digrammer level we run out of new dimensions to use in the 

 cathode ray tube. The principle can, however, be extended to trigram- 

 ming and general n-gramming. For example, tetragramming could be 

 accomplished by using a bank of t digrammers all in parallel, and all 

 deflected by the present and previous samples. Only one of these tubes 

 would be turned on at a time however. Which one this was would depend 

 on the other two previous symbols of the tetragram. These (by addi- 

 tional delays) would be applied to the deflecting plates of a master s\\dtch- 

 ing tube having an array of target plates in place of a mask. Depending 

 on the particular combination of signal samples applied to this tube, 

 the beam would strike a particular target. The target current would then 

 be used to turn on the beam of a particular digrammer tube, namely the 

 one with the proper mask for that particular combination of two past 

 symbols. 



The complete array of equipment is admittedly rather staggering, but 

 then, rather efficient coding should result. In practice it would probably 

 be found that the masks of many of the tubes would be so similar that 

 little gain resulted from differentiating between them. That is, the state 

 of the message source might be nearly equivalent for several past com- 

 binations. In these cases, the group of tubes could be replaced with one 

 having the best average mask, and the corresponding targets on the 

 switching tube then tied together. This compromise would be particu- 

 larly warranted for those tubes which were rarely used anyway. By these 



