250 THE BELL SYSTEM TECHNICAL JOURNAL, JANUARY 1957 



However, at 20 kc the total noise is some 8 db below the specification 

 limit, and therefore again several amplifiers could fail on one path be- 

 fore the noise exceeded the specification limit. Actually it was discovered 

 that the predominant source of third-order intermodulation on the re- 

 peater was in the nickel-iron/ceramic seals on high-voltage capacitors 

 and followed a square law with input levels. 



From a more detailed examination of the factors briefly mentioned 

 above it was decided that the initial line-up should be based on a nominal 

 flat —3.5 dbr point at the amplifier output and the final working levels 

 decided upon as the results of tests on the completed link. 



With equal loading on the grid of the output tube at all frequencies 

 the worst signal/noise ratio exists at 552 kc; some pre-emphasis of the 

 transmit signal should therefore prove to be beneficial. In fact, after 

 completing the tests on the link it was decided to improve the margin 

 on noise by raising the level at 552 kc by 2 db, thus giving a sloping level 

 response at the amplifier output in the high-frequency band. To main- 

 tain the same total power loading, the low-frequency band levels were 

 decreased bj^ 1 db, still retaining a flat response. 



Laying 



It was proposed to use laying methods with continuous testing similar 

 to those employed successfully on the Anglo-Norwegian project. The 

 complete link with a temporary splice at the equalizer would be assem- 

 bled and tested on board H.M.T.S. Monarch and laying would proceed 

 from Terrenceville to Sydney Mines in the high-frequency direction of 

 transmission. A detailed description of the actual laying operation is 

 given elsewhere.^ After completion of tests on the submarine section the 

 land section to Clarenville would be connected with appropriate equaliza- 

 tion at Terrenceville. 



DESIGN OF ELECTRICAL UNIT OF SUBMERGED REPEATER 



General 



The equipment is contained in a hermetically sealed brass cylinder 

 (filled with dry nitrogen) 7| inches in diameter and 50 inches long, which 

 is bolted at one end to one of the bulkheads of the housing. A flexible 

 coaxial cable emerges through an 0-ring seal at each end, and these are 

 ultimately jointed to the cable glands. The various imits forming the 

 complete electrical unit are mounted within a framework of Perspex 

 (polymethylmethacrylate) bars w^hich forms the main insulation of the 



