REPEATER DESIGN — NEWFOUNDLAND-NOVA SCOTIA LINKS 249 



the equalization at the 552-kc point could be largely corrected for laying 

 and temperature-coefficient errors. It is not, in practice, easy to separate 

 these two factors. 



(d) Repeater characteristic. — The repeater was designed to equalize 

 the original cable-attenuation characteristic to ±0.2 db, as this was 

 possible with a reasonable number of components. This ^-ariation ap- 

 peared as a roll in the gain/frequency characteristic, which was expected 

 to be systematic and would therefore lead to a ±3 db roll in the overall 

 response. It was proposed that equalization for this should be provided 

 at the receive terminal. Manufacturing tolerances were expected to be 

 small and random. 



(e) Repeater interaction. — At the lower frecjuencies where the loss of 

 a repeater section is comparatively small, a roll in the overall frequency 

 response will arise due to changes in the interaction loss between re- 

 peaters. The design aimed at providing a loop loss greater than 50 db 

 which would reduce rolls to less than ±0.03 db per repeater section and 

 therefore to about 0.5 db at 20 kc with systematic addition on the whole 

 route. 



Planning of Levels 



From a critical examination of all these variables it was concluded that 

 the repeater should be designed to have an overload margin of 4 db above 

 the nominal mean annual temperature condition. It was also desirable 

 for the system to be able to operate within its noise allowance if one 

 path of a twin amplifier failed. Tests on a model amplifier gave overload 

 values of +24 dbm and +19 dbm for two- and one-path operation, 

 respectively, so that with a single-tone overload requirement of 18 

 dbm"* at a zero-level point, the maximum channel level at the amplifier 

 output would be —3 dbr for a single amplifying path. 



Thermal-noise considerations (i.e. resistance plus tube noise) fixed 

 the minimum channel level at the repeater input at — 69 dbr in order to 

 meet the allowable S3^stem noise limit of +28 dba at a zero-level point. 

 At 552 kc the amplifier gain is 65 db, so that the minimum level at the 

 amplifier output is —4 dbr. A system slope of ±4 db due to temperature 

 variations, corrected by similar networks at the transmit and receive 

 terminal, would, however, degrade the noise by 0.5 db. Intermodulation 

 noise was estimated^ on an average busy -hour basis, and it was concluded 

 that the increase in noise at 552 kc from this source was negligible — 

 less than 1 db, even with several repeaters in which the amplifier had 

 failed on one path. At lower frequencies the contribution from inter- 

 modulation noise is greater, and at 20 kc it exceeds resistance noise. 



