78 THE BELL SYSTEM TECHNICAL JOURNAL, JANUARY 1957 



back the peak gain is proportional to the mutual conductance of the 

 three tubes. 



At frequencies well off resonance the impedance of the crystal is high 

 so that no reduction in feedback results. Periodic measurements of gain 

 at the resonant frequency relative to measurements made at a frequency 

 off resonance will show any changes in the tubes. The crystal frequency 

 is different for each repeater so that by measuring the gain from the 

 shore stations at the various crystal frequencies it is possible to monitor 

 the performance of the individual repeaters. 



The increase in gain at the peak is approximately 25 db. The crystal 

 frequencies, spaced at 100-cycle intervals, are placed above the normal 

 transmitted band between 167 and 173.4 kc. 



Thermal noise always present at the input to the repeater, is also am- 

 plified over the narrow band of frequencies corresponding to the peak 

 gain in each repeater so that at the receiving end of the line there are a 

 series of noise peaks, one for each repeater. Should a repeater fail, the 

 noise peaks of all repeaters between the faulty repeater and the receiving 

 end will be present and those from repeaters ahead will be missing. By 

 determining which peaks are missing the location of the failed repeater 

 can be determined. It is obvious that to locate a faulty repeater the power 

 circuit must be intact. To guard against power interruption owing to 

 an open electron-tube heater, a gas tube V4, Fig. 1, is connected across 

 the heater string as a bypass. 



Loop Feedback 



The design of the feedback loop follows conventional practice. The 

 restrictions that limit the amount of feedback that can be obtained in 

 the transmitted band are well known. ^ Broadly speaking, the figure of 

 merit of the electron tubes and the incidental circuit capacitances de- 

 termine the asymptotic cutoff which limits the amount of feedback that 

 can be obtained in the band. With the flexible repeater circuit, capaci- 

 tances are rather large because of the severe space restrictions and physi- 

 cal length of the structure. Transit time of 1.8° per megacj^cle per tube 

 and a like amount for the physical length of the feedback loop reduced 

 the available feedback by 2 db. 



Margins of 10 db at phase cross-over and 30° at gain cross-over were 

 set as design objectives. While these may seem to be ultraconservative 

 in view of the tight controls placed on components and the mechanical 

 assembly, it should be borne in mind that the repeaters are inaccessible 

 and repairs would be costly. 



Modulation and tube aging considerations require a minimum feed- 



