108 THE BELL SYSTEM TECHNICAL JOURNAL, JANUARY 1951 



In the design of the El repeater it is desirable for the network to have 

 control of the impedance presented at the Hne terminals over the voice 

 frequency range (/2 and /a of Fig. 11). To accomplish this all impedances 

 shunting the converter circuit are made as large as possible and all impe- 

 dances in series are made as small as possible at these frequencies. Further- 

 more, Ml is made as close to fi2 as practical, and large enough so that the 

 factor (ni — 1)/(m2 + 1) is made as close to unity as possible. If this factor 

 approximates unity and if the series term Rp/{1 + 112) is relatively small 

 with respect to impedances between which the converter is operating, it 

 can be seen from the equations that battery supply variations and tube 

 changes should have little effect upon the negative impedance presented 

 by the converter at these frequencies. 



The retard coil and transformer inductances should be considered at low 

 frequencies as a possible source of instability. At low frequencies the in- 

 ductance of the retard coil shunting terminals 3 and 4, and the inductance 

 of the transformer shunting terminals 1 and 2 materially affect the im- 

 pedances facing the ideal converter. 



Another important consideration in the design is the ratio of the line 

 transformer. The ratio must be such that the tubes will operate efficiently 

 with the normal network, line load, and plate supply voltages. This part of 

 the design follows conventional methods. 



The distributed capacitances of the line transformer windings should be 

 taken into account. In general, these are large compared to the tube capac- 

 itances between cathodes and between ground and each cathode. These 

 interelectrode capacitances can be neglected at voice frequencies, but at 

 the higher frequencies all of them should be considered, and conventional 

 methods of suppression of parasitic oscillations should be applied. Also, the 

 distributed capacitance of the retard coil must be considered from the 

 standpoint of stability at the higher frequencies. 



Conclusion 



A vacuum tube arrangement for producing negative impedance can be 

 represented by an equivalent four-terminal network consisting of an ideal 

 converter having a ratio of transformation of —k'A and two positive im- 

 pedance networks located one on each side of the ideal converter. In these 

 two networks some of the elements can be cancelled in effect by balancing 

 those in one network against corresponding elements in the other. Elements 

 which cannot be balanced can be made either relatively large or relatively 

 small compared to the impedances between which the converter is designed 

 to operate. Across one pair of the four terminals of a practical converter can 

 be seen a reverse voltage type of negative impedance; across the other pair 



