Thus 



The solutions for A:i , A: 2 and kz take the form 



kn = a5i + 652 + c53, (11) 



where a, h and c depend solely on the shapes F„»(/). 



Thus the values of the 5's may be decoded into the values of the equiva- 

 lent fc's by the simple process of multiplying each "5" by some fraction 

 that is a function of the equalizer shapes; or, more precisely, by a frac- 

 tion that is a function of the values of the various shapes at the fre- 

 quencies /i , /2 , etc. 



The circuit of the computer is quite simple; consisting of about A''^ 

 resistors for the control of "iV" networks by the deviations that are 

 measured at M = A^ pilot frequencies. This simplicity is valuable for 

 its own sake, but, as previously noted, there is considerable value in 

 the fact that substitution of new equalizer shapes requires only that 

 changes be made in the values of some of the resistors. 



Fig. 2 illustrates the principle by showing the computer circuit re- 

 quired for the previous example of three shapes for which information 

 is given at three frequencies, /i , /2 and /a . The three dc voltages repre- 

 senting the deviations 5i , 52 and 53 are decoded by simply cross-connect- 

 ing them to the three pairs of output terminals through fixed resistors 

 chosen to satisfy the relationships of equations (8), (9) and (10). The 

 output voltages will be proportional to the desired equaUzer correction 

 quantities fci , k2 and ^3 . 



In general the calculation of some of these resistors will give negative 

 values. Thus it will generally be necessary to require that the dc voltages 

 representing the errors 5i , 52 , etc., be available in both polarities. Al- 

 ternately, the errors can be provided in only one polarity and the cir- 

 cuits to which the computer outputs connect can provide the push-pull 

 circuit. This latter course has been used in the L3 regulators as indicated 

 on Fig. 2. 



The effect of using the computer is as follows. If one pilot changes, 

 all regulating networks correct but in such proportions and polarities 

 as to produce no gain change at any pilot frequency except at the one 

 originally disturbed. If all pilots deviate in proportions corresponding 



[ 



