DISTORTION CORRECTION 467 



This is generally a difficult problem and it becomes necessary to resort 

 to simplifying methods somewhat similar to those employed in the 

 design of electric wave-filters. One reason for this difficulty is that 

 we are limited to physical resistance, inductance, and capacity ele- 

 ments, all of which must, in general, be positive. We would, therefore, 

 begin with known forms of networks whose general propagation 

 characteristics have been determined and choose from them one or 

 more whose combination offers the possibility of giving a satisfactory 

 desired result. A number of points which are applicable in the general 

 case may be noted as follows: 



1. First, determine the desired propagation characteristics of the 

 distortion correcting network corresponding to formula (8). 



2. If necessary, divide this propagation characteristic into several 

 parts each of which has the approximate characteristic belonging to a 

 known network structure. 



3. Assume one of these networks physically capable of having such 

 an alloted characteristic and attempt a design to approximately 

 fit it according to one of the methods of Section 2.7. Where there is 

 an attenuation characteristic, Method 1 is usually best, as attenuation 

 is generally of more importance than phase and hence its simulation 

 requires greater accuracy. The network will introduce a phase con- 

 stant which will necessarily have to be taken into account. Of the 

 two or more possible solutions for the lattice type network, the one 

 with the most desirable phase constant would obviously be chosen and 

 in some cases this may be close to requirements. Another reason for 

 usually following this order of simulating the attenuation first and 

 the resultant phase later is furnished as a consequence of Theorems 1 

 and II of Section 2.2. From them we see the physical possibility of 

 introducing certain phase characteristics without attenuation (ideally), 

 but not varying attenuation characteristics without phase. Method 3 

 imposes a rather severe requirement on a single network. 



4. If the network design comes out to be unphysical with the 

 particular characteristic values assumed, small variations from these 

 values should be tried, since the natural varying curvatures in the 

 propagation characteristic of the network must sometimes be allowed 

 for. Otherwise, a different kind of network should be used, or a 

 composite one, which has a similar characteristic. 



5. In designing successive sections of the complete transducer, the 

 effects of previous parts must be considered. 



To facilitate the application of this method of distortion correction, 

 general propagation characteristics together with formulae have been 

 derived for a representative number of lattice type structures. These 



