STUDIES IN RADIO BROADCAST TRANSMISSION 169 



The character of this skewing, then, and its magnitude depend upon 

 the rate at which the object being photographed in vertical sections 

 moves, and the direction of the movement. 



In Fig. 18 is shown an imaginary night-time band fading record in 

 the "assembled" form. Since such a record contains frequency as a 

 third dimension, in addition to amplitude and time as shown in the 

 ordinary fading record, our simple fading curve has assumed the 

 broader aspect of a surface, the selective fading making more or less 

 parallel "valleys" running across it. The step-frequency system of 

 recording the points amounts to photographing sections of this solid. 

 The important point to be kept in mind is that these sections are 

 not perpendicular to the time axis. If they were, the skewing previ- 

 ously described would not be present. By setting these sections up in 

 their true relation to the time axis, however, and filling in to produce 

 a continuous surface such as is shown in Fig. 18 the result is correctly 

 represented. In order to make a detailed and accurate study of the 

 band fading records, therefore, it is desirable to construct from the 

 oscillograph sections the complete surface by the method suggested. 



In Fig. 18 the trace of minima crossing the band is shown by M, 

 M' and M". Picture sections obtained as our recording apparatus 

 literally moves back and forth across this frequency band are shown 

 as (a-b-c-d), ih-c-a'), {a'-h'~c'), etc. It will be evident that the 

 section Pi, for example, will, in case a minimum is crossing rapidly, 

 appear entirely unrelated to section P2. When the minima run 

 nearly parallel to the time axis (slow changes in transmission condi- 

 tions) the successive pictures Pi, P-i, P3, etc., will reveal their rela- 

 tion by direct comparison. 



Actually to obtain frequency-amplitude sections perpendicular 

 to the time axis in Fig. 18 would require the simultaneous trans- 

 mission and reception of a large number of frequencies spaced at 

 short intervals along the frequency axis. A more practical thought 

 is to speed up the process and though this seems very simple at first 

 consideration, it will be shown later to involve a particular kind of 

 distortion which cannot be separated out as easily as the skewing 

 encountered by the more deliberated method. 



Now that we are familiar with the data, Fig. 19 showing, partially 

 superimposed in vertical strips, the outlines of successive built-up 

 pictures of the frequency traverse will be of greater significance. 

 During the steady periods there appears within the 2,280-cycle 

 band covered by these data approximately one complete cycle of 

 selective fading. The lack of flatness in the audio-frequency-trans- 

 mission characteristic of terminal apparatus has caused the suppression 



