SELECTIVE FADING OF SUPER-HIGH FREQUENCY SIGNALS 1199 



These data have been applied practically to the design of frequency- 

 diversity measures for minimizing circuit outages caused by fading in 

 TD-2 radio relay systems. 



Fig. 9 shows an illustration of the practical use of the data. Curve H 

 (which is the same as Curve H on Fig. 8) shows the distribution with 

 time of the fading on a typical radio relay path without frequency 

 diversity measures. Curve J shows the distributions of fading if certain 

 specific frequency diversity measures are used (based solely on fading 

 considerations). The difference between Curves H and J shows the 

 improvement gained by using that specific kind of frequency diversity. 

 For example, Curve H shows that without frequency diversity fading 

 in excess of 30 db will occur 0.075 per cent of the time and fading in ex- 

 cess of 40 db will occur 0.02 per cent of the time. But, if the kind of fre- 



Table II 



quency diversity to which Curve / corresponds is used, fades deeper 

 than 30 db will occur only 0.0012 per cent of the time, and fades deeper 

 than 40 db will occui only 0.00008 per cent of the time. Thus the improve- 

 ment resulting from this type of frequency diversity is a reduction of 

 fades deeper than 30 db from 0.075 to 0.0012 per cent of the time, and 

 fades deeper than 40 db from 0.02 to 0.00008 per cent of the time. 



To explain how Curve J was derived, let us assume that during any 

 time when there is a fade of 30 db or more on the operating frequency of 

 a given channel the signal Avill be switched to another channel fre- 

 quency. Let us further assume that the alternate frequency will be sepa- 

 rated by 160, 240, 320, or 400 mc from the assumed operating frequency 

 and that the choice of alternate frequency is random. If we also assume 

 that the conditions of August, 1950, on the Iowa path prevail, we will 

 find from Curve H on Fig. 8 that the assumed operating frequency will 

 have a fade of 30 db or more 0.075 per cent of the time. Then a new dis- 

 tribution curve can be prepared, based on (1) Curve B (on Fig. 7) for 



