PRODUCTION OF TELEVISION SIGNALS 



585 



systems such as have been used for still pictures and television has 

 proven a useful index of phase characteristics. 



After a preliminary estimate from experience with still pictures that 

 the limit on db/doo should be dz 10 microseconds, an electrical network 

 consisting of five sections of a simple lattice structure was used for 

 testing the effect of phase distortion with television apparatus. This 

 network introduced negligible amplitude distortion and a drift in the 

 value of db/dco of 50 microseconds over the frequency range of to 

 20,000 cycles. Its effect was perceptible in blurring the image of a face 

 and it decidedly affected a sharp pattern of two parallel lines of such 

 width and spacing as to be just within the resolving power of the 

 apparatus. This variation of db/dw was about 23^2 times greater than 

 that postulated. Hence ±10 microseconds was agreed on as a 

 desirable limit for db/dco, though it was felt that this limit might be 

 exceeded by a factor of two in restricted parts of the frequency band. 



When this network was combined with a filter the slope of whose 

 envelope delay curve was in the opposite direction so that over the 

 greater part of the frequency range the combined delay of the two 





f2 



TIME 



Fig. 20 — Signal details of concentrated frequency spectrum for illustrating the 



effect of envelope delay 



circuits was constant and equal to 140 microseconds, this time delay 

 effect was very graphically brought out. Every time the combined 

 circuit was cut in, the undistorted received image jumped to a new 

 position a little over 10 per cent of the width of the picture to one side 

 in the direction of scanning. 



To see why dbjdo: should be maintained at a constant value, consider 

 two sharply defined details near together in the picture which would 

 produce a variation in signal intensity with time as indicated in Fig. 20. 

 Imagine each to be cyclically continued so that the small detail defines 

 a frequency /i and the other defines a frequency fi. It is then known 

 from Fourier analysis that the frequency spectra of the two details 

 are chiefly concentrated around the frequencies /i and /o. If dbjdw is 

 appreciably different at the frequencies /i and /o for any part of the 



