584 BELL SYSTEM TECHNICAL JOURNAL 



reproduction of the image, the matter of the second and third trans- 

 mission requirements on received ampHtude and phase characteristics 

 over the frequency scale is one which had to be decided largely on the 

 basis of the experimental results and judgment based on general 

 considerations. We have already seen that the removal of the very 

 lowest frequencies simply changes the tone value of the whole picture. 

 It may be similarly reasoned that departures from the average 

 efficiency of transmission in the lower part of the frequency range 

 would result in the appearance of diffuse shadows or high lights. 

 Likewise, it may be concluded that broad deviations from the average 

 efficiency of transmission in the uppermost part of the signal frequency 

 range would result in the accentuation or the fading out of the finer 

 detail of the scene. Steep slopes in the amplitude-frequency curve 

 would result in the superposition of oscillations upon signals repre- 

 senting sudden changes in intensity. To reduce these effects every 

 reasonable effort was made to keep the variations in the amplitude 

 characteristic with frequency as slight as possible, aiming to hold 

 these characteristics for the separate parts of the demonstration 

 system to within ± 2 T U or better. 



In addition to transmitting the component frequencies with the 

 same relative efficiency as regards amplitude, it is also particularly 

 essential in television to send them through the system with small 

 relative phase shifts; that is, with constant velocity or what is equiva- 

 lent, a phase shift proportional to frequency. It has long been known 

 in optical theory that the envelope of a group of waves of nearly the 

 same wave-length and nearly the same frequency may travel along 

 with a "group velocity" somewhat different from the phase velocities 

 of the component elements. If the system has but small departures 

 from a flat amplitude-frequency characteristic and from a linear 

 phase shift frequency characteristic, it can be shown that the time of 

 group transmission or "envelope delay" is given by db/dco^, the slope 

 of the curve obtained by plotting the phase shift, b, for the system, 

 against the angular velocity, co = lirf. The time of transmission of 

 a crest for any sine wave component of frequency w/Itt is, of course, 

 given by b/w. U b = ceo, bjoi =*c and dbjcLw = c. Then the phase 

 and envelope times of transmission are equal and all frequencies as 

 well as their group envelopes get over in the same time. If b is given 

 in radians, db/dco is given in seconds. In general a knowledge of b 

 as a function of co is necessary and sufficient to determine the phase 

 distortion. A knowledge of dbfdco as a function of co is not sufficient 

 to determine all factors in signal distortion. It is, however, often 

 easier to measure with the needed accuracy and in transmission 



