CALIBRATION FOR WIDE FREQUENCY RANGES 



67 



usual to measure the overall signal level and to ob- 

 tain a frequency analysis of the noise. Often the time 

 variation of the noise is of interest. This can be ob- 

 tained in the form of a time-level distribution or as 

 the crest factor of the noise. This latter is defined as 

 a ratio of the crest value to the effective value of the 

 quantity. 



The overall signal level is determined with a wide- 

 band measuring circuit. This circuit must carry the 

 highest noise peak without overloading and include 

 a square-law measuring device, which is the only type 

 that adds up the contributions of the various fre- 

 quencies in the signal in such a way as to give the rms 

 signal level. 



A frequency analysis can be obtained by sweeping 

 over the frequency range of the noise with a narrow 

 heterodyne band-pass filter. The design of this filter 

 must be carefully considered from the standpoint of 

 transient response. The requirements for the rest of 

 the system are the same as those discussed above for 

 measuring the overall signal. 



The crest factor can be measured by obtaining a 

 wide-band response with a square-law rectifier (ther- 

 mocouple) and by obtaining the peak response on an 

 oscilloscope. 



One of the difficulties in all of these measurements 

 is the requirement of using a square-law measuring 

 device. The only true square-law device is the ther- 

 mocouple, but this is so slow-acting that it can furnish 

 only a long time average and has a limited dynamic 

 range. Therefore, vacuum-tube detectors are usually 

 used. These are fast, but they follow the square law 

 only over a limited range of input levels. The ques- 



tion then arises as to whether or not the noise can be- 

 satisfactorily measured witli the particular rectifier 

 available. One method that has been used to answer 

 this question consists of measuring the crest factor 

 and determining whether or not the rectifier is still 

 square law for the highest peaks in the noise. While 

 this method is helpful, there is theoretically some 

 question as to whether it is a sufficient criterion. It is 

 therefore desirable to consider the problem from 

 other angles as well. Sometimes previous measure- 

 ments on similar noises which have been satisfactory 

 are available. At other times an assurance can be de- 

 veloped from a study of the data itself. 



The above requirements on the measuring system, 

 namely, that it (I) shall not overload on the highest 

 noise peaks, (2) shall adequately cover the frequency 

 range of the noise, and (3) shall have a square-law 

 rectifier, are sufficient to obtain a level measurement 

 and a frequency analysis of the noise. In some cases 

 it is desirable also to obtain a graph of t He- noise or to 

 view it on the oscilloscope. In such cases, it is neces- 

 sary (hat the phase relations between different fre- 

 quencies be maintained as well as the magnitudes. 

 This adds another requirement, namely, that the sys- 

 tem have its phase shift linear with frequency. 



The requirements of the pick-up device used in 

 the tests are similar to those on the circuits of the 

 system. If the hydrophone has a uniform response 

 over the frequency range of interest, which usually 

 implies a phase shift linear with frequency, it will be 

 satisfactory for use in these tests both for obtaining a 

 level measurement and also from the standpoint of 

 maintaining phase relations. 



