17 

 NOISE 



It is not possible to record an indefinitely small signal by the employment of 

 indefinitely large amounts of ampHfication. This is because there are generated 

 within the source of the signal itself, and within the amphfying apparatus, 

 random electrical effects of a definite statistical magnitude. If the effect 

 produced at the indicating device, meter, cathode ray tube, etc., by this 

 random activity is of such a size as to obscure the effect produced by the 

 signal, the latter will at least require special measurement techniques, and 

 may be impossible to measure at all. 



This electrical disturbance is called 'noise', a most apt description for it 

 when heard over a loudspeaker: the sound produced may be described as 

 'rushing'. Because of its random nature nothing can be said about its 

 instantaneous amplitude or frequency. It is, however, possible to allot it a 

 root mean square amplitude and a frequency spectrum, which is the frequency 

 spectrum or band-width of the recording gear. 



The classical causes of noise are 'resistance noise' and 'valve noise'. The 

 former is due to the mechanical oscillations of the molecules of which the 

 resistance is composed ; the latter is due to the inhomogeneous nature of the 

 electron stream. When one tries to frame a definition in general terms, such 

 as 'noise is produced wherever there is molecular agitation, or where a 

 flow of current is of an irregular nature', a great many other deficiencies of 

 recording gear become difficult to exclude — for example, who can say that 

 the current from a dying HT battery or the supply mains at winter rush hour 

 is not irregular? In this way the concept of 'noise' has tended to broaden, so 

 that in some quarters 'noise' is held to be 'anything which is not signal'. 

 We shall confine the term noise to its narrow sense, and where other spurious 

 effects have a name we shall use it. 



HOMOGENEOUS RESISTANCE NOISE 

 (OTHERWISE THERMAL OR JOHNSON NOISE) 



The R.M.S. noise power delivered into a matched load by any homogeneous 

 piece of material between two terminals is given by 



kTB 



Power delivered 



I 



I 



.JL. 



kTB T" 



Figure 17.1 



where k is Boltzmann's constant, T is the absolute temperature and B is the 

 band-width of the device in c/s —the upper frequency in this case being 

 limited by the shunting effect of stray capacitance {Figure 17.1). It is worth 



253 



