NOISE 



frequencies ; the flicker noise voltage in a small band bco has been shown by 

 Macfarlane^ to be proportional to (I I coy where x lies between -| and 1. It is 

 also proportional to (4)^ where j is between 1 and 1-5, 4 being the cathode 

 current of the valve. 



Microphony 



Microphony is a modulation of valve anode current caused by small 

 variations in the geometry of the valve structure. These variations may be 

 produced by the valve itself by 'settling' of its structure as it warms up, 

 or may be caused by vibrations from an external source — usually the experi- 

 menter — being communicated to the valve via its mounting arrangements 

 or through the air. Microphony is combated by keeping the anode current 

 low, choosing valves specifically made for non-microphonicity (by giving them 

 a specially braced structure), keeping still and, when all else fails, resilient 

 mounting and enclosure in a wadded container. 



Hum and Drift — These will be dealt with in Part IV. 



ORDERS OF MAGNITUDE 



It is not hard to see that as the noise generated in the low-level part of an 

 amplifier receives the most subsequent amplification, efforts to reduce noise 

 must primarily be directed at the first amplifying stage and, where there is 

 any freedom of choice, the signal source itself. It is clear from the expressions 

 that have been given that valve noise may be minimized by low heater voltage 

 (shot noise) and low anode and screen voltages (partition noise, flicker, 

 microphony) and by correct choice of valve (triode if possible, otherwise 

 low ratio of screen to anode current, and of anti-microphonic construction). 



Whether or not valve noise is important depends on the signal source. 

 If the latter is an E.E.G. subject of internal resistance 10,000 ohms, and the 

 band-width of the apparatus is 100 cycles, as it well might be, then at 24°C the 

 R.M.S. noise voltage produced by the patient is 0-128 microvolt, and if the 

 first stage gain is 30 the effect produced at the anode is 30 times greater, i.e. 

 3-84 microvolts. Correcting the valve noise in the examples given for the 

 reduced band-width, we get a triode noise of about 0-3 microvolt and a 

 pentode noise of about 0-7 microvolt, plus the effects of flicker. Valve noise 

 is comparable in magnitude to the Johnson noise and worthy of attention. 



On the other hand, suppose one is recording from a 10 megohm glass micro 

 electrode over a band of 10 kc/s. Then the source noise cannot be less than 

 40 microvolts at 24°C at the first amplifying stage grid, or 1,200 microvolts 

 at the anode. This is clearly enormously greater than valve noise over the 

 same band, leaving one more free to choose the valve with other points in 

 mind. 



NEGATIVE FEEDBACK AND NOISE 



The effect of negative feedback applied to an amplifier is to reduce the noise 

 from all causes. Thus in Figure 17.2 we have an amplifier of gain A, the noise 

 content of whose output is supposed to be generated by the external generator 

 NG; n is the instantaneous noise voltage. Suppose now a fraction B be 



256 



