256 HOWARD J. CURTIS 



stage of the amplifier. The next is to keep the plate potential and 

 plate current as low as possible. If carbon potentiometers are used 

 in the first stages they should be kept very clean with an organic 

 solvent, while wire-wound potentiometers should be oiled on the rub- 

 bing surfaces with petroleum jelly. The tubes should be insulated 

 from noise and vibration, although this precaution is necessary only 

 in extreme cases with modern tubes. 



When these precautions are taken, and using a type 12SJ7 tube 

 as a triode, the noise in the plate circuit of the tube is roughly of the 

 same magnitude as is caused by an applied potential in the grid cir- 

 cuit of 3 /xv. Thus we speak of the equivalent noise level as being 

 3 (XV. It is obvious from this that if a potential of less than 3 ^uv. were 

 applied to the grid of the tube, it would not be possible to detect it, 

 since it would be lost in the noise of the tube. If additional stages of 

 amplification are added, they amplify the noise as much as the signal, 

 so nothing is gained. The situation is analogous to that of empty 

 magnification with a microscope — where an object possesses struc- 

 ture too small to be resolved by the objective, no amount of magnifica- 

 tion will help. 



This limitation can be partially overcome in special cases. If 

 alternating current of a special frequency is amplified by an amplifier 

 tuned to that particular frequency, only that fraction of the noise 

 having that particular frequency will come through and the equiva- 

 lent noise level can be drastically reduced. Again, if only direct cur- 

 rents need be amplified, it is convenient to put a capacitor across the 

 output to short-circuit the high frequency components of the noise. 



Another troublesome feature of direct current amplifiers is drift. 

 As indicated above, this can be minimized by a number of precau- 

 tions but never eliminated. A careful selection of tubes in the first 

 stage of a balanced amplifier is probably the most important single 

 factor. It will usually help, in a balanced amplifier, if there is a 

 separate control on the filament current of each tube, and a value for 

 each tube can usually be found by trial that will very considerably 

 reduce the drift. It is not difficult to construct a d.c. amplifier in 

 which drift is not troublesome provided potentials of a few hundred 

 microvolts are to be measured. If the potentials are only slightly 

 larger than the noise level, it is quite difficult to construct an amplifier 

 in which drift is not ver3^ troublesome. 



It should be realized that there are certain situations in which a 

 galvanometer can be used more a(:lvantageously than an amplifier. 



