134 DRS. GUY BARLOW AND H. B. KEENE ON THE ANALYSIS OF SOUND. 
nmmm- 
vmm/imz 
Fig. 
wmmx- -mm, 
intervals shown in fig. 2 it has been found possible to eliminate the subharmonics \n, 
Ttn , Ytn, &c., in addition to the even orders. The 
fundamental response is reduced to three-fourths 
of its usual value, and there are certain other 
disadvantages suggested' by the few experiments 
which have so far been made. 
In order to analyse by this method of periodic interruption mechanical vibration of 
a solid body or of sound waves in air or water, the vibration must be converted into an 
electrical current which in wave-form faithfully represents the original motion. Some 
-distortion of the wave-form would not be a serious objection, provided it followed a 
simple relation allowing correction to be made. Actually very few methods of con¬ 
verting vibration into current are available, and none of these is free from objection. 
Among the most practicable are —• 
(1) Variation of electrical resistance by pressure, e.g. carbon microphone. 
(2) Variation of electrical resistance by change in thermal conditions, e.g. Tucker 
Hot Wire Microphone. 
(3) Electromotive force generated by induction, magnetophones, &c. 
It may be pointed out that all these methods depend on induction (assuming a trans¬ 
former is used in (1) and (2)), and the final current therefore represents the velocity of 
the vibration under investigation, but this is not an objection from the point of view 
of analysis. 
For, suppose the original vibration is resolved into simple harmonic components— 
<h sin (2 irn v t -f oti) + a 2 sin (2 7 rn 2 t + a 2 ) + &c. ; 
then, assuming no other form of distortion, the current will be proportional to 
2 x ^1 n x cos (2t rn x b -f a x ) -|- 2tt a 2 n 2 cos (2t rn 2 t -f- a 2 ) -f &e. 
The analysis of this current will then give correctly the frequencies of all the com¬ 
ponent vibrations, but in each case the amplitude is magnified in proportion to the corre¬ 
sponding frequency. The product 2 nan, representing the maximum velocity, is itself 
an appropriate measure of the importance of the component, as the relative energies 
for different components are proportional to (an) 2 . 
In the present experiments the determination of the frequencies of the components 
has been effected with all the accuracy desired, but as it has not yet been found possible 
to avoid selective action due to resonance of diaphragms, the amplitudes of the com¬ 
ponents are not faithfully represented. No attempt has been made to deduce the 
absolute amplitudes of motion of the original vibration. 
When the components of a vibration have strictly commensurable frequencies, as 
in a harmonic series, the phase relations of the components are quite definite, and the 
determination of the relative phases might be of value—in fact it would be necessary 
