obtained from a seven minute record are not very reliable. 
Wave analysis 
The wave analyzer constructed by Klebba [1946, 1949] and the 
one used by Deacon [1949] find some function related to [a(H) 1°, 
possibly its square root. Tukey and Hamming [1949] give some de- 
Sign criteria for such devices. One extremely important point is 
that they can be just as much in error as the numerical method 
estimates for low values of f, and if the tuned circuit in the de- 
vice is too sharply peaked they can be even more in error. The 
modifications which could improve the design of Klebba's instrument 
will be described in a later chapter. They are not too difficult 
to make. Even as it is now constructed it is a very valuable instru- 
ment, and it can be made even more valuable by these modifications. 
Very Slowly varying 
Equation (10.39) gives the number of degrees of freedom per 
elemental band of the derived power spectrum if_the spectrum is 
very Slowly varying over that band. If the spectrum varies rapidly, 
then the true number of degrees of freedom is less. For additional 
details see Tukey and Hamming [1949]. The considerations given 
above serve as design criteria and the real errors may be some- 
what worse than are indicated by the table. 
The work involved 
ee 
Tukey and Hamming [1949] have estimated that (m + 1)(N + 2 + 2) 
Nl 
additions, (m+ 1)(N + 5 + 4) + 2 multiplications, and m+ 3 
divisions have to be made to carry out one numerical analysis with- 
out checks for accuracy. Simplifications of procedure are given 
in the above reference. With a desk calculator, and with N equal 
- 286 - 
