variations in the course of a test, the data set was considered to be o£ 

 unsatisfactory quality. Variations in W(a) or W(B) over the 360° 

 range of orientation angle were always much larger than the variation in 

 W. Some tests required 4 hours to obtain a complete data set, but meas- 

 urements at the beginning and end were usually the same for the same 

 pile orientation. Measurement drift in time was a common problem only 

 with the CERC strip gages (d and e in Fig. A-2); this was evidently due 

 to soaking of the adhesive attaching the strips to the pile face (see 

 Fig. A-3). Thus, 'temporal wave variability had insignificant effects on 

 the 96-foot tank data. 



However, spatial wave variability was marked in the 96-foot tank. 

 Figure B-2 shows the waveform at seven locations along the tank for the 

 highest wave at T = 1.55, 2.32, and 3.10 seconds; measured wave dimen- 

 sions are listed in Table B-1. In each case, there are notable changes 

 in crest height and curvature and in the secondary details of the wave- 

 form; the wave must be regarded as transforming as it propagates in con- 

 stant water depth. Two possible causes of this are nonlinear wave propa- 

 gation effects and departures from the ideal in the wave tank (e.g., 

 variations in tank cross section). The 96-foot tank had considerable 

 imperfections but these were not precisely documented. The tank was 

 dismantled after the tests, so this possible cause of spatial wave varia- 

 bility cannot be evaluated. 



Figures B-3 to B-6 present waveforms for all the generator settings 

 commonly used in the 96-foot tank tests; measured wave dimensions for 

 Figures B-3, B-4, and B-5 are listed in Tables B-2, B-3, and B-4. Nearly 

 sinusoidal waves of small amplitude and virtually permanent form could be 

 generated; the more complicated waveforms with secondary crests result 

 from nonlinear effects associated with finite wave amplitude. Features 

 of these waveforms agree with the findings of Galvin (1971), except that 

 Figure B-3 shows a secondary crest can exist for d/L = 0.124, which is 

 contrary to his report. However, the secondary soliton (in Galvin 's 

 terminology) is very weak, explaining how it could have been overlooked 

 in analyzing motion picture records. Because the generation and propaga- 

 tion of nonlinear shallow-water laboratory waves is a subject of continu- 

 ing interest , these data will be compared to analyses in available 

 literature. 



Madsen (1971) presented a second-order analysis of the waves created 

 by sinusoidally moving piston, in the case of d/L < 0.1 and the Stokes 

 parameter, S = (HL^/2d^) < 4Tr2/3, where H is the height of the primary 

 generated wave, a Stokes second-order progressive wave at the frequency 

 of the piston motion. His solution predicts observable secondary waves 

 for S > 217^/3, caused by a second harmonic free wave propagating more 

 slowly than the primary wave; the interference between the primary and 

 secondary waves causes an approximately sinusoidal wave height variation 

 away from the generator with a wavelength of L/2. Mei and Unluata (1972) 

 analyzed the resonant interaction between the first and second harmonics 

 off a sinusoidally moving piston in shallow water, finding the harmonics 

 vary periodically in amplitude, with dimensionless beat length, B/L, 

 exhibiting a complicated dependence on S. Their preliminary experimental 



83 



