Smith and Cummins 



In principle, the experiment is beautifully simple. In practice, there are a 

 number of difficulties to overcome. First and most obviously, we are dealing 

 with a system with six degrees of freedom, and there is strong coupling among 

 some of the modes of oscillation. A much more serious and subtle problem 

 arises from the fact that we obtain the response of the ship for all frequencies 

 from a small set of relatively short records. Thus, the desired information is 

 highly compressed in the time scale. The resolution of this information re- 

 quires records of very high quality and an analysis procedure which degrades 

 the data as little as possible. 



Prior to the presentation of Ref. 1, experiments were performed to test 

 this procedure as a practical tool. Declining oscillations were used instead of 

 impulsive excitation, but most of the troubles encountered would be even more 

 characteristic of the latter type of test. The measurement system was some- 

 what superior to those typical of seakeeping work at that time. In the process 

 of analysis it became quite clear that major improvements were necessary in 

 order for the technique to be other than a curiosity. 



There were several sources of difficulty, and as the method of overcoming 

 these are key factors in the present paper, they will be mentioned here. First 

 is the question of accuracy. It is clear that when desired data is superimposed, 

 the accuracy to which it can be separated is certainly no higher than the net ac- 

 curacy of the system. The original system had an accuracy of perhaps 5 per- 

 cent and this was not good enough. The second major difficulty was noise, as it 

 is evident that the real objective is a high signal to noise ratio. By noise we 

 mean here all unwanted disturbances such as wall reflections and true electrical 

 noise. The input for the declining oscillation experiment is a step function, 

 which is completely suitable theoretically, but has undesirable qualities practi- 

 cally. These arise from the fact that the step function has harmonic content at 

 all frequencies, all the way to infinity, and such an excitation not only causes 

 the model to oscillate, but in addition it vibrates as a beam at its natural fre- 

 quency. Further, all instruments, attachments, etc., are excited in their various 

 natural frequencies. In consequence, the signal to noise ratio was well below 

 that which is necessary. 



As the potential value of the transient experiment is great, much effort has 

 been devoted to upgrading our measurement, and analysis system since these 

 early tests. The present paper is a progress report on the present state of this 

 program. The details will be discussed in the subsequent sections, but the most 

 significant accomplishments will be mentioned here. 



The first is a technique of towing the model, rather than self propelling it. 

 This is contrary to the current trend toward powered models for seakeeping 

 work. However, we feel that this technique offers real advantages. Specifically, 

 we measure all restraints on the model imposed by the towing, guidance, and 

 excitation system. The sum of these is the net input to the model. Thus, towing 

 gear inertias and frictions are of no concern, as their effects are included in 

 the measured input. 



The second achievement is the use of an excitation pulse of controlled har- 

 monic content. The technique is an analog of that used by Davis and Zarnick for 



440 



