Volume 76, Issue S1, October 1984
Index of content:
- PROGRAM OF THE 108TH MEETING OF THE ACOUSTICAL SOCIETY OF AMERICA
- Session E. Underwater Acoustics I: Computation Intensive Ocean Acoustics I
- Invited Papers
76(1984); http://dx.doi.org/10.1121/1.2021699View Description Hide Description
Ocean acoustic problems are very complicated by nature and require large scale computations. Recent computertechnology advances have produced very fast, inexpensive pipelined array processors. These processors allow the efficient, cost effective solution of complicated problems. This presentation begins with a demonstration of a two‐dimensional modelocean acoustic long range propagation problem and discusses its solution on both a conventional sequential computer and a pipelined array processor. The next step of the application of multi‐array processors is presented. A three‐dimensional modelocean acoustics problem is chosen as a sample problem for such multiprocessors. [Work supported by ONR.]
Radiation and scattering from large axisymmetric structures in an infinite or semiinfinite fluid medium76(1984); http://dx.doi.org/10.1121/1.2021700View Description Hide Description
A complex axisymmetric structure immersed in an infinite or semiinfinite fluid medium, excited by a plane acoustic wave or by a spherical or cylindrical acoustic wave emanating from a source in the vicinity of the structure or by a mechanical force acting on the structure, scatters and/or radiates the acoustic waves in the fluid medium. At a lower frequency range of excitation, vibrations of the structure and the radiating acoustic pressure strongly couples, whereas at higher frequency of excitation the structure can be regarded as rigid. A program called FIST (Fluid Interacting with STructures) is developed to analyze these problems. Using the Helmholtz integral, one can write a linear relation between the particle velocity normal to a cavity surface and the corresponding acoustic pressure. For efficiency and economy of calculations and for the consistency of velocity distribution between the structural elements and the corresponding fluid elements in contact, the distribution of pressure and velocity on the cavity surface is described by a cubic polynominal. These fluid equations couple with the equations of motion of the structure. Combining the fluid and structural equations we get . (1) Since this is a steady‐state excitation problem, differential equation (1) reduces to a set of algebraic equations (2) Here matrix C is full and complex, K is real and banded, and M is real and diagonal. Further, the elements of K are several orders of magnitude larger than those of C. This ill conditioning causes a severe degradation in computational accuracy when one attempts to decompose the dynamic matrix. Two alternatives are available. (a) Iterate Eqs. (2) using the initial vector given by the solution of the in vacuo response of the structure. (b) Transform Eqs. (2) using in vacuo modes of the structure. We opted for the second approach and using that we have calculated the signature of a full scale marine structure. Construction of fluid and the modal matrices are computationally quite intensive. An original attempt to use a UNIVAC 1108 was dropped because it would have taken approximately 250 h of CPU time to calculate one bistatic plot. A UNIVAC 1110, although faster, took 50 h of CPU time to calculate one bistatic plot. Using a CRAY computer with its large core and speed of computations, it took only 20 min of CPU time to calculate a complete monostatic plot.
A numerically efficient global matrix approach to the solution of the wave equation in stratified environments76(1984); http://dx.doi.org/10.1121/1.2021701View Description Hide Description
Solution of wave propagation problems in horizontally stratified environments arises in many fields, including underwater acoustics,seismology, and ultrasonics. When the environment is considered range‐independent the wave equation can be separated in depth and range by standard integral transform techniques. The depth‐dependent solution is then found by matching boundary conditions at horizontal interfaces, and the field as a function of range is found by evaluating the inverse integral transforms. Several numerical methods have been developed for this purpose, in underwater acoustics known as the fast‐field technique and in seismology as full wave field and reflectivity methods. These methods have generally been based on propagator matrix solutions for the multilayer Green's function. In contrast to these techniques, the use of a global matrix approach to solve the depth‐separated wave equation automatically yields the possibility of treating problems with several sources and receivers without requiring separate Green's function calculations. Unconditionally stable solutions are obtained in a computationally efficient fashion, leading to a code that is an order of magnitude faster than existing models. The generality and efficiency of this global matrix method makes it well suited to a wide class of propagation problems, as demonstrated by selected examples from underwater acoustics. Total wave fields in depth and range are calculated for both cw and pulsed sources.
- Contributed Papers
76(1984); http://dx.doi.org/10.1121/1.2021702View Description Hide Description
The reflection and transmission of narrow sound beams at the interface between two fluid media was studied experimentally by Muir et al. Sound Vib. 64, 539–551 (1979)] who found that narrow beams impinging on a sedimentary bottom at grazing angles below the critical angle will not be totally reflected as predicted by Snell's law. Here a numerical model yielding an exact solution to the wave equation in horizontally stratified environments is used to analyze the observed phenomenon. A beam of any realistic width is generated by introducing a vertical source array and properly phasing the single sources. It is shown that the deviation from Snell's law is due to the finite width of the angular spectrum of narrow beams, and the results given are in good qualitative agreement with the experimental results.
A hybrid numerical/analytic technique for the computation of wave fields in stratified media based on the Hankel transform76(1984); http://dx.doi.org/10.1121/1.2021703View Description Hide Description
A hybrid numerical/analytic technique for computing the field due to a monochromatic point source in a horizontally stratified medium was developed. This procedure is extremely accurate for all ranges. It is particularly appropriate when the field is composed of few dominant modes and a significant contribution from the continuous spectrum. This is the case for long‐range propagation in the deep ocean when the source and receiver are near the bottom and there is a low speed layer at the water‐bottom interface. The method is based upon a numerical evaluation of the Sommerfeld integral, which is in the form of a Hankel transform. Both computational speed and high accuracy are obtained by treating the singularities in the kernel of the Sommerfeld integral with a new technique that allows the singular portions to be handled analytically but which keeps the remaining portion of the integral well behaved numerically. The treatment of these singularities was motivated by a study of the effects of aliasing in the Hankel transform. Both speed and accuracy in the calculation of the Hankel transform are obtained by applying a new fast (N * log N) Hankel transform algorithm that requires its input on a square root grid. This grid is more suitable for representation of the kernel of the Sommerfeld integral than more conventional linear grids. The algorithm has significant speed advantages over quadrature and adaptive integration techniques.
76(1984); http://dx.doi.org/10.1121/1.2021704View Description Hide Description
A numerical scheme for generating both the trapped mode and continuum portions of acoustic fields for a horizontally stratified ocean and bottom is presented. The technique is based on the fact that the branch line integral corresponding to the continuum portion of the field can be performed by Hankel transforming a modified Green's function. The modified Green's function is obtained by removing the pole contributions from the actual Green's function. The continuum portion is then added to the trapped mode portion to form the total synthetic field. The technique has the advantage that it is fast, accurate, and can be used for a more general geoacoustic model than the Pekeris waveguide. Synthetic results for several examples are presented and discussed.
Numerical implementation of intrinsic mode Green's functions for oceans with weakly sloping penetrable bottom76(1984); http://dx.doi.org/10.1121/1.2021705View Description Hide Description
Two recently developed related spectral theories [A. Kamel and L. B. Felsen, J. Acoust. Soc. Am. 73, 1120–1130 (1983); J. M. Arnold and L. B. Felsen, J. Acoust. Soc. Am. 73, 1105–1119 (1983)] have provided potentially new options for calculating source‐excited sound fields in a weakly range‐dependent ocean environment. These theories have so far been applied to a homogeneous two‐dimensional ocean and bottom separated by a plane sloping interface. It has been recognized that their common building blocks are what have been referred to as “intrinsic modes” [J. M. Arnold and L. B. Felsen, J. Acoust. Soc. Am. (to appear)]. Intrinsic modes have spectral integral representations that reduce in a lowest order approximation to adiabatic modes, where these can be defined, but which remain uniformly valid in their integral form through the cutoff transition in upslope propagation. An efficient numerical algorithm has been developed for calculating the intrinsic mode Green's function in the ocean and in the bottom. In a sense, this algorithm may be regarded as a range‐dependent generalization of the range‐independent Fast Field Program (FFP), but with the important difference that angular wave spectra replace the conventional rectilinear spectral decomposition. Numerical results are compared with those from the parabolic equation [F. B. Jensen and W. Kuperman, J. Acoust. Soc. Am. 67, 1564–1566 (1980)] and the augmented adiabatic mode theory [A. Pierce, 3. Acoust. Soc. Am. 74, 1837–1847 (1983)]. Also discussed and compared are asymptotic approximations of the spectral integral, which do not require the patching of the augmented adiabatic model. [Work supported by ONR Ocean Acoustics.]
76(1984); http://dx.doi.org/10.1121/1.2021706View Description Hide Description
The method of finite differences is applied to the elastic waveequation to generate synthetic seismograms for laterally varying seafloor structures. The results are compared with borehole seismic data from the Gulf of California (Deep Sea Drilling Project Site 485) in which lines were shot over flat and rough topography. The significant new phenomenon observed in both the synthetic seismograms and the field data is the generation of a “double head wave” due to the interaction of the incident wavefront with the side of a hill and the flat sea floor adjacent to the hill. In these models the hills are on the order of a seismic wavelength in height and steep velocity gradients occur over distances comparable to wavelengths. Ray theoretical methods would not be suitable for studying such structures. True amplitude record sections are obtained by the finite difference method, which show for these models that the head wave generated at the flat sea floor adjacent to the hill is lower in amplitude than if the hill were not present and is lower in amplitude than the head wave generated at the hill. A second feature which is important for borehole receivers is the existence of the “direct wave root” in the upper basement. This energy occurs below the sharp interface when the direct wave impinges on the interface from above. There is no corresponding Snell's law ray path for this energy and the energy is evanescent with depth in the lower medium. The properties of both the double head wave and the direct wave root are clearly demonstrated in the finite difference “snapshot” displays.
76(1984); http://dx.doi.org/10.1121/1.2021707View Description Hide Description
A model of acoustic propagation in solid media has been derived. It is a one‐way wave equation based on a high‐order Padé approximation to the square root function. The physical properties of the environment are modeled as thin stratified layers. A generalization of the equations to range‐dependent environments is easily implemented by allowing the material properties of the layers to vary in range. Furthermore, reflecting interfaces of variable depth can be approximated by using an equivalent reflector, consisting of two thin layers whose material properties vary in range.
- Session F. Psychological Acoustics II: Intensity, Frequency, and Pattern Discrimination
76(1984); http://dx.doi.org/10.1121/1.2021708View Description Hide Description
Using classical respiratory conditioning, the smallest detectable changes in the intensity of a continuous sound were measured in the goldfish as functions of signal duration and overall level. Gap detection thresholds for noise were also obtained. Neural correlates were studied in single auditory (saccular) nerve units. Noise increment detection shows perfect temporal summation (10–160 ms) in that equivalent signal levels for in‐phase addition fall by 3 dB per doubling of duration. Thresholds are independent of overall level, in accordance with Weber's law. Noise decrement thresholds grow more rapidly toward a minimum gap of 35 ms. Tone increment and decrement thresholds are identical (0.1–0.2 dB), and show no duration effect. Increment thresholds decline significantly with overall level between 15 and 40 dB SL (Weber's law does not hold). Auditory nerve units vary widely in adaptation patterns and in sensitivity to intensity changes. Recovery from adaptation (a burst of spikes evoked as the sound intensity returns to the adapting level following a decrement) has a different time course for tones and noise and may account for gap detection thresholds. These results are discussed in the context of current theories on hair cell‐nerve fiber synaptic mechanisms. [Supported by the NSF and the NIH (NINCDS).]
76(1984); http://dx.doi.org/10.1121/1.2021709View Description Hide Description
Intensity DLs were measured as a function of stimulus duration for tones at 250, 1000, and 8000 Hz. At each frequency three levels were tested: 85, 65, and approximately 40 dB SPL. Ten or eleven stimulus durations ranging from 2 ms (4 ms at 250 Hz) to 2 s were tested. An adaptive two‐alternative forced‐choice procedure with feedback was used. The rise and fall times were 1 ms and the interstimulus interval was 250 ms. Results for three young normal listeners show that DLs decrease with increasing duration up to at least 2 s, except at 250 Hz. At 250 Hz, the maximum integration time is between 0.5 and 1.0 s. In a double logarithmic plot of DL (in dB) versus duration the data are well described by straight lines (r < − 0.97). The slope of these lines is approximately independent of level, but appears steeper at 1000 Hz (slope≈ −0.38) than at 250 Hz (slope≈ −0.29) and 8000 Hz (slope≈ − 0.26). At all three frequencies, the slopes are shallower than the slope of −0.5 predicted by an optimum detector. [Supported by NIH.]
Intensity discrimination of pure tones is not a monotonic function of sensation level for frequencies above 6 kHz76(1984); http://dx.doi.org/10.1121/1.2021710View Description Hide Description
Some recent studies of intensity discrimination at high frequencies have reported that measures of intensity discrimination are nonmonotonically related to sensation level (SL) for frequencies between 6 and 10 kHz [Carlyon and Moore, J. Acoust. Soc. Am. Suppl. 1 73, S93 (1983); Cullen and Long, Abstr. Seventh Midwinter Research Meeting of ARO, 5 (1984)]. At these frequencies, difference limens estimated from amplitude modulation detection and pulsed tonediscrimination are largest for stimuli near 30 to 40 dB SL. We have extended our studies to 14 kHz and obtained a more detailed characterization of this nonmonotonicity by measuring amplitude modulation detection thresholds for stimuli in 5‐dB steps between 10 and 65 dB SL. Our data indicate that the larger difference limens reported by Florentine [J. Acoust. Soc. Am. 74, 1375–1379 (1983)] for high‐SL high‐frequency signals are found when the frequencies tested are near the upper limit of a subject's hearing. [Supported, in part, by Kam's Fund and Eye, Ear, Nose, and Throat Foundation of New Orleans.]
Detection of tones in the absence of external masking noise. II. A reconsideration of the relation between d′ and the ISO standard76(1984); http://dx.doi.org/10.1121/1.2021711View Description Hide Description
Watson, Franks, and Hood [J. Acoust. Soc. Am. 52, 633–643 (1972)] investigated the detectability of tones in the absence of external masking noise in normally hearing listeners and compared their data with the ISO standard for audiometric zero. They concluded that the ISO standard represented thresholds roughly equivalent to d′ = 1.0 performance. Watson and his colleagues used stimuli 150 ms in duration. Pure‐tone temporal‐integration data show that longer signal durations yield lower threshold estimates up to a frequency‐dependent limit. The ISO‐standard data were collected with signals designed to exceed the limits of temporal integration. Psychometric functions for tone detection in quiet are steep; consequently, when longer duration signals are used, a small threshold change of the magnitude expected from temporal integration would alter the performance relation suggested by Watson et al. Re‐evaluation suggests that the ISO standard probably represents a performance level greater than d′ = 2.0. [Work supported by NINCDS.]
76(1984); http://dx.doi.org/10.1121/1.2021712View Description Hide Description
Detectability of a 100‐ms 1000‐Hz sinusoid was measured in quiet and in broadband noise having a spectral level of 20 dB SPL. Intensity discrimination of an in‐phase increment was studied for standards that were −9, −6, −3, 0, 3, 6, 9, 12, and 18 dB relative to threshold. When discrimination thresholds were plotted in terms of signal amplitude versus standard level, all subjects showed nonmonotonic functions (negative masking). Smallest thresholds occurred when the standard was just detectable. These discrimination thresholds were about 10 dB lower than detection thresholds. Psychometric functions were measured for simple detection (no standard) and for intensity discrimination with a standard at threshold. For intensity discrimination, d′ was proportional to signal amplitude. For simple detection, d′ was proportional to signal amplitude to the fourth power. The results in noise agreed well with those in quiet, although there was a tendency to obtain more “negative masking” in quiet and psychometric functions for detection were slightly steeper in quiet. [Work supported by NSF.]
76(1984); http://dx.doi.org/10.1121/1.2021713View Description Hide Description
This experiment examines how stimulus bandwidth and variability affect the difference limen, , for a change in intensity. DLs were measured at 30, 60, and 90 dB SPL for a 3‐kHz tone and for nine noise bands with bandwidths ranging from 50 Hz to 12 kHz, using an adaptive two‐interval forced‐choice procedure. Results for three normal listeners show that the DLs are the same for frozen and random noise, except for the narrowest noise bands with steep skirts. For 50‐Hz bands, DLs for frozen noise decrease from about 3.5 to 1.2 dB as level increases from 30 to 90 dB, similar to DLs for the tone, but for random noise they are roughly constant, between 3 and 4 dB. Bandwidth has essentially no effect on the DLs at 90 dB, but at the lower levels the DLs decrease as bandwidth increases. An anomalous result is that the DLs are larger at 60 dB than at 30 dB for all bandwidths tested. The interaction of bandwidth and level is consistent with the excitation pattern model, but the effect of level between 30 and 60 dB suggests modifications. [Supported by NIH.]
76(1984); http://dx.doi.org/10.1121/1.2021714View Description Hide Description
Early results on synthetic‐vowel formant amplitude discrimination by Flanagan [J. Acoust. Soc. Am. 27, 1223–1225 (1955); J. Speech Hear. Disord. 22, 205–212 (1957)] indicated that ΔI for overall vowel level was about ± 1.5 dB, but that ΔI for F2 was about ± 4 dB. The latter result was for F2 ≃ − 12dB re: F1 for the “comparison” vowel. Subjects made a same‐different judgment which may have affected the interpretation of these results. Here we report ΔI for several “comparison” vowels which differ in amplitude relations among F1, F2, and F3. In addition, we report on perception of extreme shifts in formant amplitude, Implications for computer recognition of speech will be discussed.
The coupling of sub‐audio amplitude and frequency modulafions and the perception of spectral envelope characteristics76(1984); http://dx.doi.org/10.1121/1.2021715View Description Hide Description
One possible cue for the relative perceptual invariance of a resonant source with pitch change is the tracing of its spectral envelope by its modulating frequency components. This tracing would reduce the ambiguity concerning the resonant structure of the source at higher F 0's where the formants are not filled by partials. This paper investigates the role that spectral tracing (FM‐induced AM) plays in the perception and identification of vowels with high F 0. Two five‐formant vocal spectra were selected to give nearly identical spectra when not modulated. When modulated by vibrato, a single component differed in the sign of its amplitude‐frequency slope while the behavior of the rest was identical in the two cases. At larger frequency modulation depths, a difference in vowel quality is perceived. The threshold depth at which the stimuli were distinguished and the threshold depth for identification were determined. Discrimination thresholds were lower than identification thresholds. The discrimination thresholds for sine tones with amp‐freq slopes of opposite sign were also determined, and were found to be greater than the thresholds for the complex tones. The results will be discussed in terms of intensity discrimination results and of recent work in spectral profile analysis.
76(1984); http://dx.doi.org/10.1121/1.2021716View Description Hide Description
Young chickens will suppress their regular peeping when they hear a change in the frequency of pulsing pure tones. The duration of this peep suppression grows with increasing amounts of frequency change. Various adaptive tracking procedures can thus use this response to estimate frequency difference limens in neonatal chickens. These estimates in chickens as young as 4 days of age are close to others available in the literature from mature subjects. Despite large variability in the results from these “staircases,” frequency difference limens generally decrease by approximately one percent over the first 4 days of post‐natal life. As expected from human studies, variability in this discrimination task is greater than previously reported in a comparable detection task [L. Gray and E. W. Rubel, J. Acoust. Soc. Am. Suppl 1 71, S31 (1983)]. In conclusion, there is a significant increase in responsiveness to small changes in the frequency of pure tones as animals mature over the first few post‐natal days. [Work supported by NIH.]
Frequency DL duration functions and estimated excitation‐pattern slopes in normal and hearing‐impaired listeners76(1984); http://dx.doi.org/10.1121/1.2021717View Description Hide Description
When the frequency DL is plotted as a function of tonal duration, an abrupt change in the slope of the DL duration function is often observed [Liang and Chistovich, Soy. Phys. Acoust. 6, 75–80 (1961)]. Their explanation was that the DL is normally limited by the auditory filter bandwidth, but when the signal is shortened so that its bandwidth is broader than that of the auditory filter, the broader spectral bandwidth of the signal determines the size of the frequency DL. We tested a similar hypothesis using excitation‐pattern slopes and spectral slopes rather than bandwidths, and extended the hypothesis to predict DLs from hearing‐impaired listeners. DL duration functions and estimated excitation‐pattern slopes were obtained from five normal‐hearing and six hearing‐impaired listeners. The break points in DL duration functions predicted by equating excitation‐pattern and signal spectral slopes were in close agreement with the observed break points. As predicted by their more gradual excitation‐pattern slopes, the break points occured at shorter durations, or not at all, in the hearing‐impaired subjects. [Work supported by NINCDS Grants NS15451 and NS12125.]
76(1984); http://dx.doi.org/10.1121/1.2021718View Description Hide Description
Untrained subjects judged whether two sequences of tones were different or the same. Each sequence consisted of three tonal items (A = filtered pulse train, B = sinusoidal tone,C = filtered square wave) which were arranged either in the form [ABCABC⋯A] followed by [ACBACB⋯A] (different) or [ABCABC⋯A] followed by [ABCABC⋯A] (same). Each tonal item had the same waveform repetition frequency (800 Hz) and pitch, but possessed a distinctive quality. Duration of items was the same within sequence pairs, and was varied systematically from item durations of several seconds (permitting easy verbal labeling of components in the proper order) down to 10 ms (components could not be recognized). Accuracy of judgments was well above chance at each item duration. It is suggested that discrimination of differences in the order of items was accomplished through a naming of individual components in order of occurrence for longer item durations, and through recognition of qualitative differences between sequences for brief item durations. Implications for perception of sequences of sounds in speech and music will be discussed. [Work supported by NIH and NSF.]