Project Topics
Get Complete Project Material File(s) Now! »
Microphone Distance to Termination
According to the ASTM standard for the two-microphone impedance tube measurement, the distance from the end microphone to the surface sample must be known to an accuracy of 0.1mm. [3] In the standard this requirement is waived for highly absorptive, roughly textured surfaces. This could apply to hair, but denitely not to skin measurements.
For calibration with a rigid termination, and for skin which is assumed to be reasonably rigid at this point, this exception does not hold. Therefore, it is necessary to determine the distance from this microphone to the rigid to semirigid termination. The requirement is precision to 0.1mm. The diaphragm was aligned parallel to the walls of the tube, and had a diameter on the order of 1.0mm. Therefore, even if it was possible to measure the physical distance with an optical measurement method, such as a caliper or laser scheme, the precision required was on a scale smaller than the size of the diaphragm. It was necessary to determine the distance from the acoustic center of the closer diaphragm to the termination, for which this method would not suce.
A standard way to determine this position is to move the position of the microphone, and scan for pressure minima. This method contains possible errors in determining the exact acoustic position of the end of the tube, requiring end corrections to be determined and included. From this position, it is possible to calculate the distance quite easily. In designing the system, the position of the microphones was determined, and then xed. In addition, due to the small tube diameter, the insertion of a probe microphone inside the tube would most likely disturb the sound eld in the tube such that the plane wave assumptionwould be compromised. Therefore, a method which allows for the xed position of the microphone must be used. Taking the traditional method of scanning the sound eld for the pressure minima at a given frequency and transforming the problem to the frequency domain, it should be possible to perform the same measurement by sweeping frequency and determining the exact frequency at which the xed microphone is at a pressure minima. The necessary frequency precision was determined as follows: the approximate distance to the termination was measured to be 5 cm. Using 343 m/s as an approximate sound speed, this would result in a rst null (for a quarter wavelength) to be at 1715 Hz. A variation of 0.1mm would yield a distance of 5.01 cm, and a quarter wavelength frequency of 1711.6 Hz. With the required minimal accuracy being 0.1mm, precision of 0.01mm was attempted. Introducing this variation to the approximate measurement of 5 cm yields 5.001 cm, which has a corresponding quarter wavelength frequency of 1714.7 Hz. From these two rough calculations, a frequency resolution of 0.3 [0.01mm resolution] to 3.4 Hz [0.1mm resolution] is necessary, depending on the condence in the measurement and the desire to remain within the standard requirements.
Spacing Between Microphones
Using the distance to termination data from both microphones, it is also possible to determine the spacing between the microphones, s; a distance necessary for the acoustical calculations. Using the results from Tbl. 3.1, the separation distance was determined to be 2:365 0:004 cm, taking an average for the distances determined from each null. This distance can also be calculated by using an anechoic termination and determining the phase delay between the two sensors, again using the transfer function. [8] This method of measurement is totally independent of the previous method and was therefore used as a reference check.
Using an anechoic termination, the transfer function was measured for the sensors in their normal position, and also with the positions switched. Using anspace.m (see section A.1) to calculate the spacing, the sensors were rst calibrated as per Eq. (3:6), and the acoustic transfer function was determined using Eq. (3:7). The microphone spacing was then calculated using the phase delay between the two sensors, normalized with respect to frequency, as dened in Eq. (3:8). H = Hanechoic Hcal = H12 pH12 H21.
Planar Surface Assumption
For semirigid to rigid impedance terminations the distance to the termination is crucial, as described in the previous section. Therefore, it is important that the sample be at to roughly the same degree of accuracy at the standard requires. In measuring the properties of skin, the measurement tube was placed against the area of skin to be measured. It was quickly observed that a curvature of the skin was induced which was proportional to the pressure applied with the tube. As discussed in the literature, and as would be expected, the impedance characteristics of the skin will vary with the applied pressure as the tension on the area will vary, creating a dierent mechanical system. It would be dicult to create a measurement system which would reliably use the same applied force for each trial, simulate normal skin condition, and provide a good acoustic seal. The following procedure was therefore conceived to minimize these problems. The impedance tube was attached to the skin area with a thin layer of nontoxic rubber cement, providing a good acoustic seal. Once cured, the tube was pulled slightly so that the skin sample was visibly raised from the skin surface area as represented in Fig. 3.2. The result of this procedure was that the skin sample was under little tension in the plane of the skin, keeping close to the normal skin position situation, without eects due to the applied pressure of the tube. Having the skin slightly pulled away from the normal surface removed the eect of bowing or curvature due to the pressure of the tube against the skin, or any normal curvature present on the sample area. The result was a planar skin sample with reasonably typical skin tension in the planar direction. It was acknowledged that the tension in the direction normal to the surface would not be the same as the typical position, but this was assumed
to have much less of an eect than the other conditions the procedure was designed to alleviate. To calibrate for the distance to the sample, the rigid sensor termination was attached to the tube in the same manner, by using rubber cement. To verify the adequacy of the seal, additional distance measurements were made by including an external clay ring with the reference termination. Results showed no appreciable eect and therefore the rubber cement seal was assumed to be sucient for use in the skin measurement (see section 3.2.2).
HRTF Equalization and Labelling
In order to facilitate the plotting of such a large number of experimental conditions, a labelling scheme for experimental data was used which utilized three parameters. Data were identied by subject number, measurement method, and ear. There were a total of six subjects (no. 1{6) used throughout the work, though not all measurements were made with all subjects. For example, only subject 1 was used for the computational work, so for the simulations no subject number was necessary. Measurement methods were as follows; WPAFB real subject = R, WPAFB pinna mold = P, and CRE real subject = C: Finally, ear designation was L or R for left and right ear measurement respectively. All simulated measurements used only the left ear, so again no designation was necessary. An example of the labelling scheme would be 1PL; which would correspond to subject 1, WPAFB pinna mold measurement, left ear. This scheme may also be divided among dierent parts of a gure, so that the main plot could be of 1L, where dierent data sets in the plot could be P; R; and C.
Individual HRTF
For the individual HRTF measurement, the subjects were seated inside the dome. The head was positioned so that the center of the interaural axis was located at the center of the sphere. This was accomplished using two low powered lasers to align the head correctly. The subject’s location was stabilized using a chin rest (see Fig. 4.6), thereby keeping the head reasonably immobile during the measurement. This rest was covered with acoustically absorptive foam in order to reduce any eects to the HRTF. Allowing for settling time and data averaging, this procedure lasted about 45 minutes. This time included several rest periods for the subjects. Due to the time involved in this measurement, only a subset of the entire source array was used. Out of the 272, only 70 source positions were measured. These positions consisted of the horizontal plane, the median plane, and three points in the middle of each of the eight octants. This provided ample data for comparison.
Table of contents :
List of Figures
List of Tables
Acknowledgements
Chapter 1. Introduction
1.1. Research Overview
1.2. Thesis Organization
Chapter 2. Literature Review
2.1. Localization
2.1.1. Justication
2.1.2. Historical Review { Up to 1960
2.1.3. Historical Review { 1960 to 1973
2.1.4. Current Literature
2.1.5. Anatomical and Neurological Processes in Localization
2.1.6. Anatomical and Neurological Processes of Localization for a Monaural Listener
2.2. Absorption and Impedance Measurements
2.3. HRTF Calculations
Chapter 3. Acoustical Parameter Measurements
3.1. Measurement Technique
3.2. Apparatus
3.2.1. Microphone Calibration
3.2.2. Microphone Distance to Termination
3.2.3. Spacing Between Microphones
3.2.4. Planar Surface Assumption
3.3. Measurements
3.3.1. Reference Terminations
3.3.2. Eect of distance errors
3.3.3. Skin
3.3.4. Hair
Chapter 4. HRTF Measurement
4.1. HRTF Equalization and Labelling
4.2. Session WPAFB
4.2.1. Apparatus
4.2.2. Individual HRTF
4.2.3. Pinna Replica HRTF
4.2.3.1. BAMAR Head
4.2.3.2. Pinna Replica Construction
4.2.3.3. Measurement
4.2.4. WPAFB Results
4.3. Session CRE
4.3.1. Apparatus
4.3.2. Individual HRTF
4.4. Measured HRTF Data Comparison
Chapter 5. HRTF Calculation
5.1. Boundary Element Method Approach
5.2. BEM Model Generation
5.2.1. Head Geometry Acquisition
5.2.2. Mesh Modication
5.2.2.1. Data Format
5.2.2.2. Closure
5.2.2.3. Coarsen Mesh
5.2.2.4. Rene Mesh
5.2.3. Field Mesh
5.3. Calculation
5.3.1. Calculation Speed
5.3.2. Various Meshes
5.4. Results
5.4.1. Spherical model
5.4.2. Rigid models
5.4.3. Impedance Condition
Chapter 6. HRTF Comparisons
Chapter 7. Conclusion
References
Appendix A. MatLab Scripts
A.1. anspace.m
A.2. polynul2.m
A.3. whatc.m
Appendix B. C Code
B.1. decimate.c
B.2. edgealign.c
Appendix C. Sysnoise Related Command Files
C.1. Example setup command le for head mesh
C.2. Hair denition command le
C.3. Example setup command le for sphere mesh
C.4. Example computation command le
C.5. Unix sysmon script
C.6. Example sysmon log le
Appendix D. Data
D.1. WPAFB source locations
D.2. CRE source locations
