Appendix


Appendix A- Full Data Set - Listening Tests I            Appendix B- Full Data Set - Listening Tests II             Appendix C- Matlab Files

Appendix D - Impulse Response            Appendix E - Loudspeaker Frequency Response                TOC or Beginning

Appendix A- Full Data Set - Listening Tests I
 
 



 
 


 
 


 
 


 
 
 


 
 


 
 


 
 
 

 

Top

 

Appendix B- Full Data Set - Listening Tests II 


 
 
 

 

 

 

Top  

 

 

Appendix C- Matlab Files

1) Energy Distribution Matlab File

2) Power Spectral Density Matlab File

3) Impulse Response Matlab File

4) Spectrogram Matlab File

5) Loudspeaker Frequency Response Matlab File
 

 

 

Top

 

Appendix D - Impulse Response

        To ensure a thorough definition of the acoustic space, the impulse response of the test room was determined.  Specifically, an impulse response of the Gusman “dead” room was recorded and compared to a “free field” impulse response.  This was performed using the following equipment:


        In the “dead” room, the microphone was setup to be at the center-head position of the test subjects at a height of 1.22m from the floor.  The impulse generator was then mounted atop the speaker stands, which would be used during the experiments.  The stand was located directly in front of the microphone at the same height and at a distance of 2m (see Figure 41).  A string was tied to the release of the rattrap, which ran outside of the room and allowed remote triggering.  

Figure 41: “Dead” room Impulse Response Setup

        The “free field” impulse responses were generated by using the same setup previously described, this time located in the approximate middle of the University of Miami’s Intramural (IM) fields (south of the percussion studio).  The test was performed late in the evening to minimize interference of other sounds.  The approximate location can be seen in Figure 42.

Figure 42: “Free field” Impulse Response Setup


 

In both cases, several test trials were run in order to maximize the recorded signal level without clipping.  Five impulse responses were then recorded in each acoustic space.  These recordings were transferred from the Digital Audio Tape (DAT) to a CD audio track and then converted to a mono WAV digital audio file at 44,100 Hz and 16 bits.  The WAV files were truncated to 100 ms lengths using Sound Forge in preparation for the frequency analysis using Matlab (see Appendix C for code).

The mathematics and theory behind this exercise should be quickly reviewed.  Particularly, recording the resulting sound of an impulsive sound source can capture the impulse response of an acoustic space.  However, this recording would contain both the response of the room and the additional unwanted response of the recording equipment.  It is desirable to remove the equipment’s response from the overall recorded response.  Therefore an additional recording in an anechoic environment can be used to obtain the equipment’s response.  With both frequency responses, the equipment’s response can be removed from the room’s recording, resulting in the desired frequency response of the room.

In these recordings, the impulse response generator is assumed to generate an ideal impulse.  Additionally, the “field” recordings are assumed to represent the needed anechoic condition.  Consider the discussed analysis in a mathematical context with the output (Y), input (X) and impulse response (H):
 

Assuming an ideal anechoic environment allows  , and an ideal impulse response source gives , which results in:
 

and


         The recorded impulse responses were input to Matlab, which was used to calculate the frequency responses (H) using the Discrete Fourier Transform (DFT).This analysis showed that one recording from each location seemed to be outside the expected DFT, and therefore was eliminated.The remaining four responses were averaged and normalized to create the plots shown in Figure 44 (see Appendix C for Matlab code).Also shown is a temporal plot of the recorded impulse response (Figure 43) and the resulting impulse response (Figure 45).

Figure 43: Temporal Plots of Impulse Responses
 
 
  

Figure 44: Spectral Plots of “Room” (top) and “Field” Impulse Response
 
 
 
 

Figure 45: Resulting Impulse Responses of “Room”

Top

Appendix E - Loudspeaker Frequency Response

            In addition, it was necessary to verify the frequency response of the loudspeakers.  The following equipment was used for this portion of the experiment:

            ·Microphones - B&K 4003 with black diffusion cap
·Recording - Portable DAT machine, TASCAM model DA-P1
·Loudspeakers - M&K model MPS-1610 loudspeakers previously discussed.
·Speaker Stands - Studio Tech SN-A adjustable metal speaker stands

            A CD audio track was played; having one hundred sinusoidal signal bursts ranging from 20-20,000 Hz was played.  Each frequency was held for 1.75 seconds, and there was .25 seconds of silence between successive tone increments.  The recording was played through the loudspeaker under test and recorded, as shown in Figure 46.The microphones were setup at the same height as the center of the woofer and at a distance of 1 meter.

  

Figure 46:  Frequency response measurement setup

            As mentioned, each of the recorded microphone tracks was converted from DAT to CD and then to a WAV 16 bit, 44.1 kHz mono digital audio file.  The WAV files were then analyzed using Matlab (see Appendix C for code).  The frequency response can be shown by plotting the data of the microphone directly in front of the speaker versus frequency (see Figure 47).

Figure 47:  Frequency Response of MPS-1610 Loudspeakers

 
 
 

Top or TOC or Beginning

 

 

Created  February 2003 by Rob Hartman

Copyright (C) 2003