Appendix

Specific processing steps.

A) Cassette tape digitizing and processing steps were as follows to produce MP3 file:

Recorded with Steinberg WaveLab Lite: 24-bit, 96,000, monophonic into
a Microsoft® Windows® RIFF format PCM WAV file:
Pioneer CW-650R dual-cassette tape deck to
Behringer EurorackMX802A mixer to
Terratech EWX-2496 master recording sound card.

Engulf Audio from EXE Consulting to create high-quality stereo image:
Concert Hall room (22.51m X 37.51m X 15.01m)
Binaural simulation: Solutions obtained at (11.135m, 15.0m, 3.0m)
(Left) and (11.375m, 15.0m, 3.0m) (Right) in left-handed
coordinate system. Point source placed at (11.135m, 6.0m, 3.0m)
in same system, 9.0m from Left solution. Solves 3-D wave
equation, providing stereo separation, echo, and reverberation in
a self-consistent manner.
Absorption: T60 = 0.30 seconds at 100 Hz, 0.20 seconds at 4000 Hz.
Reverb times kept low to maintain speech comprehension.
Designed for headphone listening.
The reason for performing this step first was to create a stereo
image so that following processing steps could be more readily
monitored using studio-quality headphones.
During this process, the resulting file is normalized to very nearly
0 db while the data are being maintained as double-precision
floating point.

CoolEdit 2000 from Syntrillium for the following steps:
Dynamics Processing:
flat 1.00 : 1 above -20 dB
cmp 2.00 : 1 below -20 dB
exp 10.5 : 1 below -96 dB
Noise Reduction (primarily removed noise produced by tape deck):
Noiseprint from section at end of digitized recording of
cassette tape
4096-point FFT
40 db reduction
Smoothing Amount = 2
Flat, 100%
Trimmed end and beginning to remove non-cassette tape audio
sections. These sections are NOT part of the cassette tape but
were produced between the time the digitizing program was started
and the tape deck was started AND between the time that the tape
deck was stopped and the digitizing program was stopped.
Downsampled from 24-bit, 96,000 samples per second stereo to CD
quality (16-bit, 44,100 samples per second stereo),
Compressed as MP3 using algorithms licensed from Fraunhofer
Institute. 256KBits per second, constant bit rate, 22,050 Hz
maximum band width, "High Quality" Codec.

B) The CD track was ripped from the audio CD using Ahead Nero into a
CD-quality, Microsoft Windows RIFF format PCM WAV file (16-bit,
44,100 samples per second, monophonic). Then the following steps
were performed:

Conversion of two-channel identical monophonic to single-channel
monophonic by removing right channel data using Syntrillium's
CoolEdit 2000.

Engulf Audio from EXE Consulting, same conditions as for the tape
(above, part A).

CoolEdit 2000 from Syntrillium for the following steps:
Dynamics Processing, same settings as for tape (above, part A).
Noise Reduction:
Noiseprint taken from approximately time = 8.95 seconds to 9.85
seconds
4096-point FFT
40 db reduction
NR Level 53
Sloped Noise Reduction Level: 0 Hz -30.2 %, 48,000 Hz -100%
Precision Factor = 5
Smoothing Amount = 2
No trimming was performed.
Noise samples for part C were extracted into separate files.
Downsampled from 24-bit, 96,000 samples per second stereo to CD
quality (16-bit, 44,100 samples per second stereo),
Compressed as MP3 using algorithms licensed from Fraunhofer
Institute. 256KBits per second, constant bit rate, 22,050 Hz
maximum band width, "High Quality" Codec.

C) Four noise samples were taken from the 24-bit, 96,000 samples per
second, stereo file that was upsampled from the CD track (see part
B above) and were each further compressed according to the
following:
flat 1.00 : 1 above -10 dB
cmp 2.99 : 1 below -10 dB
exp 15.3 : 1 below -96 dB

The samples were:

1) From position at beginning of CD track, time = 8.95 seconds to 9.85
seconds.

2) From time = 1:18.67 to time = 1:20.10.

3) From time = 1:23.69 to time = 1:24.67.

4) From approximately time = 1:28.6 to approximately time = 1:32.9.

Noise samples #1, 3, and 4 were edited so that only the left channel
remained.

Noise sample #2 was edited so that only the right channel remained.

Each of noise samples 1, 3, and 4 were overlaid one at a time to noise
sample 2. Noise sample 2 remained in the right channel while each
of the others remained in the left.

Noise samples were then adjusted in overlay so that a repetitive
background noise common to both the left and right channels was
synchronized. Although there were other noises in the recordings,
the common noise, somewhat like that of a dishwasher in operation,
could be heard in both left and right stereo channels. The
offsets used were: -0.2 seconds for #2 overlaying #1; -0.4 seconds
for #2 overlaying #3; +0.7 seconds for #2 overlaying #4. The
offsets are defined as the offset of the zero for #2 relative to
the zero of the section to which it is being overlaid, with "zero"
being defined as the beginning times listed above for each audio
section.

The FFT filter "Mackie Mid Boost" from Syntrillium's CoolEdit 2000 was
applied so that the repetitive noise common to both channels could
be heard more clearly.

After trimming, the three overlays were each looped eight times for
ease of listening, then they were concatenated into a single file
with a 0.3 second silence between them.

The resulting concatenated file was then downsampled to CD quality,
then compressed into an MP3 file with the same settings as used
previously.

D) Two noise samples were extracted from the digitized version of the
cassette tape. These two samples correspond to #2 and #3 above
(see part C above), the only corresponding ones that are available
on the cassette tape recording:

After Engulf Audio processing, samples were extracted.

Overlaid with 0.3 seconds offset --- times not same for cassette
recording as for CD track due to difficulty of locating zero.

Trimmed to area of overlap only.

Dynamics processing:
flat 1.00 : 1 above -10 dB
cmp 2.99 : 1 below -10 dB
exp 15.3 : 1 below -96 dB

Filter FFT Mackie Mid Boost

Quick Filter:
hi: 187, 375, 750, 1500, 3000, 48,000 Hz sections of filter: -30
db; 6,500 and 16,000 Hz + 3.7 db
lo: 187, 1500, 3000, 6500, 16,000, 48,000 Hz sections of filter: -
30 db; 375 and 750 Hz + 3.7 db

Amplified: +6.0 db boost.

Created loops, then MP3 files:
high: looped eight times, then normalized, downsampled to CD
quality, compressed to MP3 at 256 KBits per second.
low: downsampled, looped eight times, compressed to MP3 (not
normalized) at 256 KBits per second.

MP3's used "High Quality" codecs as before.

E) An MP3 file was created from the purported conversation on the
audio cassette recording. This section begins shortly after the
fourth time that the 911 dispatcher says, "Patsy?" It was
processed with the following:

Normalized to 0db

Dynamics processing (to bring up low-level sounds):
flat 1.00 : 1 above -10 dB
cmp 2.99 : 1 below -10 dB
exp 15.3 : 1 below -96 dB

FFT Filter, bandpass (lessen hum):
220 Hz, 0%; 656 Hz, 100%; 2350 Hz, 100%; 8106 Hz, 0%
2048-point, Blackman-Harris windowing

Dynamics processing (to bring volume of keystrokes down):
cmp 3.00 : 1 above -30 dB
flat 1.00 : 1 below -30 dB

Amplify 400%

Engulf Audio from EXE Consulting, same conditions as for the tape
(above, part A).

© Copyright "Dave" on Jameson's Webbsleuths 2003. All rights reserved.