>From - Sat Mar 02 00:57:28 2024
Received: by 10.35.61.2 with SMTP id o2mr309767pyk;
Sun, 23 Jul 2006 23:51:58 -0700 (PDT)
Return-Path: <jm..._at_tscm.com>
Received: from lvs00-fl-n06.ftl.affinity.com (lvs00-fl-n06.ftl.affinity.com [216.219.253.152])
by mx.googlegroups.com with ESMTP id c17si3029638pyc.2006.07.23.23.51.56;
Sun, 23 Jul 2006 23:51:58 -0700 (PDT)
Received-SPF: neutral (googlegroups.com: 216.219.253.152 is neither permitted nor denied by best guess record for domain of jm..._at_tscm.com)
Received: from [141.149.184.232] ([141.149.184.232]:34570 "EHLO
nikola.tscm.com") by ams006.ftl.affinity.com with ESMTP
id S391101AbWGXGv4 (ORCPT <rfc822;T..._at_googlegroups.com>);
Mon, 24 Jul 2006 02:51:56 -0400
Message-Id: <7.0.1.0.2.20060724024425.115a3818_at_tscm.com>
Message-Id: <7.0.1.0.2.20060723132208.1077b218_at_tscm.com>
X-Mailer: QUALCOMM Windows Eudora Version 7.0.1.0
Date: Mon, 24 Jul 2006 02:44:40 -0400
To: TSCM-L <TSCM-..._at_googlegroups.com>
From: "James M. Atkinson" <jm..._at_tscm.com>
Subject: Personal Electronics for Law Enforcement Solid State Recorders
and Body Wires
Mime-Version: 1.0
Content-Type: multipart/alternative;
boundary="===========_256428750=.ALT"
--===========_256428750=.ALT
Content-Type: text/plain; charset="iso-8859-1"; formatowed
Content-Transfer-Encoding: quoted-printable
The following text was extracted from the NIJ
document from July 2005. The original file can
also be cfound on my website, and is also
attached to this document as an enclousure.
http://www.tscm.com/NIJ-210488.DigitalRecorders.July2005x.pdf
-jma
Personal Electronics for Law Enforcement
Solid State Recorders and Body Wires
Table 2-1. Audio Recorder Feature Summary
Table 6-1. Body Wire Types and Feature Comparison
Feature
NBFM
Digital
Voice quality
Scheme
Table 6-2. Figure of Merit for Body Wire Audio Bandwidth
Sound
Table 6-3. Relative Audio Dynamic Range – Sound Pressure Le
Rating
Table 6-4. Figure of Merit for Audio Dynamic Range
Rating
Table 6-5. Figure of Merit for Audio Signal to Noise Ratio
Rating
Good
Table 6-6. Figure of Merit for Bit Error Rate
Power Level
20 mW to 100 mW
Table 6-7. Transmit Power Uses
System Type
Digital systems
Table 6-8. Receiver Sensitivity
Table 6-9. Figure of Merit for Size (Thickness)
MP3
MINIDISC
MINIDISC
Table B-1. Solid State Recorder Spreadsheet
APPENDIX D
-
-
Personal Electronics for Law Enforcement
Solid State Recorders and Body Wires
-
William Butler, Georgia Tech Research Institute
Scott Crowgey, Georgia Tech Research Institute
William Heineman, Tektron
Susan Gourley, Tektron
Prepared Under:
Contract Number N65236-00-K-7805
Submitted to:
Attention: Mr. Richard Baker, Code 741
Mr. Jerry Owens, Code 741JO
Commanding Officer
SPAWARSYSCEN Charleston
PO Box 190022
North Charleston, SC 29419-9022
July 2002
CONTENTS
x Introduction
x Current Commercial Solid State Recorder Products
x Overview of Commercial Audio Recorder Products
x Commercial Audio Recorder Issues
x Current and Projected State of the Art in Solid
State Recorder Technology Areas
x Block Diagram of a Typical Solid State Recorder
x Microphones
x Delta Sigma Analog to Digital Converters
x Audio Compression Algorithms
x Audio Compression Hardware
x Flash Memory
x Batteries
x Projected Commercial Solid State Recorder Products (in 2 years)
x Solid State Recorder Conclusions
x Current Body Wire Products
x Overview of Body Wire Products
x Body Wire Issues
x Current and Projected State of the Art in Body Wire Technology Areas
x Block Diagram of a Typical Body Wire
x Projected Commercial Body Wire Products (in 2 years)
x Body Wire Conclusions
x APPENDIX A – Solid State Recorder Product Matrix
x APPENDIX B – Solid State Recorder Components
x APPENDIX C – Body Wire Product Source Matrix
x APPENDIX D – Survey of Recorder and Body Wire Use by Law Enforcement Ag=
encies
1. INTRODUCTION
-
This report summarizes the work performed by the
Communications Networking Division (CND) of the
Information and Telecommunications Technology
Laboratory (ITTL) of Georgia Tech Research
Institute (GTRI) under the "Personal Electronics
for Law Enforcement" program. This program is
being performed for the SPAWARSYSCEN Charleston.
The report covers work done as part of a joint
effort between GTRI, and Tektron, Inc. GTRI’s
efforts are focused on solid state audio
recorders that could be used for law enforcement
applications, and Tektron’s efforts are focused
on body wires for law enforcement applications.
This report includes information that is intended
to assist the law enforcement community in the
evaluation and purchase of audio recorders and
body wires. It includes a market survey of
commercially available audio recorder and body
wire products, and it includes a brief review of
key technologies used in these products. The
first section of the report covers audio
recorders, and the second section covers body
wires. In addition, an appendix contains the
results of a survey of law enforcement agencies
that deals with the use of recorders and body
wires for law enforcement applications.
2. COMMERCIAL SOLID STATE AUDIO RECORDER PRODUCTS
-
Throughout this program, data has been collected
on commercial-off-the-shelf (COTS) audio
recorders that could be used for law enforcement
applications. An incredible variety of recorders
are available, including solid state audio
recorders based on flash memory. Since solid
state recorders have no moving parts, they can
offer higher fidelity recordings than
conventional cassette recorders. The solid state
recorder does not suffer from background tape
hiss or tape speed variations that degrade the
fidelity of cassette recorders. For these
reasons, special emphasis has been placed on
solid state recorders in this study. For
comparison with solid state recorders, data has
also been collected on MP3 recorder/players, mini
disc recorder/players, and digital audio tape recorder/players.
2.1 Overview of Commercial Audio Recorder Products
-
Table 2-1 presents a summary of the performance
of various kinds of audio recorders. A solid
state flash recorder (made by Olympus), a MP3
recorder (made by Creative Labs), a mini disc
recorder (made by Sony), and a microcassette
recorder (made by Sony) are all compared in the
table. This table does not include all the
devices reviewed in the survey, but instead,
includes devices that typify the performance of
commercially available audio recorders that would
be suitable for law enforcement applications.
Data in the table is current as of July 2002.
Table 2-1. Audio Recorder Feature Summary
-
From the table, it is seen that all the devices
are available in similar sizes, and all devices
are capable of at least 2 hours of record time.
The Olympus voice recorder and the MP3
player/recorder have similar frequency response
to the microcassette. The ATRAC3 compression used
by the minidisc recorder and the digital audio
tape have the best bandwidth. In the cost
category, the MP3 player/recorder is the next
lowest cost after the microcassette. In the media
cost category, the minidisc is the lowest after the microcassette.
2.1.1 Flash Audio Recorders
-
The flash based audio recorder is the main
subject of this report. It offers a number of
potential advantages: high fidelity, high
reliability, small size, and reasonable cost
(cost of both the recorder and the recording
medium). Flash recorders have benefited from the
proliferation of the use of flash memory for
digital cameras and MP3 players over the past few
years, and the cost of flash audio recorders has
come down as a result. The material to follow
describes the features of several representative
commercially available flash audio recorder products.
The Olympus flash audio recorders are available
in several models. The DS2000 is listed in the
table. The DM-1 is also available for
approximately the same cost, and has the added
ability to play back MP3 music recordings. The
DM-1 does not provide protection against
accidental erasure. The Olympus DW-90 flash audio
recorder costs approximately $90, has a
non-removable 8MB flash memory, uses ADPCM
compression, and can record from 22 (5.8kHz) to
90 (1.7kHz) minutes of audio. The DS2000 and DM-1
Olympus flash recorders use a file format called
Digital Speech Standard (DSS). Files stored in
this format occupy 12 times less memory space
than uncompressed WAV files, while achieving
roughly the same audio quality. Olympus flash
voice recorders feature voice-activated recording
that can be switched off. Olympus voice recorders
use a standard USB interface to transfer data
from the recorder to a PC. The Olympus recorders
can record in monaural mode, but not stereo.
Further information on these products may be
obtained at the manufacturer’s web site:
http://www.olympusamerica.com/cpg_section/cpg_vr_digitalrecorders.asp .
The Panasonic RR-XR320 is another example of a
flash audio recorder. The RR-XR320 is 1 7/8” x 3
9/16” x ½” in size, uses ADPCM recording and uses
SD flash memory. It has a battery life of 11
hours when recording, and uses two AAA batteries.
The MSRP of the RR-XR320 is $329, and street
prices around $280 are common. This flash
recorder uses a standard USB interface to
transfer data from the recorder to a PC. It can
record up to 150 minutes in “LP” mode with a 16MB
SD flash memory card. High quality (HQ), standard
play (SP), and long play (LP) recording modes are
available. Further information on this product
may be obtained from the manufacturer’s web site:
http://www.prodcat.panasonic.com/shop/NewDesign/ModelTemplate.asp?ModelID=
=13081
.
The Sony ICD-MS515 is another audio recorder that
uses flash memory (in the form of a memory
stick). It is 1/3/8” x 4 1/8” x 23/32” in size,
and uses 2 AAA batteries. The MSRP is $250. It
can record for 10 hours in SP mode, and 12 hours
in LP mode on a single set of batteries. It has
voice activated recording, and uses a standard
USB interface to transfer data from the recorder
to a PC. It can record 64 minutes in SP mode
(using 16kHz sampling), and 150 minutes in LP
mode (using 8kHz sampling). It features a built
in omnidirectional microphone, and is a monaural
recorder. Sony also makes a less expensive flash
recorder (ICD-B25) without removable media for
$100. Further information on these products may
be obtained from the manufacturer’s web site:
http://www.sonystyle.com/electronics/prd.jsp?hierc=8627x8667x8668&catid=
=8668&pid=31982&type=p
.
The DIALOG4/ORBAN SOUNTAINER MP3 player/recorder
is another example of a compact audio recorder
that uses flash memory in a multimedia card (MMC)
format. Instead of ADPCM or DSS, it uses MP3
recording of audio. It is comparable in size and
features to other recorders. This manufacturer
prefers that detailed information on its recorder
specifications should not be reproduced. So, for
more information on this recorder, the reader is
referred to the manufacturer’s web site:
http://www.dialog4.com/products/sountainer/supp_snt1.html .
Please note that solid state audio recorders from
Adaptive Digital Systems (EAGLE/FBIRD8) are
available for law enforcement purposes. For
specifications on these products, please see
http://www.adaptivedigitalsystems.com . A
password, which may be obtained from the
manufacturer, is required to access the specifications for these recorders.
Another manufacturer of solid state audio
recorders for law enforcement purposes is Digital
Audio Corporation. The product made by this
corporation is the SSABR, which is described as a
“state of the art, body worn digital recorder,
specifically designed for collecting accurate,
covert recordings.” Details on this product may
be found at
http://www.dacaudio.com . A password,
which may be obtained from the manufacturer, is required to access this dat=
a.
Yet another manufacturer of solid state audio
recorders for law enforcement applications is
Geonautics. This company makes a very small
“Whisper” flash based recorder that is available
in both mono and stereo configurations. Details
on these products may be found at
http://www.geonautics.com . A password, which may
be obtained from the manufacturer, is required to access this data.
2.1.2 MP3 Player/Audio Recorders
-
Another class of commercial product with
potential application for covert recording is the
MP3 player. MP3, or MPEG Layer 3, is a lossy
compression format that allows CD-quality music
recordings to be compressed into files
significantly reduced in size to facilitate
transfer over the internet and to and from PC’s.
MPEG formats accomplish this reduction in size
partly by eliminating components of the recording
that would be masked by the human hearing process
based on a psychoacoustic model of hearing.
Many, but not all, MP3 players have voice
recording capability in addition to MP3 playback
capabilities. The MP3 playback frequency response
is listed as a very high quality range of 20Hz to
20kHz. Unfortunately, the claimed frequency
response of 20Hz – 20kHz applies only to the
playback of MP3 files, not to recorded voice. The
portable MP3 recorder/players reviewed to date
use ADPCM for recording voice. The ADPCM
implementations used have a bandwidth of 3 to
4kHz, which is much worse than the 20-20kHz
achieved when playing back MP3 recordings. The
ADPCM used in the voice recordings is based on 8
bit PCM samples, and has an upper limit of
approximately 50dB for its signal to noise ratio.
The Creative Labs Nomad IIc MP3 player/recorder
is a widely available MP3 player/recorder that
can be used for recording audio onto flash memory
(Smartmedia format flash). It is 3.7” x 2.6” x
0.9” in size, and uses 32kbps G721 (an ITU
standard) ADPCM recording. It features a USB
interface for transferring files to a PC. Further
information on this device may be obtained from
the manufacturer’s web site:
http://www.americas.creative.com/products/category.asp?category=2&maincat=
egory=2
.
The Sensory Science Rave MP2200 samples voice at
8kHz, and requires approximately 1MB of flash
memory space for every 4 minutes of voice
recording. So, for a built in flash memory of
64MB, this unit can store over 4 hours of voice.
The specification of approximately 4 minutes of
voice per 1MB indicates that some compression is
being used to store the voice (approximately a
2:1 compression), which is consistent with 32kbps
G721 ADPCM. Unfortunately, the Rave MP2200 does
not store voice files on removable media, but
only on the built in flash. Cost of the Rave
MP2200 is approximately $200. More information on
the Rave MP2200 may be obtained at the following
URL:
http://www.sonicblue.com/support/goVideo/downloads/MP2200manual.pdf .
A few of the MP3 recorder/players use 40 MB
Iomega Clik! disks as the storage media, which
are much cheaper than the removable flash cards.
However, these disks are susceptible to shock and
vibration, which could be a disadvantage for
certain law enforcement applications.
2.1.3 Minidisc Player/Audio Recorders
-
A third interesting class of commercial products
with potential for covert recording applications
is the minidisc recorder/player. The minidisc is
the most compact of the removable memory storage
media, capable of storing approximately 160 MB of
audio data on a disc that is 64 mm in diameter
and approximately 1 mm thick. A typical minidisc
device is not much bigger than the minidisc
itself, with typical dimensions of 70 mm x 67.5
mm x 5 mm and being very similar in size to the
MP3 player/recorders discussed above. Only Sharp
and Sony currently produce minidisc
recorder/players. These are the only COTS
products reviewed so far that can make voice
recordings in stereo and that can record voice
using the full 44.1 kHz, 16 bit sampling that is
a standard for audio CD’s, allowing the full
recording bandwidth for music or voice of 20 Hz
to 20 kHz. However, to store audio with this
large a bandwidth on the limited amount of memory
space available, all minidiscs utilize a
proprietary ATRAC3 compression scheme for the
storage of data that is lossy, compressing the
audio files by a ratio of approximately 4.83:1.
Both Sharp and Sony have plans to produce higher
density minidiscs and drives with a capacity of about 650 MB.
Pre-recorded minidiscs are fabricated using the
same plastic-aluminum structure as CD’s. The
minidisc is read by focusing a laser on pits and
valleys within the transparent polycarbonate
substrate backed by a coating of aluminum that
then reflects or disperses the beam to produce a
series or 1’s and 0’s which can then be
translated back into either the original data or
sound. Recordable minidiscs have a pre-groove
instead of the CD-type pits and valleys and a MO
(magneto-optical) coating instead of the aluminum
one. While recording, the laser focuses on the
pre-groove and heats the MO recording layer at
that point to its Curie point while a magnetic
field from a head in contact with the other side
of the disc aligns magnetic dipoles within the
heated spot on the MO layer. During playback, the
laser focuses on the pre-groove again, but at a
lower power, allowing the measurement of changes
in the polarization of the light reflected from
the previously magnetized layer. All minidisc
players have a dual function optical assembly
that detects the disc type and switches between
the measurement of reflectivity for pre-recorded
minidiscs or polarization for recordable
minidiscs. Sony claims recordable minidiscs can
handle up to 1 million recordings. The minidiscs
have a user table of contents that can be damaged
if the minidisc is abused and render the minidisc
unusable. Sony claims that data using
magneto-optical technology can be stored for more
than 30 years without loss or degradation.
However, strong magnets placed directly against the minidisc can destroy da=
ta.
Minidiscs use a buffer memory that temporarily
stores recorded audio, thereby helping to prevent
vibrations from affecting either the recording
onto or playback from the minidisc. However,
problems have been reported with recording when
the minidisc recorder is subjected to shock and
vibration, apparently due to the laser beam
“skipping” and accidentally erasing previously
recorded data on adjacent tracks. Therefore, it
is recommended that the recorder should be
immobile and not subjected to shock or vibration
while recording. In addition, because of the
400-900 rpm rotation of the minidisc, all such
devices produce a humming noise when recording or
playing audio. Although this humming noise
reportedly does not degrade the recording or
playback process, it could possibly interfere
with the covert recording process. Because the
laser beam must heat the disk while recording,
the minidisc device is the only portable
recording device that consumes more power during
recording than during playback. And even during
playback, the devices still consume 50-100% more
power than any other class of recording device.
Until recently, none of the minidisc
recorder/players have had a convenient means to
connect to a PC to allow the rapid transfer of files.
The Sony MZ-N707 minidisc recorder offers some of
the advantages of flash recorders. It records
onto a digital medium (the minidisc), and is not
subject to the tape hiss that is present in
cassettes. The minidisc must spin to work, so,
unlike flash recorders, there are moving parts
inside the minidisc recorder. The size of the
MN-N707 is 3 ¼” x 3” x 1 1/8”. It comes with a
rechargeable battery, and records in a high
fidelity ATRAC3 format. An external microphone is
needed to record audio, since the unit does not
come with a built in microphone. It uses Sony’s
ATRAC3 compression technique for storing audio
(and music). The ATRAC3 compression technique
achieves relatively high fidelity, but it is not
lossless. Another model, the MZ-N1, is available
for $350, and it is somewhat smaller in size: 3”
x 2 7/8” x ½”. The MZ-N1 features a higher
capacity battery than the MZ-N707. Further
information on these devices may be found at the
following URL:
http://www.sonystyle.com/electronics/ssctypg.jsp?hierc=8627x8650x8647&cat=
id=8647
.
2.1.4 Digital Audio Tape (DAT) Recorders
-
One DAT device, a TCD-D100 produced by Sony, is
included in this survey for comparison purposes.
This DAT recorder, which lists for $900, can
provide up to 4 hours of stereo recording on two
AA batteries. This DAT device can sample at
48kHz, 44.1kHz or 32kHz, and uses 16 bit
quantization. At a 48kHz sample rate, it has a
20-22 kHz frequency response (within 1 dB), which
is greater than the full range of human hearing
(20-20kHz). At 44.1 kHz and 32 kHz sample rates,
it has a 20-14.5 kHz frequency response (within 1
dB). The signal to noise ratio is 87dB, and the
total harmonic distortion is 0.008%. The wow and
flutter is less then 0.001 percent. All of these
specifications are excellent, and stack up
favorably against the solid state recorders. DAT
tapes are available providing 60 minute and 120
minute recording times. A digital output is
available, but it is not known how easily a
digitized recording could be transferred to a PC
using this output. Recordings can be transferred
to a PC in real-time using the Line In/Line Out
connections. A microphone must be purchased
separately. Further information on this device
may be found at the following URL:
http://www.sonystyle.com/home/item.jsp?hierc=9687&catid=8662&itemid=5=
91&telesale=null&hidden=null&cps=null&type=s
. A related product, the NT-2 Digital Micro
Recorder is also available from Sony. The NT-2 is
smaller than the TCD-D100, but it has slightly
worse specifications. Further information on the
NT-2 may be found at:
http://www.sonystyle.com/home/item.jsp?hierc=9687&catid=8668&itemid=5=
63&telesale=null&hidden=null&cps=null&type=s
.
2.1.4 Microcassette Recorders
-
Microcassette and cassette recorders are used by
numerous law enforcement agencies. These devices
typically have relatively poor frequency response
(250 Hz to 4 kHz for microcassettes, somewhat
better for cassettes), have relatively high wow
and flutter (due to the mechanical tape
transport), and have poor signal to noise and
distortion characteristics when compared to solid
state recorders. Often they have voice activated
recording and automatic level control that cannot
be turned off. In many law enforcement
applications, it is an advantage to be able to
turn off voice activated recording and automatic
level control. Some advantages of the
microcassette and cassette recorder are they are
small, they are low cost, they use batteries that
are widely available, and they use cassettes that are widely available.
The Sony M850V is a typical microcassette
recorder. It has a frequency response of 250 Hz
to 4 kHz, which is well below the human hearing
range of 20 Hz to 20 kHz. It has an 11 hour
battery life, and uses 2 AA batteries. It is
relatively small, with dimensions of 4” x 2.25” x
7/8”. It has voice operated recording and
automatic level control. This recorder is
monaural, and has its own built in microphone and
speaker. It features two recording speeds. The
frequency response quoted is for the higher
recording speed. Further information on this
recorder may be found at the following URL:
http://www.sonystyle.com/home/item.jsp?hierc=9687x8667x8671&catid=&item=
id=34003
.
2.2 Commercial Audio Recorder Issues
-
One issue (potential shortcoming for law
enforcement use) with the flash recorders and MP3
player/recorders is the bandwidth achieved in the
audio recording. For the recorders with LP (long
play) mode, relatively low sampling rates are
used in recording the data (longer recordings can
be made in a given memory size when lower sample
rates are used). Unfortunately, the Nyquist
sampling criteria limits the bandwidth of the
recording to one-half of the sampling frequency.
Thus, the audio bandwidths for recordings made in
the LP mode are relatively low. A typical sample
rate for LP recording is 8kHz, and, by the
Nyquist sampling criterion, the resulting audio
bandwidth must be less than 4kHz (typically
3kHz). This audio quality is approximately the
same as telephone voice quality. The 32kbps ADPCM
recording scheme used by the MP3 player/recorder
in Table 2-1 also achieves a telephone voice
quality bandwidth of approximately 3kHz. A
somewhat better frequency response is achieved by
the SP (standard play) mode of the flash
recorders. Sampling rates used in this mode are
12kHz, and the resulting audio bandwidth must be
less than 6kHz (typically 5kHz). Although these
bandwidths are adequate for speech recognition
purposes, they do not compare favorably to the
human hearing bandwidth of approximately 20 kHz,
and they may not be suitable for all law enforcement applications.
Another issue with flash recorders and MP3
player/recorders is the loss in fidelity caused
by compression schemes used by the recorders. The
recorders attempt to maximize recording time for
the available memory by compressing the sampled
audio using proprietary compression schemes. The
higher compression algorithms (greater than 4:1),
which conserve the most memory space, turn out to
be lossy; that is, they degrade the fidelity of
the recording. The 32kbps ITU G.711 ADPCM
compression used by the MP3 player/recorder in
Table 2-1 is a relatively low loss algorithm. The
minidisc recorder uses ATRAC3 compression, which
has a compression ratio of 4.8:1. Manufacturers
that use the MP3 compression standard have a
compression ratio of greater than 10:1. The DSS
(digital speech standard) compression scheme used
by the Olympus DS2000 stores 120 minutes (7200
sec) of highest quality voice in 16Mbytes of
flash. A rough calculation of the seconds of
uncompressed speech that can be stored in 16Mbytes is:
16M bytes x 1sec/12k samples x 1 sample/1 byte = 1333 sec
Comparing the compressed seconds of storage to
the uncompressed seconds of storage gives a
compression ratio of 5.4 for DSS. Although
playback quality may not be affected
significantly by lossy compression schemes for
most purposes, one of the concerns in using
nonlinear compression for law enforcement
recordings is the legal question that might be
raised over the accuracy and faithfulness of the recording.
A third issue with the commercial audio recorders
has to do with dynamic range and signal to noise
ratio of the recorded audio. The 8 bits per
sample used in these recorders provides for a
signal to noise ratio that will not exceed 50dB
(i.e., 6.02n + 1.76). This signal to noise ratio
may not be adequate for all law enforcement
applications. For example, if we try to recover
low level audio that is more than 50dB below some
high level audio, it will be buried in noise.
Another issue in using commercial audio recorders
for law enforcement purposes is whether or not to
use automatic gain control. It would be desirable
to be able to defeat the automatic gain control feature for some applicatio=
ns.
Another issue is whether or not to use voice
activated recording. Voice activated recording
conserves room on the recording medium, and it
extends battery life by shutting down the process
of recording the audio onto the storage medium
when no audio is present. However, a threshold
must be set to activate the voice recording. If
the threshold is set too high, some weaker
signals that are desirable evidence may be
missed. So, for recorders that do have voice
activation, it is desirable to be able to turn off the voice activation.
Another issue is storage of original evidence at
a reasonable price. Unfortunately, there is often
a significant delay of months or even years
before a case comes to trial. It would be costly
to have to remove the recorder from use while
waiting on a trial. And downloading the original
recording to a PC may not be accepted as original
evidence. The original flash memory module used
to record the audio may be the only recording
accepted as original evidence. The ability to
have removable flash memory that can be saved as
original evidence is a desirable feature of flash
based audio recorders. Using removable flash
memory allows the recorder to continue to be used
(with a new flash module) while the original
evidence flash module is saved for trail.
3.0 CURRENT AND PROJECTED STATE OF THE ART IN
SOLID STATE RECORDER TECHNOLOGY AREAS
3.1 Solid State Recorder Block Diagram
-
Figure 3-1 shows a block diagram of a typical
solid state voice recorder. Starting at the top
left of the figure, the audio signal is received
by one or more microphones. Next, an
amplifier/filter increases the voltage of the
signal from the microphone to the correct level
for the input to the analog to digital converter
(ADC). Some filtering (removal) of unwanted
signals may also occur in this block. The ADC
converts the analog input signal to a digital
word that is fed to audio compression hardware.
The audio compression hardware (for example, a
DSP) implements an audio compression algorithm,
which reduces the number of bytes needed to store
the audio signal. The resulting compressed audio
bytes are saved in flash memory. The flash memory
is removable, for convenient storage of evidence.
The user interface is implemented by the DSP
sensing the switch positions on the recorder.
Figure 3-1. Solid state voice recorder block diagram.
-
Playback of the audio signal may be provided in a
number of ways. Commercial items, such as the
SanDisk ImageMate USB CompactFlash/SmartMedia
media reader (cost is approximately $30), may be
used to transfer the compressed audio from the
flash memory card to a PC. The PC would run
software to decompress the audio, and play back
the result on the PC sound system. The flash
recorder manufacturer could provide custom audio
decompression PC software, or, if a standard
compression/decompression algorithm is used, a
third party could provide the audio decompression PC software.
In the design of the flash based audio recorder,
the resulting product is only as good as the
worst individual part. The various parts of the
recorder will be discussed in the following sections.
3.2 Microphones
-
Two major classes of microphones that could be
used in flash based audio recorders are dynamic
and electret condenser. The dynamic microphone
transforms sound into an electrical signal by the
movement of a diaphragm with a coil of wire
attached to it. This coil of wire is located
close to a magnet, and when the diaphragm/coil
moves in the magnetic field, a current is
produced in the coil. This current corresponds to
the audio signal that moves the diaphragm.
Electret condenser microphones operate somewhat
differently. The diaphragm and the “back plate”
in an electret condenser microphone form two
surfaces of a capacitor. Either the diaphragm or
the back plate contains a permanently charged
electret material. When the diaphragm moves, the
distance between the surfaces of the capacitor
changes, inducing a current that corresponds to
the audio input. Since the electret condenser
microphone diaphragm does not have a coil
attached to it, it can be fairly light when
compared to the dynamic microphone. As a result,
the electret condenser microphone generally has
better sensitivity and high frequency response than a dynamic microphone.
One characteristic of a microphone is its
directivity, which is its sensitivity to sound
arriving from different directions. A microphone
that picks up sound equally well in all
directions has an “omnidirectional” pattern. A
microphone that is more sensitive to sounds in
front of the microphone that behind the
microphone is “unidirectional” (cardioid). A
microphone that is sensitive to sounds in front
and behind, but not to the sides, is “bi-directional” (noise canceling)=
.
One typical electret condenser microphone is the
Panasonic WM-61 series. This back electret
condenser omnidirectional microphone has a
frequency response from 20 to 20kHz, a signal to
noise ratio of better than 62dB, and a
sensitivity of –35dB (0dB = 1V/Pascal). The WM-61
has a low power consumption of 0.5 mA at 2V.
For law enforcement applications, the superior
frequency response and sensitivity of the
electret condenser microphone is preferred over
the dynamic microphone. For monophonic
recordings, an omnidirectional microphone is
preferred. For stereo applications, either
cardioid or omnidirectional pattern microphones may be used.
3.3 Delta Sigma Analog to Digital Converters
-
In solid state audio recorders, audio signals
must be converted from the analog domain to the
digital domain so they can be stored in a digital
format in flash memory. This function is
performed by a high quality analog to digital
converter (ADC). One type of ADC that is
particularly well suited to this task is the
delta sigma ADC. The delta sigma ADC samples the
input waveform at a much higher rate than is
normally required (often 128X oversampling is
seen in these devices). Oversampling distributes
the quantization noise all the way up to the
sampling frequency, thereby reducing the amount
of quantization noise in the audio band. The
delta sigma ADC also uses a noise shaping filter.
This noise shaping filter effectively moves
quantization noise from the audio band to higher
frequencies. The delta sigma converter then uses
a digital filter to remove the higher frequencies
(and quantization noise), and retain the audio
frequencies. The resulting digitized signal from
the delta sigma converter has very little
quantization noise, and is highly accurate
representation of the input analog waveform.
Delta sigma converters are typically inexpensive,
have low power requirements (suitable for a
battery operated voice recorder), and are highly
accurate. One device, the CS5333, converts two
inputs (for stereo operation), provides 24 bits
of output, requires only 11mW, and costs less
than $5. As seen in the specifications of the
CS5333, the current generation of delta sigma
analog to digital converter has more than enough
performance to meet law enforcement needs of high
dynamic range and full audio bandwidth, and it is reasonably priced.
Delta sigma ADCs and digital to analog converters
(DACs) are used in audio sound cards for PCs. The
demand for these products may roughly be expected
to follow the demand for PCs in the future. Which
is to say, the high demand for delta sigma
converters (both digital to analog and analog to
digital) in computer audio sound systems makes
the continued availability of these devices highly likely.
3.4 Audio Compression Algorithms
-
Audio compression algorithms are used in flash
based audio recorders to reduce the amount of
flash memory required to record a specified
duration of audio. If 20kHz audio is sampled at
the Nyquist rate of 40kHz, then, in the absence
of compression, each second of audio requires
40,000 audio samples to be stored in flash
memory. A compression algorithm that achieves a
compression ratio of 4:1 would reduce the flash
memory storage requirements from 40,000 down to
10,000. A fixed amount of flash memory can store
4 times as much audio when a 4:1 compression algorithm is used.
Audio compression algorithms may be divided into
two categories: lossless and lossy. When
recordings made using lossless compression are
played back, the original signal is reproduced
exactly, and no compression artifacts are
present. When recordings made using lossy
compression are played back, the original signal
is not exactly reproduced, but a slightly
degraded version of the original signal is
reproduced. Lossless audio compression schemes
typically achieve compression ratios in the range
of 1.5:1 to 3:1. Lossy audio compression schemes
typically achieve compression ratios in the range
of 4:1 to 12:1 and higher (Windows Media Audio claims 24:1).
The state of the art lossless audio compression
process can be divided into three stages:
framing, decorrelation, and entropy coding.
Framing divides the audio signal into equal
duration frames. Optimum duration frames appear
to be in the range of 13 to 26ms. Audio signals
exhibit a high degree of autocorrelation; that
is, the current sample can be predicted from
previous samples. To take advantage of this
characteristic, the original signal is
decorrelated (and the correlation characteristic
is remembered). It is more efficient to store the
correlation characteristics in the encoded
waveform than it is to store the audio samples.
Several techniques are available for performing
decorrelation: coding with linear prediction,
coding with approximation, and transform coding.
Once the correlation in the waveform has been
removed, the remaining decorrelated waveform must
be encoded. Entropy coding is used for this
purpose. Some standard entropy coding methods
include: Huffman coding, run length coding, and
Rice coding. Some representative state of the art
lossless audio encoders include the following:
- AudioPAK (integer)
- MUSICompress (fixed point)
- Sonarc (fixed point)
- Shorten (floating point)
- Ogg Squish (floating point)
- LTAC (lossless transform audio compression - floating point)
- Waveform Archiver (floating point)
The objective of the AudioPAK algorithm is to
reduce the complexity of implementing lossless
audio compression while maintaining compression
ratios that are comparable to the more complex
lossless audio compression algorithms. Notice
that the AudioPAK uses integer operations, while
the other algorithms mentioned use fixed point or
floating point operations. Much of the material
in this section has been derived from:
Optimization of Digital Audio for Internet
Transmission by Mat Hans – Georgia Institute of
Technology PhD thesis, 1998 (
http://users.ece.gatech.edu/~hans/ ). This thesis
describes the AudioPAK algorithm.
Another lossless compression technique worth
mentioning is bit plane encoding. With this
technique, a particular bit of each PCM sample is
encoded over a frame of samples. This process is
repeated until all bits have been encoded. This
technique expects that the most significant bits
of audio PCM samples will not change very often,
and can be efficiently run length encoded. The
least significant bit is expected to change
frequently, and can use Huffman entropy encoding.
Before leaving the subject of lossless audio
compression, it is interesting that lossless
audio compression can be achieved by using the
PKzip (Winzip) algorithm that is so familiar to
today’s computer users. Unfortunately, the PKzip
algorithm does not achieve very good compression
ratios for audio files (typically 1.1:1). One
reference that discusses PKzip (Winzip)
performance relative to other lossless audio
compression techniques, and lossless audio
compression performance in general is “Digital
Audio Gets an Audition, Part 1 Lossless Compression,” EDN, January 4, 200=
1.
A number of lossy audio compression techniques
are available. We will briefly describe two here:
MP3 and ATRAC. These two compression techniques
are used in MP3 players and mini disc recorders respectively.
Although the MP3 players do not use MP3
compression to record audio (they typically use
32kbps ADPCM instead), the compression scheme is
one dominant form of lossy audio compression that
is used today (the Soundtainer product mentioned
above uses MP3 to record audio). MP3 stands for
MPEG Audio Layer 3 (as opposed to the lower
compression MPEG Audio Layers 1 and 2). MP3 can
achieve compression ratios of 10:1 to 14:1 with a
bandwidth of over 15kHz. One method the MP3
algorithm uses to reduce the amount of
information to be encoded (thereby compressing
the size of the audio file) is to omit audio that
is not perceptible to humans. One example is
called frequency masking. In this situation a
loud sound present in one frequency band masks
softer sounds present in an adjacent frequency
band. In this instance, humans will not notice a
difference when the soft sounds are completely
removed. Similarly, MP3 observes a minimum audio
threshold, and will not record sounds below a
certain level at certain frequencies (2 to 5kHz),
since these will not be perceptible to humans.
MP3 “borrows” from a “reservoir of bytes” to
encode more complex audio, and “replenishes” the
reservoir during less complex audio passages. MP3
uses a discrete cosine transform that has 384
coefficients to decorrelate the audio signal. It
then throws away data that would not be noticed
by the listener. Finally, MP3 uses Huffman
entropy encoding, once the audio that will not be
encoded has been subtracted from the signal.
ATRAC (Adaptive Transform Acoustic Coding) is the
compression technique used in mini disc
recorders. It typically achieves a 5:1
compression ration on CD audio, and, like MP3, it
uses a psychoacoustic model of human hearing to
determine what sounds may be subtracted form the
original signal without being detected by human
hearing. ATRAC divides the audio frequency band
into 3 subbands (0-5.5kHz, 5.5-11kHz, and
11-22kHz) using Quadrature Mirror Filters (QMFs –
prevents aliasing when reconstructing). A
discrete cosine transform is performed on each
subband using an adaptive block length (long or
short). Long block lengths provide superior
frequency resolution, but are subject to “pre
echo” during “attack” portions of the audio
signal. Short block lengths are used to prevent
pre echo. and transforms these subbands into the
frequency domain. Signals that would be masked by
psychoacoustic effects are subtracted from the
resulting frequency domain coefficients, and the
coefficients are encoded into BFUs (block
floating units). (reference: ATRAC: Adaptive
Transform Acoustic Coding for Mini Disc, Tsutsui
et al, 93rd Audio Engineering Society Convention,
Oct 1-4, 1992.) A newer version of ATRAC called is ATRAC3 is now available.
Both MP3 and ATRAC (as well as many other lossy
audio compression algorithms – AC3 for example)
use a psychoacoustic model of human hearing to
remove signals that would not be perceived by
humans as a method of reducing the amount of
audio that must be saved (as a method of
compressing audio). For law enforcement
applications, this practice may not be acceptable
in some situations. For example, if a soft sound
contains information needed by law enforcement
personnel, and a loud sound “masks” it, the soft
sound will be removed from the encoded audio when
either MP3 or ATRAC is used. For this reason,
audio compression algorithms that rely on the
psychoacoustic model of human hearing to delete
audio signals are not recommended for law
enforcement applications (at least not all law enforcement applications).
Other lossy audio compression algorithms include:
- AAC (Advanced Audio Coder www.aac-audio.com )
- ATELP ( www.softsound.com/ATELP.html )
- DTS ( www.dtsonline.com )
- ePAC ( www.lucent.com/ldr )
- Indeo ( www.ligos.com )
- Ogg Vorbis ( www.vorbis.com )
- Qdesign ( www.qdesign.com )
- Real Audio ( www.real.com )
- TAC (kk-research.hypermart.net)
- TwinVQ (sound.splab.ecl.ntt.co.jp/twinvq-e)
- Windows Media Audio ( www.microsoft.com/windows/windowsmedia )
A reference that gives an overview of these
algorithms and their performance is “Digital
Audio Gets an Audition, Part 2 Lossy Compression,” EDN, January 18, 2001.
One further audio compression algorithm worth
mentioning is apt-X 4:1. This algorithm uses
ADPCM to achieve a 4:1 compression ratio, with
very little loss in audio quality. This algorithm
uses four frequency subbands, but does not rely
on psychoacoustic models of human hearing to
throw away audio information that is not
perceptible to humans. More information may be
found on this technique at
http://www.aptx.com .
3.5 Audio Compression Hardware
-
Audio compression algorithms are implemented on
audio compression hardware, which includes
Digital Signal Processors (DSPs) and custom
Application Specific Integrated Circuits (ASICs).
DSPs are specialized computer chips that have
features that facilitate the implementation of
audio compression algorithms. Like any computer,
DSPs may be reprogrammed to perform different
functions. ASICs are not reprogrammable. The
Field Programmable Gate Array (FPGA) may be used
to develop algorithms that are then readily transferred into an ASIC.
Highly capable, low cost DSPs have become
available in the past few years. For example, the
TMS320VC5402 DSP from Texas Instruments is
capable of 100 million instructions per second
(MIPS), and has 16K words of on chip RAM, and 4K
words of on chip ROM. The cost of this part is
approximately $6 in quantities of 1000. A part
specifically designed for low power consumption,
the TMS320VC5502, is also becoming available. It
has 32K words of on chip RAM and 4K words of on
chip ROM, and features 400 MIPS performance. The
5502 part will sell for approximately $10 in
quantities of 1000. Both the 5402 and the 5502 are fixed point processors.
The trends toward lower core voltages, smaller
geometry devices, and higher processing
capabilities in DSPs and ASICs can only benefit
flash based audio recorders. The current
capabilities of DSPs like the 5502 are more than
adequate for implementing fixed point and integer
lossless audio compression algorithms for flash based audio recorders.
3.6 Flash Memory
-
Flash memory is used to store the compressed
audio in the solid state recorder. Flash memory
is used in cell phones, digital cameras and MP3
players. The cost of the flash memory is the
dominant cost of the recorder. A 16kHz Obviously,
any reductions in the cost of flash will reduce
the cost of the flash based audio recorder.
In mid 1998, an Intel 28F640J5 8 M byte flash
part cost $65. Today (2002) a comparable part,
the 28F640J3A, costs $13.42, a reduction of
nearly 5 times. Perhaps even more important than
the cost of the individual flash chips is the
cost of removable flash media. Driven by the
proliferation of digital cameras and MP3 players,
the cost of removable flash media has dropped
significantly in the past years. Today’s street
prices for SanDisk flash products are as follows:
CompactFlash 1 G Byte: $631
CompactFlash 512 M Byte: $269
CompactFlash 256 M Byte: $115 (if 1000 units are purchased, the cost is $10=
2)
CompactFlash 128 M Byte: $63
Memory Stick 128 M Byte: $70
Secure Digital 256 M Byte: $161
MultiMedia 64 M Byte: $52
Ultra CompactFlash 128 M Byte: $77
Flash memory provides a method of storing digital
audio data that is non volatile; that is, data
stored in flash memory is not lost when the power
is turned off to the device. Flash memory may be
NAND based or NOR based. NAND based technology is
considered well suited for high capacity data
storage applications, such as storage of audio
files. Current flash memory for file storage
often uses 2 bit per cell storage, an improvement
over the older single bit per cell flash technology.
Flash memory that uses 0.25, 0.16, and 0.13
micron semiconductor process technology is
currently available, and smaller process
technology is being planned. Parts that operate
on 3V and 1.8V are commonplace, and lower
voltages are being planned. These developments
are expected to reduce cost for given storage
size devices, and lower power consumption, which
would both benefit flash based audio recorders.
SanDisk is expecting prices to drop approximately 30% over the coming year.
3.7 Batteries
-
Batteries for mobile electronic applications such
as digital cameras and MP3 players may be divided
into two groups: rechargeable and
non-rechargeable. Within the rechargeable group,
the most popular technologies today are: nickel
metal hydride (NiMH) and lithium ion (Li-Ion). In
the non-rechargeable group, the most popular
technologies are the alkaline and carbon zinc batteries.
Nickel metal hydride batteries require recharging
more often than lithium ion batteries, but they
cost less than lithium ion batteries. Lithium ion
batteries provide a better energy density than
nickel metal hydride batteries. An energy density
figure of 75 Watt hours per kilogram is provided
by nickel metal hydride batteries, versus 135
Watt hours per kilogram for lithium ion
batteries. The output voltage of lithium ion
batteries is typically higher (3.0V) than the
output voltage of nickel metal hydride batteries
(1.2V). When compared to nickel cadmium
rechargeable batteries, both nickel metal hydride
and lithium ion batteries offer the advantage of
not having any memory effect. (
http://www.nec-tokin.net/now/english/product/me/chisiki/li3.html ).
The alkaline battery in a AA size can provide
3000mAh at 1.5V, and an energy density of 140
Watt hours per kilogram. In comparison, a AA
carbon zinc battery in AA size can provide only
950mAh at 1.5V and an energy density of 50 Watt hours per kilogram.
Alkaline and carbon zinc batteries have sloping
discharge curves. That is, as the battery is
discharged, the voltage goes down over time. In
contrast, the nickel metal hydride and lithium
ion batteries have flatter discharge curves. When
these batteries are discharged, the voltage does
not go down over time as much as with the alkaline and carbon zinc batterie=
s.
The popularity of laptop computers, cell phones,
cordless phones, digital cameras, MP3 players,
and personal digital assistants has spurred the
demand for rechargeable batteries. In 2000, the
market for rechargeable batteries was $1.75
billion. This market is projected to grow to
$2.19 billion by 2006. The technology that will
account for most of the battery demand in 2005 is
the Lithium Ion. Lithium Ion batteries are
expected to grow from 25% of the battery market
in 1999 to 55% of the battery market in 2005.
(source:
http://www.eetimes.com/myf00/ao_batt.html )
One emerging battery technology is Lithium Ion
Polymer. This battery technology has the
potential to greatly increase the energy density
when compared to current Lithium Ion products.
Another Lithium Ion emerging battery technology
replaces the cobalt in the battery with a
different cathode material. The problem with
cobalt is that it requires protection circuits
inside the battery to prevent thermal runaway
when the battery is being charged. One company,
Valence Technology, claims that using Saphion for
the cathode will reduce the cost of lithium ion
batteries (
http://www.valence.com/saphion.asp ).
The development of a low cost, widely available,
lithium ion polymer battery with high energy
densities could reduce the size required for
batteries in the flash based audio recorder. A
reduced size recorder has obvious advantages for
law enforcement purposes. It is uncertain when
lithium ion polymer batteries will reach this stage of development.
4.0 PROJECTED COMMERCIAL SOLID STATE RECORDER PRODUCTS
(IN 2 YEARS)
-
Solid state recorder products are in a state of
rapid development and improvement, with models
constantly being discontinued and replaced by
newer, improved models. Since this program and
evaluation of devices began, many of the models
initially in the product matrix (see Appendix A)
had to be dropped and replaced by more current
models. The trend has been toward recorders/MP3
players with larger memory capacity and lower
costs. This trend is expected to continue.
But the demand for higher bandwidth portable
voice recorders has not been seen yet. There is a
strong demand for MP3 players (i.e., there are
lots of MP3 player products being sold), which
feature voice recorders as a secondary feature.
And there is a strong demand for voice recorders
used for business dictation applications. But
these kinds of voice recorders do not need to
have high bandwidth, and no mass market
commercial voice recorder has bandwidths up to 16
kHz or higher that would be useful for law enforcement applications.
As removable flash memory cards continue to
increase in memory storage space and decrease in
cost, the solid-state MP3 recorder/players may
evolve to take advantage of this storage space
and become true music recorders as opposed to
simply voice recorders. Commercial motivation may
encourage these recorder/players to take
advantage of the 44.1 kHz sampling frequencies
currently used to decompress and playback the MP3
files and begin recording audio files in stereo
with MP3 compression as opposed to simply
decompression. These devices could then compete
very favorably with the minidisc recorder/players
in the marketplace, being slightly smaller in
size and less power hungry. But the market is not
seen for MP3 recorders. Most consumers are not
interested in recording their own music, via a
microphone. Instead, they are interested in
transferring music tracks from CDs or from the
web to their PC and then storing them on MP3
players. And even if a MP3 recorder did evolve,
the lossy compression used in MP3 is not always
suitable for law enforcement purposes.
The market for high fidelity portable audio
recorders would seem to be pretty much the same
as the market for portable digital audio tape
(DAT) recorders. Today, this market is a low
volume, relatively high cost niche. For example,
the Sony TCD-D100 DAT recorder lists for $900 and
it was difficult to find a dealer that sells this
device. It may be possible that DAT will evolve
into a flash based product, but the demand to
make it a low cost item sold in large quantities is not seen.
5.0 AUDIO RECORDER CONCLUSIONS
-
Desired characteristics of audio recorders for
law enforcement purposes are as follows:
- wide and flat frequency response (20-20kHz)
- high signal to noise ratio and dynamic range
- low wow and flutter
- low harmonic distortion
- lossless compression
- defeatable automatic level control
- defeatable voice activated recording
- stereophonic recording
- combined microphone response that is omnidirectional
- record times of at least 60 minutes, with 120 minutes and higher availabl=
e
- removable media
- low cost media
- wide availability of media
- wide availability of batteries
- small size
- low cost
Large volumes are projected in MP3 player
recorders, and, to a lesser extent, in solid
state voice recorders. But large volume (low
cost) products (current and projected) fall short in several key areas:
Current and projected future commercial flash recorder products weaknesses:
1. lossy compression
2. poor frequency response
3. poor dynamic range and signal to noise ratio
Current and projected MP3 player voice recorder combination product weaknes=
ses:
1. poor frequency response
2. poor dynamic range and signal to noise ratio
Current and projected mini disc weaknesses:
1. lossy compression
2. Susceptibility to shock and vibration
Many commercial products use automatic level
control and voice activated recording features
that cannot be defeated. The ability to defeat
these features is desirable for many law enforcement applications.
To record 120 minutes of uncompressed 16 bit PCM
audio with a bandwidth of 16 kHz (sampling at 32
kHz) requires 460 Mbytes of flash. State of the
art lossless compression algorithms achieve 2:1
to 3:1 compression ratios. So, when lossless
compression is used, only 230 Mbytes of flash are
needed (instead of 460). Four years ago, the cost
of flash was approximately $8 per Mbyte, making
the cost of the flash memory required in the
above situation about $1200. Today, the cost of
flash has dropped to less than $0.50 per Mbyte
(street price). A 256 Mbyte CompactFlash plug in
card can be purchased for $128 or less. So the
falling cost of flash has improved the
affordability of the flash based audio recorder.
$128 is a significant amount of money to spend on
a recording medium that may be put on the shelf
while waiting on a trial. But it is much better
than the $1200 or so it would have cost 4 years
ago. And the expected future improvements in the
cost of flash memory will reduce this $128 to an even lower figure.
Falling prices and larger sizes of flash memory
make the solid state recorder a very practical
idea. Improved wow and flutter, increased
immunity from shock and vibration, and the
elimination of tape hiss (improved signal to
noise ratio) result from the use of flash. But
widespread commercial demand for improved
frequency response, high dynamic range, high
signal to noise small solid state recorders is
not seen. Instead, business uses of recorders for
dictation purposes, which do not require high
bandwidth, high dynamic range, and high signal to
noise ratios are seen as the driving factor in
future solid state audio recording products.
There is a market for high bandwidth, high
dynamic range, high signal to noise ratio
playback products (MP3 players), but only for playback, not for recording.
To get the characteristics of lossless
compression, high bandwidth, high dynamic range,
and high signal to noise ratio, law enforcement
personnel must continue to purchase specialty
products such as the FBIRD or the SSABR. No
projected future high volume commercial product
will provide all the capabilities of these devices.
It would be possible to make a product that would
satisfy today’s law enforcement demands at a
reasonable cost. Appendix B shows the major
components of such a product, and that the cost
would be around $140 in parts (excluding circuit
boards and cases, and assuming parts are
purchased in quantities of 1000). The major cost
factor in this product is the flash memory, which
accounts for $102 out of the $140 total component
cost. The product would feature lossless
compression (2:1), a bandwidth of 16kHz, a signal
to noise ratio approaching 90 dB, and over 120
minutes of record time. It would feature
removable flash media, and not have automatic
level control or voice activated recording. It
would have a size of approximately 9 square inches.
6.0 COMMERCIAL BODY WIRE PRODUCTS
-Law enforcement agencies utilize body-wires for
officer security, and to obtain evidence. Audio
quality, transmitted power, and price vary with
different systems. Typically, the transmission
range can be between 30 and more than 3,000 feet
depending on the environment and the quality of
the equipment. In addition, there are many
different frequencies utilized for transmission.
The purpose of the body wire is to transmit audio
in the form of a radio signal to be received,
understood and/or recorded at a remote location.
The person wearing the body wire can be moving
and turning in locations that range from outside
to inside a building and from ground level to
many stories up. While the transmitter is often
mobile, the receiver is generally in one
location. In the case of vehicle audio
surveillance, both the transmitter and receiver
are in motion, but the transmission distance is generally constant.
The transmitter may be required to cover
thousands of feet or a few yards. The different
environments of the signal propagation path will
cause different attenuation levels: the signal
will become attenuated and the range, therefore,
reduced if it is required to travel through
numerous buildings. Noise conditions, present at
the time the audio signal is recorded, will vary
from those of an outdoor, urban environment
(which has many possible levels of background
noise) to that of a quiet indoor room.
Throughout this program, data has been collected
on body wires that could be used for law
enforcement applications under the conditions
described above. A wide variety of body wires are
available. These products are discussed in the following sections.
-6.1 Overview of Body Wire Products
-
Table 6-1 presents a summary of the performance
of various kinds of body wires. Three types of
body wires cover the majority of body wire
products: Narrow Band FM (NBFM), Digital, and
Spread Spectrum Digital. Actual representative
products were evaluated to fill in this table,
but the band names of the products have been
omitted. Strengths and weaknesses of each body
wire category are listed in the table.
Table 6-1. Body Wire Types and Feature Comparison
-
-6.1.1 Narrow Band FM Body Wires
-
Narrow band frequency modulation (FM) body wire
transmitters use the output signal of the
microphone to frequency modulate a radio
frequency (carrier) to form the transmitted
waveform. This modulation technique is an analog
modulation technique, since the microphone signal
was not first digitized before modulating the carrier.
Frequency modulation leaves the amplitude of the
carrier constant, but changes the “instantaneous”
frequency of the carrier in accordance with the
amplitude of the signal from the microphone. Loud
audio signals from the microphone correspond to
relatively large changes in the frequency of the
carrier. Soft audio signals from the microphone
correspond to relatively small changes in the
frequency of the carrier. Since FM signals use
the frequency instead of the amplitude of the
carrier to carry the audio from the microphone,
they are inherently immune to amplitude noise.
Commercial FM radio stations use a form of FM
called wideband FM. In this case, the frequency
of the carrier can change up to 75kHz due to loud
audio from the microphone. In contrast, body
wires use narrowband FM. For narrowband FM the
frequency deviation caused by loud audio from the
microphone is much less than the wideband case.
In the case of narrowband FM, the frequency of
the carrier can change up to 5 or 7kHz.
-6.1.2 Digital Body Wires
-
A digital body wire passes the output of the
microphone through an analog to digital converter
(ADC), the output of which is a series of 1’s and
0’s referred to as bits of digital data. This
digital data then modulates a radio frequency
(carrier) using a digital modulation technique.
One example of a digital modulation technique
that is commonly used in digital body wires is
phase shift keying (PSK). In the case of phase
shift keying, the phase of the carrier is changed
according to whether a 1 or a 0 data bit is being
transmitted. For example, transmitting a 0 data
bit could correspond to no change in the phase of
the carrier, and transmitting a 1 could
correspond to a 180 degree change in the phase of
the carrier. There are many variations on this
simple example of PSK that could be used in body
wires. Other digital modulation techniques, such
as differential phase shift keying (DPSK),
frequency shift keying (FSK), or amplitude shift
keying (ASK), are also possible to use in digital
body wires. A comparison of some of the more
common digital modulation techniques is shown in the table below.
Table 6-1. Comparison of Binary Digital
Modulation Schemes (from Digital and Analog
Communication Systems by K. Shanmugam)
-
As seen in the table, the PSK and DPSK techniques
achieve the best bit error rate performance.
Digital modulation schemes with better bit error
rate performance will require less transmit power
to communicate over a fixed range, which is
beneficial for battery life. Or, equivalently, a
system with better bit error rate performance can
communicate over a longer range using a fixed transmit power.
The “S/N” term in the table refers to signal to
noise ratio. This term is the ratio of the
received signal power to the received noise
power. This ratio is often measured in decibels
(dB). The formula for S/N in dB is 10 x LOG10
(signal power/noise power), where LOG10 is a base
10 logarithm. In the table, notice that the ASK
modulation takes a S/N of 18.33 dB to achieve a
bit error rate of 10-4. In comparison, PSK
modulation only takes a S/N of 8.45 dB to achieve
the same bit error rate. So it takes a lower
signal power for PSK than for ASK to achieve a
given bit error rate performance, which is an
advantage of using PSK modulation instead of ASK modulation.
-6.1.3 Digital Spread Spectrum Body Wires
-
The digital spread spectrum body wire passes the
output of the microphone through an analog to
digital converter. Next, the resulting digital
data is further encoded by another (higher rate)
sequence of 1’s and 0’s referred to as a PN
(Pseudorandom Noise) sequence. The resulting high
rate digital sequence is then used to modulate
the carrier. This form of spread spectrum is
referred to as “direct sequence” spread spectrum.
The rate of the PN sequence is referred to as the
“chip rate,” and the rate of data bits from the
microphone’s analog to digital converter is
referred to as the data rate (rb). The processing
gain of the spread spectrum signal is
approximately the ratio of the chip rate to the data rate.
The effect of encoding the microphone digital
data with the higher rate PN sequence is to
spread the energy of the transmitted signal over
a wider band of frequencies than would otherwise
be used. One advantage of spreading the
frequencies in this manner is that the signal
becomes harder to detect than non spread signals.
When a direct sequence spread spectrum signal is
received, the first operation is to “despread”
the received signal. This dispreading operation
is performed by multiplying a time aligned
version of the PN sequence with the PN sequence
in the received waveform. As a result of this
despreading, any narrowband interferers present
in the received signal will be spread out, and
less energy from the interferer will be passed
into the data demodulation process. The amount of
rejection of the narrowband interferers
corresponds to the processing gain of the signal.
So a second advantage of a digital spread
spectrum body wire is its ability to reject narrowband interference.
One further possible advantage of direct sequence
spread spectrum worth mentioning is its secure
communication capability. It is necessary to know
the transmitter’s PN sequence in order to
despread the signal and listen to the audio.
Using a long PN sequence, and keeping the PN
sequence confidential can achieve secure communications.
In addition to direct sequence spread spectrum,
another form of spread spectrum, called frequency
hopping, is also possible. Instead of using a PN
sequence, frequency hopping spread spectrum
systems change the carrier frequency of the
transmitted waveform periodically. Frequency
hopping spread spectrum systems have similar
advantages to direct sequence spread spectrum systems.
-6.2 Body Wire Issues For Law Enforcement
-
When considering acquiring or using a body wire
system, the user should be cognizant of six performance features:
1. Voice Quality of the Received Audio
2. Transmission Range
3. Battery Lifetime
4. Physical Disguise
5. Electronic Security
6. Cost
Understanding the role played by these six
attributes will prove to be an asset in
determining the suitability of a body wire system
under consideration. Each of the features 1
through 5 is discussed in this section. The most
significant manufacturer specifications for each
feature are provided along with the associated Figures of Merit.
6.2.1 Voice Quality of the Received Audio
-
Voice quality is a very important aspect of body
wire performance. Poor voice quality could
prevent the listener from hearing words in a
conversation, from understanding words in a
conversation, or it could prevent the listener
from determining which person was speaking in a
conversation. Any of these problems could place
the agent in danger, or prevent the collection of
information needed to solve a case.
Voice Quality can be assessed primarily from
specifications of AUDIO BANDWIDTH, AUDIO DYNAMIC
RANGE and, if available, AUDIO SIGNAL TO NOISE RATIO.
AUDIO BANDWIDTH (audio frequency response) is an
important measure of voice quality. A body wire
system with the highest audio bandwidth
performance will cover the entire range of
frequencies that can be heard by the human ear
(20 Hz to 20 kHz). A high quality music Compact
Disc (CD) has a frequency response of 20 Hz to 20
kHz. A body wire system that only covers the
frequency response of a telephone, which is 400
Hz to 4 kHz, would be considered to have
relatively poor audio bandwidth performance.
There is clearly a difference between the sound
quality of a voice on a telephone and a music CD.
Table 6-2 below presents useful information for
evaluating the specifications and performance of body wire audio bandwidth.
Table 6-2. Figure of Merit for Body Wire Audio Bandwidth
-
AUDIO DYNAMIC RANGE (ADR) is another important
measure of voice quality. ADR is a measure of the
systems ability to handle loud and soft sounds.
It is the ratio of the loudest undistorted signal
that the system can handle compared to its
internal noise. Ideally, a body wire system with
the best ADR would have the same dynamic range as
the human ear, which has a dynamic range of over
120 dB. However, contemporary digital recording
techniques can only achieve a dynamic range of about 90 dB.
Table 6-3 shows relative volume levels for
different sounds. The levels in dB are relative
to the threshold of hearing that is taken to be
0dB. From the table, it is seen that the audio
dynamic range necessary to capture audio from
whispers to a shout must be greater than 72 dB (90-18 dB).
Table 6-3. Relative Audio Dynamic Range – Sound Pressure Level
-
From the above table, figures of merit may be
determined for body wire systems. Table 6-4 shows
the figures of merit for the audio dynamic range of a body wire system.
Table 6-4. Figure of Merit for Audio Dynamic Range
-
Many body wire radio systems do not have
sufficient dynamic range to handle full audio
sound levels. Some may be limited to as little as
30-40 dB. When audio dynamic range is limited,
sometimes automatic gain control (AGC) is used to
position the dynamic range window in the most
advantageous place to accurately pick up the most
critical audio levels. The AGC shifting of the
dynamic range window may produce undesirable
audio artifacts. A carefully crafted AGC will
reduce or eliminate these artifacts.
AUDIO SIGNAL TO NOISE RATIO (SNR) is another
important measure of voice quality. SNR is the
ratio of the audio signal power to the noise
power. Noise, which is undesired audio that was
not present at the transmitter’s microphone, may
come from a number of sources. These sources
include the radio frequency energy in the path
from the transmitter to the receiver, and also
include noise from the electronics in the
transmitter and the receiver. Figures of merit
for SNR in body wires are given in the table below.
Table 6-5. Figure of Merit for Audio Signal to Noise Ratio
-
BIT ERROR RATE (BER) is another specification
that affects audio quality. This specification
applies to digital and digital spread spectrum
systems, but not to narrowband FM systems. Bit
error rate is the number of bits in the digital
stream that have been received with the wrong
value, compared to the total number of bits received.
Each bit in the digital (or digital spread
spectrum) body wire received data stream has a
value of 1 or 0. These bits taken together in
preset groups (usually 8, 16 or 32 bits) form the
‘words’ which correspond to the digital
representation of the audio waveform being
transmitted. If one of these bits is somehow
assigned the wrong value, the sound from the
receiver will be distorted. The more bits that
are assigned the wrong value, the worse the resulting audio.
Typically, bit error rate is expressed as the
frequency of a single erroneous bit. For example,
a bit error rate of 10-6 means that for every
1,000,000 bits sent, one of them will be received
incorrectly, and the audio will be distorted for
that instant. The table below shows figures of
merit for bit error rates in body wire systems.
Table 6-6. Figure of Merit for Bit Error Rate
-
Bit error rate has a relationship to receiver
sensitivity. (Sensitivity is the weakest received
signal power that can be successfully received.)
In general, receiver sensitivity is influenced by
a change in bit error rate (and vice versa). A
receiver with –100dBm sensitivity at a BER of
10-5 could also have a sensitivity of –103 dBm at
a BER of 10-4. In the latter specification, the
audio is worse, but the apparent range is better
when compared to the former specification.
6.2.2 Transmission Range
-
The ability to receive the body wire signal at
relatively long distances from the transmitter
(agent) can be very useful in some law
enforcement applications. High ranges allow the
receiver to be located further from the agent,
reducing the likelihood of physical discovery of
the operation. In addition, for mobile
applications, high ranges reduce the likelihood
that an agent will move out of range of the receiver.
However, high ranges imply high transmit power.
And high transmit power would reduce the
electronic security of the system (increase the
likelihood that wiretap detection equipment would
see the signal coming from the transmitter). For
this reason, the operation should use transmit
power that is sufficient for the range of the operation, but not excessive.
Range evaluation is dependent on two manufacturer
specifications: Transmitter (TX) power, and
Receiver (RX) sensitivity. Path loss represents
loss in signal power due to the transmission from
body wore to receiver, and is an environmental
factor which determines range for a given TX
power and RX sensitivity. Path loss can not be
specified by the manufacturer and needs to be
accounted for by the user. There are many
environmental factors that will increase path
loss over a specific distance. Urban
environments, with buildings and crowds of people
may experience much greater path loss than in
open terrain. From specifications of TX output
power and RX sensitivity, the maximum path loss
for received audio that can be accommodated by the system can be calculated=
.
In general, increasing transmitter output power
will increase range for a particular terrain and
RX sensitivity specification. The power should be
expressed in mW or dBm (dBm = 10 x log base 10 of
signal power in mW). There is no figure of merit
for transmit output power since each operation
will accommodate different equipment with different power ratings.
The table below presents some recommended
transmit power values for various law enforcement applications.
Table 6-7. Transmit Power Uses
-
The receiver sensitivity defines the lowest
received power level of the transmitted signal
that can be detected by the receiver at its
antenna. A signal received at this level should
provide audio output at the receiver. Receiver
sensitivity should be quoted for a specific SNR.
Sensitivity is usually given in dBm. Some radio
frequency (FM) systems use a receiver sensitivity
notation of microvolts (uV) for a specific SINAD
(signal to noise plus distortion).
Typically specifications will be lower (better)
for narrow bandwidth signals such as narrowband
FM and higher for wider bandwidth signals such as
digital spread spectrum. Note that sensitivity is
stated with negative numbers since the power is less than 1 mW.
Table 6-8. Receiver Sensitivity
-
In addition to transmit power and receiver
sensitivity, path loss determines the range of
the body wire system. Path loss is very dependent
on physical conditions present in the locale of
the transmitter and receiver and on the carrier
frequency. Building structures, the number of
people between transmitter and receiver,
interfering vehicles, metal wall studs, etc. (the
local operating conditions) will server to
attenuate the transmitted signal by varying
amounts. It is not uncommon to see a requirement
of a 8-fold increase in required power to double
the range. Engineering studies have shown that in
some cases, the attenuation at ground level is so
great that output power must be increased 16
times in order to double the range. The
expression often stated of 4 times the power to
double the range is mainly applicable for line of
sight conditions. Terrain is very important when
looking at power outputs of different systems
when trying to determine whether the equipment will meet range expectations=
.
A related concept to path loss is multipath.
Multipath effects are due to portions of the
signal arriving at the receiver at different
times from the main signal, caused by reflections
within the environment. These multiple signals
arriving at the receiver may be highly disruptive to communications.
6.2.3 Battery Lifetime
-
Battery related specifications of body wires
include maximum and minimum operating DC voltage
and current drain. These specifications have a
direct bearing on the type of battery most suitable for the equipment.
The current drain specification determines the
amount of current required to run the equipment.
The lower the current drain, the longer the
equipment will run on a battery. Since many
battery manufacturers list the battery capacity
in mAh (milli ampere hours), it is quite easy to
determine how long the equipment will run on a
given battery. The current drain of the body wire
equipment should be provided in the specifications.
The minimum operating voltage of a body wire is
the minimum voltage that the battery must supply
for the equipment to function properly.
Generally, if the battery falls below this level,
the equipment will cease operating. There may
also be a maximum (not to exceed) voltage.
Voltages beyond this figure will probably damage
the equipment. The minimum operating voltage is
often not given, but it is important in determining battery requirements.
In general, the less current consumed by the body
wire, the longer the batteries will last. The
lower the minimum operating voltage, the fewer
batteries that are needed. The wider the
operating voltage range (maximum voltage –
minimum voltage), the longer the system will operate on a given battery pac=
k.
NOTE: Always use new batteries. If the batteries
have been taken out of their storage package,
don’t use them for field operation.
6.2.4 Physical Disguise
-
Two physical features are important for body
wires: package size and antenna type. Dimensions
should be given for length, width, and thickness.
For body wire usage, the transmitter must be as
thin as possible. It is also advantageous to
remote the antenna from the transmitter. Being
able to move the antenna away from the
transmitter allows a greater choice of concealment options.
Table 6-9. Figure of Merit for Size (Thickness)
-
6.2.5 Electronic Security
-
Narrowband scanners can easily detect Narrowband
FM (NBFM) systems, since most detectors are of
the narrowband sweep type. The fact that the
signal is digitized is significant for electronic
security, since it reduces the probability of
interception. Digital spread spectrum systems are most secure.
Detection is defined as the ability of an outside
person to discover the presence of the body wire
signal. Detection can be accomplished with
frequency counters, spectrum analyzers, or scanning receivers.
Interception is defined as the ability of an
outside person to acquire the body wire signal
and obtain understandable audio. A tunable
receiver is necessary for this purpose. Spread
spectrum systems have very good immunity to
interception, since it is necessary for the
outside person to know the PN spreading sequence
used in order to successfully receive the audio
signal. If the PN sequence is kept confidential,
randomly selected, and is long enough, it will be
very difficult for an outside person to obtain
understandable audio from the signal.
-7. CURRENT AND PROJECTED STATE OF THE ART BODY WIRE TECHNOLOGY AREAS
-
-7.1 Block Diagrams of Typical Body Wire Transmitters
-
This section will discuss the block diagrams of
the digital, digital spread spectrum, and
narrowband FM body wires. The block diagrams are
intended to give the reader a general idea of the
types of components used in body wires. Later
sections will then briefly describe the state of
the art for some of the key components used in body wires.
Figure 7-1 shows a simplified block diagram of a
typical digital body wire transmitter. Starting
at the top left of the figure, the audio signal
is received by the microphone. Next, an
amplifier/filter increases the voltage of the
signal from the microphone to the correct level
for the input to the analog to digital converter
(ADC). Some filtering (removal) of unwanted
signals may also occur in this block. The ADC
converts the analog input signal to a digital
word. The digital word output of the ADC is fed
into a coding block, which adds bits to the ADC
words that will serve to detect and correct
errors at the receiving end of the body wire
link. After these error detection and correction
bits are added, the resulting bit stream is
differentially encoded for differential phase
shift key (DPSK) modulation. The resulting bit
stream is fed to a binary phase shift keyed
(BPSK) digital modulator. (Other digital
modulation methods could also be used, but this
particular block diagram uses BPSK). The BPSK
digital modulator changes the phase of the radio
frequency (RF) carrier (from the RF synthesizer),
according to whether a 0 or 1 bit is input. The
resulting phase modulated carrier then goes to
the power amplifier block. The power amplifier
block increases the power of the modulated RF
carrier to a level that is suitable for
transmission. The power amplifier output goes to
an antenna matching network, which assures that
the power of the amplified phase modulated RF
signal from the power amplifier is efficiently
transferred to the antenna. The matching network
feeds the antenna, which sends the signal through the air to the receiver.
Batteries and voltage regulators provide power
for the body wire. The battery voltage is applied
to voltage regulators, which provide the voltages
needed by various components in the body wire.
-
Figure 7-1. Digital body wire block diagram.
-
Figure 7-2 shows a block diagram of a digital
spread spectrum body wire. The diagram is very
similar to the diagram for the digital body wire,
except for the addition of a PN sequence
generator and an exclusive or (XOR) block. The PN
sequence generator generates a random sequence of
bits (chips) that is fed to the XOR block. The
XOR block combines the PN sequence with the data
bits from the differential encoding block. The
resulting bit stream is fed to the digital
modulator (a BPSK modulator is shown).
Adding the PN bit sequence increases the bit rate
going into the BPSK modulator, and thereby
increases the bandwidth of the modulated signal
around the RF carrier (the bandwidth of a phase
shift keyed signal is approximately twice the bit
rate). This increased bandwidth can make a spread
spectrum signal more difficult to detect, since
the energy of the signal is spread out over a wider band of frequencies.
-
Figure 7-2. Direct Sequence Spread Spectrum body wire block diagram.
-
Figure 7-3 shows a block diagram of a typical
narrowband FM body wire transmitter. The main
difference in this system and the digital body
wire is that analog modulation is used instead of
digital modulation. The analog signals from the
microphone’s amplifier/filter are fed to the FM
input on the frequency synthesizer, which
frequency modulates the carrier. The output of
the frequency synthesizer goes to a power amp.
The power amplifier output goes to a matching
network, and the matching network feeds the antenna.
The frequency synthesizer consists of several
components: a voltage controlled oscillator
(VCO), a divide by N counter, a phase/frequency
detector, and a loop filter. A temperature
compensated crystal oscillator (TCXO) serves as a
frequency reference for the frequency
synthesizer. The frequency synthesizer is used in
the narrowband FM, the digital, and the digital spread spectrum body wires.
-
Figure 7-3. Narrowband FM body wire block diagram.
-
Please note that many variations on all three of
the above block diagrams are possible. The intent
is to give typical block diagrams, and to give
the reader a general idea of the types of components used in body wires.
7.2 Microphones
-(See the discussion on microphones in the solid state audio recorder secti=
on)
7.3 Analog to Digital Converters
-(See the discussion on delta sigma analog to
digital converters in the solid state audio recorder section)
-7.4 Frequency Synthesizer
-
The frequency synthesizer provides the RF carrier
used by the transmitter. A typical frequency
synthesizer consists of several components: a
temperature compensated crystal oscillator
(TCXO), a phase-frequency detector, a loop
filter, a voltage controlled oscillator (VCO),
and a divide by N circuit (prescaler). The
phase-frequency detector compares the phase of
the divided VCO with the phase of the TCXO. The
loop filter filters the output of the phase
frequency detector, and the output of the loop
filter is applied to the voltage control input of
the VCO. If the phase of the divided VCO is
different than the phase of the TCXO, the voltage
applied to the VCO will change the VCO frequency
until the phase of the TCXO is aligned with the
phase of the divided VCO. In this manner, the
frequency of the VCO is controlled so that it is
N times the frequency of the TCXO. By adjusting
the value of N, it is possible to generate different frequencies.
Low cost, low power integrated circuits are
available to perform one or more of the functions
needed by the frequency synthesizer. For example,
the National Semiconductor LMX2346 provides the
phase-frequency detector and divide by N
functions. It consumes 6mA of current and costs
$2.05 in quantities of 1000. It is available in
small surface mount packages including a 0.25” x
0.2” package and an even smaller chip scale
package. A very good TCXO part is the ECS 39SM
series made by ECS, Inc. This part consumes 1.5mA
at 3.3V and has a very good frequency accuracy of
1.5 ppm. It is available in a surface mount
package that is 0.45” x 0.38”, and costs $7.70 in
quantities of 1000. An example of a VCO that will
cover the the 915MHz ISM band (for a direct
conversion transmitter) is the Maxim 2623. This
part consumes 9mA at 3.3V, is available in a
3.0mm x 4.9mm (uMax) package, and sells for $1.80
in quantities of 1000. The loop filter may be
constructed from passive components (resistors
and capacitors) for only a few cents in cost and with very small packages.
Future frequency synthesizer components should
benefit from the overall trend in electronics
toward smaller die geometries. Smaller geometries
may be operated with lower supply voltages,
resulting in lower power consumption. Smaller
geometries can also lead to smaller IC sizes or
greater part densities. The use of synthesizer
components in high volume cell phone and cordless
phone markets should ensure the continued
development and availability of low cost, power
efficient, small size frequency synthesizer components.
-7.5 Power Amplifiers
-
In most body wire designs, the power amplifier
consumes more power than any other single
element. For that reason, operating times for the
body wire are dictated largely by the power needed by the power amplifier.
One of the biggest decisions in selection of a
power amplifier is dictated by whether or not a
constant envelope waveform type of modulation is
used. If a constant envelope modulation is used,
then one of the more power efficient amplifier
classes may be used (Class B, AB, or even C). In
contrast, if a modulation technique is selected
that is not constant envelope, then a less power
efficient amplifier class (Class A) must be used.
One problem with using the more power efficient
amplifiers is they do not operate in the linear
region, and are subject to spectral regrowth. A
current area of research in power amplifiers for
wireless applications is how to improve the
spectral regrowth problems in power efficient
amplifiers. Several techniques to prevent
spectral regrowth and preserve efficiency are
being investigated. One such technique is to
adaptively bias the power amplifier so it
operates in the most efficient region of class A
operation all the time. Another technique is to
predistort the waveform, so that it is relatively
undistorted after it is amplified by a Class C amplifier.
An example of a state of the art, low cost
amplifier for constant envelope waveforms in the
900MHz ISM band is the Maxim MAX2235. It features
a +30dBm (1W) power output, 47 percent
efficiency, has a footprint of 6.4 x 6.5mm, and
sells for $2.07 in quantities of 1000.
The continued growth in wireless commercial
applications (such as cell phones, wireless
phones, PCS, Bluetooth, HomeRF, and wireless
LANs) is expected to spur future development of
more efficient power amplifiers.
8. PROJECTED COMMERCIAL BODY WIRE PRODUCTS
(IN TWO YEARS)
-
It can be anticipated that body wire equipment
capabilities will undergo a steady change in the
next two years. The forecast is that digital
technology will be assume as more prominent role
in wireless operations. The benefits of security
and audio quality will become more in demand. The
challenge to the designers and manufacturers is
to bring digital equipment into the same range
performance standard as analog and still keep
costs down. Digital, because of its additional
complexity, is inherently more expensive than
analog. Unfortunately, digital operation will be
faced with a range penalty and the way to
increase range is to engineer more efficiency and
range enhancements into the digital equipment.
This extra engineering comes with an added cost
burden. Operational requirements demand the best
quality audio for evidentiary and investigative
purposes, which is of course directly in the
province of digital technology. Better education
and training is the way around the seemingly
contradictory desires of performance and cost.
Properly trained personnel will understand the
benefits and drawbacks of digital equipment and
will be able to get the results needed in the
difficult investigations involving foreign
translations and poor audio environments. Well
trained technicians will be able to get good
audio at the ranges required, making the
change-over from analog to digital much less painful.
-9. BODY WIRE CONCLUSIONS
-
Body wire equipment is currently available in a
wide variety of size, technology and operating
lifetime. The old adage ”You get what you pay
for” is ever applicable. Good analog equipment is
not cheap, neither are good digital transmitters
and receivers. Cheap equipment manifests itself
in shoddy performance and poor reliability, in
spite of claims of high quality . Written
specifications can give a distorted picture to
the unknowing. The best advice one can get is to
become knowledgeable about the meaning of
equipment specifications and their operational
impact. One must evaluate equipment prospects as
to performance and operational tradeoffs – then
consider cost. It has been the case that a law
enforcement user has said that a certain piece of
equipment is the only thing that can be afforded,
only to discover that the product is quite
useless. The small savings in equipment cost can
cause a much larger loss of funds when the entire
case is destroyed because the jury could not
understand the spoken word or the translator
could not properly comprehend the idiomatic speech.
-
APPENDIX A
SOLID STATE RECORDER PRODUCT MATRIX
-
The following spreadsheet contains a complete
listing of the data collected on the various
types of recorders investigated. Many of the
desired specifications were not available or were
considered too proprietary to release,
particularly related to details regarding the
type of encoding used for the voice recording.
[snip, see original pdf files on
http://www.tscm.com/ for these tables and graphics]
APPENDIX B
SOLID STATE RECORDER COMPONENTS-
The following spreadsheet shows the major
components used to make a flash based audio
recorder. The recorder features a dynamic range
approaching 90dB, a bandwidth of 16kHz, and has
over 120 minutes of record time. The size of the
recorder is dictated largely by the size of the
removable compact flash, and by the size of the batteries.
Table B-1. Solid State Recorder Spreadsheet
-
APPENDIX C
BODY WIRE PRODUCT SOURCE MATRIX
-
APPENDIX D
SOLID STATE RECORDER AND BODY WIRE SURVEY RESULTS
The following tables contain the results of a
survey submitted to a cross section of US law
enforcement agencies. The populations served by
these agencies range from less than 1,000 to over
1,000,000. The locations of these agencies are
all across the continental United States. 77
responses were received from the survey. The
survey asked a number of questions regarding the
typical use of recorders and body wires for law
enforcement applications. The answers to these
questions are summarized in the tables below.
[snip, see original pdf files on
http://www.tscm.com/ for these tables and graphics]
We Hunt Spies, We Stop Espionage, We Kill Bugs, and We Plug Leaks.
James M. Atkinson, President and Sr. Engineer
Granite Island Group
127 Eastern Avenue #291
Gloucester, MA 01930-8008
Phone: (978) 546-3803
Fax: (978) 546-9467
Web: <
http://www.tscm.com/>
http://www.tscm.com/
E-Mail: <mailto:jm..._at_tscm.com>jm..._at_tscm.com
--===========_256428750=.ALT
Content-Type: text/html; charset="iso-8859-1"
Content-Transfer-Encoding: quoted-printable
<html>
<body>
The following text was extracted from the NIJ document from July 2005.
The original file can also be cfound on my website, and is also attached
to this document as an enclousure.<br><br>
<a href="
http://www.tscm.com/NIJ-210488.DigitalRecorders.July2005x.pdf" e=
udora="autourl">
http://www.tscm.com/NIJ-210488.DigitalRecorders.July2005x.pdf<br><br>
</a>-jma<br><br>
<br><br>
<br>
Personal Electronics for Law Enforcement<br>
Solid State Recorders and Body Wires<br>
Table 2-1. Audio Recorder Feature Summary<br>
Table 6-1. Body Wire Types and Feature Comparison<br>
Feature<br>
NBFM<br>
Digital<br>
Voice quality<br>
Scheme<br>
Table 6-2. Figure of Merit for Body Wire Audio Bandwidth<br>
Sound<br>
Table 6-3. Relative Audio Dynamic Range – Sound Pressure Le<br>
Rating<br>
Table 6-4. Figure of Merit for Audio Dynamic Range<br>
Rating<br>
Table 6-5. Figure of Merit for Audio Signal to Noise Ratio<br>
Rating<br>
Good<br>
Table 6-6. Figure of Merit for Bit Error Rate<br>
Power Level<br>
20 mW to 100 mW<br>
Table 6-7. Transmit Power Uses<br>
System Type<br>
Digital systems<br>
Table 6-8. Receiver Sensitivity<br>
Table 6-9. Figure of Merit for Size (Thickness)<br>
MP3<br>
MINIDISC<br>
MINIDISC<br>
Table B-1. Solid State Recorder Spreadsheet<br>
APPENDIX D<br>
-<br>
-<br><br>
<br>
Personal Electronics for Law Enforcement<br>
Solid State Recorders and Body Wires<br>
-<br><br>
<br>
William Butler, Georgia Tech Research Institute<br>
Scott Crowgey, Georgia Tech Research Institute<br>
William Heineman, Tektron<br>
Susan Gourley, Tektron<br><br>
<br><br>
Prepared Under:<br>
Contract Number N65236-00-K-7805<br><br>
<br><br>
<br>
Submitted to:<br>
Attention: Mr. Richard Baker, Code 741<br>
Mr. Jerry Owens, Code 741JO<br>
Commanding Officer<br>
SPAWARSYSCEN Charleston<br>
PO Box 190022<br>
North Charleston, SC 29419-9022<br><br>
<br><br>
<br>
July 2002<br><br>
<br><br>
<br>
CONTENTS<br><br>
x Introduction<br>
x Current Commercial Solid State Recorder Products<br>
x Overview of Commercial Audio Recorder Products<br>
x Commercial Audio Recorder Issues<br>
x Current and Projected State of the Art in Solid State Recorder
Technology Areas<br>
x Block Diagram of a Typical Solid State Recorder<br>
x Microphones<br>
x Delta Sigma Analog to Digital Converters<br>
x Audio Compression Algorithms<br>
x Audio Compression Hardware<br>
x Flash Memory<br>
x Batteries<br>
x Projected Commercial Solid State Recorder Products (in 2 years)<br>
x Solid State Recorder Conclusions<br>
x Current Body Wire Products<br>
x Overview of Body Wire Products<br>
x Body Wire Issues<br>
x Current and Projected State of the Art in Body Wire Technology
Areas<br>
x Block Diagram of a Typical Body Wire<br>
x Projected Commercial Body Wire Products (in 2 years)<br>
x Body Wire Conclusions<br>
x APPENDIX A – Solid State Recorder Product Matrix<br>
x APPENDIX B – Solid State Recorder Components<br>
x APPENDIX C – Body Wire Product Source Matrix<br>
x APPENDIX D – Survey of Recorder and Body Wire Use by Law Enforcement
Agencies<br><br>
<br><br>
1. INTRODUCTION<br>
-<br>
This report summarizes the work performed by the Communications
Networking Division (CND) of the Information and Telecommunications
Technology Laboratory (ITTL) of Georgia Tech Research Institute (GTRI)
under the "Personal Electronics for Law Enforcement" program.
This program is being performed for the SPAWARSYSCEN Charleston. The
report covers work done as part of a joint effort between GTRI, and
Tektron, Inc. GTRI’s efforts are focused on solid state audio recorders
that could be used for law enforcement applications, and Tektron’s
efforts are focused on body wires for law enforcement
applications.<br><br>
This report includes information that is intended to assist the law
enforcement community in the evaluation and purchase of audio recorders
and body wires. It includes a market survey of commercially available
audio recorder and body wire products, and it includes a brief review of
key technologies used in these products. The first section of the report
covers audio recorders, and the second section covers body wires. In
addition, an appendix contains the results of a survey of law enforcement
agencies that deals with the use of recorders and body wires for law
enforcement applications.<br><br>
<br>
2. COMMERCIAL SOLID STATE AUDIO RECORDER PRODUCTS<br>
-<br>
Throughout this program, data has been collected on
commercial-off-the-shelf (COTS) audio recorders that could be used for
law enforcement applications. An incredible variety of recorders are
available, including solid state audio recorders based on flash memory.
Since solid state recorders have no moving parts, they can offer higher
fidelity recordings than conventional cassette recorders. The solid state
recorder does not suffer from background tape hiss or tape speed
variations that degrade the fidelity of cassette recorders. For these
reasons, special emphasis has been placed on solid state recorders in
this study. For comparison with solid state recorders, data has also been
collected on MP3 recorder/players, mini disc recorder/players, and
digital audio tape recorder/players.<br><br>
2.1 Overview of Commercial Audio Recorder Products<br>
-<br>
Table 2-1 presents a summary of the performance of various kinds of audio
recorders. A solid state flash recorder (made by Olympus), a MP3 recorder
(made by Creative Labs), a mini disc recorder (made by Sony), and a
microcassette recorder (made by Sony) are all compared in the table. This
table does not include all the devices reviewed in the survey, but
instead, includes devices that typify the performance of commercially
available audio recorders that would be suitable for law enforcement
applications. Data in the table is current as of July 2002.<br><br>
Table 2-1. Audio Recorder Feature Summary <br>
-<br><br>
From the table, it is seen that all the devices are available in similar
sizes, and all devices are capable of at least 2 hours of record time.
The Olympus voice recorder and the MP3 player/recorder have similar
frequency response to the microcassette. The ATRAC3 compression used by
the minidisc recorder and the digital audio tape have the best bandwidth.
In the cost category, the MP3 player/recorder is the next lowest cost
after the microcassette. In the media cost category, the minidisc is the
lowest after the microcassette.<br><br>
2.1.1 Flash Audio Recorders<br>
-<br>
The flash based audio recorder is the main subject of this report. It
offers a number of potential advantages: high fidelity, high reliability,
small size, and reasonable cost (cost of both the recorder and the
recording medium). Flash recorders have benefited from the proliferation
of the use of flash memory for digital cameras and MP3 players over the
past few years, and the cost of flash audio recorders has come down as a
result. The material to follow describes the features of several
representative commercially available flash audio recorder
products.<br><br>
The Olympus flash audio recorders are available in several models. The
DS2000 is listed in the table. The DM-1 is also available for
approximately the same cost, and has the added ability to play back MP3
music recordings. The DM-1 does not provide protection against accidental
erasure. The Olympus DW-90 flash audio recorder costs approximately $90,
has a non-removable 8MB flash memory, uses ADPCM compression, and can
record from 22 (5.8kHz) to 90 (1.7kHz) minutes of audio. The DS2000 and
DM-1 Olympus flash recorders use a file format called Digital Speech
Standard (DSS). Files stored in this format occupy 12 times less memory
space than uncompressed WAV files, while achieving roughly the same audio
quality. Olympus flash voice recorders feature voice-activated recording
that can be switched off. Olympus voice recorders use a standard USB
interface to transfer data from the recorder to a PC. The Olympus
recorders can record in monaural mode, but not stereo. Further
information on these products may be obtained at the manufacturer’s web
site:
<a href="
http://www.olympusamerica.com/cpg_section/cpg_vr_digitalrecorder=
s.asp" eudora="autourl">
http://www.olympusamerica.com/cpg_section/cpg_vr_digitalrecorders.asp</a>
.<br><br>
The Panasonic RR-XR320 is another example of a flash audio recorder. The
RR-XR320 is 1 7/8” x 3 9/16” x ½” in size, uses ADPCM recording and=
uses
SD flash memory. It has a battery life of 11 hours when recording, and
uses two AAA batteries. The MSRP of the RR-XR320 is $329, and street
prices around $280 are common. This flash recorder uses a standard USB
interface to transfer data from the recorder to a PC. It can record up to
150 minutes in “LP” mode with a 16MB SD flash memory card. High quality
(HQ), standard play (SP), and long play (LP) recording modes are
available. Further information on this product may be obtained from the
manufacturer’s web site:
<a href="
http://www.prodcat.panasonic.com/shop/NewDesign/ModelTemplate.as=
p?ModelID=13081" eudora="autourl">
http://www.prodcat.panasonic.com/shop/NewDesign/ModelTemplate.asp?ModelID=
=13081</a>
.<br><br>
The Sony ICD-MS515 is another audio recorder that uses flash memory (in
the form of a memory stick). It is 1/3/8” x 4 1/8” x 23/32” in size, =
and
uses 2 AAA batteries. The MSRP is $250. It can record for 10 hours in SP
mode, and 12 hours in LP mode on a single set of batteries. It has voice
activated recording, and uses a standard USB interface to transfer data
from the recorder to a PC. It can record 64 minutes in SP mode (using
16kHz sampling), and 150 minutes in LP mode (using 8kHz sampling). It
features a built in omnidirectional microphone, and is a monaural
recorder. Sony also makes a less expensive flash recorder (ICD-B25)
without removable media for $100. Further information on these products
may be obtained from the manufacturer’s web site:
<a href="
http://www.sonystyle.com/electronics/prd.jsp?hierc=8627x8667x8=
668&catid=8668&pid=31982&type=p" eudora="autourl">
http://www.sonystyle.com/electronics/prd.jsp?hierc=8627x8667x8668&cat=
id=8668&pid=31982&type=p</a>
.<br><br>
The DIALOG4/ORBAN SOUNTAINER MP3 player/recorder is another example of a
compact audio recorder that uses flash memory in a multimedia card (MMC)
format. Instead of ADPCM or DSS, it uses MP3 recording of audio. It is
comparable in size and features to other recorders. This manufacturer
prefers that detailed information on its recorder specifications should
not be reproduced. So, for more information on this recorder, the reader
is referred to the manufacturer’s web site:
<a href="
http://www.dialog4.com/products/sountainer/supp_snt1.html" eudor=
a="autourl">
http://www.dialog4.com/products/sountainer/supp_snt1.html</a> .<br><br>
Please note that solid state audio recorders from Adaptive Digital
Systems (EAGLE/FBIRD8) are available for law enforcement purposes. For
specifications on these products, please see
<a href="
http://www.adaptivedigitalsystems.com/" eudora="autourl">
http://www.adaptivedigitalsystems.com</a> . A password, which may be
obtained from the manufacturer, is required to access the specifications
for these recorders.<br><br>
Another manufacturer of solid state audio recorders for law enforcement
purposes is Digital Audio Corporation. The product made by this
corporation is the SSABR, which is described as a “state of the art, body
worn digital recorder, specifically designed for collecting accurate,
covert recordings.” Details on this product may be found at
<a href="
http://www.dacaudio.com/" eudora="autourl">
http://www.dacaudio.com</a> . A password, which may be obtained from the
manufacturer, is required to access this data.<br><br>
Yet another manufacturer of solid state audio recorders for law
enforcement applications is Geonautics. This company makes a very small
“Whisper” flash based recorder that is available in both mono and stere=
o
configurations. Details on these products may be found at
<a href="
http://www.geonautics.com/" eudora="autourl">
http://www.geonautics.com</a> . A password, which may be obtained from
the manufacturer, is required to access this data.<br><br>
2.1.2 MP3 Player/Audio Recorders<br>
-<br>
Another class of commercial product with potential application for covert
recording is the MP3 player. MP3, or MPEG Layer 3, is a lossy compression
format that allows CD-quality music recordings to be compressed into
files significantly reduced in size to facilitate transfer over the
internet and to and from PC’s. MPEG formats accomplish this reduction in
size partly by eliminating components of the recording that would be
masked by the human hearing process based on a psychoacoustic model of
hearing.<br><br>
Many, but not all, MP3 players have voice recording capability in
addition to MP3 playback capabilities. The MP3 playback frequency
response is listed as a very high quality range of 20Hz to 20kHz.
Unfortunately, the claimed frequency response of 20Hz – 20kHz applies
only to the playback of MP3 files, not to recorded voice. The portable
MP3 recorder/players reviewed to date use ADPCM for recording voice. The
ADPCM implementations used have a bandwidth of 3 to 4kHz, which is much
worse than the 20-20kHz achieved when playing back MP3 recordings. The
ADPCM used in the voice recordings is based on 8 bit PCM samples, and has
an upper limit of approximately 50dB for its signal to noise
ratio.<br><br>
The Creative Labs Nomad IIc MP3 player/recorder is a widely available MP3
player/recorder that can be used for recording audio onto flash memory
(Smartmedia format flash). It is 3.7” x 2.6” x 0.9” in size, and uses
32kbps G721 (an ITU standard) ADPCM recording. It features a USB
interface for transferring files to a PC. Further information on this
device may be obtained from the manufacturer’s web site:
<a href="
http://www.americas.creative.com/products/category.asp?category=
=2&maincategory=2" eudora="autourl">
http://www.americas.creative.com/products/category.asp?category=2&mai=
ncategory=2</a>
.<br><br>
The Sensory Science Rave MP2200 samples voice at 8kHz, and requires
approximately 1MB of flash memory space for every 4 minutes of voice
recording. So, for a built in flash memory of 64MB, this unit can store
over 4 hours of voice. The specification of approximately 4 minutes of
voice per 1MB indicates that some compression is being used to store the
voice (approximately a 2:1 compression), which is consistent with 32kbps
G721 ADPCM. Unfortunately, the Rave MP2200 does not store voice files on
removable media, but only on the built in flash. Cost of the Rave MP2200
is approximately $200. More information on the Rave MP2200 may be
obtained at the following URL:
<a href="
http://www.sonicblue.com/support/goVideo/downloads/MP2200manual.=
pdf" eudora="autourl">
http://www.sonicblue.com/support/goVideo/downloads/MP2200manual.pdf</a>
.<br><br>
A few of the MP3 recorder/players use 40 MB Iomega Clik! disks as the
storage media, which are much cheaper than the removable flash cards.
However, these disks are susceptible to shock and vibration, which could
be a disadvantage for certain law enforcement applications.<br><br>
2.1.3 Minidisc Player/Audio Recorders<br>
-<br>
A third interesting class of commercial products with potential for
covert recording applications is the minidisc recorder/player. The
minidisc is the most compact of the removable memory storage media,
capable of storing approximately 160 MB of audio data on a disc that is
64 mm in diameter and approximately 1 mm thick. A typical minidisc device
is not much bigger than the minidisc itself, with typical dimensions of
70 mm x 67.5 mm x 5 mm and being very similar in size to the MP3
player/recorders discussed above. Only Sharp and Sony currently produce
minidisc recorder/players. These are the only COTS products reviewed so
far that can make voice recordings in stereo and that can record voice
using the full 44.1 kHz, 16 bit sampling that is a standard for audio
CD’s, allowing the full recording bandwidth for music or voice of 20 Hz
to 20 kHz. However, to store audio with this large a bandwidth on the
limited amount of memory space available, all minidiscs utilize a
proprietary ATRAC3 compression scheme for the storage of data that is
lossy, compressing the audio files by a ratio of approximately 4.83:1.
Both Sharp and Sony have plans to produce higher density minidiscs and
drives with a capacity of about 650 MB.<br><br>
Pre-recorded minidiscs are fabricated using the same plastic-aluminum
structure as CD’s. The minidisc is read by focusing a laser on pits and
valleys within the transparent polycarbonate substrate backed by a
coating of aluminum that then reflects or disperses the beam to produce a
series or 1’s and 0’s which can then be translated back into either the
original data or sound. Recordable minidiscs have a pre-groove instead of
the CD-type pits and valleys and a MO (magneto-optical) coating instead
of the aluminum one. While recording, the laser focuses on the pre-groove
and heats the MO recording layer at that point to its Curie point while a
magnetic field from a head in contact with the other side of the disc
aligns magnetic dipoles within the heated spot on the MO layer. During
playback, the laser focuses on the pre-groove again, but at a lower
power, allowing the measurement of changes in the polarization of the
light reflected from the previously magnetized layer. All minidisc
players have a dual function optical assembly that detects the disc type
and switches between the measurement of reflectivity for pre-recorded
minidiscs or polarization for recordable minidiscs. Sony claims
recordable minidiscs can handle up to 1 million recordings. The minidiscs
have a user table of contents that can be damaged if the minidisc is
abused and render the minidisc unusable. Sony claims that data using
magneto-optical technology can be stored for more than 30 years without
loss or degradation. However, strong magnets placed directly against the
minidisc can destroy data.<br><br>
Minidiscs use a buffer memory that temporarily stores recorded audio,
thereby helping to prevent vibrations from affecting either the recording
onto or playback from the minidisc. However, problems have been reported
with recording when the minidisc recorder is subjected to shock and
vibration, apparently due to the laser beam “skipping” and accidentally
erasing previously recorded data on adjacent tracks. Therefore, it is
recommended that the recorder should be immobile and not subjected to
shock or vibration while recording. In addition, because of the 400-900
rpm rotation of the minidisc, all such devices produce a humming noise
when recording or playing audio. Although this humming noise reportedly
does not degrade the recording or playback process, it could possibly
interfere with the covert recording process. Because the laser beam must
heat the disk while recording, the minidisc device is the only portable
recording device that consumes more power during recording than during
playback. And even during playback, the devices still consume 50-100%
more power than any other class of recording device. Until recently, none
of the minidisc recorder/players have had a convenient means to connect
to a PC to allow the rapid transfer of files.<br><br>
The Sony MZ-N707 minidisc recorder offers some of the advantages of flash
recorders. It records onto a digital medium (the minidisc), and is not
subject to the tape hiss that is present in cassettes. The minidisc must
spin to work, so, unlike flash recorders, there are moving parts inside
the minidisc recorder. The size of the MN-N707 is 3 ¼” x 3” x 1 1/8=
”. It
comes with a rechargeable battery, and records in a high fidelity ATRAC3
format. An external microphone is needed to record audio, since the unit
does not come with a built in microphone. It uses Sony’s ATRAC3
compression technique for storing audio (and music). The ATRAC3
compression technique achieves relatively high fidelity, but it is not
lossless. Another model, the MZ-N1, is available for $350, and it is
somewhat smaller in size: 3” x 2 7/8” x ½”. The MZ-N1 features a hi=
gher
capacity battery than the MZ-N707. Further information on these devices
may be found at the following URL:
<a href="
http://www.sonystyle.com/electronics/ssctypg.jsp?hierc=8627x86=
50x8647&catid=8647" eudora="autourl">
http://www.sonystyle.com/electronics/ssctypg.jsp?hierc=8627x8650x8647&=
;catid=8647</a>
.<br><br>
2.1.4 Digital Audio Tape (DAT) Recorders<br>
-<br>
One DAT device, a TCD-D100 produced by Sony, is included in this survey
for comparison purposes. This DAT recorder, which lists for $900, can
provide up to 4 hours of stereo recording on two AA batteries. This DAT
device can sample at 48kHz, 44.1kHz or 32kHz, and uses 16 bit
quantization. At a 48kHz sample rate, it has a 20-22 kHz frequency
response (within 1 dB), which is greater than the full range of human
hearing (20-20kHz). At 44.1 kHz and 32 kHz sample rates, it has a 20-14.5
kHz frequency response (within 1 dB). The signal to noise ratio is 87dB,
and the total harmonic distortion is 0.008%. The wow and flutter is less
then 0.001 percent. All of these specifications are excellent, and stack
up favorably against the solid state recorders. DAT tapes are available
providing 60 minute and 120 minute recording times. A digital output is
available, but it is not known how easily a digitized recording could be
transferred to a PC using this output. Recordings can be transferred to a
PC in real-time using the Line In/Line Out connections. A microphone must
be purchased separately. Further information on this device may be found
at the following URL:
<a href="
http://www.sonystyle.com/home/item.jsp?hierc=9687&catid==
8662&itemid=591&telesale=null&hidden=null&cps=null&=
amp;type=s" eudora="autourl">
http://www.sonystyle.com/home/item.jsp?hierc=9687&catid=8662&it=
emid=591&telesale=null&hidden=null&cps=null&type==
s</a>
. A related product, the NT-2 Digital Micro Recorder is also available
from Sony. The NT-2 is smaller than the TCD-D100, but it has slightly
worse specifications. Further information on the NT-2 may be found at:
<a href="
http://www.sonystyle.com/home/item.jsp?hierc=9687&catid==
8668&itemid=563&telesale=null&hidden=null&cps=null&=
amp;type=s" eudora="autourl">
http://www.sonystyle.com/home/item.jsp?hierc=9687&catid=8668&it=
emid=563&telesale=null&hidden=null&cps=null&type==
s</a>
.<br><br>
2.1.4 Microcassette Recorders<br>
-<br>
Microcassette and cassette recorders are used by numerous law enforcement
agencies. These devices typically have relatively poor frequency response
(250 Hz to 4 kHz for microcassettes, somewhat better for cassettes), have
relatively high wow and flutter (due to the mechanical tape transport),
and have poor signal to noise and distortion characteristics when
compared to solid state recorders. Often they have voice activated
recording and automatic level control that cannot be turned off. In many
law enforcement applications, it is an advantage to be able to turn off
voice activated recording and automatic level control. Some advantages of
the microcassette and cassette recorder are they are small, they are low
cost, they use batteries that are widely available, and they use
cassettes that are widely available.<br><br>
The Sony M850V is a typical microcassette recorder. It has a frequency
response of 250 Hz to 4 kHz, which is well below the human hearing range
of 20 Hz to 20 kHz. It has an 11 hour battery life, and uses 2 AA
batteries. It is relatively small, with dimensions of 4” x 2.25” x 7/8=
”.
It has voice operated recording and automatic level control. This
recorder is monaural, and has its own built in microphone and speaker. It
features two recording speeds. The frequency response quoted is for the
higher recording speed. Further information on this recorder may be found
at the following URL:
<a href="
http://www.sonystyle.com/home/item.jsp?hierc=9687x8667x8671&am=
p;catid=&itemid=34003" eudora="autourl">
http://www.sonystyle.com/home/item.jsp?hierc=9687x8667x8671&catid=&=
amp;itemid=34003</a>
.<br><br>
2.2 Commercial Audio Recorder Issues<br>
-<br>
One issue (potential shortcoming for law enforcement use) with the flash
recorders and MP3 player/recorders is the bandwidth achieved in the audio
recording. For the recorders with LP (long play) mode, relatively low
sampling rates are used in recording the data (longer recordings can be
made in a given memory size when lower sample rates are used).
Unfortunately, the Nyquist sampling criteria limits the bandwidth of the
recording to one-half of the sampling frequency. Thus, the audio
bandwidths for recordings made in the LP mode are relatively low. A
typical sample rate for LP recording is 8kHz, and, by the Nyquist
sampling criterion, the resulting audio bandwidth must be less than 4kHz
(typically 3kHz). This audio quality is approximately the same as
telephone voice quality. The 32kbps ADPCM recording scheme used by the
MP3 player/recorder in Table 2-1 also achieves a telephone voice quality
bandwidth of approximately 3kHz. A somewhat better frequency response is
achieved by the SP (standard play) mode of the flash recorders. Sampling
rates used in this mode are 12kHz, and the resulting audio bandwidth must
be less than 6kHz (typically 5kHz). Although these bandwidths are
adequate for speech recognition purposes, they do not compare favorably
to the human hearing bandwidth of approximately 20 kHz, and they may not
be suitable for all law enforcement applications.<br><br>
Another issue with flash recorders and MP3 player/recorders is the loss
in fidelity caused by compression schemes used by the recorders. The
recorders attempt to maximize recording time for the available memory by
compressing the sampled audio using proprietary compression schemes. The
higher compression algorithms (greater than 4:1), which conserve the most
memory space, turn out to be lossy; that is, they degrade the fidelity of
the recording. The 32kbps ITU G.711 ADPCM compression used by the MP3
player/recorder in Table 2-1 is a relatively low loss algorithm. The
minidisc recorder uses ATRAC3 compression, which has a compression ratio
of 4.8:1. Manufacturers that use the MP3 compression standard have a
compression ratio of greater than 10:1. The DSS (digital speech standard)
compression scheme used by the Olympus DS2000 stores 120 minutes (7200
sec) of highest quality voice in 16Mbytes of flash. A rough calculation
of the seconds of uncompressed speech that can be stored in 16Mbytes
is:<br><br>
16M bytes x 1sec/12k samples x 1 sample/1 byte = 1333 sec<br><br>
Comparing the compressed seconds of storage to the uncompressed seconds
of storage gives a compression ratio of 5.4 for DSS. Although playback
quality may not be affected significantly by lossy compression schemes
for most purposes, one of the concerns in using nonlinear compression for
law enforcement recordings is the legal question that might be raised
over the accuracy and faithfulness of the recording.<br><br>
A third issue with the commercial audio recorders has to do with dynamic
range and signal to noise ratio of the recorded audio. The 8 bits per
sample used in these recorders provides for a signal to noise ratio that
will not exceed 50dB (i.e., 6.02n + 1.76). This signal to noise ratio may
not be adequate for all law enforcement applications. For example, if we
try to recover low level audio that is more than 50dB below some high
level audio, it will be buried in noise.<br><br>
Another issue in using commercial audio recorders for law enforcement
purposes is whether or not to use automatic gain control. It would be
desirable to be able to defeat the automatic gain control feature for
some applications.<br><br>
Another issue is whether or not to use voice activated recording. Voice
activated recording conserves room on the recording medium, and it
extends battery life by shutting down the process of recording the audio
onto the storage medium when no audio is present. However, a threshold
must be set to activate the voice recording. If the threshold is set too
high, some weaker signals that are desirable evidence may be missed. So,
for recorders that do have voice activation, it is desirable to be able
to turn off the voice activation.<br><br>
Another issue is storage of original evidence at a reasonable price.
Unfortunately, there is often a significant delay of months or even years
before a case comes to trial. It would be costly to have to remove the
recorder from use while waiting on a trial. And downloading the original
recording to a PC may not be accepted as original evidence. The original
flash memory module used to record the audio may be the only recording
accepted as original evidence. The ability to have removable flash memory
that can be saved as original evidence is a desirable feature of flash
based audio recorders. Using removable flash memory allows the recorder
to continue to be used (with a new flash module) while the original
evidence flash module is saved for trail.<br>
3.0 CURRENT AND PROJECTED STATE OF THE ART IN SOLID STATE RECORDER
TECHNOLOGY AREAS<br><br>
3.1 Solid State Recorder Block Diagram<br>
-<br>
Figure 3-1 shows a block diagram of a typical solid state voice recorder.
Starting at the top left of the figure, the audio signal is received by
one or more microphones. Next, an amplifier/filter increases the voltage
of the signal from the microphone to the correct level for the input to
the analog to digital converter (ADC). Some filtering (removal) of
unwanted signals may also occur in this block. The ADC converts the
analog input signal to a digital word that is fed to audio compression
hardware. The audio compression hardware (for example, a DSP) implements
an audio compression algorithm, which reduces the number of bytes needed
to store the audio signal. The resulting compressed audio bytes are saved
in flash memory. The flash memory is removable, for convenient storage of
evidence. The user interface is implemented by the DSP sensing the switch
positions on the recorder.<br><br>
<br>
Figure 3-1. Solid state voice recorder block diagram. <br>
-<br>
Playback of the audio signal may be provided in a number of ways.
Commercial items, such as the SanDisk ImageMate USB
CompactFlash/SmartMedia media reader (cost is approximately $30), may be
used to transfer the compressed audio from the flash memory card to a PC.
The PC would run software to decompress the audio, and play back the
result on the PC sound system. The flash recorder manufacturer could
provide custom audio decompression PC software, or, if a standard
compression/decompression algorithm is used, a third party could provide
the audio decompression PC software.<br><br>
In the design of the flash based audio recorder, the resulting product is
only as good as the worst individual part. The various parts of the
recorder will be discussed in the following sections.<br><br>
3.2 Microphones<br>
-<br>
Two major classes of microphones that could be used in flash based audio
recorders are dynamic and electret condenser. The dynamic microphone
transforms sound into an electrical signal by the movement of a diaphragm
with a coil of wire attached to it. This coil of wire is located close to
a magnet, and when the diaphragm/coil moves in the magnetic field, a
current is produced in the coil. This current corresponds to the audio
signal that moves the diaphragm.<br><br>
Electret condenser microphones operate somewhat differently. The
diaphragm and the “back plate” in an electret condenser microphone form
two surfaces of a capacitor. Either the diaphragm or the back plate
contains a permanently charged electret material. When the diaphragm
moves, the distance between the surfaces of the capacitor changes,
inducing a current that corresponds to the audio input. Since the
electret condenser microphone diaphragm does not have a coil attached to
it, it can be fairly light when compared to the dynamic microphone. As a
result, the electret condenser microphone generally has better
sensitivity and high frequency response than a dynamic
microphone.<br><br>
One characteristic of a microphone is its directivity, which is its
sensitivity to sound arriving from different directions. A microphone
that picks up sound equally well in all directions has an
“omnidirectional” pattern. A microphone that is more sensitive to sound=
s
in front of the microphone that behind the microphone is “unidirectional=
”
(cardioid). A microphone that is sensitive to sounds in front and behind,
but not to the sides, is “bi-directional” (noise canceling).<br><br>
One typical electret condenser microphone is the Panasonic WM-61 series.
This back electret condenser omnidirectional microphone has a frequency
response from 20 to 20kHz, a signal to noise ratio of better than 62dB,
and a sensitivity of –35dB (0dB = 1V/Pascal). The WM-61 has a low power
consumption of 0.5 mA at 2V.<br><br>
For law enforcement applications, the superior frequency response and
sensitivity of the electret condenser microphone is preferred over the
dynamic microphone. For monophonic recordings, an omnidirectional
microphone is preferred. For stereo applications, either cardioid or
omnidirectional pattern microphones may be used.<br><br>
<br>
3.3 Delta Sigma Analog to Digital Converters<br>
-<br>
In solid state audio recorders, audio signals must be converted from the
analog domain to the digital domain so they can be stored in a digital
format in flash memory. This function is performed by a high quality
analog to digital converter (ADC). One type of ADC that is particularly
well suited to this task is the delta sigma ADC. The delta sigma ADC
samples the input waveform at a much higher rate than is normally
required (often 128X oversampling is seen in these devices). Oversampling
distributes the quantization noise all the way up to the sampling
frequency, thereby reducing the amount of quantization noise in the audio
band. The delta sigma ADC also uses a noise shaping filter. This noise
shaping filter effectively moves quantization noise from the audio band
to higher frequencies. The delta sigma converter then uses a digital
filter to remove the higher frequencies (and quantization noise), and
retain the audio frequencies. The resulting digitized signal from the
delta sigma converter has very little quantization noise, and is highly
accurate representation of the input analog waveform.<br><br>
Delta sigma converters are typically inexpensive, have low power
requirements (suitable for a battery operated voice recorder), and are
highly accurate. One device, the CS5333, converts two inputs (for stereo
operation), provides 24 bits of output, requires only 11mW, and costs
less than $5. As seen in the specifications of the CS5333, the current
generation of delta sigma analog to digital converter has more than
enough performance to meet law enforcement needs of high dynamic range
and full audio bandwidth, and it is reasonably priced.<br><br>
Delta sigma ADCs and digital to analog converters (DACs) are used in
audio sound cards for PCs. The demand for these products may roughly be
expected to follow the demand for PCs in the future. Which is to say, the
high demand for delta sigma converters (both digital to analog and analog
to digital) in computer audio sound systems makes the continued
availability of these devices highly likely.<br><br>
3.4 Audio Compression Algorithms<br>
-<br>
Audio compression algorithms are used in flash based audio recorders to
reduce the amount of flash memory required to record a specified duration
of audio. If 20kHz audio is sampled at the Nyquist rate of 40kHz, then,
in the absence of compression, each second of audio requires 40,000 audio
samples to be stored in flash memory. A compression algorithm that
achieves a compression ratio of 4:1 would reduce the flash memory storage
requirements from 40,000 down to 10,000. A fixed amount of flash memory
can store 4 times as much audio when a 4:1 compression algorithm is
used.<br><br>
Audio compression algorithms may be divided into two categories: lossless
and lossy. When recordings made using lossless compression are played
back, the original signal is reproduced exactly, and no compression
artifacts are present. When recordings made using lossy compression are
played back, the original signal is not exactly reproduced, but a
slightly degraded version of the original signal is reproduced. Lossless
audio compression schemes typically achieve compression ratios in the
range of 1.5:1 to 3:1. Lossy audio compression schemes typically achieve
compression ratios in the range of 4:1 to 12:1 and higher (Windows Media
Audio claims 24:1).<br><br>
The state of the art lossless audio compression process can be divided
into three stages: framing, decorrelation, and entropy coding. Framing
divides the audio signal into equal duration frames. Optimum duration
frames appear to be in the range of 13 to 26ms. Audio signals exhibit a
high degree of autocorrelation; that is, the current sample can be
predicted from previous samples. To take advantage of this
characteristic, the original signal is decorrelated (and the correlation
characteristic is remembered). It is more efficient to store the
correlation characteristics in the encoded waveform than it is to store
the audio samples. Several techniques are available for performing
decorrelation: coding with linear prediction, coding with approximation,
and transform coding. Once the correlation in the waveform has been
removed, the remaining decorrelated waveform must be encoded. Entropy
coding is used for this purpose. Some standard entropy coding methods
include: Huffman coding, run length coding, and Rice coding. Some
representative state of the art lossless audio encoders include the
following:<br><br>
- AudioPAK (integer)<br>
- MUSICompress (fixed point)<br>
- Sonarc (fixed point)<br>
- Shorten (floating point)<br>
- Ogg Squish (floating point)<br>
- LTAC (lossless transform audio compression - floating point)<br>
- Waveform Archiver (floating point)<br><br>
The objective of the AudioPAK algorithm is to reduce the complexity of
implementing lossless audio compression while maintaining compression
ratios that are comparable to the more complex lossless audio compression
algorithms. Notice that the AudioPAK uses integer operations, while the
other algorithms mentioned use fixed point or floating point operations.
Much of the material in this section has been derived from: Optimization
of Digital Audio for Internet Transmission by Mat Hans – Georgia
Institute of Technology PhD thesis, 1998 (
<a href="
http://users.ece.gatech.edu/~hans/" eudora="autourl">
http://users.ece.gatech.edu/~hans/</a> ). This thesis describes the
AudioPAK algorithm.<br><br>
Another lossless compression technique worth mentioning is bit plane
encoding. With this technique, a particular bit of each PCM sample is
encoded over a frame of samples. This process is repeated until all bits
have been encoded. This technique expects that the most significant bits
of audio PCM samples will not change very often, and can be efficiently
run length encoded. The least significant bit is expected to change
frequently, and can use Huffman entropy encoding.<br><br>
Before leaving the subject of lossless audio compression, it is
interesting that lossless audio compression can be achieved by using the
PKzip (Winzip) algorithm that is so familiar to today’s computer users.
Unfortunately, the PKzip algorithm does not achieve very good compression
ratios for audio files (typically 1.1:1). One reference that discusses
PKzip (Winzip) performance relative to other lossless audio compression
techniques, and lossless audio compression performance in general is
“Digital Audio Gets an Audition, Part 1 Lossless Compression,” EDN,
January 4, 2001.<br><br>
A number of lossy audio compression techniques are available. We will
briefly describe two here: MP3 and ATRAC. These two compression
techniques are used in MP3 players and mini disc recorders
respectively.<br><br>
Although the MP3 players do not use MP3 compression to record audio (they
typically use 32kbps ADPCM instead), the compression scheme is one
dominant form of lossy audio compression that is used today (the
Soundtainer product mentioned above uses MP3 to record audio). MP3 stands
for MPEG Audio Layer 3 (as opposed to the lower compression MPEG Audio
Layers 1 and 2). MP3 can achieve compression ratios of 10:1 to 14:1 with
a bandwidth of over 15kHz. One method the MP3 algorithm uses to reduce
the amount of information to be encoded (thereby compressing the size of
the audio file) is to omit audio that is not perceptible to humans. One
example is called frequency masking. In this situation a loud sound
present in one frequency band masks softer sounds present in an adjacent
frequency band. In this instance, humans will not notice a difference
when the soft sounds are completely removed. Similarly, MP3 observes a
minimum audio threshold, and will not record sounds below a certain level
at certain frequencies (2 to 5kHz), since these will not be perceptible
to humans. MP3 “borrows” from a “reservoir of bytes” to encode more
complex audio, and “replenishes” the reservoir during less complex audi=
o
passages. MP3 uses a discrete cosine transform that has 384 coefficients
to decorrelate the audio signal. It then throws away data that would not
be noticed by the listener. Finally, MP3 uses Huffman entropy encoding,
once the audio that will not be encoded has been subtracted from the
signal.<br><br>
ATRAC (Adaptive Transform Acoustic Coding) is the compression technique
used in mini disc recorders. It typically achieves a 5:1 compression
ration on CD audio, and, like MP3, it uses a psychoacoustic model of
human hearing to determine what sounds may be subtracted form the
original signal without being detected by human hearing. ATRAC divides
the audio frequency band into 3 subbands (0-5.5kHz, 5.5-11kHz, and
11-22kHz) using Quadrature Mirror Filters (QMFs – prevents aliasing when
reconstructing). A discrete cosine transform is performed on each subband
using an adaptive block length (long or short). Long block lengths
provide superior frequency resolution, but are subject to “pre echo”
during “attack” portions of the audio signal. Short block lengths are
used to prevent pre echo. and transforms these subbands into the
frequency domain. Signals that would be masked by psychoacoustic effects
are subtracted from the resulting frequency domain coefficients, and the
coefficients are encoded into BFUs (block floating units). (reference:
ATRAC: Adaptive Transform Acoustic Coding for Mini Disc, Tsutsui et al,
93rd Audio Engineering Society Convention, Oct 1-4, 1992.) A newer
version of ATRAC called is ATRAC3 is now available.<br><br>
Both MP3 and ATRAC (as well as many other lossy audio compression
algorithms – AC3 for example) use a psychoacoustic model of human hearing
to remove signals that would not be perceived by humans as a method of
reducing the amount of audio that must be saved (as a method of
compressing audio). For law enforcement applications, this practice may
not be acceptable in some situations. For example, if a soft sound
contains information needed by law enforcement personnel, and a loud
sound “masks” it, the soft sound will be removed from the encoded audio
when either MP3 or ATRAC is used. For this reason, audio compression
algorithms that rely on the psychoacoustic model of human hearing to
delete audio signals are not recommended for law enforcement applications
(at least not all law enforcement applications).<br><br>
Other lossy audio compression algorithms include:<br><br>
- AAC (Advanced Audio Coder www.aac-audio.com )<br>
- ATELP (
<a href="
http://www.softsound.com/ATELP.html" eudora="autourl">
www.softsound.com/ATELP.html</a> )<br>
- DTS (
<a href="
http://www.dtsonline.com/" eudora="autourl">www.dtsonline.com<=
/a>
)<br>
- ePAC (
<a href="
http://www.lucent.com/ldr" eudora="autourl">
www.lucent.com/ldr</a> )<br>
- Indeo (
<a href="
http://www.ligos.com/" eudora="autourl">www.ligos.com</a> )<br=
>
- Ogg Vorbis (
<a href="
http://www.vorbis.com/" eudora="autourl">www.vorbis.com</a>
)<br>
- Qdesign (
<a href="
http://www.qdesign.com/" eudora="autourl">www.qdesign.com</a>
)<br>
- Real Audio (
<a href="
http://www.real.com/" eudora="autourl">www.real.com</a> )<br>
- TAC (kk-research.hypermart.net)<br>
- TwinVQ (sound.splab.ecl.ntt.co.jp/twinvq-e)<br>
- Windows Media Audio (
<a href="
http://www.microsoft.com/windows/windowsmedia" eudora="autourl=
">
www.microsoft.com/windows/windowsmedia</a> )<br><br>
A reference that gives an overview of these algorithms and their
performance is “Digital Audio Gets an Audition, Part 2 Lossy
Compression,” EDN, January 18, 2001.<br><br>
One further audio compression algorithm worth mentioning is apt-X 4:1.
This algorithm uses ADPCM to achieve a 4:1 compression ratio, with very
little loss in audio quality. This algorithm uses four frequency
subbands, but does not rely on psychoacoustic models of human hearing to
throw away audio information that is not perceptible to humans. More
information may be found on this technique at
<a href="
http://www.aptx.com/" eudora="autourl">
http://www.aptx.com</a>
.<br><br>
3.5 Audio Compression Hardware<br>
-<br>
Audio compression algorithms are implemented on audio compression
hardware, which includes Digital Signal Processors (DSPs) and custom
Application Specific Integrated Circuits (ASICs). DSPs are specialized
computer chips that have features that facilitate the implementation of
audio compression algorithms. Like any computer, DSPs may be reprogrammed
to perform different functions. ASICs are not reprogrammable. The Field
Programmable Gate Array (FPGA) may be used to develop algorithms that are
then readily transferred into an ASIC.<br><br>
Highly capable, low cost DSPs have become available in the past few
years. For example, the TMS320VC5402 DSP from Texas Instruments is
capable of 100 million instructions per second (MIPS), and has 16K words
of on chip RAM, and 4K words of on chip ROM. The cost of this part is
approximately $6 in quantities of 1000. A part specifically designed for
low power consumption, the TMS320VC5502, is also becoming available. It
has 32K words of on chip RAM and 4K words of on chip ROM, and features
400 MIPS performance. The 5502 part will sell for approximately $10 in
quantities of 1000. Both the 5402 and the 5502 are fixed point
processors.<br><br>
The trends toward lower core voltages, smaller geometry devices, and
higher processing capabilities in DSPs and ASICs can only benefit flash
based audio recorders. The current capabilities of DSPs like the 5502 are
more than adequate for implementing fixed point and integer lossless
audio compression algorithms for flash based audio recorders.<br><br>
3.6 Flash Memory<br>
-<br>
Flash memory is used to store the compressed audio in the solid state
recorder. Flash memory is used in cell phones, digital cameras and MP3
players. The cost of the flash memory is the dominant cost of the
recorder. A 16kHz Obviously, any reductions in the cost of flash will
reduce the cost of the flash based audio recorder.<br><br>
In mid 1998, an Intel 28F640J5 8 M byte flash part cost $65. Today (2002)
a comparable part, the 28F640J3A, costs $13.42, a reduction of nearly 5
times. Perhaps even more important than the cost of the individual flash
chips is the cost of removable flash media. Driven by the proliferation
of digital cameras and MP3 players, the cost of removable flash media has
dropped significantly in the past years. Today’s street prices for
SanDisk flash products are as follows:<br><br>
CompactFlash 1 G Byte: $631<br>
CompactFlash 512 M Byte: $269<br>
CompactFlash 256 M Byte: $115 (if 1000 units are purchased, the cost is
$102)<br>
CompactFlash 128 M Byte: $63<br>
Memory Stick 128 M Byte: $70<br>
Secure Digital 256 M Byte: $161<br>
MultiMedia 64 M Byte: $52<br>
Ultra CompactFlash 128 M Byte: $77<br><br>
Flash memory provides a method of storing digital audio data that is non
volatile; that is, data stored in flash memory is not lost when the power
is turned off to the device. Flash memory may be NAND based or NOR based.
NAND based technology is considered well suited for high capacity data
storage applications, such as storage of audio files. Current flash
memory for file storage often uses 2 bit per cell storage, an improvement
over the older single bit per cell flash technology.<br><br>
Flash memory that uses 0.25, 0.16, and 0.13 micron semiconductor process
technology is currently available, and smaller process technology is
being planned. Parts that operate on 3V and 1.8V are commonplace, and
lower voltages are being planned. These developments are expected to
reduce cost for given storage size devices, and lower power consumption,
which would both benefit flash based audio recorders. SanDisk is
expecting prices to drop approximately 30% over the coming year.<br><br>
3.7 Batteries<br>
-<br>
Batteries for mobile electronic applications such as digital cameras and
MP3 players may be divided into two groups: rechargeable and
non-rechargeable. Within the rechargeable group, the most popular
technologies today are: nickel metal hydride (NiMH) and lithium ion
(Li-Ion). In the non-rechargeable group, the most popular technologies
are the alkaline and carbon zinc batteries.<br><br>
Nickel metal hydride batteries require recharging more often than lithium
ion batteries, but they cost less than lithium ion batteries. Lithium ion
batteries provide a better energy density than nickel metal hydride
batteries. An energy density figure of 75 Watt hours per kilogram is
provided by nickel metal hydride batteries, versus 135 Watt hours per
kilogram for lithium ion batteries. The output voltage of lithium ion
batteries is typically higher (3.0V) than the output voltage of nickel
metal hydride batteries (1.2V). When compared to nickel cadmium
rechargeable batteries, both nickel metal hydride and lithium ion
batteries offer the advantage of not having any memory effect. (
<a href="
http://www.nec-tokin.net/now/english/product/me/chisiki/li3.html=
" eudora="autourl">
http://www.nec-tokin.net/now/english/product/me/chisiki/li3.html</a>
).<br><br>
The alkaline battery in a AA size can provide 3000mAh at 1.5V, and an
energy density of 140 Watt hours per kilogram. In comparison, a AA carbon
zinc battery in AA size can provide only 950mAh at 1.5V and an energy
density of 50 Watt hours per kilogram.<br><br>
Alkaline and carbon zinc batteries have sloping discharge curves. That
is, as the battery is discharged, the voltage goes down over time. In
contrast, the nickel metal hydride and lithium ion batteries have flatter
discharge curves. When these batteries are discharged, the voltage does
not go down over time as much as with the alkaline and carbon zinc
batteries.<br><br>
The popularity of laptop computers, cell phones, cordless phones, digital
cameras, MP3 players, and personal digital assistants has spurred the
demand for rechargeable batteries. In 2000, the market for rechargeable
batteries was $1.75 billion. This market is projected to grow to $2.19
billion by 2006. The technology that will account for most of the battery
demand in 2005 is the Lithium Ion. Lithium Ion batteries are expected to
grow from 25% of the battery market in 1999 to 55% of the battery market
in 2005. (source:
http://www.eetimes.com/myf00/ao_batt.html )<br><br>
One emerging battery technology is Lithium Ion Polymer. This battery
technology has the potential to greatly increase the energy density when
compared to current Lithium Ion products. Another Lithium Ion emerging
battery technology replaces the cobalt in the battery with a different
cathode material. The problem with cobalt is that it requires protection
circuits inside the battery to prevent thermal runaway when the battery
is being charged. One company, Valence Technology, claims that using
Saphion for the cathode will reduce the cost of lithium ion batteries (
<a href="
http://www.valence.com/saphion.asp" eudora="autourl">
http://www.valence.com/saphion.asp</a> ).<br><br>
The development of a low cost, widely available, lithium ion polymer
battery with high energy densities could reduce the size required for
batteries in the flash based audio recorder. A reduced size recorder has
obvious advantages for law enforcement purposes. It is uncertain when
lithium ion polymer batteries will reach this stage of
development.<br><br>
<br>
4.0 PROJECTED COMMERCIAL SOLID STATE RECORDER PRODUCTS<br>
(IN 2 YEARS)<br>
-<br>
Solid state recorder products are in a state of rapid development and
improvement, with models constantly being discontinued and replaced by
newer, improved models. Since this program and evaluation of devices
began, many of the models initially in the product matrix (see Appendix
A) had to be dropped and replaced by more current models. The trend has
been toward recorders/MP3 players with larger memory capacity and lower
costs. This trend is expected to continue.<br><br>
But the demand for higher bandwidth portable voice recorders has not been
seen yet. There is a strong demand for MP3 players (i.e., there are lots
of MP3 player products being sold), which feature voice recorders as a
secondary feature. And there is a strong demand for voice recorders used
for business dictation applications. But these kinds of voice recorders
do not need to have high bandwidth, and no mass market commercial voice
recorder has bandwidths up to 16 kHz or higher that would be useful for
law enforcement applications.<br><br>
As removable flash memory cards continue to increase in memory storage
space and decrease in cost, the solid-state MP3 recorder/players may
evolve to take advantage of this storage space and become true music
recorders as opposed to simply voice recorders. Commercial motivation may
encourage these recorder/players to take advantage of the 44.1 kHz
sampling frequencies currently used to decompress and playback the MP3
files and begin recording audio files in stereo with MP3 compression as
opposed to simply decompression. These devices could then compete very
favorably with the minidisc recorder/players in the marketplace, being
slightly smaller in size and less power hungry. But the market is not
seen for MP3 recorders. Most consumers are not interested in recording
their own music, via a microphone. Instead, they are interested in
transferring music tracks from CDs or from the web to their PC and then
storing them on MP3 players. And even if a MP3 recorder did evolve, the
lossy compression used in MP3 is not always suitable for law enforcement
purposes.<br><br>
The market for high fidelity portable audio recorders would seem to be
pretty much the same as the market for portable digital audio tape (DAT)
recorders. Today, this market is a low volume, relatively high cost
niche. For example, the Sony TCD-D100 DAT recorder lists for $900 and it
was difficult to find a dealer that sells this device. It may be possible
that DAT will evolve into a flash based product, but the demand to make
it a low cost item sold in large quantities is not seen.<br><br>
<br>
5.0 AUDIO RECORDER CONCLUSIONS<br>
-<br>
Desired characteristics of audio recorders for law enforcement purposes
are as follows:<br>
- wide and flat frequency response (20-20kHz)<br>
- high signal to noise ratio and dynamic range<br>
- low wow and flutter<br>
- low harmonic distortion<br>
- lossless compression<br>
- defeatable automatic level control<br>
- defeatable voice activated recording<br>
- stereophonic recording<br>
- combined microphone response that is omnidirectional<br>
- record times of at least 60 minutes, with 120 minutes and higher
available<br>
- removable media<br>
- low cost media<br>
- wide availability of media<br>
- wide availability of batteries<br>
- small size<br>
- low cost<br><br>
Large volumes are projected in MP3 player recorders, and, to a lesser
extent, in solid state voice recorders. But large volume (low cost)
products (current and projected) fall short in several key
areas:<br><br>
Current and projected future commercial flash recorder products
weaknesses:<br>
1. lossy compression<br>
2. poor frequency response<br>
3. poor dynamic range and signal to noise ratio<br><br>
Current and projected MP3 player voice recorder combination product
weaknesses:<br>
1. poor frequency response<br>
2. poor dynamic range and signal to noise ratio<br><br>
Current and projected mini disc weaknesses:<br>
1. lossy compression<br>
2. Susceptibility to shock and vibration<br><br>
Many commercial products use automatic level control and voice activated
recording features that cannot be defeated. The ability to defeat these
features is desirable for many law enforcement applications.<br><br>
To record 120 minutes of uncompressed 16 bit PCM audio with a bandwidth
of 16 kHz (sampling at 32 kHz) requires 460 Mbytes of flash. State of the
art lossless compression algorithms achieve 2:1 to 3:1 compression
ratios. So, when lossless compression is used, only 230 Mbytes of flash
are needed (instead of 460). Four years ago, the cost of flash was
approximately $8 per Mbyte, making the cost of the flash memory required
in the above situation about $1200. Today, the cost of flash has dropped
to less than $0.50 per Mbyte (street price). A 256 Mbyte CompactFlash
plug in card can be purchased for $128 or less. So the falling cost of
flash has improved the affordability of the flash based audio
recorder.<br><br>
$128 is a significant amount of money to spend on a recording medium that
may be put on the shelf while waiting on a trial. But it is much better
than the $1200 or so it would have cost 4 years ago. And the expected
future improvements in the cost of flash memory will reduce this $128 to
an even lower figure.<br><br>
Falling prices and larger sizes of flash memory make the solid state
recorder a very practical idea. Improved wow and flutter, increased
immunity from shock and vibration, and the elimination of tape hiss
(improved signal to noise ratio) result from the use of flash. But
widespread commercial demand for improved frequency response, high
dynamic range, high signal to noise small solid state recorders is not
seen. Instead, business uses of recorders for dictation purposes, which
do not require high bandwidth, high dynamic range, and high signal to
noise ratios are seen as the driving factor in future solid state audio
recording products. There is a market for high bandwidth, high dynamic
range, high signal to noise ratio playback products (MP3 players), but
only for playback, not for recording.<br><br>
To get the characteristics of lossless compression, high bandwidth, high
dynamic range, and high signal to noise ratio, law enforcement personnel
must continue to purchase specialty products such as the FBIRD or the
SSABR. No projected future high volume commercial product will provide
all the capabilities of these devices.<br><br>
It would be possible to make a product that would satisfy today’s law
enforcement demands at a reasonable cost. Appendix B shows the major
components of such a product, and that the cost would be around $140 in
parts (excluding circuit boards and cases, and assuming parts are
purchased in quantities of 1000). The major cost factor in this product
is the flash memory, which accounts for $102 out of the $140 total
component cost. The product would feature lossless compression (2:1), a
bandwidth of 16kHz, a signal to noise ratio approaching 90 dB, and over
120 minutes of record time. It would feature removable flash media, and
not have automatic level control or voice activated recording. It would
have a size of approximately 9 square inches.<br><br>
<br>
6.0 COMMERCIAL BODY WIRE PRODUCTS<br><br>
-Law enforcement agencies utilize body-wires for officer security, and to
obtain evidence. Audio quality, transmitted power, and price vary with
different systems. Typically, the transmission range can be between 30
and more than 3,000 feet depending on the environment and the quality of
the equipment. In addition, there are many different frequencies utilized
for transmission.<br><br>
The purpose of the body wire is to transmit audio in the form of a radio
signal to be received, understood and/or recorded at a remote location.
The person wearing the body wire can be moving and turning in locations
that range from outside to inside a building and from ground level to
many stories up. While the transmitter is often mobile, the receiver is
generally in one location. In the case of vehicle audio surveillance,
both the transmitter and receiver are in motion, but the transmission
distance is generally constant.<br><br>
The transmitter may be required to cover thousands of feet or a few
yards. The different environments of the signal propagation path will
cause different attenuation levels: the signal will become attenuated and
the range, therefore, reduced if it is required to travel through
numerous buildings. Noise conditions, present at the time the audio
signal is recorded, will vary from those of an outdoor, urban environment
(which has many possible levels of background noise) to that of a quiet
indoor room.<br><br>
Throughout this program, data has been collected on body wires that could
be used for law enforcement applications under the conditions described
above. A wide variety of body wires are available. These products are
discussed in the following sections.<br><br>
-6.1 Overview of Body Wire Products<br>
-<br>
Table 6-1 presents a summary of the performance of various kinds of body
wires. Three types of body wires cover the majority of body wire
products: Narrow Band FM (NBFM), Digital, and Spread Spectrum Digital.
Actual representative products were evaluated to fill in this table, but
the band names of the products have been omitted. Strengths and
weaknesses of each body wire category are listed in the table.<br><br>
<br>
Table 6-1. Body Wire Types and Feature Comparison <br>
-<br><br>
-6.1.1 Narrow Band FM Body Wires<br>
-<br>
Narrow band frequency modulation (FM) body wire transmitters use the
output signal of the microphone to frequency modulate a radio frequency
(carrier) to form the transmitted waveform. This modulation technique is
an analog modulation technique, since the microphone signal was not first
digitized before modulating the carrier.<br><br>
Frequency modulation leaves the amplitude of the carrier constant, but
changes the “instantaneous” frequency of the carrier in accordance with
the amplitude of the signal from the microphone. Loud audio signals from
the microphone correspond to relatively large changes in the frequency of
the carrier. Soft audio signals from the microphone correspond to
relatively small changes in the frequency of the carrier. Since FM
signals use the frequency instead of the amplitude of the carrier to
carry the audio from the microphone, they are inherently immune to
amplitude noise.<br><br>
Commercial FM radio stations use a form of FM called wideband FM. In this
case, the frequency of the carrier can change up to 75kHz due to loud
audio from the microphone. In contrast, body wires use narrowband FM. For
narrowband FM the frequency deviation caused by loud audio from the
microphone is much less than the wideband case. In the case of narrowband
FM, the frequency of the carrier can change up to 5 or 7kHz.<br><br>
-6.1.2 Digital Body Wires<br>
-<br>
A digital body wire passes the output of the microphone through an analog
to digital converter (ADC), the output of which is a series of 1’s and
0’s referred to as bits of digital data. This digital data then modulates
a radio frequency (carrier) using a digital modulation
technique.<br><br>
One example of a digital modulation technique that is commonly used in
digital body wires is phase shift keying (PSK). In the case of phase
shift keying, the phase of the carrier is changed according to whether a
1 or a 0 data bit is being transmitted. For example, transmitting a 0
data bit could correspond to no change in the phase of the carrier, and
transmitting a 1 could correspond to a 180 degree change in the phase of
the carrier. There are many variations on this simple example of PSK that
could be used in body wires. Other digital modulation techniques, such as
differential phase shift keying (DPSK), frequency shift keying (FSK), or
amplitude shift keying (ASK), are also possible to use in digital body
wires. A comparison of some of the more common digital modulation
techniques is shown in the table below.<br><br>
<br>
Table 6-1. Comparison of Binary Digital Modulation Schemes (from Digital
and Analog Communication Systems by K. Shanmugam) <br>
-<br>
As seen in the table, the PSK and DPSK techniques achieve the best bit
error rate performance. Digital modulation schemes with better bit error
rate performance will require less transmit power to communicate over a
fixed range, which is beneficial for battery life. Or, equivalently, a
system with better bit error rate performance can communicate over a
longer range using a fixed transmit power.<br><br>
The “S/N” term in the table refers to signal to noise ratio. This term =
is
the ratio of the received signal power to the received noise power. This
ratio is often measured in decibels (dB). The formula for S/N in dB is 10
x LOG10 (signal power/noise power), where LOG10 is a base 10 logarithm.
In the table, notice that the ASK modulation takes a S/N of 18.33 dB to
achieve a bit error rate of 10-4. In comparison, PSK modulation only
takes a S/N of 8.45 dB to achieve the same bit error rate. So it takes a
lower signal power for PSK than for ASK to achieve a given bit error rate
performance, which is an advantage of using PSK modulation instead of ASK
modulation.<br><br>
-6.1.3 Digital Spread Spectrum Body Wires<br>
-<br>
The digital spread spectrum body wire passes the output of the microphone
through an analog to digital converter. Next, the resulting digital data
is further encoded by another (higher rate) sequence of 1’s and 0’s
referred to as a PN (Pseudorandom Noise) sequence. The resulting high
rate digital sequence is then used to modulate the carrier. This form of
spread spectrum is referred to as “direct sequence” spread
spectrum.<br><br>
The rate of the PN sequence is referred to as the “chip rate,” and the
rate of data bits from the microphone’s analog to digital converter is
referred to as the data rate (rb). The processing gain of the spread
spectrum signal is approximately the ratio of the chip rate to the data
rate.<br><br>
The effect of encoding the microphone digital data with the higher rate
PN sequence is to spread the energy of the transmitted signal over a
wider band of frequencies than would otherwise be used. One advantage of
spreading the frequencies in this manner is that the signal becomes
harder to detect than non spread signals.<br><br>
When a direct sequence spread spectrum signal is received, the first
operation is to “despread” the received signal. This dispreading
operation is performed by multiplying a time aligned version of the PN
sequence with the PN sequence in the received waveform. As a result of
this despreading, any narrowband interferers present in the received
signal will be spread out, and less energy from the interferer will be
passed into the data demodulation process. The amount of rejection of the
narrowband interferers corresponds to the processing gain of the signal.
So a second advantage of a digital spread spectrum body wire is its
ability to reject narrowband interference.<br><br>
One further possible advantage of direct sequence spread spectrum worth
mentioning is its secure communication capability. It is necessary to
know the transmitter’s PN sequence in order to despread the signal and
listen to the audio. Using a long PN sequence, and keeping the PN
sequence confidential can achieve secure communications.<br><br>
In addition to direct sequence spread spectrum, another form of spread
spectrum, called frequency hopping, is also possible. Instead of using a
PN sequence, frequency hopping spread spectrum systems change the carrier
frequency of the transmitted waveform periodically. Frequency hopping
spread spectrum systems have similar advantages to direct sequence spread
spectrum systems.<br><br>
-6.2 Body Wire Issues For Law Enforcement<br>
-<br>
When considering acquiring or using a body wire system, the user should
be cognizant of six performance features:<br><br>
1. Voice Quality of the Received Audio<br>
2. Transmission Range<br>
3. Battery Lifetime<br>
4. Physical Disguise<br>
5. Electronic Security<br>
6. Cost<br><br>
Understanding the role played by these six attributes will prove to be an
asset in determining the suitability of a body wire system under
consideration. Each of the features 1 through 5 is discussed in this
section. The most significant manufacturer specifications for each
feature are provided along with the associated Figures of Merit.<br><br>
6.2.1 Voice Quality of the Received Audio<br>
-<br>
Voice quality is a very important aspect of body wire performance. Poor
voice quality could prevent the listener from hearing words in a
conversation, from understanding words in a conversation, or it could
prevent the listener from determining which person was speaking in a
conversation. Any of these problems could place the agent in danger, or
prevent the collection of information needed to solve a case.<br><br>
Voice Quality can be assessed primarily from specifications of AUDIO
BANDWIDTH, AUDIO DYNAMIC RANGE and, if available, AUDIO SIGNAL TO NOISE
RATIO.<br><br>
AUDIO BANDWIDTH (audio frequency response) is an important measure of
voice quality. A body wire system with the highest audio bandwidth
performance will cover the entire range of frequencies that can be heard
by the human ear (20 Hz to 20 kHz). A high quality music Compact Disc
(CD) has a frequency response of 20 Hz to 20 kHz. A body wire system that
only covers the frequency response of a telephone, which is 400 Hz to 4
kHz, would be considered to have relatively poor audio bandwidth
performance. There is clearly a difference between the sound quality of a
voice on a telephone and a music CD. Table 6-2 below presents useful
information for evaluating the specifications and performance of body
wire audio bandwidth.<br><br>
<br>
Table 6-2. Figure of Merit for Body Wire Audio Bandwidth <br>
-<br>
AUDIO DYNAMIC RANGE (ADR) is another important measure of voice quality.
ADR is a measure of the systems ability to handle loud and soft sounds.
It is the ratio of the loudest undistorted signal that the system can
handle compared to its internal noise. Ideally, a body wire system with
the best ADR would have the same dynamic range as the human ear, which
has a dynamic range of over 120 dB. However, contemporary digital
recording techniques can only achieve a dynamic range of about 90
dB.<br><br>
Table 6-3 shows relative volume levels for different sounds. The levels
in dB are relative to the threshold of hearing that is taken to be 0dB.
>From the table, it is seen that the audio dynamic range necessary to
capture audio from whispers to a shout must be greater than 72 dB (90-18
dB).<br><br>
<br>
Table 6-3. Relative Audio Dynamic Range – Sound Pressure Level <br>
-<br><br>
From the above table, figures of merit may be determined for body wire
systems. Table 6-4 shows the figures of merit for the audio dynamic range
of a body wire system.<br><br>
<br>
Table 6-4. Figure of Merit for Audio Dynamic Range <br>
-<br><br>
Many body wire radio systems do not have sufficient dynamic range to
handle full audio sound levels. Some may be limited to as little as 30-40
dB. When audio dynamic range is limited, sometimes automatic gain control
(AGC) is used to position the dynamic range window in the most
advantageous place to accurately pick up the most critical audio levels.
The AGC shifting of the dynamic range window may produce undesirable
audio artifacts. A carefully crafted AGC will reduce or eliminate these
artifacts.<br><br>
AUDIO SIGNAL TO NOISE RATIO (SNR) is another important measure of voice
quality. SNR is the ratio of the audio signal power to the noise power.
Noise, which is undesired audio that was not present at the transmitter’s
microphone, may come from a number of sources. These sources include the
radio frequency energy in the path from the transmitter to the receiver,
and also include noise from the electronics in the transmitter and the
receiver. Figures of merit for SNR in body wires are given in the table
below.<br><br>
<br><br>
Table 6-5. Figure of Merit for Audio Signal to Noise Ratio <br>
-<br>
BIT ERROR RATE (BER) is another specification that affects audio quality.
This specification applies to digital and digital spread spectrum
systems, but not to narrowband FM systems. Bit error rate is the number
of bits in the digital stream that have been received with the wrong
value, compared to the total number of bits received.<br><br>
Each bit in the digital (or digital spread spectrum) body wire received
data stream has a value of 1 or 0. These bits taken together in preset
groups (usually 8, 16 or 32 bits) form the ‘words’ which correspond to
the digital representation of the audio waveform being transmitted. If
one of these bits is somehow assigned the wrong value, the sound from the
receiver will be distorted. The more bits that are assigned the wrong
value, the worse the resulting audio.<br><br>
Typically, bit error rate is expressed as the frequency of a single
erroneous bit. For example, a bit error rate of 10-6 means that for every
1,000,000 bits sent, one of them will be received incorrectly, and the
audio will be distorted for that instant. The table below shows figures
of merit for bit error rates in body wire systems.<br><br>
<br>
Table 6-6. Figure of Merit for Bit Error Rate <br>
-<br>
Bit error rate has a relationship to receiver sensitivity. (Sensitivity
is the weakest received signal power that can be successfully received.)
In general, receiver sensitivity is influenced by a change in bit error
rate (and vice versa). A receiver with –100dBm sensitivity at a BER of
10-5 could also have a sensitivity of –103 dBm at a BER of 10-4. In the
latter specification, the audio is worse, but the apparent range is
better when compared to the former specification.<br><br>
<br>
6.2.2 Transmission Range<br>
-<br>
The ability to receive the body wire signal at relatively long distances
from the transmitter (agent) can be very useful in some law enforcement
applications. High ranges allow the receiver to be located further from
the agent, reducing the likelihood of physical discovery of the
operation. In addition, for mobile applications, high ranges reduce the
likelihood that an agent will move out of range of the receiver.<br><br>
However, high ranges imply high transmit power. And high transmit power
would reduce the electronic security of the system (increase the
likelihood that wiretap detection equipment would see the signal coming
from the transmitter). For this reason, the operation should use transmit
power that is sufficient for the range of the operation, but not
excessive.<br><br>
Range evaluation is dependent on two manufacturer specifications:
Transmitter (TX) power, and Receiver (RX) sensitivity. Path loss
represents loss in signal power due to the transmission from body wore to
receiver, and is an environmental factor which determines range for a
given TX power and RX sensitivity. Path loss can not be specified by the
manufacturer and needs to be accounted for by the user. There are many
environmental factors that will increase path loss over a specific
distance. Urban environments, with buildings and crowds of people may
experience much greater path loss than in open terrain. From
specifications of TX output power and RX sensitivity, the maximum path
loss for received audio that can be accommodated by the system can be
calculated.<br><br>
In general, increasing transmitter output power will increase range for a
particular terrain and RX sensitivity specification. The power should be
expressed in mW or dBm (dBm = 10 x log base 10 of signal power in mW).
There is no figure of merit for transmit output power since each
operation will accommodate different equipment with different power
ratings.<br><br>
The table below presents some recommended transmit power values for
various law enforcement applications.<br><br>
<br><br>
Table 6-7. Transmit Power Uses <br>
-<br>
The receiver sensitivity defines the lowest received power level of the
transmitted signal that can be detected by the receiver at its antenna. A
signal received at this level should provide audio output at the
receiver. Receiver sensitivity should be quoted for a specific SNR.
Sensitivity is usually given in dBm. Some radio frequency (FM) systems
use a receiver sensitivity notation of microvolts (uV) for a specific
SINAD (signal to noise plus distortion).<br><br>
Typically specifications will be lower (better) for narrow bandwidth
signals such as narrowband FM and higher for wider bandwidth signals such
as digital spread spectrum. Note that sensitivity is stated with negative
numbers since the power is less than 1 mW.<br><br>
<br><br>
Table 6-8. Receiver Sensitivity <br>
-<br>
In addition to transmit power and receiver sensitivity, path loss
determines the range of the body wire system. Path loss is very dependent
on physical conditions present in the locale of the transmitter and
receiver and on the carrier frequency. Building structures, the number of
people between transmitter and receiver, interfering vehicles, metal wall
studs, etc. (the local operating conditions) will server to attenuate the
transmitted signal by varying amounts. It is not uncommon to see a
requirement of a 8-fold increase in required power to double the range.
Engineering studies have shown that in some cases, the attenuation at
ground level is so great that output power must be increased 16 times in
order to double the range. The expression often stated of 4 times the
power to double the range is mainly applicable for line of sight
conditions. Terrain is very important when looking at power outputs of
different systems when trying to determine whether the equipment will
meet range expectations.<br><br>
A related concept to path loss is multipath. Multipath effects are due to
portions of the signal arriving at the receiver at different times from
the main signal, caused by reflections within the environment. These
multiple signals arriving at the receiver may be highly disruptive to
communications.<br><br>
6.2.3 Battery Lifetime<br>
-<br>
Battery related specifications of body wires include maximum and minimum
operating DC voltage and current drain. These specifications have a
direct bearing on the type of battery most suitable for the
equipment.<br><br>
The current drain specification determines the amount of current required
to run the equipment. The lower the current drain, the longer the
equipment will run on a battery. Since many battery manufacturers list
the battery capacity in mAh (milli ampere hours), it is quite easy to
determine how long the equipment will run on a given battery. The current
drain of the body wire equipment should be provided in the
specifications.<br><br>
The minimum operating voltage of a body wire is the minimum voltage that
the battery must supply for the equipment to function properly.
Generally, if the battery falls below this level, the equipment will
cease operating. There may also be a maximum (not to exceed) voltage.
Voltages beyond this figure will probably damage the equipment. The
minimum operating voltage is often not given, but it is important in
determining battery requirements.<br><br>
In general, the less current consumed by the body wire, the longer the
batteries will last. The lower the minimum operating voltage, the fewer
batteries that are needed. The wider the operating voltage range (maximum
voltage – minimum voltage), the longer the system will operate on a given
battery pack.<br><br>
NOTE: Always use new batteries. If the batteries have been taken out of
their storage package, don’t use them for field operation.<br><br>
6.2.4 Physical Disguise<br>
-<br>
Two physical features are important for body wires: package size and
antenna type. Dimensions should be given for length, width, and
thickness. For body wire usage, the transmitter must be as thin as
possible. It is also advantageous to remote the antenna from the
transmitter. Being able to move the antenna away from the transmitter
allows a greater choice of concealment options.<br><br>
<br>
Table 6-9. Figure of Merit for Size (Thickness) <br>
-<br><br>
6.2.5 Electronic Security<br>
-<br>
Narrowband scanners can easily detect Narrowband FM (NBFM) systems, since
most detectors are of the narrowband sweep type. The fact that the signal
is digitized is significant for electronic security, since it reduces the
probability of interception. Digital spread spectrum systems are most
secure.<br><br>
Detection is defined as the ability of an outside person to discover the
presence of the body wire signal. Detection can be accomplished with
frequency counters, spectrum analyzers, or scanning receivers.<br><br>
Interception is defined as the ability of an outside person to acquire
the body wire signal and obtain understandable audio. A tunable receiver
is necessary for this purpose. Spread spectrum systems have very good
immunity to interception, since it is necessary for the outside person to
know the PN spreading sequence used in order to successfully receive the
audio signal. If the PN sequence is kept confidential, randomly selected,
and is long enough, it will be very difficult for an outside person to
obtain understandable audio from the signal.<br><br>
<br>
-7. CURRENT AND PROJECTED STATE OF THE ART BODY WIRE TECHNOLOGY
AREAS<br>
-<br><br>
-7.1 Block Diagrams of Typical Body Wire Transmitters<br>
-<br>
This section will discuss the block diagrams of the digital, digital
spread spectrum, and narrowband FM body wires. The block diagrams are
intended to give the reader a general idea of the types of components
used in body wires. Later sections will then briefly describe the state
of the art for some of the key components used in body wires.<br><br>
Figure 7-1 shows a simplified block diagram of a typical digital body
wire transmitter. Starting at the top left of the figure, the audio
signal is received by the microphone. Next, an amplifier/filter increases
the voltage of the signal from the microphone to the correct level for
the input to the analog to digital converter (ADC). Some filtering
(removal) of unwanted signals may also occur in this block. The ADC
converts the analog input signal to a digital word. The digital word
output of the ADC is fed into a coding block, which adds bits to the ADC
words that will serve to detect and correct errors at the receiving end
of the body wire link. After these error detection and correction bits
are added, the resulting bit stream is differentially encoded for
differential phase shift key (DPSK) modulation. The resulting bit stream
is fed to a binary phase shift keyed (BPSK) digital modulator. (Other
digital modulation methods could also be used, but this particular block
diagram uses BPSK). The BPSK digital modulator changes the phase of the
radio frequency (RF) carrier (from the RF synthesizer), according to
whether a 0 or 1 bit is input. The resulting phase modulated carrier then
goes to the power amplifier block. The power amplifier block increases
the power of the modulated RF carrier to a level that is suitable for
transmission. The power amplifier output goes to an antenna matching
network, which assures that the power of the amplified phase modulated RF
signal from the power amplifier is efficiently transferred to the
antenna. The matching network feeds the antenna, which sends the signal
through the air to the receiver.<br><br>
Batteries and voltage regulators provide power for the body wire. The
battery voltage is applied to voltage regulators, which provide the
voltages needed by various components in the body wire.<br><br>
<br>
-<br><br>
Figure 7-1. Digital body wire block diagram. <br>
-<br>
Figure 7-2 shows a block diagram of a digital spread spectrum body wire.
The diagram is very similar to the diagram for the digital body wire,
except for the addition of a PN sequence generator and an exclusive or
(XOR) block. The PN sequence generator generates a random sequence of
bits (chips) that is fed to the XOR block. The XOR block combines the PN
sequence with the data bits from the differential encoding block. The
resulting bit stream is fed to the digital modulator (a BPSK modulator is
shown).<br><br>
Adding the PN bit sequence increases the bit rate going into the BPSK
modulator, and thereby increases the bandwidth of the modulated signal
around the RF carrier (the bandwidth of a phase shift keyed signal is
approximately twice the bit rate). This increased bandwidth can make a
spread spectrum signal more difficult to detect, since the energy of the
signal is spread out over a wider band of frequencies.<br><br>
-<br><br>
Figure 7-2. Direct Sequence Spread Spectrum body wire block diagram.
<br>
-<br>
Figure 7-3 shows a block diagram of a typical narrowband FM body wire
transmitter. The main difference in this system and the digital body wire
is that analog modulation is used instead of digital modulation. The
analog signals from the microphone’s amplifier/filter are fed to the FM
input on the frequency synthesizer, which frequency modulates the
carrier. The output of the frequency synthesizer goes to a power amp. The
power amplifier output goes to a matching network, and the matching
network feeds the antenna.<br><br>
The frequency synthesizer consists of several components: a voltage
controlled oscillator (VCO), a divide by N counter, a phase/frequency
detector, and a loop filter. A temperature compensated crystal oscillator
(TCXO) serves as a frequency reference for the frequency synthesizer. The
frequency synthesizer is used in the narrowband FM, the digital, and the
digital spread spectrum body wires.<br><br>
-<br><br>
Figure 7-3. Narrowband FM body wire block diagram. <br>
-<br>
Please note that many variations on all three of the above block diagrams
are possible. The intent is to give typical block diagrams, and to give
the reader a general idea of the types of components used in body
wires.<br><br>
7.2 Microphones<br><br>
-(See the discussion on microphones in the solid state audio recorder
section)<br><br>
7.3 Analog to Digital Converters<br><br>
-(See the discussion on delta sigma analog to digital converters in the
solid state audio recorder section)<br><br>
-7.4 Frequency Synthesizer<br>
-<br>
The frequency synthesizer provides the RF carrier used by the
transmitter. A typical frequency synthesizer consists of several
components: a temperature compensated crystal oscillator (TCXO), a
phase-frequency detector, a loop filter, a voltage controlled oscillator
(VCO), and a divide by N circuit (prescaler). The phase-frequency
detector compares the phase of the divided VCO with the phase of the
TCXO. The loop filter filters the output of the phase frequency detector,
and the output of the loop filter is applied to the voltage control input
of the VCO. If the phase of the divided VCO is different than the phase
of the TCXO, the voltage applied to the VCO will change the VCO frequency
until the phase of the TCXO is aligned with the phase of the divided VCO.
In this manner, the frequency of the VCO is controlled so that it is N
times the frequency of the TCXO. By adjusting the value of N, it is
possible to generate different frequencies.<br><br>
Low cost, low power integrated circuits are available to perform one or
more of the functions needed by the frequency synthesizer. For example,
the National Semiconductor LMX2346 provides the phase-frequency detector
and divide by N functions. It consumes 6mA of current and costs $2.05 in
quantities of 1000. It is available in small surface mount packages
including a 0.25” x 0.2” package and an even smaller chip scale package=
.
A very good TCXO part is the ECS 39SM series made by ECS, Inc. This part
consumes 1.5mA at 3.3V and has a very good frequency accuracy of 1.5 ppm.
It is available in a surface mount package that is 0.45” x 0.38”, and
costs $7.70 in quantities of 1000. An example of a VCO that will cover
the the 915MHz ISM band (for a direct conversion transmitter) is the
Maxim 2623. This part consumes 9mA at 3.3V, is available in a 3.0mm x
4.9mm (uMax) package, and sells for $1.80 in quantities of 1000. The loop
filter may be constructed from passive components (resistors and
capacitors) for only a few cents in cost and with very small
packages.<br><br>
Future frequency synthesizer components should benefit from the overall
trend in electronics toward smaller die geometries. Smaller geometries
may be operated with lower supply voltages, resulting in lower power
consumption. Smaller geometries can also lead to smaller IC sizes or
greater part densities. The use of synthesizer components in high volume
cell phone and cordless phone markets should ensure the continued
development and availability of low cost, power efficient, small size
frequency synthesizer components.<br><br>
<br>
-7.5 Power Amplifiers<br>
-<br>
In most body wire designs, the power amplifier consumes more power than
any other single element. For that reason, operating times for the body
wire are dictated largely by the power needed by the power
amplifier.<br><br>
One of the biggest decisions in selection of a power amplifier is
dictated by whether or not a constant envelope waveform type of
modulation is used. If a constant envelope modulation is used, then one
of the more power efficient amplifier classes may be used (Class B, AB,
or even C). In contrast, if a modulation technique is selected that is
not constant envelope, then a less power efficient amplifier class (Class
A) must be used.<br><br>
One problem with using the more power efficient amplifiers is they do not
operate in the linear region, and are subject to spectral regrowth. A
current area of research in power amplifiers for wireless applications is
how to improve the spectral regrowth problems in power efficient
amplifiers. Several techniques to prevent spectral regrowth and preserve
efficiency are being investigated. One such technique is to adaptively
bias the power amplifier so it operates in the most efficient region of
class A operation all the time. Another technique is to predistort the
waveform, so that it is relatively undistorted after it is amplified by a
Class C amplifier.<br><br>
An example of a state of the art, low cost amplifier for constant
envelope waveforms in the 900MHz ISM band is the Maxim MAX2235. It
features a +30dBm (1W) power output, 47 percent efficiency, has a
footprint of 6.4 x 6.5mm, and sells for $2.07 in quantities of
1000.<br><br>
The continued growth in wireless commercial applications (such as cell
phones, wireless phones, PCS, Bluetooth, HomeRF, and wireless LANs) is
expected to spur future development of more efficient power
amplifiers.<br><br>
<br>
8. PROJECTED COMMERCIAL BODY WIRE PRODUCTS<br>
(IN TWO YEARS)<br>
-<br>
It can be anticipated that body wire equipment capabilities will undergo
a steady change in the next two years. The forecast is that digital
technology will be assume as more prominent role in wireless operations.
The benefits of security and audio quality will become more in demand.
The challenge to the designers and manufacturers is to bring digital
equipment into the same range performance standard as analog and still
keep costs down. Digital, because of its additional complexity, is
inherently more expensive than analog. Unfortunately, digital operation
will be faced with a range penalty and the way to increase range is to
engineer more efficiency and range enhancements into the digital
equipment. This extra engineering comes with an added cost burden.
Operational requirements demand the best quality audio for evidentiary
and investigative purposes, which is of course directly in the province
of digital technology. Better education and training is the way around
the seemingly contradictory desires of performance and cost. Properly
trained personnel will understand the benefits and drawbacks of digital
equipment and will be able to get the results needed in the difficult
investigations involving foreign translations and poor audio
environments. Well trained technicians will be able to get good audio at
the ranges required, making the change-over from analog to digital much
less painful.<br><br>
-9. BODY WIRE CONCLUSIONS<br>
-<br>
Body wire equipment is currently available in a wide variety of size,
technology and operating lifetime. The old adage ”You get what you pay
for” is ever applicable. Good analog equipment is not cheap, neither are
good digital transmitters and receivers. Cheap equipment manifests itself
in shoddy performance and poor reliability, in spite of claims of high
quality . Written specifications can give a distorted picture to the
unknowing. The best advice one can get is to become knowledgeable about
the meaning of equipment specifications and their operational impact. One
must evaluate equipment prospects as to performance and operational
tradeoffs – then consider cost. It has been the case that a law
enforcement user has said that a certain piece of equipment is the only
thing that can be afforded, only to discover that the product is quite
useless. The small savings in equipment cost can cause a much larger loss
of funds when the entire case is destroyed because the jury could not
understand the spoken word or the translator could not properly
comprehend the idiomatic speech.<br><br>
<br><br>
-<br>
APPENDIX A<br><br>
SOLID STATE RECORDER PRODUCT MATRIX<br>
-<br>
The following spreadsheet contains a complete listing of the data
collected on the various types of recorders investigated. Many of the
desired specifications were not available or were considered too
proprietary to release, particularly related to details regarding the
type of encoding used for the voice recording.<br><br>
[snip, see original pdf files on
http://www.tscm.com/ for these tables
and graphics]<br><br>
APPENDIX B<br><br>
SOLID STATE RECORDER COMPONENTS-<br><br>
The following spreadsheet shows the major components used to make a flash
based audio recorder. The recorder features a dynamic range approaching
90dB, a bandwidth of 16kHz, and has over 120 minutes of record time. The
size of the recorder is dictated largely by the size of the removable
compact flash, and by the size of the batteries.<br><br>
Table B-1. Solid State Recorder Spreadsheet <br>
-<br>
APPENDIX C<br><br>
BODY WIRE PRODUCT SOURCE MATRIX<br><br>
<br>
-<br><br>
<br><br>
<br><br>
APPENDIX D<br><br>
SOLID STATE RECORDER AND BODY WIRE SURVEY RESULTS<br><br>
The following tables contain the results of a survey submitted to a cross
section of US law enforcement agencies. The populations served by these
agencies range from less than 1,000 to over 1,000,000. The locations of
these agencies are all across the continental United States. 77 responses
were received from the survey. The survey asked a number of questions
regarding the typical use of recorders and body wires for law enforcement
applications. The answers to these questions are summarized in the tables
below.<br><br>
<br>
[snip, see original pdf files on
http://www.tscm.com/ for these tables
and graphics]<br><br>
<br><br>
<br>
<x-sigsep><p></x-sigsep>
<font size=2 color="#FF0000"><i>We Hunt Spies, We Stop Espionage, We Ki=
ll
Bugs, and We Plug Leaks.<br><br>
</i></font><b>James M. Atkinson, President and Sr. Engineer<br>
Granite Island Group<br>
</b>127 Eastern Avenue #291<br>
Gloucester, MA 01930-8008<br>
Phone: (978) 546-3803<br>
Fax: (978) 546-9467<br>
Web: <a href="
http://www.tscm.com/">
http://www.tscm.com/</a><br>
E-Mail: <a href="mailto:jm..._at_tscm.com"><i>jm..._at_tscm.com<br><br>
</a></i></body>
</html>
--===========_256428750=.ALT--
Received on Sat Mar 02 2024 - 00:57:28 CST