Copyright 2000 Federal News Service, Inc.
Federal News Service
February 17, 2000, Thursday
SECTION: PREPARED TESTIMONY
LENGTH: 3830 words
HEADLINE:
PREPARED TESTIMONY OF DR. CHARLES JACKSON ON BEHALF OF THE NATIONAL ASSOCIATION
OF BROADCASTERS
BEFORE THE HOUSE COMMERCE
COMMITTEE TELECOMMUNICATIONS & FINANCE SUBCOMMITTEE
SUBJECT -
COMMERCE ON SPECTRUM INTEGRITY AND HR 3439, THE "RADIO BROADCASTING PRESERVATION
ACT OF 1999".
BODY:
Summary
Mr.
Chairman, members of the subcommittee, I am Charles Jackson. I testify here
today on behalf of the National Association of Broadcasters (NAB). I am an
engineer with an independent consulting practice. I have experience at the FCC
and have worked for more than 30 years in the electronics and communications
industry--including 4 years on the staff of this subcommittee's predecessor. I
earned my PhD in electrical engineering at MIT.
The FCC tested the
ability of consumer receivers to withstand interfering signals on adjacent radio
channels. The results of those tests were not reported properly. The FCC used an
incorrect criterion (distortion) for measuring the effects of interference and
thereby provided misleading information in their order regarding the
interference potential of LPFM stations. The FCC measured interference but
reported the results as if it had measured harmonic distortion. The FCC claims
that consumers could not hear interference measured as 1% distortion and that
interference measured as 3% distortion would not be objectionable. This is
wrong. Interference generates noise and cross-talk, not distortion. Noise and
cross-talk are far more objectionable to listeners than is distortion. The
engineering literature, FM receiver specifications, and materials from the
manufacturer of the test equipment used by the FCC support my view.
My
testimony contains demonstrations of these two impairments that allow the
listener to compare the effects of adding harmonic distortion with the effects
of adding cross-talk from interference. I note that the FCC stated in its LPFM
order that the OET tests did not use the 3% distortion level as the measure of
harmful interference. This statement is contradicted by the text of the FCC
Office of Engineering and Technology Report.
To summarize, the FCC did
not use the right criterion when assessing the performance of FM receivers in
the presence of interference. In particular, they used a measurement method that
indicated no harmful interference where in fact, harmful interference would
occur. This use of the wrong criterion has led to justification for the
authorization of LPFM stations that will result in objectionable interference to
existing radio stations-interference that the FCC does not acknowledge because
it has not used the relevant measurement tool.
*********************
Prepared Statement of Dr. Charles L. Jackson
Mr. Chairman,
members of the subcommittee, I am Charles Jackson. I testify here today on
behalf of the National Association of Broadcasters (NAB). I am an engineer with
an independent consulting practice. I have experience at the FCC and have worked
for more than 30 years in the electronics and communications industry--including
4 years on the staff of this subcommittee's predecessor. I earned my PhD in
electrical engineering at MIT.
My message is short. The FCC, as part of
its Low-Power FM (LPFM) rulemaking, tested the ability of
consumer receivers to withstand interfering signals on adjacent radio channels.
Those tests were not reported properly. The FCC used an incorrect criterion for
measuring the effects of interference and thereby provided misleading
information in their order regarding the interference potential of LPFM
stations. The fundamental problem is that the FCC measured interference but
reported the results as if it had measured harmonic distortion. Such distortion
is much harder to hear than is noise or cross-talk.
Overview Below I
establish the error in the FCC's criterion through references to the engineering
literature, FM receiver specifications, and materials from the manufacturer of
the test equipment used by the FCC. First, I describe the two measurement
criteria at issue: harmonic distortion and cross-talk. Second, I play audio
signals that meet the FCC's definition of"adequate" quality for consumers.1
After hearing these you can decide for yourself whether or not the FCC is
correct in its judgment of what is adequate.2
Measuring Audio
Impairments--Distortion versus Noise and Cross-talk
One technical
complexity intrudes. The FCC measured performance of FM receivers using a
criterion called distortion or total harmonic distortion plus noise (THD+N).3 In
contrast, the NAB recommended using a measure called signal-to-noise ratio (SNR)
as the measure of FM receiver performance. The process of measuring these two
quantities is quite similar, although the units that are used normally differ.
However, THD is normally used to measure a quantity called harmonic distortion
or nonlinearity. It is well known that listeners find it hard to notice harmonic
distortion at levels as high as 2 or 3%. In contrast, many people can hear noise
or cross-talk when it is at the level that would measure as 1% distortion (if
one were improperly measuring noise or cross-talk as distortion).
I have
attached an appendix to this testimony that goes into more detail on the
differences between measuring SNR and THD and how the choice of measurement
units can be misleading.
Examples
Let me now give you a chance
to listen to the difference between cross-talk and harmonic distortion. I will
play an audio selection with no added distortion. I will then allow you to
compare the effects of adding harmonic distortion and the effects of adding
cross-talk.
Here is a brief audio sample--one familiar to many--taken
from Bernstein's West Side Story.4 This selection is taken from track 11 on the
CD. That track has a wide dynamic
range--running up to within 1 dB of
full scale but also containing some quiet passages. This specific selection runs
to within 3 dB of full scale.
Here is that same sample, but now
transformed to pure harmonic distortion--all tones have been shifted up one
octave using signal processing software. Here is the original, but with the
distorted version added back in at the 3% level? As you probably notice, the
added distorted element is almost impossible to hear.
In contrast, here
is that same sample with cross-talk added just below the 3% level.6 The
cross-talk signal was taken from another recording.
The FCC treats these
two quite different forms of impairment as if they are the same. But, as you can
hear they are not. A central flaw in the FCC's analysis was the treatment of
added interfering signals (cross-talk) as if they were harmonic distortion.
The FCC's tests judged a signal as acceptable if interference
increased the measured distortion by no more than 3%. Consumer receivers have
distortion as high as 3.5%. Thus, the FCC's procedures would accept signals with
cross-talk just below the 6.5% level. Here is the selection from West Side Story
with cross-talk added at a level that would drive the total of cross-talk and
distortion of a signal with 3.5% distortion to just below the 6.50/0 level.
Examples from Over-the-Air Broadcasts
Here is a cut recorded
from WGMS, the number one classical music station in the Washington, DC, market.
Here it is with cross-talk just below the 6.5% level that the FCC would
judge unacceptable if the consumer were using a receiver with 3.5% audio
distortion.
Here is a cut from WHUR, the leading station in the DC
market, with cross-talk just below the FCC limit of 6.5% total measured
distortion for a radio with 3.5% audio distortion.7
The
Pickholtz/Jackson Study
Professor Ray Pickholtz of George Washington
University and ! reviewed four studies of FM receiver performance that were
before the FCC in its LPFM proceeding.8 We concluded that the tests performed by
the various parties were similar; the differences in the conclusions of the
studies reflected differences in the definition of harmful interference used in
each study. We believe that the FCC made a mistake when they reported results in
terms of distortion but they were actually measuring noise and
cross-talk--signal impairments that are much more objectionable to listeners
than is harmonic distortion. This is a serious error--roughly as bad as telling
someone to suit up for a football game in a basketball uniform.
Evidence
from Others that Interference and Distortion Are Different
The FCC's
tests used a criterion, distortion, that is appropriate for measuring how good
amplifiers perform but is not a good measure of the presence of objectionable
cross-talk or of static. FM receiver manufacturers specify both distortion
(measured in percentage just like the FCC did) and S/N ratio (SNR). For example,
Sony provides the following specifications for their STR-DE835 (the top-rated
digital receiver in the March 2000 issue of Consumer Reports).9
- FM
Frequency Response 30 -- 15 kHz, +0.5/-2 dB
- FM THD @ 1 kHz,
Mono/Stereo 0.30%/0.50% FM SIN Ratio, Mono/Stereo 76 dB/70 dB
Sony
specifies that the S/N ratio, the measure of how well the receiver pulls the
desired signal out of the natural static, is 70 dB. In contrast, Sony specifies
that the receiver's THD, a measure of how well the output stage of the receiver
reproduces signals, is 0.50%. Distortion signals at the 0.50% level correspond
to signals only 46 dB below the desired signal. If distortion and noise were
different names for the same phenomenon, then the receiver would have a 46 dB
SNR. Similarly, if distortion of 0.50% prevents one from hearing noise at levels
much below 46 dB below the desired signal, Sony is wasting its efforts in
delivering a 70 dB SNR.
Similarly, a technical paper available from
Audio Precision, the manufacturer of the test equipment the FCC used in its
tests, states, "Harmonic distortion, illustrated in Fig. 16 is probably the
oldest and most universally accepted method of measuring linearity (Cabot
1992)."10 It is well known that linearity--the degree to which an amplifier's
outputs are just bigger versions of the input signals-- measures accurately an
amplifier's performance and that small deviations from linearity are hard to
hear.
Audio Precision also says,
Most audio Total Harmonic
Distortion (THD) measurement systems are in fact Total Harmonic Distortion plus
Noise (THD+N) analyzers. They operate by removing the fundamental from the test
signal with a sharply tuned band reject or "notch" filter and measuring
everything that remains. The amplitude of this "residual" is compared to the
amplitude of the fundamental and the result is expressed as a percentage or dB
figure. This measurement technique does not discriminate between test signal
related harmonics caused by non- linearity in the device under test, broadband
noise in the device under test, crosstalk or interference from external sources,
or any other artifacts present within the measurement bandwidth. The "single
number" result may thus be ambiguous.
II The FCC's Use of the 3%
Standard
The FCC stated in its LPFM order that the OET tests did not use
the 3% distortion level as the measure of harmful interference. The FCC
specifically stated, The above conclusions of the OET report that "nearly all
the receivers in the sample appear to meet or exceed the 40 dB 2nd -adjacent
channel criterion and exceed the 3rd-adjacent channel protection criterion by a
substantial margin" reflect measurements taken at the 1% distortion level.12
This statement conflicts with the text of the OET study. It reads,
Section 73.215 of the Commission's rules provides that the predicted
field strength of a potentially interfering station can be no more than 40 dB
stronger than the protected field strength along a station's protected contour.
At the 3% distortion level all the receivers in the sample, except for two
(samples #2 and #6), appear to meet or exceed the 40 dB second adjacent channel
protection criterion and to exceed the 40 dB third adjacent channel protection
criterion by a substantial margin. 13
The text in the order also
reflects a basic confusion between distortion and other forms of signal
impairments when the FCC states, "The 1% level corresponds to a point at which
most listeners would not be able to perceive any degradation in performance. On
the other hand, the 3% distortion represents a level at which most listeners
would perceive a difference in the received signal."14 This statement is almost
a textbook discussion of the effects of harmonic distortion-- but does not apply
to noise and cross-talk. The FCC claims that a person would find it impossible
or hard to hear the effects of interference that were measured as 1% or 3%
distortion. This is incorrect. It is hard to hear 3% distortion; it is easy to
hear cross- talk at the 3% power level as I just showed you.15
The FCC
holds the entities it regulates to high standards of truthfulness (called candor
in the Commission's jargon) in their statements to the Commission. It should
hold to those same standards when it speaks to the public.
Conclusion
To summarize, the FCC used the wrong criterion when assessing the
performance of FM receivers in the presence of interference. In particular, they
used a measurement method that indicated no harmful interference where in fact,
harmful interference would occur. This use of the wrong criterion has led to
justification for the authorization of LPFM stations that will result in
objectionable interference to existing radio stations-interference that the FCC
does not acknowledge because it has not used the relevant measurement tool.
Appendix: Measuring Audio Impairments
First, subjective
testing--the use of a panel of listeners to compare and grade the performance of
alternative systems--is the gold standard of audio system evaluation. 16 Second,
although they may be the gold standard, subjective listening tests are, like
gold, very expensive-- requiring significant time and staff. Consequently, other
objective test methods have been developed. These objective measurements may or
may not be monotonically related to subjective quality, but they are close
enough for many applications. A primary measurement used to assess the
performance of analog broadcasting and recording systems is the audio or output
signal-to-noise ratio (SNR). This ratio compares the energy in the desired
signal with the energy in the obscuring or impairing noise signal. Often the SNR
is calculated using a weighting procedure that attaches more weight to noise at
the most easily heard frequencies and less weight to noise at frequencies that
are less irritating. Informally speaking, SNR is a measure of the static that
has been added to a signal.
Table 1 below shows SNR for some familiar
audio systems.
In this table, a higher number is better and SNR
is reported in dB--a logarithmic measure that matches well with the human
hearing process. A difference of about 3 dB in SNR is usually regarded as the
smallest size difference a typical observer will notice. Thus, there is not much
difference in the typical subjective evaluation of the performance of two audio
systems-one operating with 40-dB SNR and the other with 43-dB SNR. However,
there is a big difference between a system operating with 40-dB SNR and one
operating with 60-dB SNR.
TABLE 1--Signal-to-Noise Ratio for some
Familiar Audio Systems
System Approximate SNR
Compact disc 100
dB Sony Walkman digital audio tape Better than 87 dB FM broadcasting (best
conditions) 60-80 dB Consumer audio taping equipment/17 60 dB Telephone call
30-50 dB
A second measure of audio system performance is harmonic
distortion. Harmonic distortion is most often used to measure the performance of
audio devices such as amplifiers or recording systems. It is a measure of how
accurately an audio system reproduces the input signal. Harmonic distortion is
often used to characterize the performance of amplifiers. It is caused by
nonlinearity in the amplification chain that creates frequency components that
are harmonics of the original frequencies (integer multiples of the original
frequencies, also called overtones). If the output signal from an amplifier is
the same as the input signal, except bigger, then there is no distortion. With
music or pure tones, distortion can be noticed by the presence of overtones. For
example, if a real-world amplifier has as input a 1,000-Hz tone, the output will
consist primarily of a 1,000-Hz tone, but tones at 2,000 and 3,000 Hz (and other
frequencies) will also be present in the amplifier output. These unintended
overtones produced by the amplifier are called harmonic distortion. It is hard
for the human ear to hear harmonic distortion.
The human ear's response
to a 2,000-Hz tone is reduced when a strong signal is also present at 1,000 Hz.
Similarly, people often think they hear a sound at 2,000 Hz when they only hear
a sound at 1,000 Hz.18 Most music sources, such as a piano or violin note,
contain overtones that are only slightly modified by the overtones created by
distortion.
Hence, given both the reaction of the human hearing system
and the content of most music, harmonic distortion is harder to hear than
unrelated noise.19 It is generally accepted that harmonic distortion has to rise
to about 1 to 2% before people find it objectionable.20 Some people would find
1% harmonic distortion hard to notice.21 The nonlinearities in the signal
processing chain that cause harmonic distortion also cause intermodulation
distortion that produces other, unintended frequency components. The usual test
procedures for audio equipment use the measure of total harmonic distortion plus
noise (THD+N) as shorthand for all nonlinear impairments. Although it may be
possible, albeit rare, for interference to drive the signal into the nonlinear
region and cause harmonic distortion, that is not usually the principal concern
when considering the effects of interference. Interference is best treated as a
different, extraneous source of additive noise. Thus, we measure its effects by
considering the signal-to-noise plus interference ratio (SNIR). The noise we
refer to here is due to thermal, environmental, or receiver noise that we cannot
overcome and is not the interference from like signals residing in a co- or
adjacent channel. The interference of concern here is external and produced by
other emissions in the radio spectrum by other than the desired transmitter. It
is what can be controlled by regulation. It is therefore our considered opinion
that the deleterious effects caused by this interference must be measured. Other
undesirable effects, inherent in the imperfections in the signal chain may also
be present, but they are a red herring when the objective is to determine
whether controllable external additional emissions such as second and third
adjacent channel interference should be permitted to degrade expected reception
quality.
FOOTNTOES:
1 The FCC stated, "The OET and NLG studies
generally conclude that FM receivers provide for adequate rejection of
interference on 2nd- and 3rd-adjacent channels.", LPFM Order at paragraph 100.
The OET test report makes it clear that the criterion for adequacy is
performance with less than 3% added distortion.
2 The MS Word version of
this document has embedded audio objects that contain the various
demonstrations. Obviously, the printed copy cannot contain these audio objects
and the MS Word file with the objects is too large for some email systems. If
you wish a copy of the Word document with the embedded audio, it should be
available at www.jacksons.nct/H SC until at least March 1,2001.
3 The
FCC Order and the OET report consistently refer to distortion. The Audio
Precision System One manuals refer to THD+N. See. for example, System One
Description/Installation/APWIN Version 22 Guide. p. 2-6.
4 Deutche
Grammophon, 415 254-2, recorded live.
5 That is, the harmonic distortion
is reduced in volume to 31 dB below the original signal and added back in.
Because the distortion does not decline as the amplitude of the signal falls,
this process results in more distortion than would occur in a typical amplifier
with 3% measured distortion.
6 Here the 3% less a little bit level is
set to 31 dB below a full- scale sine wave as would be the case when measuring
FM receivers with a single tone as the desired signal. That is, the cross-talk
is set at the level that would measure just below 3% in the FCC's test of FM
receivers.
7 These examples, and more, together with an explanation of
the test set-up and parameters are available at www.nab.org.
8 This
study was performed on behalf of the NAB. It is available from my website at
www..jacksons.net.
9 Taken from http://www.sel..sony.com SFI. consumer
ss5 homc homeaudio/recievers.strdc835_specs.shtnal on February, 13, 2000.
10 Fundamentals of Modern Audio Measurement," by Richard C. Cabot,
Presented at the 103rd Convention of the Audio Engineering Society, New York,
USA, 1997 September 26-29, revised 1999 August 8, p. 12. Emphasis added.
11 Audio Precision Tech Note TIN-17. available at
www.audioprecision.com. Emphasis added.
12 FCC LPFM Order, footnote 156.
13 Second and Third Adjacent Channel Interference Study of FM Broadcast
Receivers, Project TRB-99-3 Interim Report, July 19, 1999, p. 31.
14
Ibid.
15 This statement requires some qualification. It is hard to hear
low- order distortion, that is 2nd- or 3rd-harmonic distortion. It is much
easier to hear higher order distortion.16 It may seem strange to some that
engineers rank a subjective test as the highest performance standard. Despite
stereotypes, engineers actually have normal endowments of common sense and they
recognize that the proper measure of a system designed to serve consumers is the
consumer reaction to that system.
17 For example, the Sony TC-KE500S.
18 See, for example, A. Gersho, "Advances in speech and audio
compression," Proceedings of The IEEE, vol. 82, pp. 900-918, June 1994. P. Noll,
"Wideband speech and audio coding," IEEE Communication Magazine, vol. 26, pp.
34-44, November 1993. J. J. N. Jayant and Y.
Shoham, "Coding of wideband
speech," Speech Communication, vol. 11, pp. 127-138, 1992.
19 It is
easier to hear someone cough at an orchestra concert than to tell that one of
the violinists is playing an octave high. Indeed, everybody in the audience can
hear the person coughing, but only audience members with unusual musical acuity
will notice that one violin is an octave high.
20 See H.F. Olson,
Elements of Acoustical Engineering, Van Nostrand, New York, 1947 as quoted in
Electronics Engineers' Handbook, 2nd Edition, Donald G. Fink and Donald
Christiansen, eds., McGraw-Hill, 1982, at p. 19-18.
21 While engineers
are good, they are not perfect. Engineers often use different units to measure
SNR and harmonic distortion. Although SNR is normally measured as a power ratio
and expressed in dB, harmonic distortion is often measured as a voltage ratio
and expressed in percent. This notational difference makes it harder for the
nonexpert to keep track of what is going on in the four studies we consider.
This confusion adds an unintended shell-game element to reading the engineering
studies in the FCC's LPFM rulemaking.
END
LOAD-DATE: February 19, 2000
