Presentedat AuDIO the 83rd Convention 1987October 16

A MUSICALLY APPROPRIATE
FOR POWER AMPLIFIERS
DYNAMIC HEADROOM
TEST
2504
(0-7)
Peter W. Mitchell
Mystic Valley Audio
Oceanside,
California
Presented at
the 83rd Convention
1987October 16-19
New York
AuDIO
®
Thispreprint has been reproduced from the author's advance
manuscript, without editing, corrections or considerationby
the Review Board. TheAES takesno responsibilityfor the
contents.
Additional preprints may be obtained by sending request
and remittance to the Audio Engineering Society, 60 East
42nd Street, New York, New York 10165 USA.
Ali rightsreserved. Reproduction of thispreprint, or any
portion thereof, is not permitted without direct permission
from the Journal of the Audio EngineeringSociety.
AN AUDIO ENGINEERING SOCIETY PREPRINT
A MUSICALLYAPPROPRIATE
DYNAMIC HEADROOM TEST
FORPOWERAMPLIFIERS
Peter
W. Mitchell
Mystic Valley Audio
500 N. Clementine
St., #4
Oceanside
CA 92054
(619) 433-4358
ABSTRACT
The EIA R8-490 (former II-IF A-202) amplifier test standard includes a "dynamic
headroom" test employing a 20-mS tone-burst. In an informal survey of musical
recordings, power bursts were found with durations from a few milliseconds up to
several hundred milliseconds, with an apparent clustering in the 80-200-mS range.
Since the practical value of an amplifier depends on its ability to reproduce musical
dynamics, a more useful power rating would be obtained by amending the dynamic
headroom test to employ a 200-miUisecond (or similar) tone-burst.
0. INTRODUCTION
Moment-to-moment
variations in level are characteristic
of all musical sounds.
Even in its most compressed forms (e.g. rock music compressed for television or FM
broadcast), music has momentary peak levels that are several decibels higher than
the long-term average level.
The audio industry has long recognized the need for a power rating standard
that would recognize the desirable ability of an amplifier to produce higher power
levels in short-term bursts. The old 'qgIA peak power" and "IHF music power" (the
latter defined by the 1966 IHF amplifier _est standard) were widely mis-used,
leading to the imposition of a steady-state power-testing rule in 1970 by the Federal
Trade Commission.
In 1976 the Institute of High Fidelity developed a new amplifier test standard
(A-202) that met the continuous-power testing requirement of the FTC but supplemented it with a new "dynamic headroom" test of short-term power output. This
test standard has been widely accepted, and when the IHF was absorbed into the
EIA in 1981, the II-IF test document became an EIA standard, EIA RS-490 [1].
The dynamic headroom test employs a 1000 Hz tone-burst, gated on at halfsecond intervals, each burst lasting for 20 milliseconds . The signal level is increased until the output waveferm begins to clip; the output voltage is converted to
equivalent power and is expressed in decibels relative to the amplifier's rated continuous power.
Since the IHF amplifier test committee did not have a mandate to study musicai dynamics, the 20-millisecond duration of the tone-burst was selected somewhat
arbitrarily and is not related to the tone-burst lengths that occur in music. The 20mS figure was a compromise: brief enough to avoid causing power-supply voltages
to decline significantly from their quiescent values, yet long enough to plausibly
represent an "attack" transient in a musical sound--and long enough that clipping
of the tone-burst is audible. (Very brief tone-bursts, only 1 or 2 milliseconds long,
are perceived by the ear as clicks rather than tones, and distortion in them is hard
to identify.)
1. TRENDS IN AMPLIFIER DESIGN
In effect, the audio industry has two methods of measuring amplifier power: a primary standard in which the signal is always "on" at full level, and a secondary
standard in which the signal is on for only 1/25th of the time. Neither resembles the
dynamics of music.
The continuous-power rating, especially with the preconditioning (one hour of
operation at one-third power) mandated by the FTC, is basically a test of an amplifier's thermal capacity--its ability to pass a constantly high level of current through
its power supply and output stage without overheating. This, in turn, depends on
the size of the power transformer, the filter capacitors, the number and size of the
output transistors, and the capacity of the heat sinks. These components substantially determine the amplifier's manufacturing cost.
The II-IF dynamic headroom test, on the other hand, places virtually no strain
on the power supply. And since the signal is off (or at a 1% power level) most of the
time, the average power level is 1/25th of the measured dynamic power, producing
very little heat. The IHF 20-mS headroom test is simply a measure of the available
output voltage swing. In theory an amplifier designer could make this voltage arbi-_
trarily large, constrained only by the breakdown-voltage
ratings of the transistors. --_In practice, for amplifiers that employ a conventional power supply, the II-IF dynamic headroom test is only a measure of how loosely regulated the supply is--i.e.
how high the supply voltage floats when it is net being pulled down by a heavy current drain.
Thanks to the development of"commutating" amplifiers, in which the effective
rail voltage is actively varied from moment to moment to meet the requirements of
' _
the signal, manufacturers
are now able to tailor their products to match the dynamic power demands of music. A commutating amplifier is one that has power
supplies at two or more voltage levels, relying on the low-voltage supply most of the
time (to minimize thermal dissipation), and instantly switching over to the high_
'_' voltage supply when greater output is needed to reproduce high-level musical
peaks.
When successful, this approach permits an amplifier to produce large bursts of
power to cope with the uncompressed dynamics of digital recordings_without
the
high cost that would be incurred if the amplifier had to produce such high power
levels continuously. Soundcraftsmen,
Hitachi, Carver, Yamaha, NAD, Proton, and
Acoustic Research are among the manufacturers
that have produced amplifiers
based on this idea.
One benefit of this approach can be shown by measuring the dynamic headroom, not just with a 20~millisecond tone burst, but with a full range of tone-burst
durations. Figure I shows the resulting "dynamic power envelope" for a commutating amplifier (top) and for several conventional amplifiers (bottom), tested with load
impedances of 8 and 4 ohms. The actual steady-state output of the tested amplifiers
ranged from 70 to 500 watts/channel,
but in order to compare the shape of the dynamic power curves, the measurements
for each amplifier have been scaled to the
same steady-state clipping level (arbitrarily set at 100 watts).
While the conventional amplifiers measured here were chosen arbitrarily from
readily available models (manufactured by Hafier, Yamaha, and NAD), they do not
differ greatly in their power-supply regulation. The power envelope curves are quite
similar in shape, though they span a seven-to-one range in power output. Tn every
case the 20-mS IHF dynamic headroom is between 1 and 2 dB when expressed relative to the steady-state clipping level, and the headroom for a 200-mS burst length
is less than I dB. The commutating amplifier, however, produces 3 to 4 dB more
power for all burst lengths up to 0.5 second.
Figure 2 compares the dynamic power envelope ofa commutating amplifier (A)
with two hypothetical amplifiers (B and C) that employ
conventional single-stage
power supplies. Amplifier B has the same steady-state
power output (and costs
nearly the same to manufacture) as Amplifier A, but produces only half as much
short-term power. Amplifier C has the same burst power as Amplifier A, but since it
produces twice as much continuous power, it costs nearly twice as much to manufacture.
As these graphs show, commutating amplifiers that are currently being manufactured not only produce a higher level of 20-mS dynamic power than amplifiers
using a single power supply. They also produce elevated power levels for much
longer burst periods than the 20 milliseconds of the IHF dynamic power test--long
enough, not just for the "attack" transients at the beginnings of notes, but for entire
notes and chords.
In this case it can be argued that the amplifier's useful output for musical
waveforms is most accurately represented by the dynamic power, not the continuous power. In that case the dynamic power measurement
should be not merely a
secondary test but the primary measure of the amplifier's performance.
2. THE DYNAMIC
POWER ENVELOPE OF MUSIC
How long should the tone-bursts in the dynamic power measurement be? How long
should a commutating amplifier be able to produce elevated power? The key to
these questions is: How lo/_g are the tone-bursts in music?
To find out, an oscilloscope was used to monitor the dynamic envelopes of several types of music (vocal and instrumental,
popular and classical), using Compact
Disc recordings. Dozens of high-level bursts were photographed. Typical results are
shown in Figures 3 to 11. All of the photographs have the same scaling: in each case
a two-second segment of music is shown, and each horizontal division in the photograph is 200 milliseconds long. The photos were selected for variety, not statistical
value. The power scale is arbitrary but realistic, assuming a high but not uncomfortable volume level.
Two points are noteworthy about these musical envelopes, in addition to the
tone-burst lengths:
1. The tone.burst
shapes. A musical note is sometimes modeled as a strong
attack transient followed by a decay (especially when played by a guitar or piano).
The 20-mS duration of the IHF tone-burst makes sense if it is meant to represent
only the attack transient at the beginning of a note. But a leading-edge transient is
found only a few of the musical tone-bursts.
2. The duty cycle. For what percentage of the time is the amplifier at or near
full power? The selection of these photos biases this question, since they were all
taken during periods when the amp was being driven to high power, and do not include the long periods when the music was softer.
Figures 3 to 5 are examples of popular music. Figure 3 (a segment of No Reply
at All by Genesis) shows repeated bursts lasting 50 to 100 milliseconds; the power
reaches 70 to 150 watts during each burst and falls to the 20 watt level between
bursts. Figure 4 (a vocal passage from Paradise by the BeeGees) shows a 100W
burst more than a half-second long, as well as shorter bursts and higher transient
spikes. During the large burst there is a brief spike, less than 20-mS long, to the
200W level. Figure 5, an instrumental
section of the same recording, shows very
brief bursts not much different from the 20-mS IHF test signal, superimposed on an
average power level of only a few watts.
It is often assumed that rock music has a much smaller ratio of peak-to-average power than classical music. That may be true after heavy processing of the sighal by an FM station, but as these photographs show, CDs of rock music have considerable dynamic variety.
Turning to classical music, Figure 6 shows a two-second excerpt from the first
movement of Mahler's Symphony No. 2, with a burst that remains above i00 watts
for 300 milliseconds (and brief spikes up to the 200-watt level). Figure 7, one of
m_ny passages for tympani and brass from Bruckner's Symphony No. 4, exhibits a
series of 200-mS bursts, repeating at half-second intervals. Figure 8, from Also
Sprach Zarathustra by R. Strauss, shows a crescendo that stays above the 100-watt
level for a full second, finally peaking over 200 watts. Figure 9, from Mussorgsky's
Pictures at an Exhibition shows a low-level passage (under 15 watts) followed by a
powerful chord that is above 100 watts for about 1/2 second.
Figure 10 shows a solo piano recording (a Chopin Polonaise); there are short
transients, as expected, but also full chords lasting longer than 200 milliseconds.
Figure 11 shows a portion of a long, sustained chord from a Bach Toccata
played on the pipe organ, traditionally assumed to represent th_best justification
for a high "continuous" power rating. It does indeed show a sustained power level of
about 60 to 100 watts, remaining nearly constant throughout the two-second duration of the photograph. Interestingly, though, there are also brief spikes to the 200watt level. Even in music that sounds as if its power level would be uniform, an
amplifier benefits from dynamic headroom above its steady-state power level.
As these photographs show, what music needs (and, therefore, what an amplifier is called upon to deliver) is not a constantly high level of steady-state power but
bursts of dynamic power that vary in duration from the brief 20-mS transient of the
IHF headroom test up to several hundred milliseconds. This should not be surprislng; at a typical metronome setting of 75 beats per minute, the length of a quarternote of music is 200 milliseconds.
3. CONCLUSION
For listeners, the value of an amplifier depends on the level at which it begins to
distort the sound, and the only reason to purchase a larger and costlier amplifier is
its superior ability to reproduce the high-level peaks in music (especially the uncompressed peaks in Compact Discs). With commutating amplifiers becoming increasingly common, there is no longer any correlation between the musically useful
output of an amplifier and its continuous-power rating.
The audio industry should adopt a dynamic-power or burst-power measurement as its primary standard, in order to provide consumers with a rating system
that relates to the real ability of amplifiers to reproduce music without distortion.
The IHF/EIA dynamic headroom rating is a step in the right direction, but its 20mS burst length is typical of only the briefest transient sounds in music. A measurement of dynamic power using an extended-length tone burst (with an "on" time
of 200 to 300 milliseconds, corresponding to the typical duration of a musical note)
would be the most accurate way to rate the musically useful output of amplifiers.
Comments are invited from reviewers and manufacturers.
APPENDIX: THE DUTY CYCLE
When the "on" time of the tone burst is lengthened, should the "off' time between
bursts be increased in the same proportion? Since our goal is to define a test that
authentically
measures the musically useful power output of amplifiers, the duty
cycle of the test signal shculd be chosen to reflect the demands that real musical
signals make on an amplifier's power supply and thermal dissipation.
Studies have shown that the ratio of long-term average power to short-term
peak power in music ranges from less than 5% to nearly 14%. Even in the worstcase situation (rock music compressed for broadcast), the ratio of average to peak
power is around 15%.
The IHF test signal (20 mS on, 480 mS off) has a duty cycle of only four percent, superimposed on a constant signal at one percent power. This resembles very
impulsive music (see Figure 5, for instance). Simply scaling it up in time by a factor
often yields a tone burst that is on for 200 milliseconds and off(or at 1% power) for
4.8 seconds. With such a long time for the amplifier to cool and re-charge between
bursts, this may be too easy a test.
Another proposal, however--that
the duty cycle be increased to 50%, with the
"off' time between bursts equalling the "on" time--is unreasonably severe and does
not represent any realistic music situation. With this signal the long-term average
power level would be equal to 50% of the short-term burst power, which never occurs in music. (If it did, many domestic loudspeakers would quickly burn out.)
If a 4% duty cycle is too easy a test and a 50% duty cycle is unrealistically
severe, a 20% duty cycle may be a sensible compromise (i.e. 100-mS bursts repeating
at 500-mS intervals, 200-mS bursts at l-second intervals, or 500-mS bursts at 2.5second intervals). This allows filter capacitors to re-charge between bursts of maximum power (as real music does), while producing a dissipation requirement that is
just slightly more demanding than worst-case musical signals.
Alternatively, the standard II-IF test signal could be made more challenging
(and more accurately representative
of loud passages in mu_c) by raising the signal
level between bursts to -10 dB. Thus the signal would be at full power 4% of the
time and at 10% power during the remaining 96% of the time, producing a dissipation requirement similar to the worst-case situations found in music.
ACKNOWLEDGEMENT
I would like to thank Bjorn-Erik
ful discussions.
Edvardsen,
Director of Research at NAD, for help-
REFERENCE
[1]. Standard Methods of Measurement for Audio Amplifiers, EIA RS-490 (1981).
Electronic Industries Association, 2001 Eye St, N.W., Washington DC 20006.
6
Figure
1.
5
275
4
250
IHF DYNAMIC
/ti
TRANSIENTS
POWER FOR
DYNAMIC
HEADROOM
dB
225
/
200
300
"BURST" POWER
FOR MUSICAL
NOTES
20
/
50
/
100
20O
POWER,
WATTS
STEADY-STATE
POWER
FOR
CONTINUOUS
TEST TONES
500
t.0
MILUSECONDS
2.0
3.0
SECONDS
BURST LENGTH
Figure1.CompletePowerEnvelope.measured
wlihIoadlmpedancesof8and4 ohms.New amp#her(top hne)
conventional
amplifiers(bottomlines)
Figure 2.
5
- 300
dB 234-'--_/_1
' ____
0
20
I
50
I
100
MILLISECONDS
I
200
I
500
I
1
--
I
2
-250__-' 120050
WATTS
100
SECONDS
Figure2.PowerEnvelopeCurveofthreeamplifiers.
Amphliar'_"Iscommutating
des_jnwithabi-level
powersupply,Amplifiers
"B"&'_C"eachemployconvenlionat
single.stage
powersupplies.
DYNAMIC ENVELOPES
OF VARIOUS MUSICAL SIGNALS.
DYNAMIC ENVELOPES
OF VARIOUS MUSICAL SIGNALS.