by Joseph V. DeMarinis [Mgr., Audio Eng. GTE Sylvania, Batavia, New York]
SINCE November 1975, the Federal Trade Commission has required that audio amplifier power ratings be based on measurements made after one hour of operation at one third of the intended full-power rating, using a 1 kHz sine wave signal.
Fig. 1 Overall view of calorimeter used to measure power dissipation.
Fig. 2 Internal view of calorimeter.
This generates a much more severe design requirement than had previously been considered necessary and has raised concern among the audio engineering fraternity regarding the technical validity of the ruling. To determine if there is a basis of realism behind the new requirement, this writer felt it appropriate to examine the peak-to-average power ratio of a wide variety of program material and conduct other experiments to see if (or how) it might be possible to develop one third of an amplifier's maximum RMS output, for one hour.
A calorimeter was used to measure average power. It consisted of an 8-ohm resistive load and an expanded-scale thermometer immersed in 200 cc of distilled water. The container was made from styrofoam with a minimum wall thickness of one inch. The load was a five-watt wirewound resistor, with the ceramic outer case chipped off. This provided rapid heat transfer and minimum undesired thermal material. To minimize heat loss, fine steel wire was the only electrical and mechanical connection to the load. An inch-thick top cover, which held the thermometer and load resistor, remained fixed, while the water cup was rotated at 78 rpm, to assure uniform heat distribution in the water. Figures 1 and 2 show the calorimeter construction.
Calibration measurements were made at 28 watts and 2.5 watts, with errors of -4% and -3.6% respectively. Appropriate correction factors have been applied to the test results.
A Tektronix Model 549 storage oscilloscope was used to record peak information. The oscilloscope was connected across the 8-ohm load and was operated in the storage mode at a sweep speed of 1 cm/sec for the duration of each program selection. The highest peak voltages were clearly discernible. Since amplifiers must be rated in terms of rms power, the peak power information was converted to rms with the equation:
While this doesn't necessarily represent the true rms power of transient peaks, it does relate the required clipping level to a sine wave power capability.
Since amplifier clipping would defeat the purpose of the program material study, a large amplifier was used. The one chosen is able to deliver 156-watts continuous rms power and has a transient-peak clipping level of 196 watts, with an 8-ohm load.
Tests were run with an assortment of classical and rock records, with a few rock FM stations, and also with white noise. The results, expressed as a percent of Maximum rms Power, can be summarized as follows:
Average Classical Record 3.5%
Worst Case Classical Record 6.0%
Average Rock Record 4.3%
Worst Case Rock Record %7.2%
White Noise 9.6%
Average Rock FM Station 10.7%
Worst Case Rock FM Station 15.0%
Figure 3 shows the distribution of energy for recorded music and for FM rock programming. The detailed data is shown in Table I. The reason for the high power content of FM stations is, of course, their extensive use of volume compression. It is probably more than coincidence that the highest average power was observed on the station which used Dolby encoding.
Examination of the oscilloscope patterns in Figs. 4 and 5 clearly show the high energy content and flatness of FM rock programming as compared to a typical rock record.
It is quite evident from this analysis of 26 musical programs and two tests with noise that the most severe continuous demand on an amplifier was 15 percent of its maximum rms power rating and with most program material less than 10 percent, provided there is no significant amount of clipping in the output stages.
Fig. 3 Average material.
Fig. 4 Emerson, Lake & Palmer's Brain Salad Surgery, as displayed on
storage oscilloscope. Two minutes at 1 cm/sec. sweep speed.
Fig. 5 FM station, "Rock 102," WBEN, Buffalo, N.Y., as displayed
on storage oscilloscope. Two minutes at 1 cm/sec. sweep speed.
Fig. 6 Heat sink temperature vs. percentage of rated power for FM Rock
station program compared to sine wave dissipation.
Table 1--Program material vs. power requirements.
The obvious question is, "What happens if an amplifier is allowed to be driven into clipping?" To investigate this, the large amplifier was replaced by a small one, which could only deliver 6.5 watts continuous rms power and had a transient-peak clipping level of 9 watts.
Using the rock FM station, which had given the worst case results in the previous study, the volume control was turned up to the highest levels at which the distortion was considered tolerable by non-critical listeners. Under these conditions, average outputs on the order of 30 to 40 percent of the maximum-continuous rms power capability were achieved. At maximum volume, readings as high as 89 percent were obtained. However, the audio was distorted virtually beyond intelligibility and this could not possibly be considered normal use of the amplifier.
It is evident, therefore, that the only way to achieve one-third of maximum rms power on a continuous basis with program material is to drive the amplifier into severe clipping distortion. The result can not, by any stretch of imagination, be classified as hi-fi. However, with small amplifiers, such clipping is a normal fact of life.
It is well known that when an audio output transistor is driven into saturation (clipping), its internal power dissipation drops considerably. In other words, deriving one third of rated output power from intermittent program material, which saturates the output transistors a high percentage of the time, will cause much less output device heating than generating the same average output power with a continuous sine wave which stays below the clipping level.
To evaluate this effect with actual program material, thermocouples were attached to the collector tabs of the output transistors of the 6.5 watt amplifier. The amplifier was then driven to various power output levels with a 1,000 Hz sine wave and then driven to several different average output levels with rock FM program material. The average power output of the program material was measured with the calorimeter. Each test ran exactly 3.5 minutes. Prior to each test, the collector tab and heat sink were allowed to cool to 100 degrees Fahrenheit, then temperature rise was measured during the 3.5 minute operation period. The resulting curves are shown in Fig. 6, and it is evident that the output transistor heating which occurs at one-third rated output with program material can be simulated with only 11 percent of rated power using a continuous sine wave signal.
Most important, note that the output transistor heating effect of one-third rated power with a continuous sine wave signal could not be achieved under any condition, with program material.
In the November, 1974 issue of Stereo Review, Larry Klein proposed that preconditioning be done at one-third of rated power by driving the amplifier to full output with a 1-kHz tone burst having a one-third duty cycle. Tests with that signal showed that the output transistor heating effect approximates that of 11 percent rated output using program material, while the power supply is stressed to one-third of rated output.
Tests were also made using a 1-kHz tone-burst signal which drove the amplifier to 50 percent of full output at a two-thirds duty cycle. This signal more closely approximates the condition of driving an amplifier to one third of rated output with program material.
Fig. 7 Heat sink temperature vs. percentage of rated power, comparing present
FTC rule with other test methods.
Figure 7 shows the tone-burst data plotted along with the previously discussed heat-vs-power curves. It also shows the test results obtained using white noise as a signal.
The major points and some conclusions of this study can be outlined as follows:
1. The worst-case program material found in this study would demand a continuous output equal to 15 percent of an amplifier's maximum rms power capability.
Most Program Material Demanded Less Than 10 Percent.
2. The only way an amplifier can be continuously driven to one-third of its rated power with program material is to disregard distortion as a limitation. If an amplifier is overdriven in this way, one third of rated power does seem to be the maximum level at which distortion is tolerable by noncritical listeners. A discerning listener would be annoyed by that much distortion.
In the opinion of this writer, driving an amplifier this hard is an abuse which must be reckoned with in terms of reliability and safety, but it is an operating condition far beyond the scope of performance specifications.
3. The power output transistor heating effect produced by preconditioning at one-third rated power with a continuous sine wave signal cannot be achieved under any condition with program material.
The data developed in this study clearly points out the need to revise the present FTC preconditioning requirement. The big question is, how? Figure 7 shows a few possible conditions. Continuous sine wave operation at 10 percent of rated rms power has been a benchmark of engineering practice for decades. Figure 7 shows that this does accurately simulate the average power demand of program material when an amplifier operates within its rated distortion. An amplifier must be reliable and safe when over-driven or abused in other ways, but those considerations ought to be kept separate and distinct from tests intended to define the power and distortion of a high fidelity instrument.
For these reasons, this writer would prefer that pre-conditioning be done at 10 percent of rated rms power. However, a 1-kHz tone burst of 50 percent rated power two thirds of the time may be the most convenient way to "repair" the present FTC requirement. It does closely simulate the true one third power overdrive situation and does fall within practical and reasonable design guidelines.
(Audio magazine, Sept. 1975)
Also see:
Power Amplifiers and the Loudspeaker Load: Some Problems and a Few Suggestions (Aug. 1977)
Bi-Amplification--Power vs. Program Material vs. Crossover Frequency (Sept. 1975)
Build an Audio Generator (Oct. 1975)
= = = =