The Peak Power Demands of Well-recorded Music
This attention-getting demonstration made many participants think twice about how much amplifier power is needed for the reproduction of well-recorded music that has good dynamic range. Participants were able to see in real time the very high peak-to-average power ratio (crest factor) of music while they were listening to it. The high crest factor of well-recorded music can place very high peak-power demands on a power amplifier even at modest listening levels when loudspeakers of average efficiency are employed. This workshop demonstrated why high-power amplifiers sometimes sound better. Amplifiers may clip more often than we think. Some participants brought their own music. Examples of MP3, iTunes and off-air f.m. broadcast material were also available for comparison.
The presentation relied on a Peak/Average power display built specifically for the workshop. The display includes two large LED displays that read power in integer Watts into 8 ohms. The box displays true-rms-derived average power on a long-term basis on the left meter. This is meant to be a rough approximation of the perceived loudness; it is the average number of watts that it took to make that average level of loudness. The function of the "Peak" display on the right side of the box is to show the momentary average-calibrated power that occurs on musical peaks. It is the power that a conventionally-rated power amplifier would have to have in order to reproduce that musical peak. Note that if one applies a continuous sinewave input to the box, both meters will read identically. The box will accurately detect and display a musical peak that lasts as little as about 20 microseconds and will hold it for one second before the reading begins to decay. More detail about how the box works is provided at the end of this section.
A variety of music was used for this demonstration, but the musical cut with the highest crest factor we have found so far is Rickie Lee Jones' "Ghetto of my Mind" on her Flying Cowboys CD. Play this cut at even remotely realistic levels and monitor it with the Peak/Average power display and you will re-think how much power your amplifier needs to have. Of course, most music (unfortunately) is not this "bad" because of compression added in the recoding process for the boom-box/car-CD primary target market.
This demonstration was a real "Wow" for the attendees. The Rickie Lee Jones (RLJ) cut was played at realistic, but certainly not unpleasant, levels in the relatively small hotel exhibit room on speakers with an estimated sensitivity of about 89 dB. The average power typically read 1-2 Watts, while the power on peaks often topped 250 Watts (the power display monitored only one channel, so these numbers should be interpreted as Watts per channel). On this cut, most peaks occurred with an aggressive thwack to a snare drum positioned dead center.
While it is true that the RLJ track has an unusually large dynamic range, this data still suggests that many listeners may be clipping their amplifiers more often than they think. This may especially be the case for those with tube amplifiers who are not using extraordinarily efficient speakers. The amount of clipping, and the way in which amplifiers handle clipping, may account for more of the perceived differences in amplifier sound than we realize. It would be nice if all amplifiers had accurate and fast clipping indicators. It might be a real eye-opener. If your amplifier is clipping, have you left the realm of high fidelity?
How the Peak-average Power Display Works
Here is a quick description of how the Peak-Average power display works. Keep in mind as we go that the semantics of peak and average and rms can get tricky.
The function of the box is to display true-rms-derived average power on a long-term basis on the left meter. This is meant to be a rough approximation of the perceived loudness, and the average power in Watts that it took to make that average level of loudness.
The function of the "Peak" side of the box is to display the momentary average-calibrated power that occurs on musical peaks. If you had an amplifier rated at 50 watts continuous average power (some call it 50 watts rms) into 8 ohms, and that amplifier has no dynamic headroom, it will clip at a sinewave output of 20V rms, or a peak output voltage of 28V peak. On our box, if the amplifier briefly hits a peak output voltage of 28V on a snare drum thwack, the right meter will display 50 Watts. It is the power that a conventionally-rated power amplifier would have to have in order to reproduce that musical peak if the amplifier had no dynamic headroom (maximum power for brief intervals = maximum continuous average sinewave power).
The box will accurately detect and display a musical peak that lasts as little as about 20 microseconds. It will accurately capture the peak of a single-cycle 10 kHz sinusoid. It will detect when a new peak has occurred that is greater than the currently-displayed value and hold it for one second before any decay begins. This gives the LED digital panel meter (which updates three times per second) plenty of time to properly display the result. The circuitry to accomplish all of this is not trivial.
There are two paths in the box: the "average" path to the left display and the "peak" path to the right display. The average path passes through a true-rms chip which produces a d.c. output that is representative of the long-term average (about 5 seconds) signal voltage. That d.c. voltage is then passed through a multiplier-based squaring circuit and scaling circuit to arrive at a d.c. level that represents average power. That voltage is presented to the digital panel meter, for a reading in integer Watts.
The "peak" path signal first passes through a full-wave rectifier and a fast peak detector. The resulting signal representing "stretched" peaks is then applied to a slower peak detector that has a gated decay. For one second after a new peak is detected, the decay is turned off, making that circuit a peak-hold circuit. The d.c. output represents the highest instantaneous peak voltage detected. That voltage is sent to a multiplier-based squaring circuit and scaled to make it represent average power on musical peaks for presentation to the digital panel meter.
Note that if one applies a continuous sinewave input to the box, both meters will read identically.
A signal is also available that indicates when a new peak arrives that is higher than the currently-displayed peak. This signal can be used to trigger a digital storage oscilloscope (DSO), if desired.
Finally, most amplifiers have dynamic headroom of 0.5 to 3 dB. The peak meter will quickly give you an idea of this. When you just clip the amplifier you are using, you'll see the dynamic headroom power displayed. Compare this to the rated continuous sinewave power and you get a dB ratio that represents the dynamic headroom capability of the amplifier on real music.