In Benchmark's listening room we recently demonstrated the importance of the first watt using two 100 watt stereo power amplifiers. One amplifier was a traditional class-AB amplifier, the other was Benchmark's AHB2 power amplifier with feed-forward error correction. Using a double-blind ABX test, we verified that there was a clearly audible difference when the amplifiers drove speakers at an output level of 0.01 watt.
When music is playing at a reasonably loud level on a typical studio monitoring system, or on a home hi-fi system, the average power into the speakers is usually only about 1 watt. With typical recordings, peaks reach 25 to 65 watts while the average music power is just 1 watt. The peaks are important, but most of the musical details are conveyed in the low power regions between the peaks.
Between the transient peaks, most of the important musical details are reproduced at power levels of much less than 1 watt. In any musical recording, the durations of transient peaks are very short relative to the time between the peaks. The first watt is the most important because the amplifier spends most of its time delivering less than 1 watt, even when the music is cranking.
In the low power region, it is our opinion that distortion can change the harmonic character of musical voices, clutter the unused frequencies between musical tones, and add a fatiguing harshness. Together these effects can detract from emotion and authenticity of the musical performance. Distortion can detach us from the performance and give us the distinct impression that we are just listening to sound coming out of a box.
Our ears need time to detect distortion. The thump of a kick drum, a note on the bass, or a loud transient from the percussion may approach the output limits of a power amplifier, but very little time is spent reproducing these transients. The short durations of these transients make it harder to hear the distortion that occurs on these musical peaks. In contrast, our ears have plenty of time to detect the distortion that occurs in the low-power spaces between loud transients. For this reason, our enjoyment of the music is largely dependent upon the quality of the delivery within the first watt. Ironically, amplifier bench measurements tend to focus on the THD+N (total harmonic distortion + noise) at maximum power. Unfortunately these high-power specifications have no direct relationship to the performance within the first watt. For this reason, it is important to separately examine the first-watt performance.
The following photo shows a pure 1 kHz tone being reproduced by a typical class-AB power amplifier at a power level of 1 watt. This amplifier has very good specifications near full power (100 watts), but like many class-AB amplifiers, it performs poorly at low power levels.
The top trace shows the amplifier output waveform at an output power of 1 watt. The lower trace shows the error waveform (distortion + noise).
The top trace may look like a perfect sine wave (pure tone) but our ears would easily tell us that it is not a pure tone. With an audio analyzer we can extract the error waveform (lower trace) and amplify it so that we can see what our ears are hearing. In this case, the error signal has been amplified by 1024 (about 60 dB). At a 1 watt output, this class-AB amplifier produces a distortion waveform that measures 70 dB below the output level of the 1 watt test tone (see graph at the bottom of this application note). This means that the power produced by the THD+N is 70 dB below 1 watt. If we drop the level of the test tone by 20 dB, the output power is 0.01 W. At this output level the amplifier was still producing distortion at a level of 73 dB below 1 watt. At 0.01 watt the distortion waveforms look virtually identical to the 1 watt waveforms. This distortion was clearly audible when we used this amplifier to drive a speaker at 0.01 watt. For our tests we used a stereo pair of Benchmark SMS1 speakers with a sensitivity of 87 dB 1 watt, 1 meter. The tone was reproduced at a sound pressure level of about 67 dB (measured 67 to 68 dB SPL) at the listening position while the amplifier distortion was reproduced at a calculated sound pressure level of about 14 dB (87 dB - 73 dB = 14 dB). With a 0.01 watt 1 kHz test tone, the amplifier distortion was clearly audible through the loudspeaker. At this low power level, the speaker distortion is lower and much more musical than the ugly crossover distortion produced by the amplifier.
Note the abrupt positive and negative spikes that occur in the distortion waveform. These occur whenever the top waveform crosses zero current. These spikes are caused by a less-than-perfect transition between the positive and negative output power transistors in the class-AB amplifier (see the simplified amplifier schematic below).
The class-AB amplifier used in our tests includes a traditional negative feedback system which is designed to correct the distortion caused by the output stage. Clearly, this feedback system is not fast enough to fully correct the high-speed transients that occur when the output stage transitions through the crossover region. These crossover distortion transients are a common occurrence in class-AB power amplifiers. The negative feedback networks used in power amplifiers are often too slow to reduce these transients to inaudible levels.
The bandwidth of any feedback system must be limited in order to achieve high-frequency stability. The large transistors used in power amplifiers are much slower than the small transistors used in line-level stages. The use of these slow devices dictates that the feedback network must also be slow. If the bandwidth of the feedback system is not limited (to properly match the speed of the power transistors) the amplifier will become a giant oscillator. Because of these speed constraints, power amplifiers often contribute more crossover distortion than the total of all the other electronic components in the signal chain.
Some people have argued that the differences in power amplifiers are inaudible. They justify this assertion by pointing out that loudspeakers produce more distortion than any reasonably good power amplifier. The fallacy of this argument is that the distortion produced by speakers is much different than the crossover distortion produced by a power amplifier. Speakers are also relatively clean at low power levels while power amplifiers may produce significant distortion at these low levels. In our listening tests, we verified that the distortion waveform shown above is audible while playing a 1 kHz test tone through speakers. If amplifier distortion is audible with a test tone, it may also be audible while playing music.
The Benchmark AHB2 power amplifier uses a patented distortion-reduction system that includes feed-forward error correction. Unlike feedback systems, the feed-forward system is inherently stable. The bandwidth of the correction signal does not need to be limited. This means that the high-speed crossover distortion transients can be fully corrected. The next photo shows the output of the AHB2 at 1 watt. The test conditions and scale factors are identical to those shown above. Like the prior photo, the distortion waveform is boosted by 60 dB, but this time, no traces of crossover distortion are visible. The feed-forward system delivers the first-watt performance of a class-A amplifier while outperforming class-A amplifiers at high power levels.
To conduct a double-blind listening test, we connected both amplifiers to a relay-controlled ABX switch box. The output of the switch box was connected directly to a Benchmark SMS1 loudspeaker. Both amplifiers were precisely adjusted to deliver a 1 kHz tone at 0.2828 Vrms into the speaker. This drive level of approximately 0.01 watt produced a sound pressure level of about 67 dB at the listening position. The test box ran a random sequence of 25 trials. The subject had a remote control with four buttons; "A", "X", "B", and "Enter". One amplifier was assigned to "A" and the other was assigned to "B". For each of the 25 trials, "X" was a random selection of either "A" or "B". In each trial, the subject was free to switch between "A", "B", "and "X" an unlimited number of times before determining that "X" is "A" or "B". At the end of 25 trials, the ABX box showed how many of the 25 trials were correctly identified. It was very easy for me to score a perfect 25 on my first attempt. The distortion created by the conventional class-AB amplifier at 0.01 watt was clearly audible while the AHB2 produced what sounded like a pure 1 kHz tone. The distortion is also audible at a power level of 1 watt, but the test tone is far too loud to conduct a sequence of 25 trials. Even at 0.01 watt, the test tone is annoyingly loud (67 dB SPL). This just goes to show how much of our music is reproduced at very low power levels. A single note on a flute may sound loud when it is delivered to the speakers at a power level of only 0.01 watt! Dick Olsher once said that "the first Watt is the most important Watt". In my opinion, "the first 10 milliwatts are the most important"!
On musical peaks, speakers may be the largest source of distortion in the entire playback system. But, in between the peaks, traditional class-AB amplifiers may produce more audible distortion than speakers. In our listening tests, it was easy to hear the difference between the two amplifiers, and score perfectly in a sequence of blind random ABX trials. Both amplifiers were playing a 1 kHz tone at 0.2828 Vrms into Benchmark SMS1 speakers (a level of about 0.01 watt). There was a clear audible difference when the amplifiers were performing this very simple task. The AHB2 reproduced a pure tone while the traditional class-AB amplifier added audible distortion. At a power level of 0.01 watt, the distortion produced by the speaker did not mask the difference between the two amplifiers.
We have proven that amplifier crossover distortion can be very audible through loudspeakers when playing pure tones at 0.01 watt. We have also shown that feed-forward error correction is very effective at eliminating the fast transients that are produced by crossover distortion in an amplifier's output stage. In contrast, traditional feedback systems are usually too slow to fully remove these transients. The remaining crossover distortion artifacts may be audible.
The following graph shows THD+N vs Power where THD+N is dB relative to 1 watt. This scale-factor makes it easy to calculate the sound pressure level of the distortion signal when it drives speakers. For example, if the speaker sensitivity is 87 dB, 1 watt, 1 meter, the distortion sound pressure level produced by the conventional amplifier at 1 W is (87-70) = 17 dB SPL. At this same 1 W power level, the distortion sound pressure level produced by the AHB2 is (87-108) = -21 dB SPL. The negative number indicates that this distortion signal should be 21 dB below the threshold of hearing when played in a quiet room. The conventional class-AB amplifier produces audible distortion, the AHB2 does not.