Most power amplifiers, and the amplifier sections within integrated amplifiers, are strikingly similar in operation. These so-called “Class A/B, push-pull” amplifiers have been around for decades and are the staple technology of nearly every amplifier manufacturer. I’ll explain how Class A/B amplifiers work later in this excerpt, but first I’d like to introduce you to some alternatives to the classic Class A/B amplifier. These include the single-ended triode amplifier, the single-ended solid-state amplifier, the switching power amplifier, and the digital amplifier.
Single-Ended Triode Amplifiers
One of the most interesting trends in high-end audio over the past 25 years has been the return of the single-ended triode power amplifier. The single-ended triode (SET) amplifier was the first audio amplifier ever developed, dating back to Lee De Forest’s patent of the triode vacuum tube in 1907 and his triode amplifier patent of 1912. SET amplifiers generally deliver very low power, sometimes just a few watts per channel.
You heard right: Large numbers of audiophiles are flocking to replace their modern power amplifiers with amplifiers based on 100-year-old technology. Have the past 100 years of amplifier development been a complete waste of time? A surprising number of music lovers and audio designers think so.
The movement back to SET amplifiers began in Japan in the 1970s, specifically with designer Nobu Shishido, who combined SET amplifiers with high-sensitivity horn-loaded loudspeakers. Many who heard SET amplifiers were startled by their goosebump-raising musical immediacy and ability to make the music “jump” out of the loudspeakers. Thus began the rage for SET amplifiers in Japan, which was about 10 years ahead of the SET trend in the United States. You can’t open a high-end
audio magazine today without seeing ads for very-low-powered single-ended triode amplifiers. The SET enthusiast’s mantra, coined by reviewer Dick Olsher, is, “If the first watt of amplifier power doesn’t sound good, why would you want 199 more of them?”
One early audio-amplifier triode tube, the Western Electric 300B, was suddenly in such demand that audiophiles were paying as much as $500 apiece for it. This shocking fact prompted Western Electric to start producing the 300B again. If you’d told Western Electric executives of the 1980s that in 1997 they would put the 300B back into production, they’d have thought you were crazy. (A single-ended triode power amplifier using the 300B output tube is shown in Fig.1 The 300B is the bulbous tube in the middle.)
The triode is the simplest of all vacuum tubes; its glass envelope encloses just three electrical elements rather than the five elements in the more common (and modern) pentode tube. Triodes have much less output power than pentodes, but more benign distortion characteristics. Virtually all modern tube amplifiers before the SET-comeback used pentodes.
In a single-ended triode amplifier, the triode is configured so that it always amplifies the entire audio signal. That’s what “single-ended” means. Virtually all other power amplifiers are “push-pull,” meaning that opposing pairs of tubes (or transistors) alternately “push” and “pull” current through the loudspeaker. (Later in this excerpt we’ll look more closely at how this works.) SET proponents believe that because the triode amplifies the entire waveform, SET amplifiers offer the ultimate in sound quality and musicality. Moreover, SET amplifiers have no need for a circuit called a phase-splitter, making them even simpler. Note that a single-ended tube amplifier can use more than one output tube; what makes it single-ended is that the tubes are configured in such a way that they always conduct current throughout the entire musical waveform.
In addition, SET circuits are extremely simple and often use very little or no negative feedback. Negative feedback involves taking some of the amplifier’s output signal and feeding it back to the input. Such feedback lowers distortion, but many feel that any feedback is detrimental to amplifier musicality.
On the test bench, SET amplifiers have laughably bad technical performance. They typically produce fewer than 25Wpc of output power and have extremely high distortion—as much as 10% total harmonic distortion (THD) at the amplifier’s rated output. Although most SET amplifiers use a single triode output tube, additional triodes can be combined to produce more output power. Some SET amplifiers, however, put out just 3Wpc. In fact, there’s a kind of cult around SET amplifiers that strives for lower and lower output power. These SET enthusiasts believe that the lower the output power, the better the sound. One SET designer told a reviewer, in all seriousness, “If you liked my 9W amplifier, wait until you hear my 3W model.”
In addition to low output power and high distortion, SET amplifiers have a very high output impedance as amplifiers go: on the order of 1.5–3 ohms. This is contrasted with the 0.1 ohm output impedance (or less) of most solid-state amplifiers, and the 0.8 ohms of many push-pull tube designs. Because a loudspeaker’s impedance isn’t constant with frequency, the SET amplifier’s high output impedance reacts with the loudspeaker’s impedance variations to produce changes in frequency response. That is, the SET amplifier will have a different frequency response with every loudspeaker it drives. These variations can range from just 0.1dB with some loudspeakers that have a fairly constant impedance, to as much as 5dB with other loudspeakers. The SET amp’s sound will therefore be highly dependent on the loudspeaker with which it is matched.
Despite these technical drawbacks, my listening experience with SET amplifiers suggests that this ancient technology has many musical merits. SET amps have a certain presence and immediacy of musical communication that’s hard to describe. It’s as though the musicians aren’t as far removed from here-and-now reality as they are with push-pull amplifiers. SET amps also have a wonderful liquidity and purity of timbre that is completely devoid of grain, hardness, and other artifacts of push-pull amplifiers. When I listen to SET amplifiers (with the right loudspeakers), it’s as though the musicians have come alive and are playing in the listening room for me. There’s a directness of musical expression that’s impossible to put into words, but is immediately understood by anyone who has listened for himself. You must hear an SET firsthand to know what the fuss is about; no description can convey how it sounds.
When auditioning an SET amplifier, it’s easy to be seduced by the midrange. That’s because SET amplifiers work best in the midband, and less well at the frequency extremes of bass and treble. If the SET demo is being run for your benefit, be sure to listen to a wide variety of music, not just small-scale music or unaccompanied voice—both of which will accentuate the SET’s strengths and hide its weaknesses.
The importance of matching an SET amplifier to the right loudspeaker cannot be overemphasized. With a low-sensitivity speaker, the SET will produce very little sound, have soft bass, and reproduce almost no dynamic contrast. The ideal loudspeaker for an SET amplifier has high sensitivity (higher than 93dB/1W/1m), high impedance (nominal 8 ohms or higher), and no impedance dips (a minimum impedance of 6 ohms or higher). Such a speaker will produce lots of sound for a small amount of input power, and require very little current. There’s been a resurgence in high-sensitivity speakers that has paralleled the popularity of SET amplifiers. Some loudspeakers designed for SET amplifiers have sensitivities of more than 100dB, which enable them to produce satisfying listening levels with 5Wpc. SET amplifiers are often coupled with horn-loaded loudspeakers, which have extremely high sensitivity but, in my experience, can introduce unacceptable levels of coloration.
The popularity and unmistakable sound quality of SET amplifiers pose a serious dilemma: How can an amplifier that performs so poorly by every “objective” measure produce such an involving musical experience? How can 100-year-old technology eclipse, in many ways, amplifiers designed in the 21st century? What aren’t we measuring in SET amplifiers that would reflect their musical magic? Why do conventional measurements fail so dismally at quantifying what’s right in SET amplifiers? Do SET amplifiers sound so good because of their high distortion or despite it? As of yet, no one has the answers to these questions.
Single-Ended Solid-State Amplifiers
Single-ended amplifiers aren’t confined to those using ancient vacuum-tube technology. Transistors can also be configured to amplify the entire musical waveform. A solid-state, single-ended amplifier is shown in Fig.2. Note the large heatsinks required to dissipate the additional heat produced by single-ended operation.
Single-ended solid-state amplifiers have better technical performance than single-ended triode amps, with a lower output impedance, more power, and the ability to drive a wider range of loudspeakers. Still, they share many of the benefits of SET amps, particularly the very simple signal path, lack of crossover distortion, and greater linearity. Although single-ended solid-state amplifiers generally produce less power than their push-pull counterparts, they generally have much more output power than single-ended tube units. Nonetheless, it’s a mistake to equate single-ended solid-state with single-ended tube amplifiers: there are so many other design variables that single-ended solid-state and single-ended tube amplifiers should be considered completely different animals.
Switching (Class D) Power Amplifiers
If single-ended triode amplifiers represent a return to fundamental technology, the switching power amplifier may represent the future of audio amplification. Switching amplifiers, also called Class D amplifiers, have been gain-ing in popularity due to their small size, low weight, high efficiency, and low cost. You can hold some 250Wpc switching amplifiers in the palm of your hand. Many of them dissipate so little heat that they can be housed in an enclosed equipment cabinet—something you’d never do with a conventional amplifier (also called a linear amplifier). That’s because a linear amplifier is typically about 15% efficient, meaning that it converts only about 15% of the power it draws from the wall into the electrical signal that drives the loudspeakers. The other 85% is dissipated as heat. By contrast, a switching amplifier is as much as 90% efficient, converting just 10% of its power draw into heat. Fig.3 shows a switching power amplifier. This monoblock delivers 175W into 8 ohms and 335W into 4 ohms. It weighs just eight pounds and measures 8.5" x 14" x 1.8".
Switching amplifiers are sometimes erroneously called “digital” amplifiers, but that appellation is reserved for a special type of switching amplifier described later in this excerpt.
At the low-end of the audio spectrum, switching amplifiers are ubiquitous in home-theater-in-a-box units and as integral subwoofer amplifiers. A home-theater-in-a-box may need to power six loudspeakers from a DVD-player-sized chassis—all for a few hundred dollars. Such a unit can output perhaps 300 watts (50Wpc x 6), yet run cool enough to be placed in a cabinet. In this application, the advantages of a switching amplifier are undeniable. But are switching amplifiers suitable for high-end systems?
Before tackling that question, let’s first look at how a switching amplifier works. In a conventional linear amplifier, the output transistors amplify an analog signal—the musical waveform. The current flow through the output transistors (or tubes) is continuously variable—a direct analog of the musical waveform. In a switching amplifier, the analog input signal is converted into a series of “on and off” pulses. These pulses are fed to the output transistors, which turn the transistors fully on or fully off. When the transistors are turned on, they conduct the DC supply voltage to the loudspeaker. When they’re turned off, no voltage is connected to the loudspeaker. The audio information is contained in the durations of these on-off cycles. The train of pulses amplified by the transistors is smoothed by a filter to recover the musical waveform and remove the switching noise. Because the signal amplitude is contained in the width of the pulses, switching amplifiers are also called pulse-width modulation (PWM) amplifiers. In fact, the Direct Stream Digital encoding system used in SACD is nearly identical to the pulse-width modulation scheme in Class D power amplifiers.
One major drawback of switching amplifiers is the need for an extremely clean power supply. Remember that the transistors switch the DC supply directly to the loudspeaker. Consequently, any noise or ripple (a small amount of AC on the DC voltage) will appear at the loudspeaker’s input terminals. Moreover, the output transistors may not turn on and off at precisely the right time, introducing distortion. Finally, switching amplifiers generate large amounts of switching noise that must be filtered by huge inductors and capacitors at the amplifier output. In practice, the sound quality of switching amplifiers seems to be highly dependant on the environment, the loudspeaker, and the loudspeaker cables. A switching amplifier that sounds reasonably good in one system might be unlistenable in another.
Nonetheless, some successful high-end amplifiers employ switching technology. The field is relatively new, and manufacturers are finding ways to get good sound from switching amplifiers. A few of the high-end switching amplifiers I’ve heard sounded excellent, suggesting that switching technology may have a future in products other than car stereos and home-theaters-in-a-box.
A related amplifier technology uses a switching output stage, but accepts digital, rather than analog, input signals. These “digital” amplifiers take in the pulse-code modulation (PCM) signal from a CD transport, music server, or other source and convert those audio data to a pulse-width modulated signal. This PWM signal drives the output transistors, just as in a switching amplifier. The difference between a switching amplifier and a digital amplifier is that the digital amplifier accepts digital data rather than analog signals.
This difference might not seem that great at first glance, but consider the signal path of a conventional playback chain with a digital source and a switching power amplifier. In your CD player, data read from the disc go through a digital filter and are converted to analog with a DAC; the DAC’s current output is converted to a voltage with a current-to-voltage converter; the signal is low-pass filtered and then amplified/buffered in the CD player’s analog output stage. This analog output signal travels down interconnects to a preamplifier with its several stages of amplification, volume control, and output buffer. The preamp’s output then travels down another pair of interconnects to the power amplifier, which typically employs an input stage, a driver stage, and the switching output stage. In addition to the D/A conversion, that’s typically six or seven active amplification stages before the signal gets to the power amplifier’s output stage.
To reiterate the contrast with a true digital amplifier, PCM data are converted by DSP into the pulse-width modulation signal that drives the output transistors. That’s it. There are no analog gain stages between the PCM data and your loudspeakers. The signal stays in the digital domain until the switching output stage, which, by its nature, acts as a digital-to-analog converter in concert with the output filter. The volume is adjusted in DSP. Digital amplifiers are usually not just power amplifiers, but also include inputs, source selection, and volume control, effectively giving them the functional capabilities of an integrated amplifier. Fig.4 shows a digital amplifier, and Fig.5 is a block diagram comparing the signal path of this amplifier with conventional system architecture.
Output Stage Topology and Class of Operation
We’ve seen that the output stage of a power amplifier can be configured either as single-ended or push-pull. To reiterate, in a singled-ended amplifier the output devices (tubes or transistors) always amplify the entire musical waveform. The single-ended amplifier cannot operate in any other way—that’s the very definition of “single-ended.” In a push-pull output stage, pairs of opposing devices are arranged so that they work alternately—one device “pushes” current through the loudspeaker and then other device “pulls” current through the loudspeaker. Multiple pairs of output devices can be grouped together to increase the power output, called parallel push-pull.
That’s a description of an amplifier’s output-stage topology—how the amplifying devices are configured. Now let’s look at a separate but related factor—the amplifier’s class of operation.
Class of operation refers to how a given output stage is driven. The two main classes of operation are Class A and Class B. In a Class A amplifier, the output stage (single-ended or push-pull) amplifies the entire musical waveform. In a Class B amplifier, opposing pairs of transistors (or tubes) are operated so that one transistor in the pair amplifies the positive half of the waveform and the opposing transistor amplifies the negative half of the waveform.
Class of operation is easily confused with single-ended and push-pull output-stage topologies. But there’s an important distinction. A single-ended amplifier always operates in Class A. But a push-pull amplifier can be operated in either Class A or Class B. It might seem like a push-pull Class A amplifier is a contradiction in terms, but it’s not. In a push-pull Class A amplifier, opposing pairs of transistors are driven in such a way that current flows through both transistors throughout the entire musical waveform. All the output devices participate in amplifying through the full cycle of the audio signal. One device pulls current through the loudspeaker and the other pushes current, but both are always turned on and conducting current.
By contrast, in a Class B push-pull output stage, one transistor amplifies the signal during the positive-going portion of the signal, and the other amplifies the negative-going signal half. When one is working, the other is turned off (and getting some needed cooling).
Once again, two key factors are the output-stage topology—how the output devices are configured—and class of operation—how those output devices are driven. To clarify the matter:
• Single-ended topology always operates in Class A
• Class A operation can be implemented in a single-ended or push-pull topology
• Push-pull topology can be operated in Class A or Class B
• Class B operation is only implemented by a push-pull topology
Let’s move on from these distinctions to the more practical aspects of amplifier class of operation.
Most power amplifiers are called Class A/B power amplifiers because they operate in Class A at very low power outputs, then default to Class B operation at higher power outputs. A 100Wpc amplifier may put out 5W of Class A power, and then switch to Class B above that level. Even the heftiest Class A/B amplifiers can put out only a small portion of their rated power in Class A. A typical value is about 1% or 2%. Although this may not sound like much, an amplifier may be running at just a couple of watts at low listening levels with high-sensitivity speakers.
How much of an amplifier’s power output is Class A is determined by the amount of bias applied to the output transistors. Bias is a DC current that flows through the output stage at idle. The higher the bias, the more current flows through the transistors when no signal is present. More bias results in more Class A output power. Class B operation has no bias current; Class A/B has moderate bias current; Class A has very high bias current. The designer can keep increasing the bias in a push-pull output stage until all of its output power capability is delivered in Class A mode. The amplifier’s power output rating would be the point at which the amplifier leaves Class A operation and begins operating in Class B. The limiting factors to increasing the bias current are the ability of the transistors to handle the greatly increased current flow through them, the power supply to keep up with the transistors’ current demands, and the heatsinking to dissipate the considerable heat caused by the high bias current.
To give you an idea of the demands placed on the output stage, power supply, and heatsinking by an output stage biased for Class A operation, let’s compare two products with identical output stages that are biased completely differently. The two products are the Pass Labs INT-150, a 150Wpc Class A/B integrated amplifier, and the Pass Labs INT-30A, a 30Wpc pure Class A integrated amplifier. The INT-150’s push-pull output stage is biased so that it produces 10Wpc of Class A power before switching into Class B to deliver its full rated output of 150Wpc. The amplifier can double its output power to 300Wpc into 4 ohms. The INT-30A is exactly the same amplifier, employing the same power supply, an identical number and type of output transistors, and the same heatsinks as the INT-150, but is rated at just 30Wpc. The difference is that the INT-30A’s 30Wpc are pure Class A watts. The amplifier biased into Class A delivers just one fifth the power output of its Class A/B counterpart.
As you can see, Class A operation is enormously inefficient. A Class A amplifier converts nearly all the power it draws from the wall outlet into heat, and consumes just as much power at idle as when it is operating at its maximum output power. Moreover, a Class A amplifier is much more expensive to build on a “watts per dollar” basis than its Class A/B counterpart.
So why would designers go to the trouble and expense of creating Class A amplifiers, and why would consumers pay such a huge premium for “Class A watts” over “Class A/B watts”?
Class A has many theoretical and practical advantages. For starters, Class B and A/B amplifiers suffer from crossover distortion, a discontinuity at the zero-crossing point where one transistor in the opposing pair “hands off” the signal to the other transistor in the pair. A waveform discontinuity can occur at this transition, and is lessened as bias current is introduced and increased in value. Crossover distortion can’t occur in Class A operation because each transistor amplifies the entire audio waveform, not just half of it. Second, the large thermal hardware capacity required by Class A has the advantage of keeping the output transistors in better thermal stability (a more constant temperature). This makes their operating characteristics more uniform, and less subject to changes resulting from the signal characteristics the transistors are amplifying. For example, if the transistors have just delivered a surge of current to the loudspeakers, they won’t behave differently—and thus sound different—immediately afterward because they are momentarily hotter. Third, increasing the bias current so that an amplifier produces more
Class A power not only reduces harmonic distortion, but more importantly, changes the nature of the harmonic distortion. As the bias is increased, the upper-order harmonics (everything above the third harmonic) that are most sonically harmful are most dramatically reduced in amplitude, leaving the predominant distortion component the more sonically benign second and third harmonics. Class A power amplifiers can sound extremely good, with a sweetness and liquidity that set them apart from Class A/B amps. In my experience (I recently lived with a pair of 100W Class A monoblocks for 18 months), Class A amplifiers have many of the virtues of a tube amplifier but without the tube amp’s technical limitations. This isn’t to say that a Class A amplifier mimics the sonic character of tubes, but rather that Class A avoids many of the characteristic nonlinear distortions of Class A/B solid-state amplifiers.