Anatomy of the Power Amplifier
Dissecting the Modern Audio Power Amplifier and Power Supply
By Robert Zeff


With the Proliferation of Different Amp Types, Which is the One for You? In the past we had essentially two types of amplifiers to choose from: Class "AB" and class "A". Today we have AB, A, D, G, H, & T, in addition to some that do not have a class name. New technology brought down the size and price while improving performance and efficiency. We'll review the various topologies of the modern amplifier, spending extra time on the aspect of efficiency (as the quest for smaller, more efficient designs have spawned the class D, G, H, & T designs). We'll also try to dispel some of the misconceptions and folklore that seem to surround amp design.

Amplifiers require circuitry for short and thermal protection, fan control, turn on delay, and over voltage protection. In the past we littered the designs with dozens of components to handle these events. Today we can use a single microprocessor to handle all of this in addition to having many more features without additional cost. The microprocessor can monitor the battery voltage, internal voltages, temperature, control volume and crossovers, and drive external displays. These embedded computer chips also allow features like compression and power limiting with little added cost. Of course, what is an amplifier without a power supply? First we'll visit the power supply designs, as every amplifier needs one.

The Power Supply
The purpose of the supply is to convert the auto's battery voltage to a higher voltage. For example, if an amplifier is to produce 100 watts into a 4 ohm speaker, we need 20 volts RMS. This implies that we need about +/-28 volts. (20 volts R.M.S. = 28.28 volts peak). We call that the "rail" voltage. Since the amplifier's output transistors cannot pull all the way up to this rail, we actually need a slightly higher voltage.

The process is to convert the 12 volts DC into AC, feed it to a transformer and convert it back to DC again.

Converting the 12 volt battery voltage to AC is simple, a PWM (pulse width modulator) IC feeds a bank of MOSFETS (MOSFETs are switching transistors perfectly suited for this task).

The 12 volt power is switched at a very high frequency, somewhere between 40 and 150 kHz. Slower switching speeds require a larger transformer, but high speeds have more switching loss. Advanced transformer core materials, faster rectifiers, and clever winding methods have enabled us to utilize very high frequencies. Some of today's better amplifiers have very small power supplies that produce enormous amounts of power.

Regulated Power Supplies
Most early audio amplifiers contained unregulated power supplies. Regulated supplies require very high quality filter capacitors (called "low ESR" capacitors), output chokes, and an optically isolated voltage feedback circuit. Regulation occurs by controlling the switching pulse width from 0 - 100% to compensate for changes in the battery and rail voltage. The same action occurs when the audio level increases. As the amplifier draws more power from the supply, the rail voltage drops. Again, the regulator circuitry senses this drop and responds with an increased pulse width.

The high frequency PWM waveform is rectified (converted to DC) and applied to the output filter choke and capacitors. This output of this circuit is the + and - DC rails that feed the power amplifier.

Unregulated Power Supplies
Unregulated power supplies are less expensive than regulated supplies. They do not require an output choke, voltage sense or isolation circuitry. Because the duty cycle is nearly 100%, capacitor ripple current is much lower in unregulated supplies. Lower ripple current requires less expensive capacitors throughout.

Often we hear that unregulated designs have more "headroom". That means that the amplifier will produce extra power during transients. Most home audio amplifiers employ unregulated power supplies. The power supplies in these amplifiers run at 60 Hz, thus the filter capacitors must be 200-500 times larger than those used in high frequency switchers. The extra capacitance in home audio amplifiers results in extra headroom. Headroom for anything other than very short transients simply doesn't exist in the unregulated designs. The following is an example of specifications for an unregulated vs. regulated amplifiers.

Unregulated designs have a higher supply voltage at low power, causing higher voltage on the output transistors. This reduces the amplifier's efficiency.

Small amplifiers (less than 100 watts) cannot justify the extra cost of the regulation circuitry, so we often see unregulated supplies in these amplifiers.

Pros and Cons of Regulated / Unregulated SuppliesSome designers try to keep their supplies regulated down to battery voltages as low as 9.5 volts. The supply compensates by increasing the current. The following table shows voltage and currents for a 500 watt over-regulated amplifier operating at full power.

The current increases dramatically at the lower voltages. Because of higher currents at the lower voltages, the supply efficiency drops further, requiring even more current.

At higher voltages, the pulse width reduces, causing increased ripple current. This high current creates heat in the filter capacitors and can destroy the capacitor's electrolyte. Some manufacturers do not use capacitors of sufficient quality for this range of regulation. These amplifiers may not perform up to specification just one year after installation. Also, the extra current at low voltages is extra hard on a battery that is already suffering! So, we recommend that amplifiers stay in regulation down to about 11 - 11.5 volts. Any properly working charging system can easily keep the battery voltage well above this.

The Amplifier Section, Class AB and AClass AB and A amplifiers are similar, so we'll discuss both here. Class AB amplifiers have transistors that pull up to the positive rail and transistors that pull down to the negative rail. This corresponds to the action of pushing the speaker cone out and in.

Class AB means that the output transistors do not always have current on them. For example, when the upper transistors are pulling up towards the positive rail (pushing the speaker out), there is no current in the lower transistors. When the output signal swings through zero, towards the negative rail, the output transistor must go through a transition from zero current to a non-zero current. The best analogy that I can think of is driving an old car with too much slop in the steering. As you go from one side of the road's crown to the other, the steering crosses a "dead" zone, and you tend to over-steer. Special temperature compensated bias circuitry reduces this dead zone, known as notch distortion. The figure below shows the output of a class AB amplifier with too little bias and the resulting distortion. Notch distortion increases at higher frequencies and low volume levels. Some modern designs have reduced this type of distortion to very low levels.

Class A means that every transistor is always conducting current. They are very similar to class AB amplifiers, but the bias circuitry is set so that there are very high currents in the output transistors. Because these amplifiers do not have this "dead zone', less feedback is required to achieve low distortion.

A 100 watt amplifier may dissipate nearly 100 watts internally even when there is no audio output. This type of design is impractical in the harsh auto environment. Many class A amplifiers pedaled for the automotive market are not really class A. They are huge power wasters in the home as well.

Input and Driver Stages
The amplifier works this way: A small audio signal is presented to the amplifier's input. Transistors are not linear, which means that the input signal will distort somewhat as it passes through the various amplifier stages. To correct this distortion, a portion of the output is compared with the input. The difference creates a correction signal reducing this distortion. The input stage is a special type, called "differential". It has a + and a - input because it must accept both the audio input and the input from the feedback circuitry. Excess feedback can lower distortion dramatically, but cause instability. Careful design rules must be followed to avoid this instability.

The output of the input stage feeds into the driver stage. The driver stage may use one, two, or three devices. Often this circuitry is referred to "Darlington", or "Triple Darlington". The driver circuit feeds the output stage, which may have two, four, six, or more transistors. The more output transistors, the better. Multiple output devices reduce distortion (requiring less negative feedback) and improve reliability.

Bipolar or MOSFET?
We have seen both MOSFET (Metal Oxide Silicon Field Effect Transistor) and Bipolar transistors used in audio amplifiers. Claims have been made that each is superior. I have seen claims that MOSFETs have a tube ("Valve" for the Brits) sound. This is more folklore. The musicians and their instruments are supposed to have "the sound", not audio equipment! MOSFETs are tougher than Bipolars, and can pull closer to the supply rail. It takes more Bipolar transistors to achieve the same power as a MOSFET, therefore Bipolar amps tend to be more expensive. But, MOSFETs are very non-linear, compared to Bipolars and require much more feedback to achieve reasonable distortion numbers. They are a great choice for bass amps, as low frequency audio is not difficult for a MOSFET. The most expensive car and home amplifiers almost always use Bipolar transistors.

What makes an amplifier get hot? Both the power supply and the power amplifier generate heat. The maximum efficiency of the power supply is nearly 100%. Good power supply designs, with the highest quality components approach 85%. The class AB amplifier efficiency at full power can approach 75%. The total efficiency, including the power supply, can be about 65%. But, efficiency drops at lower power and can typically be under 20%. A class AB amplifier actually runs cooler at full power than it does at half power. Run this amplifier into clipping and it might run even cooler! Where is all this power going? The output transistor is basically a large variable resistor. If the instantaneous output voltage should be 40 volts and the power supply is 100 volts, then 60 volts must be "wasted" in the output transistors. Driving a reactive load (like a speaker) causes the efficiency to drop ever further. This brings us to the other audio classes designed to improve efficiency.

Class D
First, let's dispel another myth: Class D does not stand for digital. The input is converted to a two-state (binary) representation of the audio waveform. That's where the similarity ends. This distinction is important because class D doesn't provide the benefits normally associated with digital components.

That being said, class D designs dramatically improve efficiency. Instead of wasting power in the output transistor, the output is switched at a very high frequency between the positive and negative supply rails. If the output is to be zero, then the waveform is at a 50% duty cycle. If the output is to be a positive voltage, then the duty cycle would be greater than 50%. Because the output devices are either completely turned on (no wasted voltage) or completely turned off, theoretically efficiency is 100%. So the audio input must be converted to a pulse width modulated waveform (PWM). The yellow trace below is the output of the amplifier; the blue trace is the PWM waveform. The blue waveform is fed to an output filter, which results in the yellow output waveform. Notice that the output looks somewhat distorted. All of the switching noise and distortion cannot be removed and the result can be seen here. Because of this process of converting the input signal to PWM and converting back to analog, a good deal of distortion is introduced. Conventional feedback like that used in class AB designs is used in these amplifiers to reduce distortion.

MOSFETs are the only choice for class D designs. Most class D designs are useful only for bass amps as they can not switch fast enough to reproduce high frequencies. Some high quality, full range class D designs exist for pro audio, but they are complex with multi-phased outputs.

Class T
Class T (Tripath) is similar to class D with these exceptions: This class does not use analog feed back like its class D cousin. The feedback is digital and is taken ahead of the output filter, avoiding the phase shift of this filter. Because class D or T amplifier distortion arises from timing errors, the class T amplifier feeds back timing information. The other distinction is that this amplifier uses a digital signal processor to convert the analog input to a PWM signal and process the feedback information. The processor looks at the feedback information and makes timing adjustments. Because the feedback loop does not include the output filter, the class T amplifier is inherently more stable and can operate over the full audio band. Most listeners can not hear the difference between class T and good class AB designs. Both class D and T designs share one problem: they consume extra power at idle. Because the high frequency waveform is present at all times, even when there is no audio present, the amplifiers generate some residual heat. Some of these amplifiers actually turn off in the absence of music, and can be annoying if there is too much delay turning back on.

Class G
Class G improves efficiency in another way: an ordinary class AB amplifier is driven by a multi-rail power supply. A 500 watt amplifier might have three positive rails and three negative rails. The rail voltages might be 70 volts, 50 volts, and 25 volts. As the output of the amplifier moves close to 25 volts, the supply is switched the 50 volt rail. As the output moves close to the 50 volt rail, the supply is switched to the 70 volt rail. These designs are sometimes called "Rail Switchers". This design improves efficiency by reducing the "wasted" voltage on the output transistors. This voltage is the difference between the positive (red) supply and the audio output (blue). Class G can be as efficient as class D or T. While a class G design is more complex, it is based on a class AB amplifier and can have the same clean characteristics as well.

Class H
Class H is similar to class G, except the rail voltage is modulated by the input signal. The power supply rail is always just a bit higher than the output signal, keeping the voltage across the transistors small and the output transistors cool. The modulating power supply rail voltage is created by similar circuitry that you would find in a class D amplifier. In terms of complexity, this type of amplifier could be thought of as a class D amplifier driving a class AB amplifier and is therefore fairly complex.

How to Choose?
Regulated or unregulated? Class AB, D, or T?If you're really into a lot of bass, the class D or T may be for you as these amplifiers will produce the highest SPL with the smallest size. If you just want to wake the neighbors, blur your vision, or make a big splash in SPL contests, maybe you just need one of the inexpensive, powerful, & dirty class D designs. Want the cleanest high frequencies? Maybe a good class AB amp would be your selection. Whatever you choose, I hope this information helps you achieve the sound you're looking for!