The Design and Function of Transistors in Audio: Past, Present, and Future

He has published articles on power amplifier design in the Journal of the Audio Engineering Society (JAES) and other publications, as well as numerous books on the subject.

1. Introduction

Transistors are the basic active building block of audio circuits, just as they are for virtually all electronics. Here we will discuss the various types, a bit about how they work, their advancements for use in audio over the years, and new developments.

The three most common categories of transistors are bipolar junction transistors (BJT), junction field effect transistors (JFETs), and metal oxide field effect transistors (MOSFET). They all play slightly different roles in audio circuits, each with different characteristics and advantages.

Transistors are the central active device in integrated circuits (IC), and ICs also play a basic role in audio circuits, most commonly as operational amplifiers (op amps), but also in digital signal processing (DSP), so the use of transistors in linear and digital ICs will also be discussed.

It is difficult to discuss the influence of transistors and their evolution on audio without explaining a bit about the different types of transistors and how they work, plus the application circuits in which they are used. So, we will include discussions about both.

The various transistor types will be discussed first. Then we will describe their use and impact on audio applications, including their role in the emergence of class-D power amplifiers and switch-mode power supplies (SMPS). Those two latter technologies play a big role in making audio equipment smaller, lighter, and less expensive. This discussion is by no means definitive, and many simplifications are made in some of the explanations. For a deep dive into Class D audio amplifiers, see Bruno Putzey’s recent article in Secrets.

2. Small-Signal Transistors

The element germanium was used for transistors in the 1950s and early 1960s. Germanium is a semiconductor which has properties that fall between a conductor and an insulator, making it useful in electronic devices like transistors. Germanium transistors were slow, could not operate at higher voltages, and were adversely affected by high temperatures. Silicon (also a semiconductor) transistors came into widespread use in the late 1960’s. They also made integrated circuits possible. Small-signal transistors largely matured by the end of the 1970s. These transistors had good current gain, good high-frequency performance, and low noise. Small-signal junction field effect transistors (JFETs) also matured, especially regarding noise performance. These developments set the stage for high-performance audio preamplifiers and power amplifiers. Metal oxide field effect transistors (MOSFETs) were yet to make much impact on audio circuits but had begun to make an enormous impact on digital circuits. The development of the microprocessor was a testament to this.

Most of the advancements in small-signal transistors for use in audio circuits had been made by the end of the 1970s. While discrete transistors are still used in power amplifiers, the use of small-signal discrete transistors has declined over the years due to progress in analog integrated circuits, most notably in operational amplifiers (op-amps). An operational amplifier is a DC-coupled electronic voltage amplifier with a differential high-impedance input (+, -, and ground), but most often with a single-ended output, and with very high gain. They were originally used to perform mathematical computations like addition and subtraction in voltage-based analog computers (a long, long time ago). They are usually small self-contained units. Rohm’s TLR377GYZ op-amp is 0.88 mm x 0.58 mm x 0.33 mm, and it is the smallest available op-amp on the market.

Op-amps are still used in applications requiring very low noise, like moving coil phono preamps. Most remaining transistor advancements over the years have been in the power transistors used in the output stages of audio power amplifiers.

3. Linear Integrated Circuits and Operational Amplifiers

Integrated circuit operational amplifiers saw widespread use in audio circuits beginning in the 1970s, but their performance left a lot to be desired. They required PNP transistors, but the lateral PNP transistors that could be fabricated in ICs at the time had very poor characteristics, including very low current gain and poor high-frequency performance. The op-amps like the μA741 were noisy, and they had poor high-frequency response, inadequate slew rate, and significant distortion, especially at the upper end of the audio band. JFET op-amps, like the TLO71, also became available, but they had fairly high noise. Apart from noise, their audio performance was better than that of the μA741, and their very high input impedance was beneficial for high-impedance circuits.

The NE5534 op-amp was introduced circa 1979 and was a game-changer. It was far superior to other audio op-amps of the time, like the 741. It saw extremely widespread use in audio circuits, both consumer and professional audio. Mixing consoles became chock full of them. The 5534 achieved remarkable performance in spite of the fact that it did not employ vertical PNP transistors. It had low noise, high slew rate, wide bandwidth, and very low distortion. It is still used to this day in audio equipment and instrumentation. The dual version, the NE5532, was also used extensively, although its performance falls short of that of the single NE5534. Today their performance/price ratio is very high.

Linear integrated circuits, like operational amplifiers, were limited by the need for lateral PNPs, which were very slow. PNP transistors in integrated circuits were difficult to make. While NPN transistors had a vertical geometry, PNP transistors could not be made that way with the processes available in the day, so they had to be made as so-called lateral transistors, where the current flowed laterally rather than vertically. Among other disadvantages, the width of the base was determined by lithography, so it could not be really small, as required for high speed and high current gain. Often, the current gain was as small as unity or less (unity gain means the output is the same as the input). Their transition frequency f_t, where current gain fell to unity, was usually only about 1 MHz. The f_t for NPN transistors was usually 100 MHz or more.

The junction-isolated vertical PNP transistor process was developed at Bell Labs and was put into commercial use in telecommunications applications. The vertical PNP is named that way because it has a vertical cross-section, just like NPN transistors. The full-complementary vertical PNP process was first used in op-amps, where high performance was needed for precision active filters. Later, many advanced audio op-amps were implemented using the full-complementary PNP process. Today’s op-amps depend on the use of vertical PNPs for their very high audio performance.

4. How Transistors Work

Before getting into the advancements, types, and applications of transistors in audio, I want to talk about how they are implemented and what is behind their operation. This will be helpful in understanding the various terms that will be used later in this article.

This first part touches on device details and a little bit of elementary physics. It can be skipped over by those who are already knowledgeable or inclined to get glassy-eyed. The great majority of transistors used in audio and all other electronics are based on silicon and are generally referred to as semiconductors. By the same token, some of those knowledgeable may take issue with the oversimplifications made here for the sake of simple explanation. Liberties have been taken in the interest of ease of understanding and the limited scope of this article. The initial focus here will be on the bipolar junction transistor (BJT), as these concepts and terms are applicable to other types of transistors that will be discussed later.

Silicon is a semiconductor that lies between conductors and insulators as a material, hence the term semiconductor. Pure (intrinsic) silicon is an insulator, while silicon doped with an appropriate impurity becomes a conductor, like a resistor. The key to its use as a semiconductor is the type of material with which it is doped. Examples of doping materials are arsenic, phosphorous, and boron. These create n-type and p-type silicon with an excess (negative) or deficiency (positive) of electrons, respectively.

Silicon has an atomic number of 14, meaning that it has 14 electrons in orbital shells around the nucleus that contain protons and neutrons. There are 4 valence electrons (4 electrons in its outer shell), and silicon belongs to Group IV in the periodic table. P-type dopants have 3 valence electrons and are Group III semiconductors. The 4 valence electrons allow silicon to form 4 covalent bonds with other atoms. Covalent bonds are where two atoms share their electrons to complete their outer shell. The number of valence electrons determines how easily an atom can form bonds with other atoms, which influences its conductivity. Atoms with 4 valence electrons are most suitable as semiconductors.

When these impurities are put into silicon there is a net deficiency of electrons, and holes are created. N-type dopants have 5 valence electrons, and when silicon is doped with them, there is an excess of electrons and free electrons are created.

If you put a layer of p-type silicon on top of a layer of n-type silicon, a so-called p-n junction is formed, in this case creating a diode. If a voltage of the correct polarity is placed across the diode, it will conduct a certain amount of current. This is referred to as a forward bias or a forward voltage.

Materials that are n-type have an excess of electrons, and these free electrons are called charge carriers. Those atoms are negatively charged. The p-type material has a deficiency of electrons (holes). Those atoms are positively charged. Holes are also charge carriers. If holes and electrons meet, recombination may occur, where the electrons fill the holes. Those atoms are no longer charge carriers and do not participate in current conduction. For this and other reasons, a thin region without charge carriers is created at the junction, and it is called the depletion region or layer. This layer acts as an insulator and constitutes a barrier that must be overcome for conduction to occur. This is why diodes do not conduct when reverse-biased, and in fact, do not conduct until a forward bias is applied that is sufficient to overcome the barrier (about 0.6 V). The electric field created by a reverse bias will actually cause the depletion region to be wider, pushing the conductive regions further apart.

The forward current in a diode increases exponentially as the forward bias voltage from the anode to the cathode increases. By the same token, the forward voltage across a diode increases as the log of the current through the diode.

The transistor that is historically most well-known is the bipolar transistor. The name transistor derives from the term transconductance resistor. A resistor conducts current, so it has conductance. Transconductance means that input voltage controls output current in an active device. In this case, the base-emitter voltage V_BE controls collector current I_c. This is a slightly different view than the one where base current I_b controls collector current via current gain β. They are both true and both views are used in circuit analysis.

Transistors have two p-n junctions, the base-emitter junction and the base-collector junction. In an NPN transistor, the emitter and collector are N-type material while the base, sandwiched between the emitter and collector, is made of P-type material. The transistor thus comprises two p-n junctions that share the base, somewhat like two back-to-back diodes. In normal operation, the base-emitter junction is forward-biased, and the base-collector junction is reverse-biased.

If a sufficient voltage is applied to the base, the base current will flow, and the emitter will be stimulated to emit free electrons into the base. Those electrons will travel through the p-type base and enter the collector, where they will cause the collector current to flow.

In reality, it is truer that the voltage applied to the base, V_BE, causes the transistor to conduct and control the collector current in the same way that the voltage across a diode causes current to flow – in the same exponential fashion. Base current is really just an unwanted consequence of the forward-biased base-emitter junction and loss of charge carriers in the base due to recombination.

It is important to note that most of the charge carriers get across the base before they have time to recombine with other carriers of the opposite charge. In an NPN transistor, free electrons are the majority carrier in the n-doped emitter, but they become so-called minority carriers when they enter the p-doped base. The now-minority negative charge carrier electrons in the base are attracted by and diffuse into the collector, causing the collector current to flow.

5. Power Amplifier Output Stages

Power amplifiers moved from vacuum tubes to transistors in the late 1960s. Power amplifiers rated at greater than 100 Watts became commonplace. The biggest influence on audio by transistor improvements over time has been on power transistors.

The only decent silicon power transistors in the 1960s and early 1970s were NPN. Even NPN power transistors were slow and had limited safe operating area (SOA), meaning that they could not safely conduct a significant amount of current at high voltages. This hobbled the design of solid-state push-pull power amplifier output stages. Decent PNP power transistors came out in the mid-1970s and enabled full-complementary output stages.

Most transistor power amplifiers employ push-pull output stages. In these circuits, a top transistor is connected to the positive power supply rail and pushes (sources) current into the loudspeaker load. A bottom transistor is connected to the negative power supply rail and pulls (sinks) current from the load.

In a class A power amplifier, both transistors are constantly on and contribute signal current to the load over the full cycle of the signal. This arrangement draws considerable current all of the time. Even when there is no signal, the top and bottom transistors each conduct half the peak current that occurs at full power. This creates a lot of heat that is dissipated in the output stage.

In a class B design only the top or bottom transistor is on at a time and contributing load current. This arrangement is much more efficient since both transistors are nearly off when there is no signal. However, significant distortion is created when the signal handling responsibility is handed off from one half of the output stage to the other since the gain of the output stage will usually decrease in this region or even fall to a very small amount if there is a gap in the handoff. This is called crossover distortion.

The most popular arrangement is the class AB output stage, where one transistor is mostly responsible for the output current for one-half of the signal cycle. Both transistors contribute current to the load during the handoff in the region of a small signal current. Both transistors are on at idle, conducting a small amount of current. This reduces crossover distortion, but usually not completely because the gain of the output stage, which is slightly less than unity, varies a bit through the crossover region [1]. When small-signal gain changes as a function of signal amplitude, distortion is created.

The amount of quiescent (idle) bias current is set at an optimum value to minimize crossover distortion. If it is too small, the behavior of the output stage will approach class B, and the gain of the output stage will decrease in the crossover region and cause distortion. If it is too high, both transistors will be contributing too much signal current to the load and gain will increase beyond its nominal value, causing distortion.

Finally, the duration of the crossover region can be quite small at high frequencies, meaning that the transistors must be able to turn on and off quickly during the transition. This is especially the case with how long it takes them to turn off. Increases in the time it takes to turn off can create what is called dynamic crossover distortion.

In a straightforward design, the top transistor in the schematic is an NPN transistor that forms the upper half of an emitter follower (EF) that provides unity voltage gain but high current gain. The output EF is usually preceded by a driver EF to provide even more current gain in a so-called Darlington configuration. A PNP power transistor implements the bottom half.

A simple push-pull output stage is illustrated in Figure 1. An NPN transistor at the top sources current into the load, and a PNP transistor sinks current from the load. These transistors are connected as emitter followers (EF), with near unity voltage gain. Their sole purpose is to provide current gain. A so-called bias spreader is illustrated with two silicon diodes. It provides enough voltage from base to base to turn on the output transistors. This voltage makes up for the two base-emitter (V_BE, about 0.6 V) turn-on voltages of the output transistors to get them to conduct an appropriate amount of quiescent current, often on the order of 100 mA. Emitter resistors allow enough current-dependent voltage drop to provide thermal stability. In practice, the bias spreader is more complex than just two diodes and also includes a diode or transistor mounted on the heat sink to monitor the heat sink temperature and track temperature changes in the heat sink. For convenience, we will often refer to the voltage drop of a forward-biased silicon diode as being one V_BE, since it is also about the same voltage.

Significant amounts of current need to be supplied to the load in a power amplifier. A 100-watt amplifier must provide a peak current of 5 Amps into an 8-Ω load. The same 100-watt amplifier driving a 4-Ω load to 200 Watts will have to supply a peak current of 10 Amps. A 200-Watts into 8 Ω amplifier driving a 4-Ω load may have to supply 14 Amps. This can get ugly fast. Power amplifiers thus require high current gain in the output stage, requiring the use of two emitter followers in tandem (a driver and an output transistor). This is often referred to as a Darlington output stage. In some cases, even three transistors are used in tandem (a Triple), with the first transistor referred to as a pre-driver. High current gain is needed in the output stage so that it does not load the driving circuits and create distortion or even a complete lack of adequate current [1].

A Darlington connection or simply a 2 EF output stage, is illustrated in Figure 2(a). The bias spreader is illustrated with 4 diodes that create 4 V_BE of voltage needed to drive the 4 V_BE of the 4 output transistors effectively connected in series. A Triple EF can be used to achieve very high current gain so as to provide a very light load on the driving circuits. Figure 2(b) shows an output Triple with a six-V_BE bias spreader.

With a transistor current gain of 50, a 2-EF design provides a total current gain of 2,500, while a Triple provides a current gain of 125,000. In the former case, the driving circuit sees a load that looks like 20,000 Ω when driving an 8-Ω load, requiring 2 mA to drive an 8-Ω load with 5 A at 100 Watts. This is a significant fraction of the typical current of 10 mA that a drive circuit can supply. The drive circuit sees a load of 1000,000 Ω when driving an 8-Ω loudspeaker load with a Triple, requiring only 5 μA to drive the 8-Ω load to 100 Watts. The current gain of power output transistors falls from their nominal value when driving high current. This effect is reduced when several output transistors are connected in parallel so that they share the load.

The simple complementary output stages described above were rarely used in the early days of solid-state power amplifiers. Good silicon PNP power transistors were largely unavailable in the late 1960s and early 1970s. Those that were available were slow and costly. For this reason, output stages back then used a circuit with an NPN power transistor to emulate a PNP power transistor for the bottom half of the output stage. Two transistors, a PNP followed by an NPN formed a functional approximation to a complementary output stage called a quasi-complementary output stage. As shown in Figure 3, this arrangement is also called a Sziklai pair, after George C. Sziklai, of RCA. It is also referred to as a complementary feedback pair (CFP). The asymmetry of the quasi-complementary output stage caused increased distortion and the CFP at the bottom sometimes had stability problems because of the local negative feedback that it used to emulate a PNP power transistor.

6. Crossover Distortion

Crossover distortion occurs in the quasi-complementary output stage due to the asymmetry of the positive and negative halves of the output stage. The Harman-Kardon Citation 12 and the Crown DC300 were examples of amplifiers that employed a quasi- complementary output stage. The large Phase Linear 700 was rated at 350 Watts per channel, had 100-V power supply rails and used 5 output transistor pairs in parallel in a quasi-complementary output stage. It required high-voltage transistors and used those found in TVs at the time.

7. Safe Operating Area

The safe operating area (SOA) limitations of the output transistors available in the late 1960s and early 1970s often required the use of aggressive protection circuits called V-I limiters. They limited the output transistor current if an unsafe condition occurred. They also provided short-circuit protection. If the protection circuit engaged, some caused the output stage to turn into a constant current source with high impedance, rather than its normal function as a voltage source with low output impedance. This was especially problematic if the load was inductive, and some loudspeakers can look inductive in certain frequency ranges, especially at low frequencies. This could hurt the sound if the amplifier entered the protection mode due to an unsafe combination of current through the transistor and voltage across the transistor. In some cases, a flyback voltage spike could occur that could damage a tweeter.

Figure 4 illustrates the safe operating area boundaries for an output stage power transistor. The device can operate in the region bounded by its allowed power dissipation up to a certain collector-emitter voltage (V_CE). At that point, its safe operating current decreases more rapidly with increased V_CE than the power dissipation boundary.

8. Output Stage Biasing

In a push-pull class AB output stage, the handoff from the top to bottom transistors as the output current goes through zero is important and determines crossover distortion. The idle bias current keeps the gain in the handoff region from being too small and leaving a gap in gain, or being too large and providing too much overlap in top and bottom transistor gain contribution to causing too much gain, causing in the limit what is called gm doubling [1].

This optimum bias current is hard to maintain because the power transistor’s V_BE for a given operating current decreases with increasing temperature. A significant issue here is the temperature of the heat sink to which the output transistors are mounted. The heat sinks heat up when music is being played. Power dissipation in the output transistors of an amplifier changes with the amount of power being delivered to the load. This changes the temperature of the transistors and thus the amount of bias current at which they are operating. It is the job of the bias spreader to change its voltage to track the temperature changes of the transistors to compensate for these bias changes, but the temperature of the heat sink changes very slowly, so dynamic temperature tracking is difficult.

9. Thermal Runaway

Transistor collector current has a strong positive temperature coefficient, making achieving bias current stability tricky. Thermal runaway can occur when the output transistors heat up too much before the temperature of the heat sink can respond and cause the temperature change to be compensated for.

Higher-powered amplifiers usually have multiple output transistors in parallel to handle the higher currents. In a paralleled group of transistors that are not well-matched, one transistor may conduct more current than the other(s). As a result, it will dissipate more power and heat up more. This in turn reduces its V_be and causes it to conduct still more current compared to the other transistor(s) in parallel with it. This is called current hogging and can lead to thermal runaway. The hot gets hotter, just as the rich get richer.

Current hogging can also occur within a transistor, where one area tends to conduct more current than another due to local process or geometry differences. This can lead to current hogging that can lead to local hot spots. A local form of thermal runaway can thus occur, leading to the destruction of the device. This is part of the mechanism that causes a second breakdown that results from the safe operating area being violated.

10. Power Dissipation and Safe Operating Area

Rated power dissipation of a power transistor mainly depends on maximum junction temperature, which in turn depends on the thermal resistance from the junction to the case, assuming that the case is held at room temperature. However, heat sinking is required to keep the case temperature from rising. At the end of the day, the usable power dissipation is governed by the total thermal resistance from the junction to the ambient air.

However, the usable power dissipation falls at higher voltages. A voltage versus current envelope is thus formed that is the safe operating area (SOA), as shown in Figure 4. For low and medium voltages, this area is bounded by a constant power dissipation line. At higher voltages, the boundary is a steeper line where the current falls more quickly than the constant power dissipation line. This is due to what is called secondary breakdown.

11. Ring Emitter Power Transistors

Ring emitter power transistors emerged in the 1980s and provided higher speed and a larger safe operating area. They were constructed in such a way that they essentially acted like multiple small devices connected in parallel. Small devices tend to be faster and much less prone to local current hogging and thermal runaway. The ring emitter transistor was introduced. The current gain of these devices was also better maintained at high currents. The use of ring emitter and similar devices led to significant improvements in amplifier performance by combining significantly higher speed with improved robustness.

12. ThermalTrak^® Power Transistors

So-called ThermalTrak^® power transistors were introduced by ON Semiconductor in the late 1990s. They were designed to reduce the problems with thermal tracking of output stage bias to reduce departures from the optimum bias current as output transistor temperatures changed with warm-up and signal levels. This was accomplished by including a temperature-sensing diode in the same package as the output transistor so that faster response to changes in transistor junction temperature could be tracked [1]. That diode could be used in the bias spreader to compensate more quickly for temperature changes. Without this approach, thermal tracking depended entirely on the very slow changes in temperature of the heat sink. The use of these transistors led to better thermal stability, which in turn reduced the chances of thermal runaway. It also resulted in reduced distortion because it was possible to keep the output stage bias current closer to its optimum value for minimum crossover distortion.

13. The Junction Field Effect Transistor (JFET)

The junction field effect transistor (JFET) operates by a different principle than bipolar transistors. A JFET includes a source, drain, and gate. These are analogous to the emitter, base, and collector of a BJT. The drain current I_d in a field effect transistor is controlled by an electric field created by the gate-to-source voltage V_GS. A voltage thus controls the drain current, so the JFET is a transconductance device. In the normal operation of a BJT, the base-emitter junction is forward-biased, so some base current flows, leading to finite current gain. The normal operation of a JFET is just the opposite; the gate junction is reverse-biased, so no gate current flows.

Recall that a current cannot flow through silicon if there are no charge carriers, as in the depletion region for a diode. Recall also that the size of a depletion region can be influenced by an electric charge, and that a reverse-biased junction includes a depletion region whose size increases as the reverse bias increases with a stronger electric field.

Consider a diode junction that is created by placing a small layer of p-type material on top of an n-type bar of silicon that normally conducts current and is basically a resistor. This is an n-channel JFET. The region under the junction is called a channel through which the current will flow. With no voltage on the gate with respect to the source (V_GS), the JFET will act like a simple resistor between the source and drain and will conduct current. However, the junction formed between the p-type gate and the n-type channel will create a depletion region in the doped silicon beneath the gate. The JFET acts based on the depletion of charge carriers in the channel that would normally conduct current, so the JFET is referred to as a depletion-mode device.

JFETs are used both as discrete transistors and in JFET operational amplifiers in audio circuits. Their extremely high input impedance makes them ideal for many applications and some people prefer their sound. Electret and condenser microphones depend on JFETs for their first amplification stage. JFETs do not draw input bias current as BJTs do. Such current can cause what is called shot noise in high-impedance circuits. In such circuits, they can have lower noise than BJTs [2].

14. The Metal Oxide Field Effect Transistor (MOSFET)

The metal oxide field effect transistor (MOSFET) is the basis for virtually all modern digital integrated circuits. In earlier days, it was also called an insulated gate field effect transistor (IGFET), a more general term that allows for the use of gates that are not made of metal. The gate is insulated from the channel by a thin layer of non-conductive silicon dioxide.

An electric field can influence the presence or absence of holes and electrons and their position in a doped semiconductor. Most MOSFETs operate based on an electric field creating a region that has charge carriers. This is the opposite of operation based on the depletion of charge carriers as in a JFET. Such a MOSFET is referred to as an enhancement-mode device. In this case, a forward bias on the gate “inverts” the polarity of the doped silicon below the gate.

Consider an n-channel MOSFET. If the source and drain regions are n-type silicon and the region below the gate is doped to be a p-type region, no current will flow. It will be like two back-to-back diodes facing in opposite directions. If a forward bias is applied to the gate, the resulting electric field will invert the region below the gate to become of the same polarity as the source and drain, thus permitting current to flow.

Although the MOSFET is called a metal oxide semiconductor, meaning a metal gate is placed over an oxide insulator which is on top of the silicon channel, the conductive gate material need not be metal. While in the early days, it was metal, virtually all modern MOSFETs employed doped conductive polysilicon as the gate element. However, there has been a return to metal gates to achieve lower resistance and higher speed, and for compatibility with newer so-called high-k dielectrics that have replaced silicon dioxide as the gate insulator in process nodes below about 45 nm. This part of the discussion is beyond the scope of this article.

Some MOSFETs operate on the basis of an electric field depleting an area of charge carriers, and these are called depletion-mode devices. In this case, the source and drain might be made of heavily doped n-type silicon and the channel region below the gate will be made of lightly doped n-type silicon of the same polarity. Thus, current can flow in this device with no voltage applied to the gate. However, a reverse bias applied to the gate will create a depletion region, just as in a JFET, that will ultimately prevent conduction if the reverse bias voltage is sufficient. Both enhancement and depletion- mode MOSFETs were often used in early very large-scale (VLSI) MOSFET logic circuits that had only n-channel transistors.

15. CMOS Integrated Circuits

There are also p-channel MOSFETs. When an IC is made that contains both n-channel and p-channel MOSFETs, this is referred to as a CMOS IC (complementary metal oxide semiconductor). The CMOS process is the technology by which virtually all logic circuits are implemented today. The CMOS process is more complex and takes a longer time to mature. While small-scale CMOS ICs were available in the 1970s, it was not until the early 1980s that the CMOS technology matured to the point where large-scale logic circuits could be implemented with it.

The first microprocessors made in CMOS emerged in the 1980s with feature sizes (lithography) in the 1–3-micron range. Today’s smallest feature sizes are less than 10 nanometers. In very rough terms, the number of transistors that can fit in a given size chip (die) changes with the square of the feature size. From 3 microns to 3 nanometers is a factor of 1000, so a very advanced die with 3-nanometer features can contain 1 million times as many transistors as one with 3-micron features. This reduction in feature size has occurred over about 40 years, from 1984 to 2024.

16. Moore’s Law and Scaling

Moore’s Law is an observation that the number of transistors doubles about every two years. As a very rough example, over 40 years there have been 20 such doublings. The ratio is thus 2 to the 20^th power. The tenth power of 2 is about 1000, so 2 to the 20th power is about 1 million. This very roughly assumes the same size die.

Processes with a given feature size are often referred to as process nodes. In the 1980s, process nodes ranged from 3 microns down to about 1 micron. Many of today’s most advanced process nodes are below 10 nanometers. As node sizes decreased, so did the operating voltages, which have progressed from 5 V to below 1 V. This change in both process feature size and operating voltage is a characteristic of what is called scaling.

17. Impact of VLSI on Audio

VLSI (Very Large Scale Integration) enabled the implementation of digital signal processing (DSP) in consumer electronics. DSP chips made it more affordable, just as VLSI enabled the implementation of powerful computer chips used in PCs. Compact Disc players were among the first audio products to benefit from DSP. Modern over- sampled sigma-delta digital-to-analog (D2A) converters and analog-to-digital (A2D) converters rely heavily on embedded DSP.

18. Power MOSFETs

The late 1970s saw the introduction of power MOSFETs. They are used in linear (analog) power amplifier output stages, as switches in Class D amplifier output stages, and in switching power supplies. One of their most important characteristics is their on resistance and transconductance. They are characterized by high immunity to secondary breakdown, with large SOA. They are much faster than BJTs, being a unipolar device with no minority carrier charge storage. Their equivalent f_T (speed) is usually in the hundreds of MHz, as compared with tens of MHz for BJTs.

For class AB audio power amplifiers, power MOSFETs are an alternative to bipolar junction transistors (BJT) for the output stages where high current must be provided. Their near-infinite current gain, high speed, and relative immunity from secondary breakdown make them attractive for this application. Power MOSFETs are available in n-channel and p-channel versions, analogous to NPN and PNP bipolar transistors. The current conduction in a MOSFET is controlled by the electric field created by a forward voltage applied between the insulated gate and the source (V_GS).

MOSFETs used in power amplifier output stages are usually configured as source followers, analogous to the emitter follower configuration often used in BJT designs. The source follower stages thus ideally operate at unity gain. Compared to BJTs, MOSFETs require a larger turn-on voltage at the gate than BJTs require at the base. They also have less transconductance than a BJT when operating at the same current. This usually means that the MOSFET source follower has a gain that is further less than unity than that of a BJT emitter follower. This can result in what is called transconductance droop in the crossover region, a source of crossover distortion [1]. When fully turned on with a large forward gate-source voltage (V_GS), a MOSFET acts as a low-value resistor whose value is referred to as R_{DS_on}. A smaller value of R_{DS_on} is desirable because it determines the amount of voltage drop between the power rail and the loudspeaker when maximum power is being delivered.

There are two types of power MOSFETs. One is called a lateral MOSFET because its current flow is largely horizontal across the die. Lateral power MOSFETs are used almost exclusively in linear class AB power amplifiers. The other is called a vertical MOSFET because its current flow is mainly in the vertical direction. The different constructions of the lateral and vertical power MOSFETs are illustrated in Figure 5 [3].

Vertical MOSFET power transistors called HEXFETs by International Rectifier Corp. emerged in the late 1970s. Their use was largely driven by switch mode power supplies (SMPS) which mainly used n-channel devices. They are also the primary power devices in class D audio amplifiers. However, p-channel devices were also made available, making them attractive for use in class AB audio power amplifiers.

19. Lateral MOSFET Power Amplifiers

Lateral MOSFET power amplifiers generally do not suffer from second breakdown (SOA) limitations. They require little or no protection circuitry – often just a speaker fuse. However, they require higher power supply voltage for a given output power due to the need for several volts of forward gate bias at high current.

Lateral MOSFET output stages have a smoother class AB crossover transition. This does not necessarily mean less distortion – just lower-order distortion, which is often preferred. They can be operated at a higher quiescent bias current and achieve a wider crossover region without causing higher distortion. A larger class A region (when both top and bottom output transistors are on) can lead to better sound, especially at low listening levels. Audio amplifiers based on lateral MOSFETs were popularized by the very successful Hafler amplifiers like the DH-220.

Lateral power MOSFETs have a negative temperature coefficient (TC) of V_GS at low values of drain current, meaning that drain current increases with temperature for a given value of V_GS. However, as the drain current becomes higher, this value becomes smaller and eventually passes through zero and goes positive. Lateral MOSFETs have a fairly low current (about 150 mA) at which the temperature coefficient of V_GS for a given drain current is zero, making biasing more stable. This is in contrast to BJTs where the TC of V_BE is always significantly negative, meaning that collector current increases with temperature for a given value of V_BE. This is accompanied by higher on resistance (R_{DS_on}), however. Lateral MOSFETs thus have good thermal stability and moderate R_{DS_on}.

20. Vertical MOSFET Power Amplifiers

Vertical MOSFETs were also introduced and used in power amplifier output stages. They were even faster than lateral MOSFETs and were capable of delivering higher current, largely due to their significantly smaller value of R_{DS_on}. However, they require a larger turn-on voltage than lateral MOSFETs, often on the order of 4 V [3]. Vertical MOSFETs are capable of delivering very high current and feature the absence of secondary breakdown.

The TC of V_GS for vertical MOSFETs does not go through zero until significantly higher values of drain current, in the Ampere range. While still better than that for BJTs, their temperature stability of bias current is not as good as that for laterals, meaning that bias circuits can be a bit more complex. They often require some modest temperature compensation by mounting a diode or transistor on the heat sink.

21. Switched-Mode Power Supplies (SMPS)

Switched-mode power supplies (SMPS) have had a tremendous impact on the electronics industry, making equipment smaller, lighter, more affordable, and less power-hungry.

This applies to audio products as well. SMPS are made possible by modern power MOSFETs.

Conventional linear power supplies are heavy and bulky due to the low line frequency of 50 Hz or 60 Hz. The power transformer is heavy and expensive. Large reservoir capacitors are required because of the low ripple frequency (50 Hz – 60 Hz AC).

Conventional supplies have poor power factor due to their highly non-linear impulsive input current. The rectifier pulses of many tens of Amperes under load cause voltage drops in the resistive mains supply that reduce the amount of power that can effectively be extracted from the mains.

The use of switched-mode power supplies in all kinds of audio equipment has become much more widespread in recent years. In power amplifiers, they are attractive because they are smaller and lighter than conventional linear power supplies that employ large power transformers and large electrolytic reservoir capacitors. This is especially the case in professional audio power amplifiers. In many cases, they are also much better regulated than traditional linear power supplies. Their operation at high frequencies greatly reduces the size and weight of passive components and also reduces power supply voltage ripple. Buck, boost, Cuk, flyback, and other variations of SMPS are covered in detail in Ref. 1. Power factor correction (PFC) is becoming increasingly important as more emphasis is placed on Green energy. PFC circuits are made possible by SMPS techniques.

SMPS depend on high-speed on-off switching, and that is made practical by the use of fast vertical power MOSFETs that can deliver high current. Power transistors dissipate very little power when they are either on or off. When on, they have high current running through them but very low voltage across them, so little power is dissipated, and little heat is generated. Similarly, when they are off, they have high voltage across them but virtually no current running through them, once again resulting in low power dissipation.

For this reason, fast vertical MOSFETs are used in switched-mode power supplies. It is important that the transistors have very little resistance when they are on since any resistance causes power dissipation, increased temperature, and loss of efficiency. In modern SMPS, so-called TrenchFET MOSFETs are being used because they have very small resistance when they are on. TrenchFETs are a relatively new type of power MOSFET that provides even lower on resistance than planar MOSFETs by employing a structure with a 3-dimensional gate cross-section. Rather than being planar, the gates are implemented in the sidewalls of trenches in the MOSFET structure. An enormous number of MOSFET structures are created along the side walls of many trenches. This leads to very high gate density and very low R_{DS_on}.

22. Class D Power Amplifiers

Throughout the 2000s class D power amplifiers have come into greater use. Early Class D amplifiers suffered from poor distortion performance, but matters have improved dramatically since the late 1990s. Their sonic performance has now improved to the point where they are competitive with traditional class AB power amplifiers while having enormous advantages in size, weight, and efficiency, just as do switch-mode power supplies.

Like switch mode power supplies, they are based on very fast on-off switching of modern power MOSFET transistors. It is important that the transistors have very little resistance when they are on since any resistance causes power dissipation, increased temperature, and loss of efficiency.

Class D amplifiers operate on an entirely different principle than linear class AB power amplifiers. The output stage in a class D amplifier comprises MOSFET switches that are either on or off. The top switches apply the positive supply to the output for one brief period and then the bottom switches connect the negative supply rail to the output for the next brief period. The process is then repeated indefinitely at a high switching frequency. This results in a square wave at the output. If these two intervals are the same, the net output is zero. If the first is longer than the second, the output has a net positive value. If the first is shorter than the second, the output has a net negative value. A low-pass filter extracts the average value to drive the loudspeaker. The cutoff of the LPF often lies in the 30 kHz to 60 kHz range, but higher low-pass frequencies are also used.

This process is referred to as pulse width modulation (PWM). These switching intervals alternate at a high frequency, often in the range of 500 kHz. Thus, the average value of the square wave drives the load. Because the switches are either on or off, they dissipate little power. Virtually all of the input power from the power rails is transferred to the load, so efficiency is very high, and power dissipation is very low. Efficiency of 85% to 95% is common. Physically small amplifiers that run cool are the result.

A big challenge in class D amplifiers is the proper driving of the output stage switches so that the on and off timing intervals accurately reflect the input signal. Numerous schemes are used to address this issue. All of these approaches depend on fast, clean switching. It is especially important to avoid both transistors being on at the same time, called totem- pole conduction, also called shoot-through current. This can happen due to timing errors in the driving circuits or to slow turn-off of the MOSFETs. For this reason, a small amount of dead time is included between the on times of the top and bottom switches. With too little dead time, too much shoot-through current results. With too much dead time, a form of crossover distortion can result [1].

Faster MOSFET transistors permit the use of less dead time and allow higher switching frequencies to be used. Higher switching frequencies usually result in better-sounding amplifiers. Among other things, higher switching frequencies allow the output low-pass filters to have higher cutoff frequencies, further above the audio band.

Although the D in class D does not stand for digital, the implementations of class D amplifiers are indeed moving more toward digital, and in fact, some approaches involve direct digital conversion from PCM input streams to class D audio outputs.

23. Gallium-Nitride (GaN) Transistors (GaN FETs)

With newer semiconductor advances, extremely fast Gallium Nitride field effect transistors (GaN FETs) have emerged as power-switching transistors, first used in SMPS and now in class D power amplifiers.

Gallium Nitride is a III-V compound semiconductor that was first used to make blue LEDs. Gallium Nitride enhancement-mode MOSFETs are relatively new and are capable of much higher speeds than silicon MOSFETs. GaN FETs are up to 10 times faster than silicon MOSFETs in these applications. This speed is called the rise time, which is the time taken for the collector current to reach from 10% of its initial value to 90% of its final value. This allows higher switching frequencies that greatly benefit both switched-mode power supplies (SMPS) and class D power amplifiers. GaN FETs are also capable of higher breakdown voltages and can operate at higher temperatures than silicon MOSFETs.

GaN transistors are much more efficient in switching applications and thus create less heat. At the same time, they can operate at significantly higher temperatures, so they provide a double benefit regarding reduced cooling requirements. Because they can operate at higher frequencies, the size of related passive devices, like capacitors and inductors in switching circuits can be smaller. They are now the premier switching device for use in switched-mode power supplies.

GaN FETs, being very well-suited to high-speed switching applications, are an outstanding technology for class D power amplifiers, where silicon MOSFETs are commonly used. The key is that the faster switching afforded by GaN FETs allows significantly higher switching frequencies to be used in class D amplifiers, leading to lower distortion and wider frequency response. Emerging class D amplifiers that employ GaN FETs are quickly becoming associated with better sound.

24. Summary

As transistors continuously advance from the past to the future, their contribution to improved audio will also continue. Recent advances in the use of new transistors like GaN FETs have already had a strong positive influence on switched-mode power supplies and class D amplifiers.

References

1. Bob Cordell, Designing Audio Power Amplifiers, 2^nd edition, Routledge, 2019.

2. Bob Cordell, Designing Audio Circuits and Systems, Routledge, 2024.

3. Bob Cordell, A MOSFET Power Amplifier with Error Correction, Journal of the Audio Engineering Society, January 1984. Available at cordellaudio.com.

Figures

1. Simple Push-pull Output Stage
2. Darlington (a) and Triple (b) Output Stages
3. Quasi-complementary Output Stage
4. Power Transistor Safe Operating Area
5. Lateral and Vertical Power MOSFETs

Bio

Bob Cordell is an Electrical Engineer who has worked in telecommunications for many years, having started at Bell Labs in Holmdel, NJ in the Seventies. He was in management for 22 of those years. His activities have included integrated circuit design, optoelectronics and optical systems design, and digital television encoding and decoding circuits. He holds 15 patents. Bob has been an audio hobbyist since his teens and has published numerous articles on power amplifier design and distortion measurement. He is a member of the Review Board of the Audio Engineering Society. His current interests include the design of MOSFET power amplifiers with error correction and the design of active loudspeakers. Bob lives in Holmdel, NJ, and can be reached through his website: https://www.cordellaudio.com/