Discrete Power Amplifier Design

Power amplifiers represent the final stage in audio and signal processing systems, converting low-power signals into outputs capable of driving loudspeakers, motors, or other high-current loads. Unlike small-signal amplifiers where thermal effects and power dissipation are secondary concerns, power amplifier design requires careful attention to efficiency, heat management, safe operating limits, and protection against fault conditions that could destroy expensive output devices or connected equipment.

Discrete power amplifier design using individual transistors offers advantages over integrated solutions in high-power applications: the ability to select optimal devices for specific requirements, easier thermal management through physical separation and dedicated heatsinking, straightforward repair and component replacement, and the flexibility to implement custom protection schemes. This article explores the fundamental amplifier classes, bias optimization techniques, thermal considerations, and protection circuits essential for building reliable high-power output stages.

Power Amplifier Classifications

Power amplifiers are classified according to their conduction angle, the portion of the input signal cycle during which the output devices conduct current. This classification determines the fundamental trade-off between linearity and efficiency that governs all power amplifier design.

Class A Operation

In Class A operation, the output device conducts current throughout the entire 360 degrees of the input signal cycle. The device is biased to a quiescent current level that exceeds the peak signal current, ensuring the device never turns off regardless of signal amplitude. This operating mode provides the lowest distortion but also the lowest efficiency.

The theoretical maximum efficiency of a Class A amplifier with resistive load is 25 percent, achieved only when the output swing equals the full supply voltage. With an inductor-coupled or transformer-coupled load that allows the output to swing above the supply voltage, efficiency can reach 50 percent. In practice, real Class A amplifiers achieve 15 to 25 percent efficiency under typical operating conditions.

Class A amplifiers dissipate maximum power at zero signal, when the full quiescent current flows through the output device with half the supply voltage across it. Power dissipation actually decreases as signal amplitude increases, since during portions of the cycle the output voltage approaches either supply rail, reducing the voltage across the output device. This thermal characteristic simplifies heatsink design since worst-case dissipation occurs with no signal applied.

Class B Operation

Class B amplifiers use two output devices, each conducting for exactly half the input cycle, or 180 degrees. When the input signal is positive, one device conducts and drives the load; when negative, the complementary device takes over. Neither device conducts during the opposite half-cycle, eliminating the quiescent current that limits Class A efficiency.

The theoretical maximum efficiency of Class B operation is 78.5 percent, a substantial improvement over Class A. This efficiency gain comes at the cost of crossover distortion, a form of non-linearity that occurs as conduction transfers from one device to the other near zero signal crossing. The small but finite turn-on voltage required by transistors creates a dead zone where neither device conducts, producing characteristic steps or glitches in the output waveform.

Pure Class B operation without crossover distortion compensation produces objectionable distortion, particularly noticeable in audio applications at low signal levels where the crossover region represents a significant portion of the signal amplitude. For this reason, Class B amplifiers without bias optimization are rarely used in precision applications.

Class AB Operation

Class AB represents a practical compromise between Class A and Class B, combining reasonable efficiency with low distortion. Each output device conducts for more than half but less than the full cycle, typically between 181 and 200 degrees. A small quiescent current flows through both devices when no signal is present, ensuring smooth crossover as conduction transfers between devices.

Class AB efficiency typically ranges from 50 to 70 percent at full power, decreasing at lower output levels as the proportion of quiescent dissipation increases. The quiescent current is set just high enough to eliminate crossover distortion without unnecessarily sacrificing efficiency, a balance achieved through careful bias circuit design.

Most discrete power amplifiers for audio and general-purpose applications operate in Class AB, as this mode offers the best combination of linearity, efficiency, and circuit simplicity for moderate to high power levels.

Class A Amplifier Design and Efficiency

Despite their low efficiency, Class A amplifiers remain important in applications demanding the highest linearity, including high-end audio, instrumentation, and radio frequency power amplification where distortion products must be minimized.

Single-Ended Class A Topology

The simplest Class A power amplifier uses a single transistor with the load connected between collector and supply or between emitter and ground. The transistor operates as a controlled current source, varying its current in response to the input signal while maintaining continuous conduction.

Design considerations for single-ended Class A stages include:

Operating point selection: The quiescent collector current must equal or exceed the peak signal current to prevent cutoff. Setting the quiescent voltage at half the supply voltage provides maximum symmetric output swing.
Load line analysis: The load resistance determines the slope of the load line on the transistor output characteristics. Maximum power transfer occurs when the load resistance equals the ratio of quiescent voltage to quiescent current.
Bias stability: The quiescent point must remain stable despite temperature variations that affect transistor parameters. Emitter degeneration and temperature-compensated bias networks help maintain consistent operation.
Power dissipation: The transistor must be rated for worst-case dissipation, which occurs at zero signal. This equals the quiescent current multiplied by the quiescent collector-emitter voltage.

Push-Pull Class A

Push-pull Class A configurations use two devices that alternately push and pull current through the load, but both devices remain conducting throughout the signal cycle. This arrangement cancels even-order distortion products and doubles the available output power for a given supply voltage and load resistance.

In a push-pull Class A amplifier, the quiescent current through each device equals the peak load current. When one device increases its current to drive the load in one direction, the other device decreases its current by an equal amount. The load current is the difference between the two device currents, allowing output current swing of twice the quiescent value while maintaining Class A operation.

Efficiency Enhancement Techniques

Several techniques can improve Class A efficiency without compromising linearity:

Sliding bias: Dynamically adjusting the quiescent current based on signal amplitude allows the amplifier to operate with lower dissipation during low-level signals while maintaining Class A operation at higher levels.
Current dumping: A parallel Class B stage carries most of the load current while a small Class A amplifier provides error correction, combining high efficiency with Class A linearity.
Supply modulation: Varying the supply voltage to track the signal envelope reduces voltage across the output devices at low signal levels, decreasing dissipation without affecting linearity.

Class B and AB Bias Optimization

The critical challenge in Class B and AB amplifier design is establishing and maintaining the optimal bias current that eliminates crossover distortion while preserving efficiency. The bias circuit must compensate for temperature variations that would otherwise cause thermal runaway or excessive crossover distortion.

Crossover Distortion Mechanism

Crossover distortion arises from the non-zero turn-on voltage of bipolar transistors and MOSFETs. For a bipolar transistor to conduct collector current, its base-emitter voltage must exceed approximately 0.6 volts. In a complementary output stage without bias, both transistors are off when the input signal is within plus or minus 0.6 volts of zero, creating a dead zone where the output cannot follow the input.

The resulting distortion produces odd-order harmonics that sound harsh and unpleasant in audio applications. The effect is most severe at low signal levels, where the crossover region represents a significant portion of the signal amplitude. At high signal levels, the relative contribution of crossover distortion decreases, but it remains audible as a grainy or rough quality in the sound.

Voltage Bias Methods

The most common approach to eliminating crossover distortion applies a fixed bias voltage between the bases of the complementary output transistors, pre-biasing them to the edge of conduction. Several circuit topologies can generate this bias voltage:

Diode bias: Two or more series-connected diodes provide a voltage drop approximately equal to the combined base-emitter voltages of the output transistors. The diodes should be thermally coupled to the output stage to track temperature variations.
Transistor bias (Vbe multiplier): A transistor configured as an adjustable voltage reference provides precise, temperature-tracking bias. The collector-emitter voltage equals Vbe multiplied by (1 + R1/R2), where R1 and R2 form a voltage divider from collector to emitter with the base connected to their junction.
Resistor divider: Simple resistive bias is less effective because it does not track temperature variations in the output transistors, leading to thermal instability or inadequate crossover compensation.

The Vbe Multiplier

The Vbe multiplier, also called a rubber diode or bias spreader, provides the most versatile bias solution for Class AB amplifiers. This single-transistor circuit creates an adjustable voltage drop with temperature coefficient that can be matched to the output stage requirements.

The Vbe multiplier consists of a transistor with a resistor divider from collector to emitter, with the base connected to the divider junction. The circuit maintains constant voltage across itself despite variations in the current flowing through it, acting as a voltage source rather than a resistance. The voltage equals the base-emitter drop multiplied by the divider ratio:

V = Vbe times (1 + R1/R2)

Where R1 connects from collector to base and R2 connects from base to emitter. Adjusting these resistor values allows precise setting of the bias voltage. The circuit inherently tracks temperature because both the Vbe of the bias transistor and the Vbe of the output transistors decrease at approximately 2 millivolts per degree Celsius.

For optimal temperature tracking, the Vbe multiplier transistor should be thermally coupled to the output stage heatsink. This ensures the bias voltage decreases as the output transistors warm up, preventing thermal runaway while maintaining consistent crossover performance.

Optimal Bias Current Setting

The ideal quiescent current for a Class AB output stage depends on several factors:

Gm doubling region: The output stage transconductance exhibits a nonlinearity where it doubles at the crossover point in pure Class B operation. Setting quiescent current so that both devices operate in their linear region around crossover eliminates this nonlinearity.
Device characteristics: Power transistors with higher current gain require less base drive and exhibit smoother crossover behavior at lower bias currents.
Thermal considerations: Higher bias current increases idle power dissipation and heatsink requirements. The optimal bias represents a compromise between linearity and thermal design constraints.
Load impedance: Lower impedance loads require higher peak currents, and the bias current should scale accordingly to maintain smooth crossover.

A practical starting point for audio amplifiers sets the quiescent current to between 20 and 100 milliamps per output device pair, depending on power level and performance requirements. Measurement of crossover distortion while adjusting bias current allows optimization for the specific design.

Crossover Distortion Elimination

Beyond proper bias setting, several circuit techniques can further reduce or eliminate crossover distortion, improving linearity to levels approaching Class A performance while maintaining Class AB efficiency.

Negative Feedback

Global negative feedback from the output to the input stage reduces all forms of distortion, including crossover distortion, by the feedback factor. A well-designed feedback amplifier can achieve distortion levels orders of magnitude lower than the open-loop stage.

However, feedback has limitations in reducing crossover distortion:

Slew rate limitations: Sharp transitions at crossover may exceed the amplifier's slew rate capability, causing temporary loss of feedback control.
High-frequency effects: The feedback loop bandwidth limits its effectiveness at high frequencies where phase shift reduces the feedback factor.
Transient intermodulation: Rapid signal changes can momentarily exceed the feedback loop's ability to correct errors, allowing brief distortion events.

For these reasons, feedback should complement rather than substitute for proper bias optimization. The open-loop stage should be as linear as practical, with feedback providing additional distortion reduction.

Error Correction

Error correction circuits measure the difference between input and output signals and inject a correction signal to cancel distortion. Unlike feedback, which operates in a closed loop, error correction operates in feedforward, avoiding stability concerns.

One implementation uses a small, highly linear Class A amplifier to measure output stage errors. The error amplifier compares the output signal (scaled appropriately) with the input signal, amplifying any difference. This error signal modulates the output stage drive to cancel the distortion.

Error correction can achieve extremely low distortion but requires careful implementation to prevent the correction circuitry itself from introducing artifacts.

Current Feedback Output Stages

Current feedback amplifiers use output stage topology that provides inherent crossover linearization. By sensing output current rather than voltage and feeding this signal back to the input stage, the amplifier can maintain linearity through the crossover region even with minimal bias current.

The current feedback approach offers wide bandwidth and fast slew rate advantages but requires careful design to maintain stability and may exhibit different noise and offset characteristics than voltage feedback designs.

Emitter Resistors

Small resistors in series with the output transistor emitters (for bipolar devices) or sources (for MOSFETs) provide local negative feedback that linearizes the output stage. These resistors reduce the effective transconductance variation at crossover, smoothing the transition between devices.

Emitter resistor values typically range from 0.1 to 0.5 ohms for power stages. Larger values provide better linearization but increase power loss and reduce efficiency. The resistors also help with current sharing when multiple devices are paralleled and provide a convenient point for current sensing in protection circuits.

Thermal Design and Heatsinking

Power transistors dissipate significant heat that must be removed to maintain junction temperatures within safe limits. Thermal design involves selecting appropriate heatsinks, ensuring good thermal interfaces, and accounting for worst-case operating conditions.

Thermal Resistance Concepts

Heat flows from the transistor junction through the package, any insulating materials, and the heatsink to ambient air. Each element in this thermal path has a thermal resistance, measured in degrees Celsius per watt, that relates temperature rise to power dissipation:

Temperature rise = Power dissipation times thermal resistance

The total thermal resistance from junction to ambient is the sum of individual thermal resistances:

Junction to case (Rth-jc): Internal to the transistor package, specified by the manufacturer, typically 0.5 to 2 degrees Celsius per watt for power devices.
Case to heatsink (Rth-cs): Depends on mounting method and interface material. Direct contact with thermal grease achieves 0.1 to 0.5 degrees Celsius per watt; insulating pads add 0.2 to 1 degree Celsius per watt.
Heatsink to ambient (Rth-sa): Depends on heatsink size, fin design, airflow, and ambient temperature. Values range from under 1 degree Celsius per watt for large fan-cooled heatsinks to 10 or more degrees Celsius per watt for small convection-cooled types.

Heatsink Selection

Selecting an appropriate heatsink begins with calculating the required thermal resistance based on maximum allowable junction temperature, ambient temperature, and power dissipation:

Required Rth-sa = (Tj-max - Ta-max) / Pd-max - Rth-jc - Rth-cs

Where Tj-max is the maximum junction temperature (typically 150 degrees Celsius for silicon), Ta-max is the maximum ambient temperature, and Pd-max is the maximum power dissipation.

Heatsink considerations include:

Convection vs. forced air: Forced air cooling from fans can reduce heatsink thermal resistance by a factor of 2 to 5, enabling smaller heatsinks or higher power dissipation.
Mounting orientation: Natural convection heatsinks work best with fins oriented vertically to promote air circulation. Horizontal mounting may require derating.
Multiple devices: When mounting several transistors on a shared heatsink, thermal coupling between devices must be considered. Total power dissipation determines the required thermal resistance.
Altitude effects: Air density decreases at altitude, reducing convective cooling effectiveness. High-altitude operation may require larger heatsinks or forced air cooling.

Thermal Interface Materials

The interface between transistor case and heatsink significantly affects thermal performance. Air gaps, even microscopic ones, dramatically increase thermal resistance. Various interface materials fill these gaps:

Thermal grease: Silicone-based compounds filled with thermally conductive particles provide the lowest thermal resistance. They require thin, even application and can be messy to work with.
Thermal pads: Pre-formed pads of thermally conductive material offer convenience and consistent thickness but typically have higher thermal resistance than grease.
Phase-change materials: Solid at room temperature but soften when heated, conforming to surface irregularities. They combine the convenience of pads with performance approaching grease.
Insulating materials: When electrical isolation is required, mica washers, silicone pads, or ceramic insulators add thermal resistance but prevent short circuits when transistor cases are at different potentials.

Thermal Runaway Prevention

Bipolar transistors exhibit positive temperature coefficient of collector current at fixed base-emitter voltage. As the device heats up, it draws more current, generating more heat in a potentially destructive feedback loop called thermal runaway.

Several design techniques prevent thermal runaway:

Emitter resistors: Resistors in series with the emitter provide local negative feedback. As current increases, the voltage drop across the resistor reduces the effective base-emitter voltage, stabilizing the current.
Temperature-compensated bias: Mounting the bias transistor or diodes on the output stage heatsink causes the bias voltage to decrease as temperature rises, automatically reducing quiescent current.
Thermal tracking networks: Thermistors or other temperature-sensitive elements in the bias network can provide additional compensation beyond what simple thermal coupling achieves.
Current limiting: Active current limiting circuits prevent excessive current regardless of temperature, providing a safety net if thermal compensation fails.

Safe Operating Area Protection

Power transistors have operating limits beyond simple current and voltage maximums. The safe operating area (SOA) defines the combinations of voltage, current, and time for which the device can operate without damage. SOA protection circuits prevent the transistor from being driven outside these limits under any operating condition.

Understanding Safe Operating Area

The SOA is typically presented as a graph with collector-emitter voltage on the horizontal axis and collector current on the vertical axis, both on logarithmic scales. Several boundaries define the safe region:

Maximum current limit: A horizontal line at the rated maximum collector current, limited by bond wire fusing or metallization current density.
Maximum power limit: A diagonal line of constant power (V times I = constant), determined by thermal limits at the specified case temperature.
Secondary breakdown limit: For bipolar transistors, a steeper diagonal line at high voltages represents secondary breakdown, a thermal instability that can destroy the device even within normal power limits.
Maximum voltage limit: A vertical line at the rated breakdown voltage.

The safe operating area shrinks for longer pulse durations. Manufacturers specify SOA curves for various pulse widths, with DC operation being the most restrictive. Short pulses allow operation at higher power levels because the thermal mass of the device absorbs the energy before the junction reaches destructive temperatures.

Secondary Breakdown in Bipolar Transistors

Secondary breakdown is a failure mode unique to bipolar transistors that occurs when current concentrates in a small region of the die, creating a localized hot spot that leads to thermal runaway and device destruction. This phenomenon is most likely under conditions of high voltage and high current occurring simultaneously, particularly during turn-off transitions.

The physics involves current crowding at the base-emitter junction edges, where the combination of high current density and the negative temperature coefficient of Vbe creates instability. As a small region heats up, its Vbe decreases, attracting more current and accelerating the heating in a destructive feedback loop.

Secondary breakdown limits are specified on a separate reverse-bias SOA (RBSOA) curve that applies during turn-off when the base-emitter junction is reverse-biased. This limit is often more restrictive than the forward-bias SOA and must be considered when designing for inductive loads that cause high voltage during turn-off.

SOA Protection Circuits

SOA protection circuits monitor the instantaneous voltage and current in the output transistors and reduce drive if the operating point approaches unsafe limits. Several approaches are used:

Current limiting: Sensing resistors in the output stage detect excessive current and reduce base drive when a threshold is exceeded. Simple current limiters use a transistor that shunts base drive when the sensing resistor voltage reaches about 0.6 volts.
Voltage sensing: Monitoring collector-emitter voltage allows the protection circuit to reduce the current limit at higher voltages, matching the SOA boundary slope.
Combined V-I limiting: Sophisticated protection circuits implement a full SOA limit by combining voltage and current sensing. The current limit decreases as voltage increases, following the SOA boundary.
Temperature-adjusted limits: Since SOA shrinks at higher temperatures, advanced protection circuits may incorporate temperature sensing to further reduce limits when the heatsink is hot.

Implementation Considerations

SOA protection must act quickly enough to prevent damage during fast transients while avoiding false triggering during normal operation. Design considerations include:

Response time: The protection circuit must respond within microseconds to catch fast transients. Propagation delay through the protection circuit and output driver must be minimized.
Stability: Protection circuits that clamp output stage drive create feedback paths that can cause oscillation. Careful compensation ensures smooth, stable operation when protection engages.
Graceful limiting: Abrupt current limiting can cause distortion or instability. Progressive limiting that gradually reduces drive as the limit is approached provides cleaner behavior.
Fault indication: LED indicators or logic outputs that signal when protection is active help with troubleshooting and warn of abnormal operating conditions.

Output Stage Protection Circuits

Beyond SOA limiting, output stages require protection against various fault conditions that could damage the amplifier or connected equipment. A comprehensive protection scheme addresses short circuits, open circuits, excessive temperature, and DC offset faults.

Short Circuit Protection

A short circuit across the amplifier output forces the output transistors to dissipate the full supply power. Without protection, this condition destroys the output stage within milliseconds to seconds, depending on device ratings and heatsink thermal mass.

Short circuit protection approaches include:

Foldback current limiting: Rather than holding constant current into a short circuit, foldback limiting reduces the current limit as voltage drops, dramatically reducing power dissipation. The current limit might be 10 amps at full output voltage but fold back to 2 amps at near-zero voltage.
Thermal shutdown: Temperature sensors on the output transistors or heatsink trigger a complete shutdown if temperature exceeds safe limits, protecting against sustained short circuits that might eventually overwhelm current limiting.
Time-limited overcurrent: Protection circuits may allow brief current excursions above normal limits to handle transient demands, then reduce limits or shut down if the condition persists.
Auto-recovery: After a fault condition clears, the amplifier should automatically resume normal operation. Some designs use thermal hysteresis, restarting only after the heatsink cools significantly below the shutdown threshold.

Open Circuit and Load Detection

An open circuit at the amplifier output presents less immediate danger than a short circuit but may indicate a fault condition worth detecting. Some applications require verification that a load is connected before enabling the amplifier:

Load presence detection: A small DC or AC test signal measures load impedance before enabling the main output stage. This prevents damage from unexpected load conditions.
Minimum load requirements: Some amplifier designs require a minimum load for stability. Load detection can warn if the connected load is too light.
Open speaker detection: Audio amplifiers may detect open speaker wiring to warn of installation problems.

DC Offset Protection

A DC offset at the amplifier output indicates a fault condition that could damage speakers or other loads not designed for DC current. Offset protection monitors the output for DC content and shuts down the amplifier if offset exceeds safe limits:

Offset detection: A low-pass filter extracts the DC component of the output signal. A comparator triggers when this DC level exceeds a threshold, typically 1 to 2 volts.
Response time: The filter time constant must be long enough to avoid false triggering on low-frequency signals but short enough to protect the load from damaging DC exposure. Time constants of 0.5 to 2 seconds are common.
Positive and negative detection: Separate detection for positive and negative offset ensures protection regardless of fault polarity.
Latching shutdown: DC offset usually indicates a component failure requiring repair. Latching shutdown requires manual reset or power cycling, preventing repeated attempts to operate a damaged amplifier.

Thermal Protection

Thermal protection complements SOA limiting by providing backup protection if the output stage overheats despite other protective measures. Implementation approaches include:

Discrete thermal sensors: Thermistors or IC temperature sensors mounted on the heatsink monitor temperature. A comparator triggers shutdown when a threshold is exceeded.
Integrated thermal shutdown: Many power transistors include internal thermal shutdown that disables the device if junction temperature exceeds approximately 150 degrees Celsius. This provides last-resort protection but indicates inadequate heatsinking if triggered during normal operation.
Proportional derating: Rather than abrupt shutdown, some protection schemes progressively reduce output power as temperature rises, maintaining operation at reduced capability rather than complete shutdown.
Warning indication: LED indicators or fault outputs can warn of approaching thermal limits before shutdown, allowing users to reduce signal levels or improve cooling.

Speaker Protection Methods

While output stage protection focuses on protecting the amplifier from load faults, speaker protection focuses on protecting the load from amplifier faults. Speakers are expensive and can be damaged by excessive power, DC offset, or ultrasonic oscillation.

Output Relay Protection

An electromechanical relay in series with the speaker connection provides complete isolation during fault conditions and prevents potentially damaging turn-on and turn-off transients from reaching the speakers:

Turn-on delay: The relay closes only after the amplifier has stabilized, typically 1 to 3 seconds after power application. This prevents power supply charging transients and initial oscillations from reaching the speaker.
Turn-off muting: When power is removed, the relay opens immediately (before the amplifier output becomes uncontrolled), muting any discharge transients.
DC offset disconnection: If DC offset is detected, the relay opens to protect the speaker from the damaging DC current.
Relay selection: The relay must handle the full output current without significant contact resistance. Contacts rated for 10 or more amps at low voltage are appropriate for most applications. Silver alloy contacts resist oxidation and maintain low resistance over time.

Electronic Muting

Electronic muting circuits can provide faster response than mechanical relays and avoid the reliability concerns of electromechanical devices. MOSFET or solid-state relay switches in the output path can mute the signal within microseconds:

Series MOSFET: A low-resistance MOSFET in series with the speaker provides electronic switching. The MOSFET on-resistance must be low enough to avoid significant power loss or damping factor degradation.
Parallel shunt: A low-resistance path across the output can short-circuit the signal during fault conditions, providing some protection without breaking the current path.
Combination approach: Electronic muting for fast response combined with a relay for complete isolation provides both speed and complete protection.

Overcurrent Detection

Monitoring output current can detect overload conditions that might damage speakers even if within the amplifier's capability:

Peak detection: Detecting peak currents above a threshold triggers warning or protection. Speakers have limited thermal mass and can be damaged by brief high-power transients.
Average power calculation: Long-term average power calculation better represents the thermal stress on speaker voice coils. Protection based on average power prevents sustained overload while allowing transient peaks.
Frequency-dependent limiting: Low frequencies cause maximum voice coil heating. Some protection schemes apply frequency-weighted limiting that restricts low-frequency power more than high-frequency content.

Subsonic and Ultrasonic Filtering

Frequencies outside the audio range can damage speakers without producing useful sound:

Subsonic filtering: A high-pass filter (typically 20 Hz or higher) removes infrasonic content that wastes amplifier power and causes excessive speaker cone excursion without producing audible output.
Ultrasonic detection: High-frequency oscillation, often caused by amplifier instability, can burn out tweeters before the problem is audible. Detection of significant ultrasonic content triggers shutdown or muting.
Slew rate limiting: Excessive slew rate indicates potential ultrasonic content or clipping. Slew rate detection can trigger protective action.

Parallel Device Current Sharing

High-power amplifiers often require multiple output transistors in parallel to achieve the necessary current handling capability. Ensuring equal current sharing among parallel devices is critical to prevent one device from carrying more than its share of the load and failing prematurely.

Causes of Current Imbalance

Several factors cause unequal current sharing among parallel transistors:

Vbe variations: Bipolar transistors exhibit manufacturing variations in base-emitter voltage. A device with lower Vbe at a given current will conduct more current when connected in parallel with higher-Vbe devices.
Thermal effects: The device initially carrying more current heats up faster, reducing its Vbe and further increasing its current share, potentially leading to thermal runaway of that device while others remain cool.
Layout asymmetries: Unequal trace resistances, thermal coupling, or physical placement can cause imbalanced operation.
Beta variations: Devices with higher current gain require less base current for the same collector current, affecting current distribution in base-driven parallel configurations.

Emitter Ballast Resistors

The most effective and widely used technique for current sharing is individual emitter resistors in series with each output transistor. These ballast resistors provide local negative feedback that forces current sharing:

When a device tries to conduct more current, the voltage drop across its emitter resistor increases, reducing the effective base-emitter voltage and limiting further current increase. The voltage drop required for effective ballasting must be significant compared to device Vbe variations, typically 100 to 500 millivolts at full current.

Design considerations for emitter ballast resistors include:

Resistance value: Higher resistance provides better current sharing but increases power loss and reduces efficiency. Values of 0.1 to 0.5 ohms are common for high-power stages.
Power rating: The resistor must dissipate its share of the output current squared times its resistance. Non-inductive types are preferred to avoid high-frequency impedance variations.
Matching: Resistors should be matched to within a few percent. Using the same resistor type and value from a single manufacturing lot helps ensure consistency.
Physical placement: Mount resistors close to their associated transistors for effective local feedback and convenient current sensing.

Thermal Coupling

Mounting all parallel devices on a common heatsink promotes thermal equilibrium. If one device runs hotter than others, heat conduction through the heatsink warms the other devices, partially compensating for the imbalance:

Single heatsink: All parallel devices should mount on the same heatsink, preferably with symmetric placement.
Intimate thermal contact: Using a single large heatsink rather than individual small heatsinks promotes thermal sharing.
Matched thermal interfaces: Use identical mounting hardware and thermal interface materials for all devices.

Device Matching

Selecting matched devices reduces initial current imbalance before temperature effects become significant:

Vbe matching: Sort devices by Vbe at a reference current. Matching to within 10 millivolts significantly improves current sharing.
Beta matching: Matching current gain ensures equal base current distribution. Less critical than Vbe matching for emitter-ballasted configurations.
Complementary pairs: For push-pull stages, match each NPN device with a corresponding PNP device of similar characteristics.
Manufacturer sorting: Some manufacturers offer matched pairs or sets of power transistors sorted for critical parameters.

Active Current Sharing

For the most demanding applications, active circuits can force precise current sharing regardless of device variations:

Current feedback: Individual current sense resistors feed back to servo circuits that adjust base drive to equalize currents.
Master-slave configuration: One master device sets the current reference; slave devices track the master through feedback loops.
Current mirrors: Using current mirror configurations to distribute base drive ensures proportional current sharing based on emitter area ratios.

Active current sharing adds complexity and potential failure modes. For most applications, properly designed passive ballasting with emitter resistors and thermal coupling provides adequate current sharing with simpler, more reliable circuits.

Design Example: 100-Watt Class AB Amplifier

This section outlines the design process for a practical 100-watt Class AB power amplifier using discrete bipolar transistors, illustrating the application of the principles discussed throughout this article.

Specifications and Requirements

Target specifications for the design example:

Output power: 100 watts RMS into 8 ohms
Frequency response: 20 Hz to 20 kHz within 0.5 dB
Total harmonic distortion: less than 0.1 percent at full power
Supply voltage: plus and minus 45 volts DC
Input sensitivity: 1.5 volts RMS for full output

Output Stage Design

For 100 watts into 8 ohms, the peak output voltage is approximately 40 volts and peak current is 5 amps. The plus and minus 45-volt supply provides adequate headroom for output swing plus driver and protection circuit requirements.

Output transistor selection criteria:

Voltage rating: At least 100 volts to handle the full supply swing plus transients
Current rating: At least 10 amps continuous for adequate margin
Power dissipation: At least 100 watts per device at 25 degrees Celsius case temperature
Safe operating area: Adequate SOA for the voltage-current combinations during transients

Devices such as the MJL21193 and MJL21194 (NPN and PNP complementary pair) meet these requirements with voltage ratings of 250 volts, current ratings of 16 amps, and power dissipation of 200 watts.

Bias Circuit Implementation

A Vbe multiplier provides temperature-compensated bias. The bias transistor mounts on the same heatsink as the output devices, ensuring thermal tracking. Adjustable resistor ratios allow setting the quiescent current to approximately 50 milliamps per output device, sufficient to eliminate crossover distortion while maintaining reasonable efficiency.

Driver Stage Design

The driver stage must provide sufficient base current to the output transistors at peak output. With output transistor beta of 25 at high current, peak base current requirement is approximately 200 milliamps. Medium-power driver transistors such as the BD139 and BD140 can supply this current with adequate beta to be driven by the voltage amplifier stage.

Local feedback through emitter resistors in the driver stage improves linearity and reduces the demands on the global feedback loop.

Protection Circuit Design

The protection scheme includes:

SOA protection using voltage and current sensing to implement foldback limiting that follows the output transistor SOA boundary
DC offset detection with a time constant of 1 second and threshold of 2 volts, triggering relay disconnection
Thermal protection using an NTC thermistor on the heatsink with shutdown threshold of 70 degrees Celsius
Output relay with 2-second turn-on delay and immediate turn-off muting

Thermal Design

Maximum power dissipation occurs at approximately one-third of maximum output power for Class AB operation, yielding worst-case dissipation of approximately 50 watts per channel. With ambient temperature of 40 degrees Celsius and maximum junction temperature of 150 degrees Celsius:

Required heatsink thermal resistance = (150 - 40) / 50 - 0.7 - 0.5 = 1.0 degrees Celsius per watt

A heatsink rated at 0.8 degrees Celsius per watt with forced air cooling provides adequate margin for the design.

Practical Considerations and Testing

Building and testing discrete power amplifiers requires careful attention to construction techniques and systematic verification procedures.

Layout Guidelines

Power amplifier PCB layout significantly affects performance and reliability:

Ground planes: Use solid ground planes where possible, with careful attention to preventing ground loops between input and output sections.
High-current paths: Wide traces or bus bars for output stage connections minimize resistive losses and inductance. The paths from supply decoupling capacitors to output devices should be as short as possible.
Thermal relief: Output device mounting pads need thermal relief to allow soldering without excessive heat sink cooling.
Decoupling: Local decoupling capacitors near each power stage supply connection prevent supply rail modulation.
Signal routing: Keep input and feedback signal traces short and away from high-current output paths to minimize pickup and oscillation risk.

Initial Testing Procedure

A systematic approach to initial power-up minimizes risk of damage from wiring errors or design problems:

Visual inspection: Check all connections, component orientations, and solder joints before applying power.
Resistance checks: Measure supply rail resistance to ground; shorts indicate wiring errors.
Current-limited supply: Initially power the amplifier through a current-limited supply set to 100 milliamps. Verify quiescent current before increasing the limit.
Bias adjustment: With the output disconnected, adjust bias for correct quiescent current. Monitor output transistor temperature during this process.
Small-signal testing: Apply a low-level sine wave and verify clean output with oscilloscope. Check for oscillation by examining the output with no input signal.
Power testing: Gradually increase drive level while monitoring output waveform, distortion, and temperature. Test at various load impedances within the specified range.
Protection testing: Deliberately trigger each protection circuit to verify proper operation. Test short-circuit protection with momentary output shorts at low power levels.

Troubleshooting Common Problems

Common issues in discrete power amplifier construction and their typical causes:

Oscillation: Usually caused by excessive feedback loop phase shift, inadequate supply decoupling, or poor layout. Reduce feedback, add compensation, or improve decoupling.
Crossover distortion: Insufficient bias current. Adjust bias circuit or check thermal coupling between bias transistor and output stage.
Thermal runaway: Inadequate emitter resistance, missing or improperly connected bias thermal sensing, or insufficient heatsinking.
DC offset: Input stage imbalance, often due to mismatched input transistors or incorrect feedback network values.
Noise: Ground loops, inadequate shielding, or noisy components in the input stage.

Conclusion

Discrete power amplifier design combines multiple disciplines including semiconductor physics, thermal management, protection circuit design, and careful construction techniques. While integrated amplifier solutions offer convenience for many applications, discrete designs remain important for high-power requirements, custom specifications, and educational purposes.

The fundamental trade-offs between efficiency and linearity, embodied in the Class A, B, and AB operating modes, underlie all power amplifier design. Understanding these trade-offs and the circuit techniques that optimize performance within each class enables engineers to select and implement appropriate solutions for specific applications.

Thermal design and protection circuits are not afterthoughts but integral parts of the amplifier design. A power amplifier without adequate thermal management and protection is simply a failure waiting to happen. The robust designs that survive real-world conditions including speaker faults, thermal stress, and years of operation result from careful attention to these often-neglected aspects of amplifier engineering.

The skills developed in discrete power amplifier design transfer directly to understanding integrated power amplifiers, switching amplifiers, and other power electronics applications. The ability to analyze heat flow, design protection circuits, and optimize bias for linearity and efficiency applies across the entire spectrum of power electronics design.