Designing a mosfet power amp offers distinct advantages over traditional Bipolar Junction Transistor (BJT) designs, primarily due to the MOSFET's high current capability, lack of minority carrier storage time, and inherent negative temperature coefficient of mobility. However, translating these semiconductor physics into a stable, high-fidelity audio amplifier requires strict attention to Safe Operating Area (SOA) limits, gate capacitance management, and precise Class AB biasing.

This guide cuts through the abstract theory to provide exact part numbers, a proven 50W application circuit, and the bench-tested procedures you need to bias, build, and troubleshoot your amplifier without blowing your output stage on the first power-up.

Selecting the Right MOSFET for Audio Power Amplification

When selecting output devices for an audio amplifier, you need complementary N-channel and P-channel pairs with matched transconductance and high voltage ratings to handle reactive speaker loads. While audiophiles often seek out expensive Lateral MOSFETs (like the Exicon ECX16N/P) for their wider SOA and lower gate capacitance, vertical power MOSFETs (HEXFETs) remain the most practical, cost-effective workbench defaults.

Below are the safe default part numbers for a standard +/- 35V to +/- 50V rail amplifier. These parts are widely available, rugged, and well-documented.

Table 1: Safe Default MOSFET Part Numbers for Audio Amps
Part Number Type Vds (Max) Id (Continuous) Pd (Max) Best Use Case
IRFP240 N-Channel 200V 20A 150W Standard +/- 35V to +/- 50V rails
IRFP9240 P-Channel 200V 12A 125W Complementary pair to IRFP240
IRFP150 N-Channel 100V 43A 200W Lower voltage rails (+/- 24V)
IRFP9150 P-Channel 100V 22A 200W Complementary pair to IRFP150
Pinout & Symbol Reminder: A standard TO-247 MOSFET has three pins: Gate (control input), Drain (main current path to the supply rail), and Source (current path to the load). Crucially, the physical construction creates an intrinsic body diode between the Drain and Source. In an N-channel device, the cathode is at the Drain and the anode is at the Source. This diode conducts if the load voltage swings past the supply rail, which is why proper power supply decoupling is mandatory.

MOSFET Operation Regions and Biasing for Class AB

One of the most common traps for builders transitioning from BJTs to MOSFETs is the terminology used for operation regions. In a BJT, 'saturation' means the device is fully turned on. In a MOSFET, 'saturation' is the active, constant-current region used for linear amplification. Confusing these will lead to fundamental misunderstandings of your bias network.

Table 2: MOSFET Operation Regions in Audio Amplification
Region Name Alternative Name Condition (N-Ch) Behavior Role in Amp
Cutoff Subthreshold Vgs < Vgs(th) No drain current flows (except leakage). Class B operation (causes crossover distortion).
Ohmic Triode / Linear Vgs > Vgs(th) AND Vds < (Vgs - Vgs(th)) Acts like a voltage-controlled resistor. Fully ON. Not used for amplification; used in switching.
Saturation Active / Constant Current Vgs > Vgs(th) AND Vds > (Vgs - Vgs(th)) Id is controlled by Vgs, largely independent of Vds. The linear amplification region.

How to Bias for Class AB: To eliminate the crossover distortion inherent in Class B (Cutoff) operation, we must bias the MOSFETs slightly into their Saturation (Active) region even when no audio signal is present. This is called quiescent current (Iq). For vertical MOSFETs like the IRFP240, the threshold voltage (Vgs(th)) is typically between 2.0V and 4.0V. To achieve a stable Iq of roughly 50mA to 100mA per device, you must apply approximately 3.2V to 3.5V to the Gate relative to the Source. Because the N-channel and P-channel devices are stacked, the total voltage between their gates must be maintained at roughly 6.5V to 7.0V by a thermal-tracking bias generator.

A Practical 50W MOSFET Power Amp Application Circuit

Below is a complete component list and topology for a robust, quasi-complementary source-follower output stage capable of delivering 50W into an 8-ohm load. This assumes a front-end voltage amplifier and differential input stage (like a classic Lin topology) is driving the bases of the driver transistors.

Power Supply & Output Stage Components

  • Power Supply: Dual-rail +/- 35V DC (unregulated is fine if capacitance is adequate, minimum 10,000µF per rail).
  • Output Devices: Q1 = IRFP240 (N-Ch), Q2 = IRFP9240 (P-Ch). Mount both on a single finned aluminum heatsink (minimum 1.5°C/W thermal resistance) using silicone thermal pads and shoulder washers for isolation.
  • Source Resistors (Rs1, Rs2): 0.22Ω, 5W wirewound. These provide local negative feedback, stabilize the bias point, and force current sharing if you parallel multiple devices.
  • Gate Stopper Resistors (Rg1, Rg2): 220Ω, 1/4W carbon film. Never omit these. They isolate the gate capacitance from the driver stage to prevent high-frequency parasitic oscillation.

Biasing and Stability Network

  • Vgs Multiplier Transistor: Q3 = BD139 (NPN). Mount this directly to the main output heatsink. It acts as a thermal sensor; as the heatsink warms up, the BD139 drops its Vce, reducing the gate drive voltage and preventing thermal runaway.
  • Bias Trim Potentiometer: 500Ω multi-turn cermet trimmer in series with a 1.5kΩ fixed resistor across the BD139 collector-emitter. Adjust this to set the exact voltage between the output gates.
  • Output Inductor (L1): 1µH. Wind 15 turns of 18AWG enameled copper wire tightly around the body of a 1Ω 1W resistor. This isolates the amplifier from capacitive speaker loads at high frequencies.
  • Zobel Network: 10Ω (1/2W) in series with 100nF (104) ceramic capacitor, placed from the amplifier output to ground.
Startup Procedure: Before applying full power, wire a 100Ω 10W power resistor in series with your main DC supply lines (a 'bulb limiter' equivalent). Connect a multimeter across the source resistors. Adjust the 500Ω trim pot until you read roughly 11mV to 22mV across the 0.22Ω source resistor (indicating 50mA to 100mA of quiescent current). Only remove the limiting resistor once the bias is stable and the heatsink is warm to the touch.

Failure Modes and Multimeter Testing Procedures

MOSFETs in audio amplifiers typically fail for three reasons: SOA Collapse (exceeding the Safe Operating Area during reactive load spikes), Thermal Runaway (failure of the bias tracking transistor), or Gate Oxide Puncture (ESD during handling or parasitic RF oscillation destroying the thin silicon dioxide gate layer).

When a MOSFET fails, it almost always fails as a dead short between Drain and Source. Here is the exact procedure to test a suspected blown MOSFET using a standard digital multimeter (DMM).

Numbered Steps for DMM Diode Testing

  1. Discharge the Gate: Set your DMM to resistance mode or use a piece of wire to short the Gate, Drain, and Source pins together for 5 seconds. The gate is a capacitor; if it holds a charge, it will turn the channel on and ruin your diode test readings.
  2. Test the Intrinsic Body Diode: Set the DMM to Diode Test mode. For an N-channel MOSFET, place the Red probe on the Source and the Black probe on the Drain. You should read a forward voltage drop of approximately 0.4V to 0.6V. Reverse the probes (Red on Drain, Black on Source); the meter should read 'OL' (Open Loop).
  3. Check for Drain-Source Shorts: Keep the DMM in Diode Test or Resistance mode. Measure between Drain and Source in both directions. If you read 0.00V or near-zero resistance in both directions, the silicon die has melted and the device is dead.
  4. Verify Gate Isolation: Measure between the Gate and Drain, and the Gate and Source. Both must read 'OL' in both directions. If you read any continuity or resistance, the gate oxide has been punctured by static or RF oscillation. Throw it away.

For a deeper dive into semiconductor testing protocols and SOA curve interpretation, refer to the comprehensive guides at All About Circuits and the audio design archives at Elliott Sound Products.

Frequently Asked Questions

Why do MOSFET power amps sound different than BJT transistor amps?

The perceived 'MOSFET sound'—often described as smoother or more 'tube-like' in the high frequencies—stems from two physical realities. First, MOSFETs are majority-carrier devices, meaning they lack the minority carrier storage time that causes BJTs to exhibit switching delays and transient intermodulation distortion (TIM) at high frequencies. Second, the transconductance curve of a MOSFET is a square-law function, which produces a different harmonic distortion profile (favoring even-order harmonics) compared to the exponential curve of a BJT when pushed into soft clipping.

Can I parallel multiple MOSFETs to increase power amp wattage?

Yes, but it requires strict matching and local feedback. Unlike BJTs, which suffer from thermal runaway when paralleled due to their negative temperature coefficient of resistance, vertical MOSFETs have a positive temperature coefficient of resistance at high currents, which naturally aids current sharing. However, at the low quiescent currents used in Class AB audio, they still exhibit a negative coefficient. To safely parallel them, you must use individual 0.22Ω source resistors for each device to force local negative feedback, and you should manually match the Vgs(th) of the devices using a curve tracer or a dedicated matching jig to within 50mV of each other.

Do I need a gate driver IC for an audio MOSFET power amp?

No. Gate driver ICs (like the IR2110) are designed for high-frequency PWM switching applications (like motor drives or SMPS) where the gate capacitance must be charged and discharged in nanoseconds to minimize switching losses. Audio signals max out at 20kHz, which is practically DC to a MOSFET. A standard discrete BJT driver stage (like a Darlington pair or a Sziklai pair) can easily source and sink the few milliamps required to charge the gate capacitance at audio frequencies without introducing the high-frequency noise and complexity of a dedicated driver IC.