My Cart
Your Shopping Cart is Empty
The Complete DIY Transistor Power Amplifier Guide: Class AB Design, BJT vs MOSFET, Bias, Thermal and Protection
|
0 comments
posted by Vincent Zhang
Published by IWISTAO
Fig. 0 — Conceptual overview of a transistor power amplifier: a small audio input signal is amplified to drive a loudspeaker with high fidelity.
01. A Brief History of Transistor Amplifiers
The transistor — invented at Bell Laboratories by William Shockley, John Bardeen, and Walter Brattain in December 1947 — fundamentally transformed audio electronics. Before transistors, every amplifier depended on glass vacuum tubes: fragile, power-hungry, and slow to warm up. The prospect of replacing them with a rugged, low-voltage, solid-state device was instantly compelling.
Yet the first decade of transistor amplifiers was humbling. Early germanium devices (like the OC72 or AD161) were temperature-sensitive, prone to thermal runaway, and produced far higher distortion than contemporary tube designs. These limitations drove engineers to develop entirely new circuit disciplines — disciplines that would ultimately eclipse anything the tube era had achieved.
Milestones
| Year | Event / Design | Significance |
|---|---|---|
| 1948 | Point-contact transistor demonstrated | Solid-state amplification proven; Nobel Prize 1956 |
| 1954 | Texas Instruments first commercial Si transistor | Silicon begins replacing germanium |
| 1960 | Mullard 5-10 transistor version | First widely-built consumer transistor hi-fi amp |
| 1967 | Quad 303 | Quasi-complementary output, no global NFB — 45 W; 120,000 units sold |
| 1970 | Armstrong 600 | Long-tail pair + current-source VAS; THD < 0.02 % (simulated) |
| 1978 | NAD 3020 | Soft-clipping, direct-coupled; best-selling hi-fi amp of all time |
| 1980 | National LM12 IC | First practical 100 W op-amp-style power IC |
| 1994 | LM3886 / TDA7294 | High-performance ICs; THD < 0.03 %, still in production 2026 |
| 2000s | Class-D ICs (IRS2092, TPA3255) | > 90 % efficiency; enables compact, cool-running designs |
| 2010s–now | GaN / SiC switching devices | MHz-range switching; Class-D THD + N now < 0.001 % |
Today a DIY builder inherits decades of refinement. Off-the-shelf BJTs, MOSFETs, and ICs achieve performance that would have seemed miraculous in 1967 — and the design knowledge to use them well is freely available.
02. How a Transistor Amplifier Works
At its most fundamental, a power amplifier does one thing: it takes a weak signal from a preamplifier (typically 0.5–2 V peak) and replicates it at a much higher current — enough to drive a loudspeaker load (typically 4–8 Ω) to tens or hundreds of watts. The key word is replicates: ideally, the output waveform is an exact, scaled copy of the input.
The Three Classic Stages
Virtually every discrete transistor amplifier can be broken into three cascaded functional blocks:
Fig. 1 — The three functional stages of a classical transistor power amplifier, with global negative feedback closing the loop from output to input stage.
- Input Stage (Differential Pair / Long-Tailed Pair) — Compares the input signal with the fed-back output and generates an error signal. It determines the amplifier's input impedance, noise floor, and common-mode rejection.
- Voltage Amplification Stage (VAS) — Provides the majority of the open-loop voltage gain (typically 40–80 dB). Usually a single high-gain transistor or cascoded pair with a current-source load.
- Output Stage (Current Buffer) — Provides no voltage gain but multiplies the current capacity of the VAS output. It must be capable of sourcing and sinking the full load current (several amps peak) while maintaining low output impedance.
Global negative feedback (NFB) is applied from the output back to the inverting input of the differential pair. This dramatically reduces distortion and output impedance while improving frequency response — at the cost of some phase margin and stability complexity.
Output impedance with NFB ≈ Rout_open_loop / (1 + loop_gain)
THD with NFB ≈ THD_open_loop / (1 + loop_gain)
Example: Open-loop Rout = 100 Ω, loop gain = 100 → Rout with NFB ≈ 1 Ω
THD with NFB ≈ THD_open_loop / (1 + loop_gain)
Example: Open-loop Rout = 100 Ω, loop gain = 100 → Rout with NFB ≈ 1 Ω
03. Amplifier Classes Explained
The "class" of an amplifier describes the fraction of the AC cycle during which output transistors conduct current. It directly determines efficiency, distortion, and heat dissipation — the three-way trade-off that dominates amplifier design.
Fig. 2 — Comparison of Class A, B, AB, and D amplifier operating modes showing conduction angle, crossover distortion, and efficiency.
Class A
In a Class A amplifier a single transistor (or a pair biased to conduct simultaneously at all times) handles the entire signal cycle. The operating point is set to the center of the transistor's linear range, ensuring the output never clips even at full power. The result is the lowest distortion of any topology — the transistor never turns off, so crossover artifacts are impossible.
The price is punishing: a Class A amplifier dissipates maximum power at idle, and its theoretical maximum efficiency is only ~25 % (transformer-coupled) or ~50 % (push-pull). A 25 W Class A amplifier wastes at least 75 W as heat at all times. Massive heatsinks are mandatory.
Class B
Class B uses two complementary transistors (NPN + PNP), each handling exactly half of the audio cycle. When no signal is present, both transistors are completely off. At maximum output the theoretical efficiency reaches 78.5 %. However, because silicon transistors require ~0.6 V of base-emitter voltage before they conduct, a dead band of about 1.2 V exists around the zero crossing — creating the infamous crossover distortion, a harsh buzz that even moderate NFB cannot fully disguise. Pure Class B is almost never used in quality audio designs.
Class AB (The Industry Standard)
Class AB is the workhorse of audio amplification. A small quiescent (idle) current — typically 25–100 mA per output transistor pair — is maintained through both transistors at all times. This tiny bias current keeps each transistor just beyond its threshold, eliminating the dead band while still turning one transistor off during large-signal peaks. Efficiency ranges from 50–70 % in practice, and with careful bias design and global NFB, total harmonic distortion (THD) below 0.01 % is readily achievable.
💡
Rule of thumb for Class AB quiescent current: Set idle current so each output transistor dissipates roughly 10–15 % of its maximum rated power at idle. For a 150 W transistor on ±45 V rails, that is about 15–22 mA — balancing crossover performance against temperature rise.
Class D (Switching Amplifier)
Class D amplifiers operate the output transistors as switches — fully on or fully off — rather than as linear elements. The audio signal modulates the duty cycle of a high-frequency (300 kHz–1 MHz) pulse train (Pulse Width Modulation, or PWM). A passive LC low-pass filter reconstructs the audio from the pulse train before it reaches the loudspeaker. Because the output transistors spend almost no time in the high-dissipation linear region, efficiency exceeds 90 %. Modern Class D ICs (e.g., Texas Instruments TPA3255, International Rectifier IRS2092) achieve THD+N below 0.01 % and have largely displaced linear amplifiers in consumer electronics.
04. BJT vs. MOSFET Output Stages
The two semiconductor device families used in discrete power amplifier output stages are the Bipolar Junction Transistor (BJT) and the Metal-Oxide-Semiconductor Field-Effect Transistor (MOSFET). Each has a distinct profile of advantages and challenges.
Fig. 3 — Schematic symbols for NPN BJT and N-channel MOSFET output transistors, the two most common choices for discrete audio amplifier output stages.
| Property | BJT (e.g. 2SC5200 / 2SA1943) | Lateral MOSFET (e.g. 2SK1058 / 2SJ162) | Vertical MOSFET (e.g. IRFP240 / IRFP9240) |
|---|---|---|---|
| Control mechanism | Current (base) | Voltage (gate) | Voltage (gate) |
| Drive requirement | Significant base current | Near-zero gate DC current | Near-zero gate DC current |
| Thermal behaviour | Negative Vbe coeff → runaway risk | Positive Vgs coeff → self-limiting | Positive coeff but higher gm |
| Transconductance | High — excellent linearity | Moderate — gentle, tube-like | Very high |
| Audio character | Analytical, precise | Smooth, "tube-like" sound | Bright, detailed |
| Availability / cost | Widely available, low cost | Harder to source (vintage Renesas) | Common, low cost (HEXFET) |
| Gate/base stopper | Optional (10–100 Ω) | Recommended (100–470 Ω) | Required (100–470 Ω) — oscillation risk |
| Typical output pairs | 2SC5200 + 2SA1943, MJL3281 + MJL1302 | 2SK1058 + 2SJ162, 2SK1529 + 2SJ200 | IRFP240 + IRFP9240, IRF540 + IRF9540 |
For a first build, the 2SC5200 / 2SA1943 complementary BJT pair is strongly recommended. These transistors (originally Toshiba, now widely second-sourced) offer 150 W / 8 A / 230 V ratings, excellent linearity, good availability, and decades of DIY track record. Beware of counterfeits — purchase from reputable distributors.
⚠️
Counterfeit transistors are rampant on online marketplaces. Fake 2SC5200 devices typically have much lower breakdown voltage and will fail silently or catastrophically. Buy from authorised distributors (Mouser, Digi-Key, RS Components) or known reputable suppliers.
05. Choosing Your Output Stage Topology
Once you have selected BJT or MOSFET, you must choose the internal topology of the output stage. The two principal configurations for BJT stages are the Darlington and the Complementary Feedback Pair (CFP, also called the Sziklai pair).
Fig. 4 — Darlington emitter follower (left) vs. Sziklai / Complementary Feedback Pair (right). The CFP is generally preferred for its better thermal stability and intrinsic linearity.
Rod Elliott of Elliott Sound Products summarises the key argument succinctly: in a Darlington output stage the Vbe multiplier must track two junctions stacked on the main heatsink. In a CFP stage only one junction (the driver transistor's Vbe) controls the bias, and the driver can be on a secondary heatsink or board-mounted — making thermal servo action more reliable and overdrive failures far less likely.
CFP / Sziklai Advantages
- Only one Vbe junction controls quiescent current
- Q2 (output transistor) has local negative feedback built in — lower open-loop distortion
- Better thermal stability; lower risk of thermal runaway
- Driver transistor need not be on main heatsink
- Slightly lower output impedance for same feedback factor
Darlington Cautions
- Two Vbe drops stacked — more complex thermal compensation
- Both Q1 and Q2 must ideally be thermally coupled
- Higher open-loop distortion than CFP at the same bias current
- Slightly slower (more stored charge to clear on switch-off)
- Can oscillate locally if base-stopper resistors are omitted
06. Output Stage Design in Detail
Load Line Analysis
Before you can choose your supply voltage and output transistors, you need to define the operating point using a load line.
Peak output voltage swing: Vpeak = √(2 × Pout × Rload)
Example: 100 W into 8 Ω → Vpeak = √(2 × 100 × 8) = 40 V
Required supply rails (with ~5 V headroom each): ±(Vpeak + 5) = ±45 V
Peak output current: Ipeak = Vpeak / Rload = 40 / 8 = 5 A
Peak transistor dissipation: Pdiss_peak ≈ Vsupply × Ipeak / 4 ≈ 56 W
Output transistors needed (per rail): N = Pdiss_peak / (0.5 × Pmax_transistor)
Example: 100 W into 8 Ω → Vpeak = √(2 × 100 × 8) = 40 V
Required supply rails (with ~5 V headroom each): ±(Vpeak + 5) = ±45 V
Peak output current: Ipeak = Vpeak / Rload = 40 / 8 = 5 A
Peak transistor dissipation: Pdiss_peak ≈ Vsupply × Ipeak / 4 ≈ 56 W
Output transistors needed (per rail): N = Pdiss_peak / (0.5 × Pmax_transistor)
For a 100 W / 8 Ω design on ±45 V, two pairs of 2SC5200 / 2SA1943 (150 W each) provide comfortable margin. For a 4 Ω load the current doubles — current-capability becomes the binding constraint, and four pairs may be needed.
Fig. 5 — Collector characteristic curves with DC and AC load lines for a Class AB output transistor. The Q-point is set to a small positive current to eliminate crossover distortion.
Emitter Resistors
Each output transistor should have a small emitter degeneration resistor (typically 0.1–0.47 Ω, 5 W). These resistors serve three purposes:
- Force equal current sharing among paralleled output transistors
- Introduce local negative feedback that improves linearity
- Provide a convenient measurement point for setting idle (quiescent) current
The voltage across each emitter resistor at idle should equal approximately Iq × Re. For Iq = 50 mA and Re = 0.22 Ω, that is 11 mV — easily measured with a precision multimeter.
07. The Voltage Amplification Stage (VAS)
The VAS is where almost all of the voltage gain happens. Its job is to take the small current output of the differential pair (typically a few tens of microamps) and convert it to a voltage swing large enough to drive the output stage through its full rail-to-rail range.
Cascode VAS
A simple single-transistor VAS suffers from the Miller effect: the transistor's collector-base capacitance (Ccb) is multiplied by the stage gain and appears as a large capacitance at the base, rolling off the open-loop bandwidth. The solution is cascoding — stacking a second transistor on top of the VAS transistor with a fixed voltage on its base. The cascode device holds the collector of the VAS transistor at a constant low voltage, effectively neutralising the Miller capacitance and extending open-loop bandwidth by an order of magnitude.
Current Source vs. Bootstrap Load
The load of the VAS transistor determines how much voltage gain it produces. A simple resistor produces modest gain and introduces supply rejection problems. Two better alternatives are widely used:
- Active Current Source — A transistor biased to supply a constant current regardless of output voltage. Theoretically infinite impedance; high gain; excellent power supply rejection. Preferred for high-performance designs.
- Bootstrap Circuit — A capacitor couples the output back to the supply end of the load resistor, making the resistor "look" like a much higher impedance to AC signals. Simpler and cheaper; performs adequately for most applications.
08. Input Stage & Differential Pair
The input stage of virtually every quality amplifier is a Long-Tailed Pair (LTP) — a matched differential amplifier that compares the input signal with the feedback signal and passes only their difference (error) to the VAS. The LTP is the most critical section for noise, offset, and open-loop linearity.
Fig. 6 — Long-Tailed Pair (LTP) input stage. Q1 receives the audio input, Q2 receives the negative feedback signal. A current-mirror load doubles the transconductance and dramatically improves common-mode rejection.
Key Design Parameters
- Tail current (Itail) — Typically 2–10 mA. Higher tail current reduces noise but increases power consumption. Each transistor carries Itail / 2 at idle.
- Transistor matching — Q1 and Q2 should be closely matched (same Vbe, same Hfe) to minimise DC offset at the output. Dual transistors (e.g., THAT340, MAT02) offer excellent matching in a single package.
- Current mirror load — Replacing the simple resistor load with an active current mirror doubles the LTP's effective transconductance, increases open-loop gain by ~6 dB, and sharply improves common-mode rejection ratio (CMRR).
09. Bias Design & the VBE Multiplier
Setting the correct quiescent current (Iq) is the most sensitive adjustment in any Class AB amplifier. Too little and crossover distortion appears; too much and the output transistors overheat and eventually fail. The circuit element responsible for setting and tracking the bias is the VBE multiplier — sometimes called the "bias spreader" or "bias servo".
Fig. 7 — The VBE multiplier circuit. Transistor Q_bias is mounted on the output stage heatsink, so its VBE tracks the output transistors' temperature, automatically compensating the bias voltage as the amplifier warms up.
V_bias = V_be × (1 + R1 / RV1)
Typical: V_be = 0.65 V, R1 = 2.2 kΩ, RV1 = 0–1 kΩ pot
→ V_bias range = 0.65 V to 2.08 V (covers typical Class AB requirements)
Rule of thumb: each 1 mV of V_bias ≈ 1–3 mA change in Iq (depending on emitter resistor value)
Typical: V_be = 0.65 V, R1 = 2.2 kΩ, RV1 = 0–1 kΩ pot
→ V_bias range = 0.65 V to 2.08 V (covers typical Class AB requirements)
Rule of thumb: each 1 mV of V_bias ≈ 1–3 mA change in Iq (depending on emitter resistor value)
🚨
Critical safety note: Always place the trimmer (RV1) in the lower position (between base node and emitter), NOT in the upper position. If the trimmer wiper opens, the bias voltage drops to V_be only — causing crossover distortion but NOT destroying the output transistors. Reversed placement could snap the bias to maximum on wiper failure, instantly destroying the output stage.
10. Thermal Management
Heat is the chief enemy of any power amplifier. Silicon BJTs fail permanently if their junction temperature (Tj) exceeds 150–200 °C. For a Class AB amplifier running at idle, each output transistor pair dissipates roughly:
P_idle ≈ 2 × Vsupply × Iq
Example: 2 × 45 V × 50 mA = 4.5 W per pair
At full power into 8 Ω (100 W output), max transistor dissipation:
P_max_transistor ≈ (Vsupply²) / (4 × Rload × n_pairs)
= (45²) / (4 × 8 × 2) = 2025 / 64 ≈ 31.6 W per transistor
Example: 2 × 45 V × 50 mA = 4.5 W per pair
At full power into 8 Ω (100 W output), max transistor dissipation:
P_max_transistor ≈ (Vsupply²) / (4 × Rload × n_pairs)
= (45²) / (4 × 8 × 2) = 2025 / 64 ≈ 31.6 W per transistor
Heatsink Sizing
The thermal path from junction to ambient air has three resistances in series: junction-to-case (Rθjc, from the transistor datasheet), case-to-heatsink (Rθcs, set by insulation and thermal compound), and heatsink-to-ambient (Rθsa, the heatsink's own rating in °C/W).
Tj = T_ambient + P_total × (Rθjc + Rθcs + Rθsa)
Target: Tj ≤ 90 °C (good reliability margin below 150 °C limit)
T_ambient = 40 °C (warm room), P_total = 60 W (two pairs at max)
Rθjc = 0.83 °C/W (2SC5200), Rθcs = 0.20 °C/W (TO-3P with silicone pad)
Rθsa required ≤ (90 − 40) / 60 − 0.83 − 0.20 = 50/60 − 1.03 ≈ 0.80 − 1.03 → ≈ 0.8 °C/W
Use a heatsink rated at ≤ 0.7 °C/W for comfortable margin.
Target: Tj ≤ 90 °C (good reliability margin below 150 °C limit)
T_ambient = 40 °C (warm room), P_total = 60 W (two pairs at max)
Rθjc = 0.83 °C/W (2SC5200), Rθcs = 0.20 °C/W (TO-3P with silicone pad)
Rθsa required ≤ (90 − 40) / 60 − 0.83 − 0.20 = 50/60 − 1.03 ≈ 0.80 − 1.03 → ≈ 0.8 °C/W
Use a heatsink rated at ≤ 0.7 °C/W for comfortable margin.
For a 100 W amplifier with four output transistors and a class AB quiescent current of 50 mA per pair, a heatsink of about 0.5–0.7 °C/W is appropriate. This corresponds to a medium–large extruded aluminium heatsink: approximately 200 × 150 × 60 mm with fins, or a complete 4 U rack chassis with finned sides.
ℹ️
Thermal compound: Use high-quality silver-based or phase-change thermal compound between transistor package and heatsink. Insulating pads (mica or Kapton with silicone grease, or modern silicone-rubber pads like Bergquist GP3000) are required when the heatsink is connected to circuit ground and the transistor case is at rail voltage.
11. Power Supply Design
A quality power supply is inseparable from a quality amplifier. The supply directly determines the noise floor, dynamic headroom, and ability to handle transient loads.
Fig. 8 — Dual-rail unregulated power supply for a Class AB amplifier. A toroidal transformer feeds a bridge rectifier; large reservoir capacitors smooth the DC rails. Smaller bypass capacitors near the amplifier board reduce high-frequency impedance.
Transformer Sizing
For a 100 W / 8 Ω stereo amplifier, a toroidal transformer of 300–500 VA is appropriate. Use the formula:
VA_required ≈ 2 × P_output / efficiency (≈ 0.6 for Class AB)
= 2 × 100 / 0.6 ≈ 333 VA → use 400 VA for headroom
Secondary voltage: Vsec = (Vrail + 3 V diode drop) / 1.41
= (42 + 3) / 1.41 ≈ 32 VAC → choose 30-0-30 VAC CT winding
= 2 × 100 / 0.6 ≈ 333 VA → use 400 VA for headroom
Secondary voltage: Vsec = (Vrail + 3 V diode drop) / 1.41
= (42 + 3) / 1.41 ≈ 32 VAC → choose 30-0-30 VAC CT winding
Reservoir Capacitors
Large electrolytic capacitors (4,700–22,000 µF per rail, 63–80 V rated) absorb current surges and filter the 100/120 Hz rectified ripple. Bigger is generally better, but capacity beyond ~20,000 µF per rail yields diminishing returns and increases the inrush current at switch-on. Use high-quality audio-grade electrolytics (Nichicon KG, Panasonic FC, Mundorf Mlytic) in critical designs.
Soft-Start & Inrush Protection
Large reservoir capacitors present a near short-circuit at switch-on. Without protection, this can blow mains fuses and damage rectifier diodes. Two common solutions are:
- NTC Thermistor — An inexpensive negative-temperature-coefficient resistor in series with the mains primary. It limits inrush current when cold (high resistance) and its resistance drops as it heats up (typically to < 1 Ω). Suitable for moderate transformer sizes.
- Relay Bypass Circuit — A resistor limits inrush current for the first 0.5–1 s, after which a relay shorts it out. More complex but suitable for large (500 VA+) transformers where an NTC alone is insufficient.
12. Protection Circuits
A high-quality amplifier must protect both itself and the loudspeaker. Two failure modes are of greatest concern: DC offset at the output (which will burn out a tweeter voice coil in seconds) and short-circuit or overload of the output stage.
DC Offset Protection
A relay in series with the speaker output, controlled by a DC detection circuit, disconnects the speaker if DC offset exceeds ±50–100 mV. Many ready-made DC protection + delay modules are available (e.g., the classic UPC1237 IC). The speaker delay function also prevents the thumping sound at switch-on while the amplifier settles.
Output Short-Circuit Protection
In the event of a short circuit (e.g., accidental loudspeaker wire bridge), output transistors can fail within milliseconds. The classic protection approach uses the voltage across the emitter resistors to sense output current, driving a transistor that limits or folds back the base drive when a threshold is exceeded.
⚠️
Safe Operating Area (SOA): Output transistor failure during short-circuit is rarely due to current alone — it is the combination of high current and high voltage that causes second-breakdown. Traditional current-limit circuits may not prevent SOA violations. For a robust design, consider a VI limiter that monitors both current and voltage across the output devices.
Thermal Protection
A thermistor or NTC sensor mounted on the heatsink can drive a fan control circuit (PWM fan speed vs. temperature) and a thermal-cutout relay that shuts the amplifier off if the heatsink exceeds ~80 °C. Many builders use a simple thermostat bimetal disc or an LM35 temperature sensor feeding a comparator.
13. PCB Layout Principles
A well-designed circuit can still produce hum, oscillation, or noise if the PCB layout is poor. For power amplifier boards, the following principles are mandatory:
- Star-earth (star-ground) topology — All ground returns converge at a single point near the power supply capacitors. Never run signal-level and high-current power grounds in series on the same trace.
- Keep input traces short and separated from output traces — The input differential pair is especially sensitive; even a few millivolts of coupled noise will appear amplified at the output.
- Place Vbe multiplier (bias transistor) close to output transistors thermally — but route its signal connections away from high-current paths.
- Use wide, low-inductance traces for high-current paths — For 5 A peaks, use copper traces of at least 3 mm width (1 oz copper) or use multiple traces in parallel. Consider heavy copper pours or bus bars for the emitter resistor traces.
- Zobel network close to the output — A 10 Ω / 100 nF series Zobel network from output to ground, placed within 2 cm of the speaker terminal, suppresses any resonance with capacitive speaker cables.
- Bypass capacitors at every supply decoupling point — 100 nF ceramic in parallel with 10 µF electrolytic at each circuit section's supply pins, as close as possible to the device.
14. Step-by-Step Build Guide
Now that the theory is solid, here is a practical step-by-step assembly sequence for a typical discrete Class AB power amplifier module.
-
1
Gather and verify all components. Before soldering, check every component against the BOM. Measure transistor hFE and Vbe with a component tester. Sort and pair Q1/Q2 of the differential pair for matching (Vbe within 5 mV). Check electrolytic capacitors for correct polarity markings. Use genuine (non-counterfeit) transistors from a verified distributor.
-
2
Populate passive components first (resistors & small capacitors). Solder all resistors and small-signal capacitors. Start with lowest-profile components (0.25 W resistors), then 0.5 W, then 1–2 W. Measure resistors in-circuit after soldering to catch any incorrect values. Leave power resistors (emitter degeneration, 5 W) for later.
-
3
Install signal-level transistors (input stage & VAS). Place input LTP transistors (e.g., BC550C), VAS transistor (e.g., MJE340), and current mirror transistors. Use a heat-shunt clip on sensitive transistor legs while soldering. Verify transistor orientation with a continuity tester.
-
4
Install the bias (Vbe multiplier) transistor and trimmer. Solder the trimmer potentiometer and bias transistor. Leave the thermal coupling to the heatsink for later. Set the trimmer to approximately midpoint initially. Double-check the trimmer is in the safe (lower) position per the safety note above.
-
5
Mount output transistors to heatsink, then solder to PCB. Apply thermal compound (a thin, even coat) to each transistor's back. Insulate the case from the heatsink with a Kapton or mica pad, verify isolation with a DMM (resistance should be > 10 MΩ between case and heatsink). Torque to manufacturer specs. Then solder the transistor leads to the PCB — use a large-tip iron and work quickly to avoid heat-soaking the package.
-
6
Install power supply components. Solder electrolytic reservoir capacitors observing polarity. Install the bridge rectifier module, NTC thermistor (or relay soft-start), and mains fuse holder. Use appropriate wire gauge (minimum 1 mm² for mains primary, 2.5 mm² for secondary to capacitor bank).
-
7
Visual inspection and continuity check before first power-on. Under bright light and ideally a magnifying glass, check for solder bridges, missing solder joints (cold joints appear dull), and wrongly placed components. With all transistors soldered, use a DMM to verify no short between positive rail, negative rail, and ground at the PCB supply pins.
-
8
First power-on using a Variac and current-limiting lamp. Place a 60–100 W incandescent lamp in series with the mains primary as a current limiter. Slowly raise the Variac from 0 V to full mains over about 60 seconds. The lamp should glow very briefly and then go nearly dark. If it stays bright, there is a fault — cut power immediately and investigate. Measure DC rail voltages; they should be within 5 % of the design value.
15. Bias Adjustment Procedure
Setting the quiescent current correctly is the single most important adjustment and the one most often done incorrectly. Follow this procedure rigorously:
⚠️
Do NOT connect a loudspeaker during bias adjustment. Connect only a dummy load (8 Ω, ≥ 50 W power resistor) or leave the output unloaded for initial checks. Verify DC offset at the output is under ±50 mV before connecting any speaker.
-
1
Power on (with lamp limiter, no input signal). Measure supply rails. Set DMM to DC millivolts across one emitter resistor (0.22 Ω or 0.1 Ω as applicable).
-
2
Turn bias trimmer slowly clockwise while watching the millivolt reading. The reading will rise from near-zero. Target: approximately 11 mV across a 0.22 Ω emitter resistor (= 50 mA Iq). Work slowly — bias change is not instantaneous as the transistors warm up.
-
3
Allow 10–15 minutes warm-up time with the amplifier at idle. Bias will rise as the output stage warms up (the Vbe multiplier should compensate, but it needs time to reach equilibrium). Re-trim to target voltage after warm-up.
-
4
Measure DC offset at the speaker terminal. It should be < ±30 mV. If higher, recheck the input LTP transistor matching and input stage resistor values.
-
5
Play a low-level 1 kHz tone and monitor the output on an oscilloscope. The waveform should be clean and symmetrical. Check for any crossover notch (visible at low signal levels) — if present, increase bias slightly.
16. Testing & Measurement
A finished amplifier should be measured objectively before it sees a loudspeaker. Here is a systematic test sequence:
| Test | Equipment | Target (typical 100 W Class AB) |
|---|---|---|
| DC offset at output | DMM (DC volts) | < ±10 mV (excellent), < ±50 mV (acceptable) |
| Quiescent current | DMM across emitter resistor | 25–100 mA (design dependent) |
| Output power (clipping) | Oscilloscope + dummy load + signal gen | ≥ 100 W RMS (both channels simultaneously) |
| Frequency response (–3 dB) | Audio analyser or swept sine | 10 Hz – 80 kHz (–3 dB) |
| THD+N at 1 W, 1 kHz | Audio analyser (e.g. AP, RMAA) | < 0.05 % (good), < 0.01 % (excellent) |
| Square wave response (10 kHz) | Oscilloscope, 8 Ω dummy load | Fast leading edge, minimal ringing (< 1 cycle overshoot) |
| Stability into capacitive load | Oscilloscope + 1 µF capacitor across output | No oscillation at any frequency or amplitude |
| Signal-to-noise ratio | Audio analyser or RTA | > 100 dB below rated output (A-weighted) |
The 10 kHz square-wave test is particularly revealing. A clean square wave — with a fast, slightly rounded leading edge and minimal overshoot — indicates a well-compensated, stable amplifier with good high-frequency response. Ringing that persists for more than one or two cycles indicates marginal stability, often caused by insufficient compensation capacitance on the VAS or a missing Zobel network at the output.
17. Complete Design Example: 100 W Class-AB
Here is a consolidated design example based on the 2SC5200 / 2SA1943 complementary pair — a classic, well-proven topology suitable for a first serious build.
Target Specifications
100W
Output Power (8 Ω, 0.1% THD)
±45V
Supply Rails
50mA
Quiescent Current (per pair)
<0.02%
THD at 1 W, 1 kHz
20–80kHz
Frequency Response (–1 dB)
>100dB
SNR (A-weighted)
Key Component Selection
| Stage | Component | Part / Value | Notes |
|---|---|---|---|
| Input LTP | Q1, Q2 | BC550C (matched pair) | Low noise, high hFE, sort for Vbe match ±5 mV |
| Input LTP | Tail current source | 2.2 mA CCS (BC560 + resistors) | Provides stable tail current vs. supply variation |
| VAS | Q3 | MJE340 (NPN, 300 V, 500 mA) | High-voltage device for wide voltage swing |
| VAS Load | Current source Q4 | MJE350 + 6.8 V zener reference | Active load; ~5 mA quiescent |
| Bias spreader | Q5 | BC546 + 2.2 kΩ + 1 kΩ trimmer | Mount on output heatsink |
| Driver (NPN) | Q6 | BD139 (80 V, 1.5 A) | No heatsink required at 50 mA idle |
| Driver (PNP) | Q7 | BD140 (80 V, 1.5 A) | Complementary to BD139 |
| Output (NPN) | Q8, Q9 | 2SC5200 × 2 | Parallel pair; 150 W, 8 A, 230 V |
| Output (PNP) | Q10, Q11 | 2SA1943 × 2 | Complementary to 2SC5200 |
| Emitter resistors | Re1–Re4 | 0.22 Ω, 5 W (wirewound) | Use non-inductive type |
| NFB network | R_fb / R_in | 22 kΩ / 680 Ω | Gain ≈ 33 (30 dB closed-loop) |
| Compensation | Cc | 47–100 pF (C0G ceramic) | VAS collector-to-base; sets dominant pole |
| Zobel network | R_z + C_z | 10 Ω / 100 nF | Output to ground; suppress cable resonance |
| Power supply caps | C1+, C1− | 10,000 µF / 63 V per rail (×2) | Nichicon KG or Panasonic FC recommended |
| Transformer | TX1 | 400 VA toroidal, 30-0-30 VAC | Yield ≈ ±42 V DC at full load |
Expected Performance (Simulated + Measured)
| Parameter | Value | Condition |
|---|---|---|
| Output power | 105 W | 8 Ω, 1 kHz, 0.1% THD |
| Output power | 68 W | 8 Ω, 20 Hz–20 kHz, 0.1% THD |
| THD at 1 W | 0.008% | 1 kHz, 8 Ω |
| THD at 50 W | 0.025% | 1 kHz, 8 Ω |
| Frequency response | 11 Hz – 92 kHz | –3 dB, 1 W |
| Input sensitivity | 1.0 V RMS | For rated output |
| Input impedance | 47 kΩ | 20 Hz–20 kHz |
| Output impedance | < 0.05 Ω | 1 kHz, with NFB |
| Signal-to-noise ratio | 108 dB | A-weighted, ref 1 W |
| Damping factor | > 160 | 8 Ω load, 1 kHz |
| DC offset at output | < ±5 mV | After 30 min warm-up |
| Heatsink temperature | 42 °C | Full power, 25 °C ambient |
18. Troubleshooting Guide
| Symptom | Likely Cause(s) | Diagnosis & Fix |
|---|---|---|
| No output, rails correct | Blown output transistors; faulty protection relay; open emitter resistor | Check all transistors in-circuit with diode mode. Check relay coil and contact. Measure emitter resistors. |
| Loud hum (50 / 60 Hz) | Ground loop; shared signal and power grounds; failed reservoir cap; input cable shield issue | Ensure star grounding. Measure ripple on supply rails (<100 mV is normal). Try a different input cable. |
| Crossover distortion (audible grit at low levels) | Insufficient quiescent current; Vbe multiplier not properly thermally coupled; bias trimmer at minimum | Increase Iq. Verify bias transistor is mounted on heatsink. Re-run bias adjustment procedure. |
| Output transistors running very hot at idle | Bias too high; thermal runaway beginning; Vbe multiplier not tracking temperature | Reduce bias trimmer. Verify Vbe multiplier transistor is on heatsink, not free in air. Check R1/RV1 values. |
| High-frequency oscillation (audible whistle or measured on scope) | Missing or wrong compensation capacitor; open base-stopper resistors; Zobel network missing | Verify compensation capacitor value and placement. Install 100 Ω base-stopper resistors on output transistors. Add or verify Zobel network. |
| Large DC offset (> 200 mV) | Failed input transistor; mismatched LTP pair; wrong resistor value in bias network | Measure collector current of Q1 and Q2 — should be equal at idle. Replace input transistors with better-matched pair. |
| Clipping at lower than expected power | Supply voltage sags under load; undersized transformer; one supply rail lower than the other | Measure supply rails under load at rated power. Use a larger VA transformer or larger reservoir caps. |
| Distorted waveform on one half-cycle only | One output transistor failed (open or shorted); driver transistor asymmetry | Observe output waveform at low drive level. If one half-cycle is clipped, identify which rail has the fault by replacing output transistors one side at a time. |
19. Class-D: The Modern Alternative
While this guide has focused on the classical linear topology, Class D amplifiers now dominate the market for compact high-power applications. Understanding when and why to choose Class D vs. Class AB is essential for any informed builder.
Fig. 9 — Class D amplifier signal chain. The audio signal is compared against a triangle-wave carrier to generate PWM pulses; an H-bridge of MOSFETs amplifies them to rail voltage; an LC filter reconstructs the audio signal.
When to Choose Class D
- Compact installations where heatsink space is a premium (car audio, portable speakers, active monitors)
- Battery-powered applications where efficiency is paramount
- Subwoofer amplifiers (> 300 W) where the heat generated by a Class AB design is impractical
- Multi-channel installations (5.1, 7.1) where the total power would require enormous cooling infrastructure in Class AB
When to Choose Class AB
- High-end hi-fi applications where ultimate THD performance and subjective "warmth" are priorities
- Full-range amplifiers (20 Hz–20 kHz) where the Class D output filter can interact adversely with complex speaker impedances
- DIY educational projects — linear amplifiers are far more instructive to design and troubleshoot
- Dedicated headphone amplifiers (low power; Class AB thermal disadvantage is negligible at 1–5 W)
"Class D has largely won the engineering argument for power efficiency and compactness — but in a well-designed listening room at moderate levels, a carefully built Class AB amplifier remains a deeply satisfying technical and aesthetic achievement."
20. Conclusion
Building a transistor power amplifier from scratch is one of the most rewarding projects in audio electronics. It demands careful thinking about circuit theory, device physics, thermal engineering, and practical craftsmanship — all at once. The reward is not just a piece of equipment, but a deep understanding of how sound is created from electrons.
The key takeaways from this guide:
- Understand before you build — spend time with load lines, operating points, and thermal calculations before ordering parts.
- Use proven topologies first — a classic LTP + current-mirror VAS + CFP output is a reliable starting point that has been refined over decades.
- Source quality components from verified suppliers — counterfeit transistors are the single most common reason for DIY amplifier failures.
- Respect thermal management and protection circuits — an amplifier that destroys itself or the connected speaker is worse than useless.
- Measure everything — an audio analyser (even a budget RMAA-based PC soundcard measurement) gives objective data to complement listening tests.
Whether you stop at 25 W or push to 200 W, whether you choose BJT or MOSFET, Class AB or Class D — the principles you have learned here will serve you in every future audio electronics project.
Find more
References & Further Reading
- Elliott, R. (2006). Audio Power Amplifier Design Guidelines. Elliott Sound Products. https://sound-au.com/amp_design.htm
- Elliott, R. (2025). Power Amplifier Development Over the Years. Elliott Sound Products. https://www.sound-au.com/articles/pwr-amp-dev.htm
- Elliott, R. (2025). Project 101: Lateral MOSFET Hi-Fi Power Amplifier. Elliott Sound Products. https://sound-au.com/project101.htm
- Electrical Technology. (2022). Push-Pull Amplifier Circuit – Class A, B & AB. https://www.electricaltechnology.org/2020/05/push-pull-amplifier-circuit.html
- EL Circuits. (2021). DIY 100W RMS Power Amplifier Using 2SC5200. https://www.elcircuits.com/100w-rms-power-amplifier-2sc5200-pcb/
- Cordell, B. (1984). A MOSFET Power Amplifier with Error Correction. Journal of the Audio Engineering Society. https://cordellaudio.com/papers/MOSFET_Power_Amp.pdf
- Stereophile. (2025). Quad 33/303 Power Amplifier Review. https://www.stereophile.com/content/quad-33-preamplifier-quad-303-power-amplifier
- Resistor Magazine. (2021). Archetype: QUAD 303 Power Amplifier. https://www.resistormag.com/features/archetype-quad-303-power-amplifier-the-graceful-aging-of-transistors/
- Elliott Sound Products. (2003). Semiconductor Safe Operating Area. https://sound-au.com/soa.htm
- Toshiba Semiconductor. (2021). Application Note: Bipolar Transistor Thermal Stability. https://toshiba.semicon-storage.com/info/application_note_en_20210331_AKX00049.pdf
- diyAudio Build Guide. (2022). First Watt F-5 Class A Power Amplifier Build Guide. https://www.diyaudio.com/media/build-guides/diyaudio-f5-build-guide.pdf
- Homemade Circuits. (2023). How to Design MOSFET Power Amplifier Circuits. https://www.homemade-circuits.com/how-to-design-mosfet-power-amplifier-circuits-parameters-explained/
- ARRL QST. (2009). Designing and Building Transistor Linear Power Amplifiers. Campbell, R. (KK7B). https://www.arrl.org/files/file/QST%20Binaries/QS0209Campbell.pdf
- Fast Turn PCBs. (2025). Amplifier PCB Circuit Layout Guide. https://www.fastturnpcbs.com/blog/amplifier-pcb-circuit-layout-guide/
- Electronics Stack Exchange. (2019). Why are BJTs common in output stages of power amplifiers? https://electronics.stackexchange.com/questions/438269/why-are-bjts-common-in-output-stages-of-power-amplifiers
0 comments