You’ll get a clear, component-level breakdown of how a Class D amplifier circuit actually works — not just textbook definitions, but the practical signal flow, critical design trade-offs (e.g., switching frequency vs. EMI), and why a 'simple' schematic often fails in real-world PCB layout. This guide answers exactly what engineers, hobbyists, and audio hardware designers search for when typing class d amplifier circuit: how PWM is generated, how MOSFETs switch without shoot-through, why LC filter placement matters more than component values alone, and where common circuit implementations go wrong.
Why Understanding the Class D Amplifier Circuit Is Non-Negotiable
Unlike Class AB or Class A amplifiers — where output transistors operate in their linear region — a Class D amplifier circuit relies entirely on switching behavior. Its core function isn’t analog amplification; it’s high-efficiency digital-like power delivery. That shift changes everything: component selection, PCB routing priorities, thermal management strategy, and even how you measure performance. Misunderstanding any single stage — especially the interplay between modulation, gate driving, and filtering — leads directly to audible artifacts, overheating MOSFETs, or electromagnetic interference that disrupts nearby microcontrollers or wireless modules1. This isn’t theoretical: over 70% of modern portable speakers, soundbars, and automotive infotainment systems use Class D topology precisely because of its >90% efficiency — but only when the circuit is correctly implemented2.
The Four Core Stages of Every Functional Class D Amplifier Circuit
A working Class D amplifier circuit isn’t a monolithic block. It’s four tightly coupled subsystems, each with distinct electrical requirements and failure modes. Let’s walk through them in strict signal order — from input to speaker — using real-world constraints, not idealized diagrams.
1. PWM Generation Stage: Where Audio Becomes Pulses
This stage converts the analog input (e.g., line-level or DAC output) into a pulse-width modulated signal. The most common method uses a comparator fed by two inputs:
- Analog audio signal (typically ±1 V to ±2 V peak)
- High-frequency carrier waveform — usually a triangle or sawtooth wave at 300–1000 kHz
When the audio voltage exceeds the carrier, the comparator outputs high; when below, it outputs low. The resulting duty cycle of the output pulses maps linearly to the instantaneous audio amplitude. Critical considerations:
- Carrier frequency choice: Must be ≥5× the highest audio frequency (≥100 kHz for full-range audio). Too low → audible switching noise; too high → increased MOSFET switching losses and harder gate driving3.
- Comparator speed & offset: Propagation delay mismatch between channels causes timing skew → differential-mode EMI. Input offset voltage >2 mV introduces DC bias → unwanted speaker cone displacement.
- No feedback in basic PWM: Open-loop modulation suffers from nonlinearity. Most production circuits add error feedback (e.g., output signal sampled and subtracted pre-filter) to correct pulse width errors caused by dead-time or supply ripple.
2. Switching Stage: MOSFETs as Binary Power Valves
Here, the PWM signal drives a pair (or two pairs, in BTL mode) of complementary MOSFETs — typically N-channel high-side + N-channel low-side, controlled via a dedicated gate driver IC. Their job: rapidly connect the load (speaker) alternately to VDD and ground, with near-zero conduction loss.
Why MOSFETs — not BJTs? Because they offer lower on-resistance (RDS(on)), faster switching (critical for >500 kHz carriers), and no base current burden. But they introduce new challenges:
- Shoot-through risk: If both high-side and low-side MOSFETs conduct simultaneously — even for nanoseconds — a low-impedance path forms between supply rails → massive current spike → device destruction. This is prevented by dead-time insertion (intentional delay between turn-off of one FET and turn-on of the other), typically 20–100 ns.
- Gate charge (Qg) & drive strength: A MOSFET with 50 nC total gate charge switched at 500 kHz requires average gate current of ~25 mA. Under-specifying the gate driver causes slow edges → increased switching loss → heat buildup.
- Parasitic inductance: Even 5 nH of source inductance in the MOSFET’s PCB trace creates voltage spikes (>20 V overshoot) during turn-off — enough to exceed gate oxide rating. Hence, short, wide, low-inductance source connections are mandatory.
3. Gate Driver Subsystem: The Critical Interface
The gate driver sits between the low-power PWM logic and the high-current MOSFET gates. It’s not optional — it’s foundational. A weak or poorly laid-out driver causes:
- Asymmetric rise/fall times → distorted pulse edges → increased THD+N
- Inadequate peak current → slow switching → excessive power dissipation in MOSFETs
- Insufficient isolation → latch-up or ground bounce coupling into control logic
Modern integrated drivers (e.g., TI UCC27524, STMicro TD350) provide matched dual-channel outputs, programmable dead time, and built-in Miller clamp protection. Discrete driver designs (using bipolar transistors or MOSFET followers) are rare outside academic projects — they lack consistency and require meticulous tuning.
4. Output LC Filter: Reconstructing Audio Without Compromise
This is where theory meets reality — and where many Class D designs fail. The LC low-pass filter removes the high-frequency carrier (e.g., 400 kHz) while preserving the audio band (20 Hz–20 kHz). A typical 2nd-order Butterworth filter uses:
- Inductor (L): 1–10 µH, rated for >2× peak output current, with low DC resistance (<5 mΩ) and shielded construction to limit magnetic coupling
- Capacitor (C): 1–10 µF ceramic or film, low ESR (<5 mΩ), placed immediately adjacent to the inductor
But component specs alone don’t guarantee success. Layout dominates:
- Filter loop area must be minimized: Any loop formed by L, C, and PCB traces acts as an antenna radiating EMI. Best practice: place L and C in series along a straight trace, with ground plane cutout *only* under the filter components — never under the entire loop.
- Capacitor placement relative to speaker: If C is placed far from the speaker terminal, trace inductance resonates with C → peaking at 1–5 MHz → audible ‘buzz’ and RF interference.
- Filter damping: Un-damped filters ring at their cutoff frequency. Adding a small resistor (0.1–0.5 Ω) in series with L suppresses ringing but reduces efficiency. Many IC-based solutions integrate active damping instead.
Bridge-Tied Load (BTL) vs. Single-Ended: Which Circuit Topology Fits Your Use Case?
Most consumer Class D ICs (e.g., TI TPA3116, Analog Devices SSM3582) use BTL configuration — two half-bridge outputs driven with inverted PWM signals across the speaker. Why?
- Double the output voltage swing: For same supply, BTL delivers 4× power vs. single-ended (P = V²/R).
- No output coupling capacitor needed: Eliminates large, expensive, non-ideal electrolytics — critical for bass response and long-term reliability.
- Reduced filter size: With differential output, common-mode switching noise cancels; thus, smaller L and C suffice.
However, BTL demands precise matching between channels. >5% gain or phase mismatch increases common-mode voltage — which the LC filter cannot reject — leading to residual carrier feedthrough and increased EMI. Single-ended topologies remain relevant for ultra-low-cost applications or when driving headphones (where BTL complexity outweighs benefit).
Three Real-World Circuit Pitfalls (and How to Avoid Them)
These aren’t hypothetical — they’re documented root causes of field failures and lab debugging sessions.
Pitfall #1: Ignoring Power Supply Decoupling at the Switch Node
The switch node (connection point between high-side and low-side MOSFETs) toggles between 0 V and VDD at full carrier frequency. Without localized, low-ESL ceramic capacitors (e.g., three 10 µF X7R 0805s in parallel) placed <1 mm from the switch node, supply rail collapse occurs during turn-on. Result: distorted PWM edges, increased THD, and potential IC reset. Rule: decouple *at the switch node*, not just at the IC VDD pin.
Pitfall #2: Treating the LC Filter as ‘Afterthought’ Layout
Many designers select L and C values correctly, then route them with 10-mm traces, shared ground pours, and no attention to current return paths. This turns the filter into a broadband radiator. Verified fix: use a dedicated ‘filter island’ — solid copper pour for power/ground, isolated by 2-mm gaps, with components mounted directly on top, and all high-di/dt currents routed on inner layers.
Pitfall #3: Assuming All ‘Class D’ ICs Are Created Equal
Two chips may share the same datasheet headline specs (e.g., “100 W, 90% efficiency”) but differ radically in internal architecture:
- Modulation type: Fixed-frequency PWM vs. self-oscillating vs. delta-sigma — affects EMI profile and loop stability.
- Feedback topology: Current-mode vs. voltage-mode vs. full-output feedback — determines THD+N floor and PSRR.
- Protection logic: Over-current detection speed (critical for short-circuit survival) varies from 100 ns to 5 µs across vendors.
Always validate with oscilloscope measurements — not just bench load tests, but real speaker loads with complex impedance curves (e.g., 4 Ω nominal, but dips to 2.8 Ω at 80 Hz).
Class D vs. Class AB: When Does Circuit Complexity Pay Off?
Don’t choose Class D solely for efficiency. Ask: what problem does it solve *in your application*?
| Criterion | Class D Amplifier Circuit | Class AB Amplifier Circuit |
|---|---|---|
| Typical Efficiency | >90% at full power (drops to ~75% at 10% power) | 50–65% (constant, regardless of output level) |
| Heat Dissipation | Small heatsink or none required (e.g., 20 W chip runs at 45°C ambient) | Large extruded heatsink mandatory (e.g., same 20 W chip reaches 85°C) |
| PCB Area | Compact (IC + 6 passive components) | Large (output transistors, driver stages, compensation network) |
| EMI Sensitivity | High — requires careful layout, shielding, and filtering | Low — minimal RF emissions |
| THD+N (1 kHz, full power) | 0.02–0.1% (modern ICs); >0.5% in discrete designs | 0.005–0.02% (well-designed discrete) |
Bottom line: Choose Class D when size, weight, or battery life constrain your design — not when ultimate linearity is the goal. For studio monitors or high-end headphone amps, Class AB or Class G remains preferred.
Verifying Your Class D Amplifier Circuit: What to Measure (and What to Ignore)
Datasheets list impressive numbers — but real validation requires targeted measurements:
- Switch node waveform (with 1 GHz probe): Check for clean edges, absence of ringing >200 MHz, and consistent dead time. Ringing indicates layout resonance — not component fault.
- Output spectrum (FFT, 1 MHz span): Look for carrier harmonics >40 dB below fundamental. Spikes at 2×, 3× carrier suggest poor grounding or asymmetrical drive.
- THD+N vs. frequency plot (20 Hz–20 kHz): Should be flat ≤0.05% up to 10 kHz, then rise smoothly — not jump at 1 kHz (indicates compensation issue).
- Load transient response (square wave input): Speaker voltage should settle within 100 µs — slow settling implies insufficient loop bandwidth or output capacitance.
Avoid relying solely on ‘efficiency at 1 kHz sine’ — it masks switching losses at music program material with rapid dynamic shifts.
Frequently Asked Questions (FAQ)
- Can I build a discrete Class D amplifier circuit without an IC?
Yes, but not practically for production. Discrete designs require precision-matched comparators, custom gate drivers, and meticulous layout — increasing BOM cost and reducing yield. ICs integrate these functions with proven reliability and thermal management. - Why do some Class D amplifiers produce audible hiss or whine?
Hiss usually stems from poor power supply rejection (PSRR) or noisy reference voltage. Whine (at carrier frequency or harmonics) indicates inadequate LC filtering, ground loop coupling, or insufficient shielding — not inherent to Class D itself. - Does PCB stackup matter for Class D circuits?
Critically. Use at least 4-layer boards: dedicated ground plane (Layer 2), power plane (Layer 3), and top/bottom for signals. Never route high-di/dt switch node traces over splits in ground plane — this forces return current through long paths → EMI. - Can I use electrolytic capacitors in the LC filter?
No. Electrolytics have high ESR and inductance, causing severe resonance and heating at carrier frequencies. Only ceramic (X7R, C0G) or polypropylene film capacitors meet the low-impedance, high-frequency requirements. - What’s the minimum switching frequency for a 20 kHz audio bandwidth?
While Nyquist says ≥40 kHz, real-world filters need ≥5× margin. 250–500 kHz is standard. Below 200 kHz, filter size balloons and EMI becomes harder to contain — making 150 kHz designs rare outside low-power applications.