The “Workhorse” 200W Power Amplifier is a high-performance Class-AB audio power amplifier designed for high-fidelity and musical instrument applications. The circuit employs a conventional three-stage amplifier architecture consisting of a differential input stage, a high-gain voltage amplification stage, and a Triple Emitter Follower (Triple EF) output stage based on principles described by Doug Self in his Audio Power Amplifier Design Handbook. By combining a constant-current-loaded input stage, active-load voltage amplification, thermally compensated biasing, and a robust multi-stage output section, the design is capable of delivering high output power with low distortion, good thermal stability, and excellent loudspeaker control.

As to the DC offset, the simulation results show around 175uV. This is with “ideal” transistors, and in the real-world, if all transistors are “matched” to each other of the same type (NPN and PNP) for hFE - DC offset should be relatively low. On breadboard tests, I found the DC offset (with matched transistors) to be about -6mV, respectably low.

The maximum DC supply rails should not exceed +/-55V; the original simulation of this amplifier (and its tests) were done with supply rails of +/-50V DC.

Note: The circuit shown below in Fig. 1 shows the use of Toshiba 2SC5200 and 2SA1943 devices for the outputs. These are considered obsolete, and even though Digikey, Mouser and RS Components shows stock of them, do NOT buy them from anywhere else (such as Temu, Amazon, eBay, Aliexpress, etc.). Even if they may be cheaper from alternative “suppliers”, they will most likely be counterfeit. The replacement alternatives are TTC5200 and TTA1943 respectively from Digikey and Mouser.

A PCB (3D render shown below) will be available soon, and this page will be updated once it has been manufactured.

Output Power: (to be determined at correct voltage)
Total Harmonic Distortion (THD): <0.05% (full power@8 ohms); <0.02% (11W@8 ohms)
Frequency Response: 10Hz - 38kHz

Fig. 1: schematic diagram of the “workhorse” amplifier. Take note of C5 marked as S.O.T. (select on test). I found through breadboarding the design, I had original selected the value as 47pF. This caused an oscillation on the output, and is probably too-high of a value. It is also probably not even needed.

This amplifier is a fairly modern, high-power Class-AB audio power amplifier built around a differential input stage, a voltage amplification stage (VAS), and a triple emitter-follower output stage, a topology discussed extensively by Doug Self in the “Audio Power Amplifier Design Handbook”. The triple emitter-follower arrangement is one of the key features of this design because it allows excellent linearity, high current capability, and reduced loading on the voltage amplification stage.

The input signal enters through coupling capacitor C1, which blocks any DC from the source equipment. R1 provides the input bias path, while R3, R4, and C2 form a low-pass input filter that reduces susceptibility to radio-frequency interference. The actual input stage consists of Q1 and Q2, which form a PNP differential pair. R2 is a “ground lift” resistor to prevent hum from interconnected equipment sharing the same earth (which can lead to ground loops).

The audio signal is applied to the base of Q1, while the base of Q5 receives the global feedback signal from the amplifier output. The amplifier therefore operates as a differential amplifier, constantly comparing the input signal with the feedback signal and amplifying only the difference between them. This is the fundamental mechanism that keeps distortion low and ensures accurate gain.

The tail current for the differential stage is supplied by Q3. Rather than simply using a resistor, Q3 operates as a constant-current source. D1, D2, R7, and the surrounding components establish a reasonably stable reference voltage, allowing Q3 to maintain a nearly constant current through the differential pair. Constant-current loading improves gain, linearity, and power supply rejection. Current through the tail pair is roughly 1mA.

Q2 and Q4 form an active load and current mirror arrangement for the differential stage. This is a common technique in modern amplifier design because it converts the differential output current into a single-ended signal while significantly increasing the voltage gain available from the input stage. Instead of wasting gain in collector load resistors, the current mirror forces signal currents to add constructively, increasing open-loop gain and reducing distortion once global feedback is applied.

The amplified signal from the differential stage is then fed into the voltage amplification stage built around Q6 and Q8. This stage provides most of the amplifier's voltage gain. Q6 and Q8 form a complementary VAS arrangement, while Q7 functions as a current source that establishes the operating current through the stage. VAS current is approximately 6mA. Since the output stage largely provides current gain rather than voltage gain, nearly all of the amplifier's voltage amplification occurs here. The gain of the amplifier, set by R13 and R12 is roughly 27, or 28dB.

Capacitor C5 is the dominant compensation capacitor, or “Miller” capacitor. Connected around the VAS, it introduces a controlled high-frequency roll-off that ensures stability when global feedback is applied. Without this capacitor, the amplifier could oscillate at ultrasonic frequencies.

The bias generator consists of Q9 together with RV1, R17, and R18. This stage establishes the correct voltage between the upper and lower driver sections. RV1 is the bias adjustment control used to set the output stage quiescent current. By adjusting this voltage, the amplifier can be operated in Class-AB with enough idle current to minimise crossover distortion while avoiding excessive heat dissipation. Q9 should be mounted on the heatsink so its temperature tracks the output devices, providing thermal compensation as the amplifier warms up.

The pre-driver stage begins with Q10 and Q11, which operate as emitter followers driven by the VAS. Their role is not to provide significant voltage gain but to supply current gain and isolate the high-gain VAS from the heavy load presented by the output transistors. These devices drive the driver transistors Q12 and Q13.

This is where the amplifier transitions into the topology that Doug Self refers to as a triple emitter-follower output stage. Looking at the positive half of the amplifier, Q10 drives Q12, Q12 drives Q14 and Q16; the latter two being paralleled output transistors. On the negative half, Q11 drives Q13, Q13 drives Q15 and Q17.

Each transistor stage operates as an emitter follower. Since an emitter follower provides current gain while maintaining approximately unity voltage gain, cascading three emitter followers creates enormous current gain without significantly loading the preceding stage.

This is the classic triple emitter-follower architecture described by Doug Self. One of its major advantages is that the VAS sees a very light load because each successive stage multiplies current gain. The VAS therefore remains in a highly linear operating region, reducing distortion. Another advantage is improved drive capability for large output transistor arrays. Emitter resistors R23 through R26 equalise current sharing, improve thermal stability, and help prevent localised current hogging.

D5 and D6 provide protection against reverse voltage conditions that can occur with highly reactive loudspeaker loads. During large transient events, the output can momentarily swing beyond the supply rails due to energy stored in the speaker's inductance. These diodes provide a safe path for that current, and they're often referred to as “flyback” diodes.

The output network consisting of L1, R28, R27, and C12 is the output stability network. L1 and R28 form the series output inductor, which isolates the amplifier from highly capacitive loads such as long speaker cables. R27 and C12 form a Zobel network that presents a predictable load at high frequencies and helps maintain stability.

Global negative feedback is returned from the output node to the differential input stage through R13. This feedback determines the closed-loop gain of the amplifier and dramatically reduces distortion, output impedance, and sensitivity to component tolerances.

Simulation of the circuit proved it has potential, with an overall total harmonic distortion of 0.05% (at full power into 8 ohms), and 0.02% at around 11W (also into 8 ohms). However, as this was simulated in TINA-TI, I'm a bit dubious as to the accuracy of those measurements.

Power output of the simulation with +/-50V supply rails provided 124W into 8 ohms, and up to 250W into 4 ohms! Not too shabby. Again, as this is was from the simulation, actual real-world power results may be slightly less - if we consider power supply sagging under load.

Below is the frequency response curve.

Fig. 2: frequency response graph

The amplifier has a decent flat frequency response of 10Hz to 20kHz, which is to be expected.

A couple of stability tests, using a square wave, were taken on the simulation by deliberately loading the output with a 1uF capacitor across it to ensure there were no “nasties”, such as high-frequency oscillation.

Fig. 3: stability test at 10kHz@8 ohms and 1uF capacitively loaded

Fig. 4: stability test at 20kHz@8 ohms and 1uF capacitively loaded

As seen by the two graphs above, the amplifier is fairly stable with only a slight “rise” (bump), then dampens out. It does not appear to continue to “ring”, indicating it's fine with capacitive loads.

Finally a total noise measurement was taken with TINA-TI simulation software, which is shown below.

Fig. 5: total noise

In short, it shows that the amplifier is fairly low-noise over the audio spectrum.

Even though, the original schematic denotes 2SC5200 and 2SA1943 Toshiba output transistors, a few substitutions can be used. We need to be aware that the aforementioned transistors have an ƒT (transition frequency or transit frequency) of 30MHz. So, fairly high-speed. MJL21193/MJL21194 would work (albeit expensive!), however they're slower in ƒT at only 4MHz, which will affect the slew rate.

TTC5200 and TTA1943 (made by EVVO Semiconductor) seem to be the genuine replacements with similar, if not better, specifications than the original in the same TO-3PL package (TO-264), which I recommend using.

Other possible substitutes (albeit package may be different) are NJW0281G for the NPN, and NJW0302G for the PNP.

Fig. 6: dimensions between transistors to assist in drilling the 200mm long 32mm x 32mm, 3mm thick aluminium heatsink bracket

Overall, this amplifier follows the architecture of many high-performance modern solid-state amplifiers: a differential input stage with current-source loading, a high-gain VAS with Miller compensation, a thermally compensated Class-AB bias network, and a Doug Self-style triple emitter-follower output stage. The use of three cascaded emitter-follower layers allows the voltage amplification stage to operate under extremely favourable conditions while still delivering the large output currents required for a 200-watt audio power amplifier.