Push-Pull Output

Part of Vacuum Tubes

Push-pull output stages use two tubes operating in antiphase to produce twice the power, cancel even-order distortion, and eliminate DC magnetization from the output transformer.

Why This Matters

A single output tube — a single-ended Class A stage — is simple and sounds musically pleasant to many listeners. But it is limited in power output and produces more distortion than a push-pull stage for the same power level. For community applications — a radio station that must be heard clearly over background noise, a public address system for outdoor gatherings, a telephone exchange serving dozens of subscribers — higher power and lower distortion matter practically.

The push-pull circuit delivers both. Two tubes, each handling one half of the audio waveform, combine their outputs in the output transformer to produce the complete waveform. Each tube’s distortion products are equal and opposite, canceling when combined. The output transformer core carries no DC magnetization current, so it can be smaller for the same power rating. The result is a more powerful, cleaner amplifier from the same tube types.

Understanding push-pull design is essential for building the audio stages of radio stations, the power amplifiers for public address systems, and the driver stages of SSB and AM transmitters. The 6L6, EL34, EL84, KT66, and similar beam power tubes and power pentodes are almost exclusively used in push-pull pairs in any serious audio application.

Phase Splitter Circuits

Push-pull requires two equal and opposite signals to drive the two output tube grids. Creating these antiphase signals from a single-ended input is the job of the phase splitter.

The cathodyne (split-load) phase splitter uses a single tube with equal plate and cathode resistors. The signal at the plate is inverted relative to the input; the signal at the cathode is in phase with the input. If plate and cathode resistors are equal, the two output signals are equal in amplitude and opposite in phase. The cathodyne has unity gain (the cathode follower action reduces gain below 1) but provides inherently well-matched outputs because both signals come from the same tube with identical component values.

The long-tailed pair phase splitter uses two tubes sharing a common large cathode resistor (the “tail”). An input signal to the first tube produces an amplified inverted signal at the first plate and an amplified in-phase signal at the second plate (through the common cathode connection). Equal plate resistors make the two outputs equal. The long-tailed pair provides gain (typically 10-30 times) and is used in higher-performance designs where the cathodyne’s unity gain is insufficient.

The transformer phase splitter uses a driver transformer with a center-tapped secondary. The two halves of the secondary produce equal and opposite signals. This was common in early push-pull amplifiers and is still used in instrument amplifiers where simplicity is preferred. The transformer’s frequency limitations (low-frequency rolloff, high-frequency rolloff from winding capacitance) are the main disadvantage.

Output Transformer Design

The output transformer is the most critical and most difficult component in a push-pull amplifier. It must:

• Present the optimum plate-to-plate load impedance to the output tubes
• Couple this impedance to the speaker (4, 8, or 16 ohms)
• Pass the full audio bandwidth (typically 30 Hz to 15 kHz)
• Carry the output tube cathode currents in opposite directions through the primary (eliminating DC core magnetization)
• Handle the full output power without saturation

Primary winding: center-tapped, with each half connected to one output tube plate. The center tap connects to B+. The winding must handle the peak plate current without resistive losses exceeding 2-3% of total power.

The plate-to-plate impedance (the total primary impedance from plate to plate across the full winding) determines the optimum load for the output tubes. Typical values: 4000-8000 ohms for pairs of EL34, 6L6, or similar tubes at standard supply voltages.

Secondary winding: wound to give the turns ratio needed to match primary impedance to speaker impedance. Turns ratio = √(Zprimary / Zspeaker). For 6000 ohm primary and 8 ohm speaker: turns ratio = √(6000/8) = √750 = 27.4:1, meaning 27.4 primary turns for every 1 secondary turn.

Core sizing: the core must not saturate at maximum power and low frequency. The maximum operating flux density for standard grain-oriented silicon steel is about 1.0-1.5 Tesla. The required core cross-section:

A = Vprimary / (4.44 × f × N × Bmax)

where f is the lowest frequency (30 Hz), N is primary turns per section, Bmax is maximum flux density. For 200V RMS (approximate plate voltage swing at maximum output), 30 Hz, 500 turns per primary half, 1.2 Tesla:

A = 200 / (4.44 × 30 × 500 × 1.2) = 200 / 79,920 = 0.0025 m² = 25 cm²

This is a large core — 25 cm² cross-section. Using laminations of width 50mm and height 50mm (2500 mm² = 25 cm²) exactly meets this requirement. Larger cores are better; smaller cores produce bass rolloff and transformer saturation at high power.

Biasing Push-Pull Stages

Both output tubes must be matched in plate current at the same operating point, otherwise the output transformer center tap carries a net DC current, partially magnetizing the core and degrading low-frequency response.

Cathode bias: a single cathode resistor shared between both tubes (connected between the common cathode point and ground) provides automatic DC balance. If one tube draws more current, it develops more voltage across the shared cathode resistor, increasing the bias on both tubes proportionally. This negative feedback mechanism helps balance the pair automatically. The shared cathode bypass capacitor must be large enough to bypass audio frequencies effectively.

Fixed bias: individual grid bias resistors from a negative supply, with individual adjustment trim pots for each tube. This allows the plate current of each tube to be set independently and precisely. Fixed bias supports higher output power (tubes can be biased closer to cutoff) but requires careful periodic adjustment as tubes age.

Matching tubes: regardless of the bias method, matched output tubes reduce transformer core DC magnetization and even-order harmonic distortion. Measure each tube’s plate current under identical conditions and select tubes with current within 5-10% of each other.

Practical Considerations

Swapping output tubes requires checking and adjusting the bias. Even matched tubes from the same manufacturer vary somewhat, and replacement tubes may have different idle current than the originals. Always measure and set bias after any tube change.

Parasitic oscillation is more common in push-pull stages than in single-ended stages because the two output tubes and the output transformer form a complex resonant structure. Anti-parasitic resistors (typically 100-470 ohms) in series with each plate lead, mounted directly at the tube socket, damp high-frequency parasitic oscillations that would otherwise cause severe distortion or tube damage.

DC offset at the speaker terminals indicates unbalanced output tubes — one tube drawing more current than the other. Measure DC voltage between speaker terminals and chassis ground. Values above a few millivolts indicate imbalance. The DC flows through the speaker voice coil, heating it and potentially damaging the speaker over time. Correct imbalance by matching tubes or adjusting individual bias trim pots.