This project originates from the recovery and complete redesign of a push-pull 2A3 vacuum tube amplifier, originally built in an approximate and technically incorrect manner. The goal was not to “fix” a poorly conceived device, but to use the chassis and part of the available components as a starting point for a properly engineered project, coherent and operating according to correct electrical criteria. The result is a parallel push-pull 2A3 amplifier with rational circuit solutions, measurable performance, and a sonic result clearly superior to the original unit.

Upon opening the chassis I was faced with the typical situation that unfortunately occurs often with self-built equipment lacking adequate expertise: flying wiring with no mechanical anchoring, randomly distributed grounds, a total absence of electrical safety criteria, and a schematic that represents exactly what should never be done in a vacuum tube amplifier. From that point on, the only viable option was the complete demolition of the original device and its reconstruction using a new schematic designed from scratch.

The adopted schematic is a parallel push-pull design with four 2A3 output tubes, biased to operate in a class AB condition very close to pure class A, with a DC-coupled input stage and a long tail phase splitter. The entire signal path was redesigned to eliminate instability, RF self-oscillation, and memory effects that heavily affected the original unit. The whole project was developed while taking into account the limits imposed by the available transformers, which unfortunately proved to be the real bottleneck of the entire amplifier.

The mechanical reconstruction was deliberately kept simple and functional. No aesthetic ambitions, but a practical approach aimed at restoring order, cleanliness, and structural solidity.

Simple spray painting was chosen to keep costs down and focus exclusively on the electrical and functional aspects of the project.

The power supply section required several compromises. The power transformer provided many secondary windings, some of which were connected in parallel to obtain adequate current for the filaments. A single 5Z3 rectifier is used instead of a pair of 5U4 tubes, and the 2A3 output tubes are powered in pairs rather than individually. Due to the available voltages and the lack of space for suitable heat sinks, it was not possible to implement a DC filament supply for the output tubes, which are therefore powered with AC. The small signal tubes, on the other hand, are powered with AC through a simple series dropping resistor on the dedicated secondary.

The input stage uses an ECF80 tube. The triode section is employed as the input amplifier, while the pentode section operates as a constant current source (CCS) for the phase splitter. The latter is implemented with a 6SN7 configured as a long tail pair, replacing the original cathodyne stage. This choice resulted in a clear improvement in symmetry, stability, and linearity of the driver stage.

During final tuning, the most serious limitations of the project emerged. The power transformer and output transformer, not designed by me, exhibit internal resistance and saturation levels much higher than expected. Under load, the anode supply voltage drops significantly, preventing full exploitation of the four 2A3 tubes. As a result, the bias was deliberately kept low to avoid excessive voltage sag.

The final result is an output power lower than what could theoretically be achieved with four well-powered 2A3 tubes, but accompanied by distortion levels significantly lower than those of the original version. A moderate amount of global negative feedback was applied, sufficient to increase the damping factor and make the amplifier stable and quiet.

The hum problem, very evident in the original version, was completely eliminated. All RF self-oscillation and thermal instability issues, as well as signal memory effects, were also removed. In the audio path only two coupling capacitors remain, between the phase splitter and the output stages. The input stage and the coupling between the input and phase splitter are fully DC coupled.

Below are the measured performance figures of the amplifier in its final configuration:

Power: 20 watt RMS peak, 15 watt continuous
Bandwidth: 20 Hz -0 dB / 80 kHz -3 dB
Damping factor: 2.85
Rout: 2.8 ohm
THD at 1 watt: 0.054%
THD at 10 watt: 0.36%
THD at 15 watt: 0.92%