Brian Powell Audio

Thoughts on Power Amplifier Design

Myself and Tim Nind designed the original Crimson power amplifiers in 1976. We were both engineers at Decca Radio and TV in Leicester. The amplifiers at Decca were modern (for that time) in the sense of using split rails and having no output capacitor. This is a clear step forward from single rail design. The IPS used a differential pair with a resistive tail. Another modern feature for that time was the use of a current source load for the VAS rather than using a bootstrap capacitor. Output stages were fully symmetrical discrete Darlingtons. Unfortunately, at that time PNP power devices were rare and expensive. And the TIP31/32 output stage was not scalable to 60W, which was our target.

We used the split supply design with a current source tail and being fans of the Quad 303 used a similar output stage i.e. compound TRIPLES. Protection circuits were VI sensing. Stability was achieved with a combination of Miller feedback at the VAS and Miller inclusive feedback to the inverting terminal. See PCB 35-01 Issue 1 – Brian Powell Audio

The design morphed to become better and better over the next few years. But it was already a long way ahead of the generic power amps of that period.

Shortcomings of the Generic design c 1980:-

1. The VAS running at 9mA will struggle to produce a final output of more than 9 Amps assuming a gain of 1000 in the output stage.

2. The VAS will distort at low frequencies due to thermal modulation of the base emitter voltage.

3. The simple Miller compensation limits the slew rate.

4. The non-linear loading on the VAS reduces the open loop gain and increases the open loop distortion. Impedance at the VAS could be 20k. The loading with a 4 Ohm load and gain of 1000 is 4k and is an order of magnitude lower than an ideal VAS load.

5. The lack of inductor at the output makes the output stage less stable and reduces the damping factor.

6. The 2 diode voltage reference without feedback has poor supply rejection

Better Topologies For Higher Performance Design

The simple LTP direct to rail with resistive loads is inadequate. Use active loads and cascodes.

A single transistor VAS is inadequate, use a Darlington or cascode.

The Darlington output stage is inadequate, use triples or FETs.

Use voltage references with feedback to ensure high PSU rejection.