This post features a project from 1981, my second to last attempt at design and construction of a low distortion audio power amplifier. It took several months on the bench, tweaking constant current sources  before etching the PCB and a few weeks cutting out the aluminium panels and wiring it all together.

Hitachi’s HMA7500 offering would have been a easier option but not as much of a challenge although it was quite expensive back in the 80s. Designed around a complementary pair of 2SK134 and 2SJ49 power Mosfets, I wanted to focus on peak current delivery, slew rate and of course avoiding the cross-over distortion that can ruin class A/B output stages.

A clue to how this was achieved to some extent is shown in the under the hood photos below. The wiring loom uses thick silver plated PTFE covered wire for all the high current paths with a star earth point to minimise earth loops.

One early limitation was the single 31-0-31v toroidal transformer I had available, so there’s a pair of heavy duty low ESR Gould 10,000µF reservoir capacitors hiding under the regulated control power supply board. The capacitors smooth the main ±42v DC to each power amplifier. An improved design would have separate toroids for left and right channels but would not fit in the box! This single toroid and caps provide 6% regulation at full output power, an acceptable compromise.

A regulated +-30v DC supply feeds a rear socket to power a separate preamplifier. This design avoids potential earth loops that could arise from a mains powered preamplifier.

Instead of analogue meters to display signal power I opted for LED displays which in the 80s seemed more state of the art. Now of course we hanker after the analogue throughout. It seems amusing looking back that the Japanese design with big moving coil meters is now seen as very cool retro but was used to disguise the bulky non-toroidal transformer used.

My initial amplifier design used Hitachi Semiconductor component application notes with modifications to improve open loop gain and linearity. It’s essentially a discrete operational amplifier with a differential pair input, constant current sources and direct coupled push-pull mosfet output stage. Douglas Self has since published several articles and designs detailing audio amplifier design if you are interested in the subject.

The Hitachi Mosfets were supplied by Ambit International, based in Brentwood Essex. Ambit sold specialist discrete devices and various kits for serious consumers that could be used to build high quality audio systems.

Ambit’s PA100 power amplifier kit design was also based on Hitachi’s application note and provided inspiration for my PCB layout. I remember Ambit’s brightly coloured catalogues full of ideas for projects. I’ll add some scans from these later…

Power output achieved just under 75 watts RMS at 20KHz into 6-8Ω at 0.05% distortion. At normal levels of around 10w the distortion was below 0.01% the limit of my analyser. The tiny ceramic preset on each power amp board allowed the quiescent current to be set to reduce distortion and run the devices warm at class A/B. Slew rate measured 40v/µS with a 5pF feedback capacitor for high frequency stability. With the capacitor removed slew rates of 116v/µS were possible measured into resistive loads.

What did I learn? You need to have patience to try to achieve perfection and to stop when you get to 80%. Audiophiles will notice a couple of relays in the box and undoubtedly sneer. Yet more resistance and more galvanic contacts in the analogue audio path between Mosfet and speaker. But without a DC offset protection the ultra-low resistance allows very high currents to flow and damage the precious speakers very quickly should a fault develop. I used Hitachi’s HA12002 power protector and a heavy duty Omron relay to ensure output devices were disconnected in a few milliseconds should such a fault occur. 5A anti-surge fuses in both A.C secondary feeds from the transformer act as backup in case the offset protection fails or should another fault occur.

Other component choices were made based on optimal audio quality. All resistors in the amplifier circuit were 2% metal oxide offering lowest noise. Low value capacitors were Suflex metal foil types. Ceramics were only used for decoupling outside the audio frequency path. Headphone output was attenuated using separate wire wound resister ladder connected via another relay.

What does it sound like? Awesome. It never sounds too loud, you can feel the power with the right set of speakers. I used Rogers Studio 1s with the thickest loudspeaker cable I could afford. In the end I did buy Hitachi’s HCA7500 control amplifier as a pre-amp so I could connect a range of analogue sources including the Ortofon Concorde moving coil cartridge and Dynavector Ruby.

I designed a second version, built for a good friend Colin, that had a slimmer toroid with higher voltage so it pumped out over 100w RMS at similar low distortion levels. It drove a pair of Bose speakers.

Ambit Catalogue Archive Extracts (PDF)

Hitachi Archive (PDF)