Following on from the 30W amplifier project, one day I paid a visit to the local audio shop where I was on quite friendly terms with the owner and announced to him that I was considering starting a new amplifier design – in the order of 120 Watt output. He asked if I would be interested in working on a joint venture on this one and suggested he would be willing to fund the project. Thus began my association with the partners of GGAcoustics which later became Vivid Audio.
My intention behind this design was to start with a clean piece of paper and put into practice all the knowledge that I had gained in my work on the 30 Watt amplifier as well as overcome the limitations that that design held. This was to be a no compromise design where audio quality took precedence over size, weight or cost.
120 Watt RMS Dual mono-block power amplifier.
8 Ohm load nominal down to 4 Ohm (120W – 8 Ohm, 240W – 4 Ohm).
Fixed voltage gain of 27 dB.
Balanced output design with balanced inputs.
Better than 100dB signal to distortion at rated power output.
The first part of the design process involved deciding on the power supply configuration, including selecting a suitable transformer. Choosing a transformer for an audio amplifier is not a trivial task and there is no exact formula to determine the required power rating of the transformer. In my experience most budget audio amplifiers and PA amplifiers incorporate transformers that are grossly underrated for their specified audio output power. You need only to try picking up a budget “stereo” that boasts a power output of 1000 Watts PMPO to discover it doesn't weigh much which implies it has a feeble transformer inside. In comparison a 1000VA transformer is quite a serious beast likely to weigh between 5 - 10kg. Transformers are rated in Volt-Amps which is only meaningful if the load is linear, however the typical bridge rectifier and smoothing capacitor arrangement used to supply power to an amplifier provides anything but a linear load to the transformer. In addition different transformer manufacturers might label the same transformer with a different VA rating depending on what software tools or methods they use to design them. The transformer VA rating is essentially defined as that load power that will cause a certain temperature rise in the transformer due to losses in the windings and core. The acceptable temperature rise being chosen as a safe value that will not cause long term damage to the transformer. With audio signals the average power demand is quite low compared to the peak power demand so this would imply that the transformer choice can be de-rated without compromising reliability. However with a bridge rectifier and smoothing circuit the RMS current drawn from the transformer can be many times greater (10 times or more) than the average DC load current, thus with high audio peak powers this will cause greater instantaneous losses in the transformer windings. The net effect is that an underrated transformer will deliver poor supply regulation in an audio amplifier even if it does not experience excessive heating.
To understand how this will affect the amplifier performance consider test driving two motor vehicles – one has a 1.3 litre engine and the other a 2.5 litre engine. You drive both down a stretch of highway at 100km/h. On a flat section of road both vehicles can maintain 100km/h but the 1.3 engine is likely to be working fairly hard whereas the 2.5 engine is barely making an effort. As you approach an incline the 1.3 really has to work and may not be able to maintain the vehicles speed even if you floor the accelerator (you can certainly hear it battling), but the 2.5 cruises up the incline effortlessly with no drop off in speed. Much the same happens in audio amplifiers, if the power supply under the hood is feeble, it might still be able to produce the same audio output level of another amplifier with a larger power supply, but just give them both a particularly demanding piece of audio to reproduce and you'll know in an instant which is more capable. This can be apparent even at low listening levels. From a technical point of view, the larger transformer presents a lower impedance to the amplifier such that when faced with a sudden demand from the amplifier, the larger transformer is better able to hold up the supply voltage. This rule can be extended more generally to the whole power supply – rectifier diodes, smoothing capacitor and even the wiring.
Now in order to assist me in choosing a transformer I wanted to be able to simulate a particular transformer and power supply under full load so I set about writing a C language PC utility using Borland BGI graphics that would simulate the charging/discharging cycles of a transformer/rectifier/smoothing capacitor circuit combination. In this simulation I incorporated the transformer winding impedances as a single lumped resistance. This utility uses a numerical method to incrementally predict voltages and currents at various points in the circuit, much like a standard circuit simulator does. The simulation can be run for any number of sine wave cycles so that one can observe when the charging/discharging of the smoothing capacitor has reached a steady state. The utility plots the waveforms of the various currents and voltages as well as reports peak, mean and RMS values. I verified the results of the simulation against a real transformer and power supply circuit, using an oscilloscope to view the waveforms. I also emulated different transformers by inserting different resistances in series with the transformer secondary windings. There was good agreement with the simulation and measured values and this also proved that the dominant factor affecting transformer regulation is the winding resistances.
My Transformer and power supply simulation utility
Unlike the 30W amplifier design which was singled ended I decided to make this 120W design a balanced design. There are various pros and cons for using a balanced design. From a power supply point of view a singled ended design requires a dual supply where the voltage difference from the positive rail to the negative rail must be in excess of the required peak to peak maximum output swing of the amplifier. This configuration also requires two smoothing capacitors which are usually big and expensive. For a balanced design a single supply voltage is required that need only be greater than the peak voltage swing of the output signal, thus a balanced design requires approximately half the supply voltage of a single ended design. From the transformers perspective this means half the number of windings on the secondary, thus half the secondary resistance. For a specific transformer with fixed VA rating if you halve the secondary voltage you double the rated secondary current, thus the secondary wire gauge size can be increased. So effectively you end up with a transformer where the secondary winding resistance is reduced by a factor of 4. In addition since the turns ratio is halved, due to the transformation of the primary impedance the loss contribution of the primary windings reflected to the secondary is also reduced by a factor of 4. Thus you win in all aspects. Using my simulation utility it was clearly obvious that the lower the transformer resistance the better. A balanced design also only requires one smoothing capacitor - another big advantage. Since the power supply is halved, quiescent heating losses in the power devices are also reduced by a factor of 4 so the heat sinking requirements for a balanced design can be reduced and component stresses are less. This also means that active devices with lower voltage ratings can be used in the design. These advantages are significant when it comes to designing high power amplifiers. Balanced designs are also inherently immune to noise and interference so mains hum is not an issue. On the down side, a balanced design requires two identical amplifier circuits, thus twice the number of components and PC board space, and the effective output impedance of the amplifier will be twice that of a single ended design. This has an effect on the damping factor for the speakers. However on the whole the advantages of a balanced design far outweigh the disadvantages.
Having decided on the supply configuration, voltages and currents I then approached a toroidal transformer manufacturer with my specifications for a custom design and along with the quote requested they supply me with details on the primary and secondary windings, turns ratio, wire gauge and calculated primary and secondary resistances. They willingly supplied me with this information which I plugged into my simulation utility. I was then able to make minor changes to the transformer design to ensure the supply voltage remained within my design specifications under all load conditions, the changes only involved adding or subtracting a turn or two on the primary and secondary. I supplied the manufacturer with the updated design details and they manufactured two transformers according to these specifications where I was to use one transformer for each channel of the stereo amplifier.
For the amplifier circuits themselves I used the same basic circuit topology I had used in the 30W design since it was tried and trusted, but with some minor changes to certain parts of the circuit for improved performance. For the output devices I used some large On-semi TO-3 audio transistors, paralleling up devices for reduced output impedance and improved drive capability. For component value and biasing calculations I made use of the programs I had written in my HP48G calculator for estimating the distortion levels produced by the various amplifier stages. I laid out the twin amplifier circuits for the balanced configuration on a single PC board, but in “mirrored” configuration and in particular was quite generous with the track width I used around the output stage circuitry and made use of copper planes for the output nodes. I designed a separate board to provide mains filtering and soft start capabilities for the toroidal transformers and power supply. The boards were manufactured commercially and I assembled them myself. I constructed the amplifier on a large piece of wood along with the toroidal transformers and smoothing capacitors. The amplifier boards were mounted on the back of large heatsinks for cooling and a heatsink/board combination constituted a single channel module.
The smoothing capacitors I purchased were 10000uF 100V low ESR EPCOS devices and so to achieve a decent storage capacity I paralleled 5 of them up on the wooden board to make 50000uF using “bus bars” made from strips of un-etched PC board. I was thus able to add or remove capacitors and assess the affect on the audio quality. Once the amplifier was complete and working I took it to the local audio shop to show the owner and give it a test drive alongside some of his other equipment. Sonically it easily matched up to the level of the best equipment he had in stock (Madrigal Laboratories Mark Levinson No. 336). However he pointed out that the bass was a bit weak. Initially I had connected up only two smoothing capacitors per channel, thus 20000uF. So I added the remaining 6 capacitors and we tried it again, this time he was impressed.
Not long after this my friend and his partner entered into negotiations with some audio contacts from the UK who they were hoping to enter into a business agreement with. They made use of my amplifier design to demonstrate our electronic design capabilities to the overseas contacts, one of whom agreed to form a business with them – this was where GGAcoustics began and the partners later created the business entity known as Vivid Audio.
November 2000 - May 2001