High Performance Power Amplifier.doc

High Performance Power Amplifier (org. design sept.1983; redesigned in the Summer of 1997)

In the beginning...

About 20 years ago I built my first pair of loudspeakers from a ready-to-build design called 'Rogers Monitor', with the famous Kef B139 as bass chassis. A huge transmission line loudspeaker, with a complex 4-way cross-over network.
Being a poor student, I owned a simple integrated amplifier, unable to drive the two giants now occupying lots of space in my tiny room. Bass control wasn't as good as I wished, and I was convinced it would be better with a more powerful amplifier.
In those days I was studying to become a electronics engineer (and succeeded) and thought I could do with some practice in building an amplifier myself.

So I began my amplifier experience by rebuilding a very good amplifier from a popular electronics magazine (Electronics Today International). The article was very well written, and the resulting amplifier was indeed very good. The amplifier had control over the speakers, and made them come alive. A very accurate and direct sound was created by the combination of amps and speakers, which I then highly appreciated.

But there were some drawbacks. Because I used a voltage of +/-45volts (higher than advised), the power dissipation in the output transistors was too high, resulting in the loss of many output transistors, and at least one very expensive bass loudspeaker. Later I built a copy of the amp for a friend, and used a lower PSU voltage. This amp still runs, after all these years.
At this time I read many magazines on the subject of high-end audio equipment and electronics, expanding my knowledge on high-end amplifier techniques. Combining many of those techniques led to a first attempt in designing my own power amplifier, a 150W per channel transistor amp, which appeared to be very successful. A few of my friends still use this amplifier, which performs very good with 'easy' loudspeakers.
The success of this amp led to a totally new design, in which I attempted to make an amplifier that could catch up with the best transistor amplifiers in the world. My great example in those days was the legendary Mark Levinson No.23 power amplifier. That amplifier is described here, in its latest version.
The amplifier is not cheap. It will cost you more than a ready-built commercial amp of good quality, but in return you will get very good, if not the best, value for money.

But now, have a look at the schematic diagram.

Input Stage.

In the original design I used electrolytic capacitors as a DC block in the input stage, but the amp sounds much better when using MKP's instead. Disadvantage here is the large capacitor value we need for near-DC frequency response. By using 270k input resistors, the input resistance of the amplifier, as seen from the preamplifier, will be greater than 100k, a value which suits most (tube-)preamplifiers very well.
To minimize DC-offset, the overall feedback resistor should also be 270k. This value is high, and might be causing high frequency response limitation, due to parallel capacitance. But, on the other hand, this causes a 'natural' high cutoff which prevents the amp from oscillating.

As you can see, the input stage is actually a differential amplifier in its most basic form. The potmeter between the to emitters of the BC556's, is used to adjust output DC-offset. The 20k potmeter in the current source (traditional current mirror configuration), is used to set the voltages on points Ua and Ub. Initial value should be around 9k. Voltage on points Ua and Ub should be around 3.1 volts, measured from negative PSU voltage.
 
Second Stage.

This is the most crucial stage of the amplifier, and should be of very high quality: the final sound quality is made here... The clipping voltage of the amplifier depends on the base-voltages of both the NPN-voltage amplifier and the PNP cascade current source. The current source clips at 3 volt from the positive supply, which means, when symmetrical clipping is preferred, Ua = Ub = 3 volt. This way the two emitter resistors and the current source can be calculated. Because the collector of the first NPN transistor is connected to the common of the psu, power dissipation in this transistor is known. This leads to only a few possible transistor combinations, of which the BD791/BD792 pair seems to be the best. All four transistors must be placed on the same heatsink to maintain linearity when heating up.
Because the BD791/792 might be hard to get, the BD139/140 combination may be used, without audible difference.
(then, why not use the 139/140 in the first place? well, thermal breakdown of the 791/792 is slightly better...)

The output stage.

The output stage is actually very common, except maybe for the darlington drivers. By using high power darlingtons, the second stage has only to deal with voltage amplification, which keeps it almost perfectly linear. The output transistors are the well known MJ15003 and MJ15004, capable of handling 300 Watts dissipation each, which means with a very large heatsink (or a low noise fan) , quiescent currents of up to 5 Amps are no problem, resulting in a medium power class A amplifier. I am using the amplifier with a quiescent current of about 1 Amp, but feel free to try it with higher currents. But beware!: the BD139 that controls quiescent current in the output stage, isn't capable of regulating the current over a large temperature range. This means the amp will only function very well when used with stable ambient temperatures. As we can not control weather circumstances, the use of a temperature controlled fan should be considered.

The resistor between the emitters of the TIP142/147 darlingtons may look a bit strange, but believe me, the amplifier sounds better with it. I experimented with different values; the 39R seems to be the best (using a quiescent current of 1 Amp, that is).
To get some more presence in the mid frequency range, parallel this resistor with a 2u2 MKP capacitor.

The emitters of the output transistors are coupled with .22R resistors to create a stable quiescent current (without the resistors the quiescent current would vary with the temperature of the output transistors...). The speaker output is best coupled with a high current silver contact relay (don't use gold; the contact resistance of silver is lower, resulting in a higher damping factor), that switches of in case of trouble with the mains supply. Always use vacuum relays, otherwise you'll have to replace them every few months.

The power supply.

The power supply is the main reason why many commercial available amplifiers sound so poor. Basically good designs can be ruined by a power supply. What goes for the output stage is even more important for the power supply; try to make it as powerful as you can. A simple rule I use is to spend at least twice as much money on the power supply as on the amplifier itself.

I now use a medium power torroidal transformer (2*33V/10A) to get a supply voltage of +/- 48 volt, enough for about 100 Watts/8ohm. The rectifier should be a bridge rectifier capable of handling at least 50A continues current, and power peaks of 150 Amps (good torroidal transformers should be capable of producing peak currents to the extent of ten times the maximum RMS current). Use ultra high speed rectifiers, or place small resistors between the transformer and the input of the rectifier.
The most expensive parts of the power supply are the electrolytic capacitors. Try to find capacitors capable of handling very large ripple currents (about 5 Amps). Never use so called 'Audio grade' capacitors. The best capacitors for audio use are so called 'computer grade' caps, made to withstand high ripple currents at high frequency. I now use 4 capacitors of 64,000 uF per channel, which is enough for very high sound levels on almost every loudspeaker. Capacitors like these are twice the size of a can of Budweiser, and cost about $150 each. (The best you can get nowadays are the Philips PED-ST 114 types; the 63V/47000uF have catalogue number 2222.114.18473, the 40V/68000uF types are 2222.114.17683; if you don't want to spend that much, you could go for the PEC-ST 154 series) Use very thick wire to connect the rectifier to the capacitors, I normally use a bundle of 4 (+/- supply) and 6 (common) twisted 2.5mm mains wires. Then the amplifier board is connected directly to the capacitors, using the same type of wires.

The PCB.

The pcb is of course a high quality double sided pcb. The top layer is used as signal common.

I have an upgraded version of the PCB available, hppa.zip for more information.