20 Watt Class-A Power Amplifier
Rod Elliott - ESP (Original Design)
Updated 30 Mar 2001

A single-ended Class-A amplifier is essentially one where there is only one active driven output device. The passive "load" may be a resistor, an inductor (or transformer) or - as in this amplifier - a current sink. Of the three basic options, the current sink offers the highest linearity for the lowest cost, so is the ideal choice.

Some esoteric (some might say idiosyncratic) designs use inductors or 1:1 transformers, but these are bulky and very expensive. Unless made to the utmost standards of construction, they will invariably have a negative effect on the sound quality, since the losses are frequency dependent and non-linear.

This amp uses the basic circuitry of the 60W power amp (see Index), but modified for true Class-A operation - it should be pretty nice! This amp has been built by several readers, and the reports I have received have been very positive.

With simulations, everything appears to be as expected, but although I have yet to actually build it and test it out thoroughly, no-one has had any problems so far. Using +/-20 Volt supplies - either conventional, regulated or using a capacitance multiplier, it should actually be capable of about 22 W before clipping, but expect to use a big heatsink - this amp will run hot.

Quiescent current has been reduced from my earlier attempts and simulations from a bit over 3A down to 2.6A - but it will still dissipate nearly 110W per amplifier!

There are a few things which must be considered - In my original article, I suggested a suitable current sink. Although this would certainly work, the dissipation actually exceeds the maximum for the MJE2955 devices. Running at 55 W each, and considering that they will be at an elevated temperature (probably around 70°C), the maximum safe power is only a little over 45W, so clearly two devices must be used. With two, the dissipation of each transistor is "only" 27.5W, and this also allows a lower thermal resistance from case to heatsink.

I strongly suggest that you use either TO-3 transistors, or large (high dissipation) plastic case devices. Heat transfer from transistors to heatsink will be the biggest problem you will face with this amplifier.

Figure 1 - Power Derating For The MJE2955

An alternative is to use bigger transistors (even reverting to the TO-3 style), but in the long run using two paralleled transistors is still a cheaper option, and provides an adequate safety margin for the MJE2955 devices. Note that TIP2955 transistors may also be used, since they are more or less direct equivalents for this design. If you want to use more robust devices, I suggest TIP36 (A, B or C).

The modifications from the original 60W amp are as follows:

• Biasing diodes and the 47 Ohm resistor removed

• Lower transistor array removed, and replaced with a current sink.

• Power supply voltage reduced to +/-25V Maximum (+/-22V recommended)

• The "tail" of the long-tailed pair has been simplified to a simple resistor

• The DC offset can be set using the trimpot - this balances the LTP

• However big the heatsink you were thinking of, use a bigger one !

Figure 2- The New 20W Class-A Amplifier

The current sink shown should have very high linearity, since it is based on the same concept as the output stage devices. The 0.25 Ohm resistor should cause little grief (4 x 1 Ohm 1W resistors in parallel), but some experimentation may be needed here, since the base-emitter voltage of the BC549 determines the current. This circuit works by using the BC549 to steal any excess base current from the compound pair. As soon as the voltage across the 0.25 Ohm resistor exceeds 0.65V, the transistor turns on and achieves balance virtually instantly.

The 1k trimpot in the collector of the first LTP transistor allows the DC offset to be adjusted. The nominal value is around 400 ohms, but making it variable allows you to set the output DC offset to within a few mV of zero.

Determining The Optimum Current

The ideal operating current for a Class-A amp will be about 110% of the peak speaker current. If the loudspeaker system has a nominal impedance of 8 Ohms (the design impedance for this amp), then with a +/- 22V supply the maximum (theoretical) speaker current is ...

I = V / R = 22 / 8 = 2.75A

In my original calculations, I decided on a quiescent current of 2.6A - this is really Ok, because the above calculation does not consider the losses in the output stage. In practice, it is likely that up to 3 Volts will be lost in the output circuit, based on the losses in the output devices, emitter resistors and driver transistors.

This now gives a maximum voltage of 19V peak (2.375A @ 8 Ohms). Applying the 110% fudge factor gives an operating current of 2.6125A, or 2.6A close enough. If these peaks are met in practice, this gives an output power of 22.5W into 8 Ohms.

Note that the current in the -ve supply rail remains constant, but that in the +ve supply rail will vary from the normal steady state current (same as the -ve supply). At signal extremes, the current will double (upper transistors turned on), or will drop to almost zero for negative peaks. This is common for single-ended Class-A amplifiers, although you will not see it stated in the text for most designs. This can complicate the design of the power supply.

Adjusting The Quiescent Current

If the current sense resistor is made a higher value than optimal (say 0.33 Ohm 5W), you can use a trimpot across the resistor with the wiper going to the base of the BC549. This will allow you to set the current more accurately. Note that the sense transistor must be kept away from heat sources (such as heatsinks and power resistors) or the current will fall as the amp gets hotter. Be very careful if you use a trimpot, because if the wiper is wound down to the -35V supply line, the current sink will attempt to sink infinite current - this is likely to cause damage (to put it mildly). Start with the wiper at the most positive end (i.e. the collectors of the output devices), and carefully increase the current until the desired setting is reached. Use of a multiturn pot is highly recommended (almost mandatory, actually).

Figure 3 - Variable Current Source

Figure 3 shows a suggested way to make the current sink variable. The 1k fixed resistor ensures that even if the pot becomes open circuit (which does happen, although rarely), the stage will not try to sink an infinite current. Remember to allow time for the temperature to stabilise - this may take 10 minutes or more, depending on the size of the heatsink. Larger heatsinks have a greater thermal mass, and take longer to reach the final operating temperature.

The heatsink is a critical part of a Class-A design, and for this amp a sink with a thermal rating of less than 0.5°C / Watt is mandatory. With a dissipation of about 110W quiescent, an 0.5°C/W heatsink will give a temperature rise (above the ambient) of 55°, so for the "British Standard" 25°C ambient temperature the transistors will operate at 80°C. This is hot. If possible, 0.25°C/W thermal rating is preferred, which will keep the temperature down to a more moderate 55°C or so - this is still hot but tolerable.

I suggest that any intending builder reads the article on heatsinks, to gain a better understanding of the difficulties involved in obtaining a good thermal transfer from transistor to heatsink. The use of TO-3 power transistors (MJ2955) will also help considerably in this respect.

---