300W Subwoofer Power Amplifier

Please note that the specification for this amp has been upgraded, and it is now recommended for continuous high power into 4 Ohms, but You will need to go to extremes with the heatsink (fan cooling is highly recommended). It was originally intended for "light" intermittent duty, suitable for an equalised subwoofer system (for example using the ELF principle - see the Project Page for the info on this circuit). Where continuous high power is required, another 4 output transistors are recommended, wired in the same way as Q9, Q10, Q11 and Q12, and using 0.33 ohm emitter resistors.

Continuous power into 8 ohms is typically over 150W (250W for ±70V supplies), and it can be used without additional transistors at full power into an 8 ohm load all day, every day. The additional transistors are only needed if you want to do the same thing into 4 ohms at maximum supply voltage! Do not even think about using supplies over ±70V, and don't bother asking me if it is ok - it isn't!

The circuit is shown in Figure 1, and it is a reasonably conventional design. Connections are provided for the Internal SIM (published elsewhere on the Project Pages), and filtering is provided for RF protection (R1, C2). The input is via a 4.7uF bipolar cap, as this provides lots of capacitance in a small size. Because of the impedance, little or no degradation of sound will be apparent. A polyester cap may be used if you prefer - 1uF with the nominal 22k input impedance will give a -3dB frequency of 7.2Hz, which is quite low enough for any sub.

300W Subwoofer Power Amplifier Skema

The input stage is a conventional long-tailed pair, and uses a current sink (Q1) in the emitter circuit. I elected to use a current sink here to ensure that the amp would stabilise quickly upon application (and removal) of power, to eliminate the dreaded turn on "thump". The amp is actually at reasonably stable operating conditions with as little as +/-5 volts! Note also that there are connections for the SIM (Sound Impairment Monitor), which will indicate clipping better than any conventional clipping indicator circuit. See the Project Pages for details on making a SIM circuit. If you feel that you don't need the SIM, omit R4 and R15.

The Class-A driver is again conventional, and uses a Miller stabilisation cap. This component should be either a 500V ceramic or a polystyrene device for best linearity. The collector load uses the bootstrap principle rather than an active current sink, as this is cheaper and very reliable (besides, I like the bootstrap principle :-)

All three driver transistors (Q4, 5 & 6)must be on a heatsink, and D2 and D3 should be in good thermal contact with the driver heatsink. Neglect to do this and the result will be thermal runaway, and the amp will fail. For some reason, the last statement seems to cause some people confusion - look at the photo below, and you will see the small heatsink, 3 driver transistors, and a white "blob" (just to the left of the electrolytic capacitor), which is the two diodes pressed against the heatsink with thermal grease.

C11 does not exist on this schematic, so don't bother looking for it. It was "mislaid" when the schematic was prepared, and I didn't notice until someone asked me where and what it was supposed to be. Sorry about that.

It is in the output stage that the power capability of this amp is revealed. The main output is similar to many of my other designs, but with a higher value than normal for the "emitter" resistors (R16, R17). The voltage across these resistors is then used to provide base current for the main output devices, which operate in full Class-B. In some respects, this is a "poor-man's" version of the famous Quad current dumping circuit, but without the refinements, and in principle is the same as was used in the equally famous Crown DC300A power amps.

Although I have shown MJL4281A and MJL4302A output transistors, because they are new most constructors will find that these are not as easy to get as they should be. The alternatives are MJL3281/ MJL1302 or MJL21193/ MJL21194.

Note: It is no longer possible to recommend any Toshiba transistors, since they are the most commonly counterfeited of all. The 2SA1302 and 2SC3281 are now obsolete - if you do find them, they are almost certainly fakes, since Toshiba has not made these devices since around 1999~2000.

Use a standard green LED. Do not use high brightness or other colours, as they may have a slighty different forward voltage, and this will change the current sink's operation - this may be a miniature type if desired. The resistors are all 1/4W (preferably metal film), except for R10, R11 and R22, which are 1W carbon film types. All low value resistors (3.3 ohm and 0.33 ohm) are 5W wirewound types.

Because this amp operates in "pure" Class-B (something of a contradiction of terms, I think), the high frequency distortion will be relatively high, and is probably unsuited to high power hi-fi. At the low frequency end of the spectrum, there is lots of negative feedback, and distortion is actually rather good, at about 0.04% up to 1kHz. My initial tests and reports from others indicate that there are no audible artefacts at high frequencies, but the recommendation remains.

Power Dissipation Considerations
I have made a lot of noise about not using this amp at ±70V into 4 ohms without the extra transistors. A quick calculation reveals that when operated like this, the worst case peak dissipation into a resistive load is 306W (4Ω/ ±70V supplies). The four final transistors do most of the work, with Q7 and Q8 having a relatively restful time (this was the design goal originally). Peak dissipation in the 8 output devices is around 70W each.

Since I like to be conservative, I will assume that Q7 and Q8 in the updated schematic shown contribute a little under 1A peak (which is about right). This means that their peak dissipation is around 18W, with the main O/P devices dissipating a peak of 70W each. The specified transistors are 230W, and the alternatives are 200W, so why are the extra transistors needed?

The problem is simple - the rated dissipation for a transistor is with a case temperature of 25°C. As the amp is used, each internal transistor die gets hot, as does the transistor case - the standard derating curves must be applied. Add to this the reactive component as the loudspeaker drives current back into the amp (doubling the peak dissipation), and it becomes all too easy to exceed the device limits. The only way that this amp can be used for continuous high power duty with ±70V supplies and a 4Ω loudspeaker load is to keep the working temperature down to the absolute minimum - that means four output devices per side, a big heatsink and a fan!

Double Output Stage

Figure 1A shows the doubled output stage, with Q9, Q10, Q11 and Q12 simply repeated - along with the emitter resistors. Each 1/2 stage has its own zobel network and bypass caps as shown, as this is the arrangement if the dual PCB version is built. When you have this many power transistors, the amp will happily drive a 4 ohm load all day from ±70V - with a big enough heatsink, and forced cooling. Over 500W is available, more than enough to cause meltdown in many speakers!

A Few Specs and Measurements
The following figures are all relative to an output power of 225W into 4 ohms, or 30V RMS at 1kHz, unless otherwise stated. Noise and distortion figures are unweighted, and are measured at full bandwidth. Measurements were taken using a 300VA transformer, with 6,800uF filter caps.

Mains voltage was about 4% low when I did the tests, so power output will normally be slightly higher than shown here if the mains are at the correct nominal voltage. Figures shown are measured with ±56V nominal, with the figure in (brackets) estimated for ±70V supplies.

8Ω 4Ω
Voltage Gain 27dB 27dB
Power (Continuous) 153W (240W) 240W (470W)
Peak Power - 10 ms 185W (250W) 344W (512W)
Peak Power - 5 ms 185W (272W) 370W (540W)
Input Voltage 1.3V (2.0V) RMS 1.3V (2.0V) RMS
Noise * -63dBV (ref. 1V) -63dBV (ref. 1V)
S/N Ratio * 92dB 92dB
Distortion 0.4% 0.4%
Distortion (@ 4W) 0.04% (1 Khz) 0.04% (1 Khz)
Distortion (@ 4W) 0.07% (10 kHz) 0.07% (10 kHz)
Slew Rate > 3V/us > 3V/us

* Unweighted

These figures are quite respectable, especially considering the design intent for this amp. While (IMO) it would not be really suitable for normal hi-fi, even there it is doubtful that any deficiencies would be readily apparent, except perhaps at frequencies above 10kHz. While the amp is certainly fast enough (and yes, 3V/us actually is fast enough - response extends to at least 30kHz, but not at full power), the distortion may be a bit too high.

Note that the "peak power" ratings represent the maximum power before the filter caps discharge and the supply voltage collapses. I measured these at 5 milliseconds and 10 milliseconds. Performance into 4 ohm loads is not quite as good, as the caps discharge faster. The supply voltage with zero power measured exactly 56V, and collapsed to 50.7V at full power into 8 ohms, and 47.5V at full power into 4 ohms.

Layout

The photo does not show the silk screened component overlay, since this is the prototype board. The final boards have the overlay (as do all my other boards). The observant reader will also see that the 5W resistor values are different from those recommended - this was an early prototype using 130W transistors.

As can be seen, this is the single board version. The driver transistors are in a row, so that a single sheet aluminium heatsink can be used for all three. Holes are provided on the board so the driver heatsink can be mounted firmly, to prevent the transistor leads breaking due to vibration. This is especially important if the amp is used for a powered subwoofer, but will probably not be needed for a chassis mounted system.

The driver and main heatsinks shown are adequate for up to 200W into 4 ohms with normal program material. The power transistors are all mounted underneath the board, and the mounting screw heads can be seen on the top of the board.