Summary:Introduction In various parts of The Audio Pages, I have said that I am not a fan of MOSFET power amplifiers. Well, this amp has changed my views, and I consider this to be a "reference" system in all respects. It uses lateral MOSFETs - not
In various parts of The Audio Pages, I have said that I am not a fan of MOSFET power amplifiers. Well, this amp has changed my views, and I consider this to be a "reference" system in all respects. It uses lateral MOSFETs - not switching types! The latter cannot be used in this circuit - they will self destruct! The performance is extremely good, with vanishingly low distortion levels, plenty of power, very wide full power bandwidth, and the "self protecting" nature of the MOSFETs themselves.
This is not to suggest that the amp is indestructible (no amplifier can make that claim successfully), but it is much more tolerant of faults than a bipolar transistor amp, and requires nothing more than a pair of zener diodes to limit the current. Having said that, I would still recommend that you avoid shorted output leads and the like - i.e. Don't push your luck ;-)
One thing that has emerged that is absolutely critical is the PCB layout. The layout of this new amplifier is similar to that used for the P68 Subwoofer amplifier, and this has some major benefits. P68 has no right to sound as good as it does, and although designed for subwoofer use, it has proven during listening and testing to be a very low distortion design - despite the Class-B output stages. All PCB tracks in the input and driver section are as short as possible, minimising the chance of noise pickup from other sections of the circuit.
This amplifier is designed to be as flexible as possible, with no bad habits. Indeed, it will operate stably with supply voltages as low as +/-5V (completely pointless, but interesting), all the way to the maximum supply voltage of +/-70V. The only change that is needed is to trim the MOSFET bias pot!
With the full supply voltage of +/-70V (which must not be exceeded!), RMS power is around 180W into 8 ohms, or 250W into 4ohms. Short term (or "music") power is typically about 240W into 8 ohms and 380W into 4 ohms. Note that depends to a very great degree on the power supply, and a very robust supply is an absolute requirement for the maximum output.
In general, unless you really need the maximum possible power, I suggest that you limit the supply voltage to ±56V using a 40+40V transformer. You will get around 150W into 8 ohms from this supply voltage (short-term), but you also relax the demands placed on the MOSFETs and heatsinks.
Since this amp probably has more power than you will normally need, even if you do skimp a little on the transformer, the loss will be very small. I must make one thing perfectly clear - this is a hi-fi amplifier. It is not designed for professional PA use, although this is possible if the supply voltage is reduced so that both peak and long term dissipation are maintained regardless of how hard the amp is driven. Although it is quite ok to use with a sub, P68 is a better proposition in that role.
It is worth noting that a MOSFET amp will always produce less power than a bipolar transistor version using the same supply voltage. Even using an auxiliary supply will make only a small difference (one reason I elected not to add the extra complexity). A bipolar design using a ±70V supply can be expected to produce something in the order of 270W into 8 ohms, and well over 500W into 4 ohms. The specified MOSFETs have a rated Vds (saturated voltage, Drain to Source) of 12V at full current, and that is simply subtracted from the DC value of the supply voltage. Using a ±70V supply with a MOSFET amp will give less power than can be obtained from a bipolar design (see below for measured figures).
The photo shows the simplicity of the PCB. The MOSFETs are mounted below the board, and are bolted down in the same way as with the P3A and P68 boards. No other mounting is needed. PCB pins or tinned copper wire pins are used as anchor points for the power ground link (the green wire along the front edge), so that the main current carrying tracks were not compromised by running a separate track (which would have required a reduction in size of the positive supply rail).
The entire front-end section is between the electrolytic caps, and is deliberately as compact as possible. This improves performance, by ensuring that there are no long tracks for the input stage, which will pick up noise and can seriously degrade the sound of the amplifier.Performance Figures
The performance of this amp is such that many measurements are very difficult. Some of the more basic measurements are as shown below, based on my custom made transformers which provide ±65V unloaded ...
In particular, the distortion figures show that amp loading causes only very small variations, with any harmonics being predominantly from my audio oscillator. There are no visible or audible high order components to the distortion waveform. Output impedance was measured on a fully built amplifier, including the internal wiring. This entails around 200mm of wire in all (per channel), so the output impedance of the amplifier itself is obviously lower than quoted. For an 8 ohm load, the damping factor at 1kHz is around 800 (8 / 10 milliohms) - completely pointless of course, since any speaker lead will ruin that very quickly.
Noise was measured with inputs open-circuited, and at -54dBV may not look too wonderful, however this figure is very pessimistic. Remember that this is the unweighted measurement, with bandwidth extending to well in excess of 100kHz. Even so, signal to noise ratio (referred to full power) is 86dB unweighted, and the amp is completely silent into typical speakers. Indeed, even connecting a pair of headphones directly to the amp outputs revealed that no noise was audible. Naturally, your methods of construction will differ from mine, and you may not be able to get the same performance.
Intermodulation distortion cannot be measured with the equipment I have available, but I have included a screen capture of the three measurements I took. Most of the harmonic content visible (not that there is a great deal anyway) is present in the two generators I used, and the amplifier contributes virtually nothing.
Click on any of the images above for a full resolution version.Description
The very first thing you will notice is that I have broken with tradition with this amp, and there are no component values shown. Given the performance of the circuit, and the fact that I have already sold a couple as completed, finished amplifiers, I am not about to give away all my secrets for the design. If you want the component values, you must purchase the PCB. There are no exceptions, so don't ask.
The schematic of the amp is shown in Fig. 1, and it is about as simple as a high power MOSFET amplifier can get - it is considerably simpler than most, but lacks nothing in performance. The circuit diagram belies the ability of the amplifier though, so do not be tempted to think that it cannot perform as well as more complex designs - it does, and exceeds the performance of many (if not most) of them. It will be seen that I elected to use a bootstrap current source rather than an active version - there is negligible cost difference, but I was unwilling to make such a radical change after testing the prototype and being so impressed with the results. (If it ain't broke, don't fix it!)
The front end is a conventional long-tailed pair (LTP) using a current mirror load and an active current sink in the "tail". Interestingly, adding the current mirror made no difference to distortion, but reduced the DC offset to less than 25mV. The improvement was such that I elected to retain the mirror.
In tests thus far (both measurement and listening), I have been unable to detect even a hint of what is commonly referred to as the "MOSFET sound". The relatively high levels of low order distortion and susceptibility to crossover (or "notch" distortion that plague many MOSFET designs are completely missing - indeed, even with zero bias on the MOSFETs, crossover distortion below 10kHz is barely measurable, let alone audible!
This is not a ploy on my part to get people to purchase my PCBs - that has already been taken care of by leaving out the component values. The simple fact is that unless the PCB layout is done with the utmost care, any amplifier can be made to have far greater distortion levels and reduced stability margins than the published figures suggest.
Low Power Version
The MOSFETs used are Hitachi lateral devices, 2SK1058 (N-Channel) and 2SJ162 (P-Channel). These are designed specifically for audio, and are far more linear than the (currently) more common switching devices that many MOSFET amps use. Unfortunately, they are not especially cheap, but their performance in an audio circuit is so much better than vertical MOSFETs, HEXFETs, etc., that there is no comparison. Note that using HEXFETs or any other vertical MOSFET type is not an option. They will fail in this circuit, as it was not designed to use them.
An alternative (and possibly marginally better than the 2SK/2SJ series) is the Exicon ECX10N20 and ECX10P20 (available from Profusion PLC in the UK). These have been used in most of the amps I have built, and they work very well. So potential constructors can verify that the semiconductors are available before purchasing a PCB, this information has now been included. All other parts are quite standard.
High Power Version
Although not shown, the transistors and MOSFETs are the same in this version as for the low power variant. The additional capacitors (C11 and C12) shown are to balance the gate capacitance. The P-Channel MOSFETs have significantly higher gate capacitance than their N-Channel counterparts, and the caps ensure that the two sides of the amp are roughly equal. Without these caps, the amp will almost always be unstable.
As noted above, the PCB is the same for both versions, but for Fig. 2 it is fully populated with 2 pairs of power MOSFETs. The high power version may also be used at lower supply voltages, with a slight increase in power, but considerably lower operating temperatures even at maximum output, and potentially greater reliability.
With both versions, the constructors' page gives additional information, and the schematics there include an enhanced Zobel network at the output for greater stability even with the most difficult load. This is provided for on the PCB, and allows the amp to remain stable under almost any conditions.
The entire circuit has been optimised for minimum current in the Class-A driver, while still providing sufficient drive to ensure full power capability up to 25kHz. The slew rate is double that required for full power at 20kHz, at 15V/us, and while it is quite easy to increase it further, this amp already outperforms a great many other amps in this respect, and faster operation is neither required nor desirable.
Note - There are actually two caps marked C5, and two marked C6. This is what is on the PCB overlay, and naturally was not found until it was too late. Since these caps cannot be mixed up, it will not cause a problem.
In both versions of the amp, R7 and R8 are selected to provide 5mA current through the voltage amplifier stage. You will need to change the value to use a different supply voltage ...