There are six mosfets per channel in the output in complementary source follower
mode.
The driver stage is also a complementary pair but with BF469/470 bjts in common
emitter mode for high gain. The collector load of each bjt is the load for the other
of the pair and gain is very high in this simple driver stage.

There are two differential input stages, one npn and the other pnp.
But they are not cascaded, but coupled to work in parallel to give wide bandwidth,
low thd and symetrical squarewave performance to a high frequency.
I found BF469 and BF470 were quite OK to use in the input and driver stages.
Each of the aluminium extruded heatsinks for the flatpack mosfets are 300mm long,
150mm high with 30 fins 40mm long, good for a 2 x 300 watt amp with no need
for a fan even when left running at 300 watts into 4 ohms with a sine wave.

The total surface area of the heatsink is 30 x 15 x 4 x 2 = 3,600 square cm.
The required surface area needed to keep T rise less than 40C approx is
about 40 sq.cm per Watt, so the extruded heatsinks I used are good for 90 Watts
of dissipation, or a lot more if one allows for a higher T rise. If anyone follows my
simple rules here they should find their amps won't run hot, but the heatsink fins
MUST be vertical, and not within a box which restricts air flow, and causes rise.
Devices should be fixed to the heatsink with machine screws and heat paste
and be spread over the area of the backing plate of the sink.
The 150 x 300 x 30 fin sink would be good for a 40W pure class A amp which
needs to dissipate 90Watts at idle to allow 40W of class A.

I have thermal fuses fitted
to each heatsink to shut the amp down if the temperature exceeeds 76C, but I have
never been able to trigger the fuse. I biased each channel it for about 42watts of idle
dissipation (300mA idle current ) which gives about 1.4 watts of class A before
resorting to class AB for the next 250 watts into 8 ohms. The high bias currents
in the mosfets stops all tendencies of the mosfets to oscillate at RF where idle
currents are low, as used in BJT amps.

The schematic is about as simple as a good 300 watt solid state amp can be.
All the input driver bjts are mounted on a strip of aluminium above the board to
keep them all at about the same temperature.
There is limiting of the input voltage to prevent excessive accidental input voltage
being applied.
There is a decent protection circuit on the output used which isn't shown, which
turns off the amp if there is a dc offset fault that exceeds about +/- 1V for longer
than 4 seconds. The 0.5 ohm "N" source resistors are non magnetic wire wound
types to prevent oscillations and force the mosfets to share the the current evenly
and remain evenly biased without great need for the mosfets to be accurately matched.

Anyone trying to build an amp with mosfet output devices should anticipate parasitic
oscillations at RF. I found that 56pF from gate to drain with short leads plus 560
ohm gate "stoppers" on each mosfet eliminated the RF oscillations. I also have four
bypassing caps. The power supply has large 75V rated Sprague 100,000uF capacitors
on each +/- 70V rail, but the leads from the caps have enough inductance to prevent
silent rails when testing with square waves so I placed 1,000 uF electros + 0.47 uF
with short leads from the drain rails to 0V, then one 0.01uF from each drain to the
chassis, and finally there was no more rail signal present at HF.

The open loop bandwidth is maximum at 200Hz but -3dB at about 5 kHz with only
180 ohms plus 100pF ( R23/C12 ) needed from the driver collectors to bases of
Q7&Q8 to tailor the open loop phase and gain to stop overshoot and HF instablity.
In addition there is the C15 across R25, 47k to compensate for the open loop phase
lag at HF. The RC zobel network R41&C28 from the commoned sources to 0V act
to provide the amplifier with a resistive load above 100kHz when the 0.22uF cap has
an impedance of 7.3 ohms. At F above 200kHz, the load on the output stage is 10
ohms if there is not other load connected so the amp is more likely to remain stable.
The value of C15 isn't given because in my case it was a pair of short insulated
wires twisted together until the stability became excellent. A trim cap of 3 to 20pF
could be used.

With 6 ohm load, each mosfet sees 36 ohms while in class A and then 18 ohms
when in class AB. So while in class A the gain reduction in the output stage with
follower action is about from 25 to about 1, and this local source follower series
voltage FB totals about 27dB.

The output stage works mainly in class AB and in fact not much different to a class
B amp where there is no idle current at all. With no source follower or global NFB
the THD of a pair of mosfets in complementary pair might be 10% at 250 watts
into 8 ohms. The 27dB of local NFB used in the source follower connection reduces
this to less than 1%. There is about 50dB of global NFB which means a total of 77dB
is operative. The driver stage, Q7/Q8, has to produce a voltage slightly higher than the
speaker output voltage and its open loop THD at about 45Vrms is about 3% of which
most is 3H followed by 2H. This adds to the output stage THD to make a grand total
of about 4% THD with no global NFB.
Most of the crossover distortion artifacts are reduced by the local output stage NFB.
The 53dB of global FB reduces the 4% open loop THD by 53dB, or by a factor of 
0.00223, so 4% becomes 0.0089%, or just under 0.01%.
The THD is approximately proportional to the output voltage, and at 4.5Vrms output
which is 2.5Watts/8ohms, THD < 0.001%, and rather difficult to measure easily.
Such low THD is typical of many solid state amplifiers. However, excellent THD, IMD
and TID measurements do not always tell the whole story about how an amplifier will or
will not alter the sound signals than pass through it.
Tube amps with ten times the THD can sound better.

(( And by the way, anyone could operate the 300 watt amp design as a pure class A
amp by having say 2 more output devices and dissipating about 80 Watts at idle,
and then using +/- 35Vdc rails instead of +/- 70Vdc. One is then able to get up to
36 Watts of pure class A, depending on the load value. The same input stages and
driver stages would produce about 3% open loop thd, and need some global NFB to
reduce it. It would not measure much better than the above low bias class AB amp.
Subjectively, The class A operation of the output mosfets may sound better than
having a 300W capability, which is never fully used. ))

With rhw class AB 300W amp as it is, any type of load can be connected and an
8 uH inductance in parallel with 8 ohms protects the output stage from over heating
if it ever had 5uF connected and the output signal had a high level signal above 20kHz.
5uF is a difficult load, but there may be some electrostatics, possibly made by
Lartin Mogan, which present such an awkward load to an amp, but usually all
awkward loads do have a series resistance in their equivalant impedance
characteristic so that the worst of phase shifts and peak currents at above 10kHz will
not bother the amplifier, and in any case the % of audio F energy above 7 kHz is
usually very small. 5uF = 4.5 ohms at 7 kHz, and 2.27ohms at 14 kHz, and in fact
most amplifiers of any type will cope quite well with such a load providing the
output voltage levels remains well below the maximum possible which is probably
going to always be the case with an amp capable of 300 Watts if it is used for
"normal hi-fi" and with average speakers, which most ppl find sound loud enough
with never any more than 10 Watts.
There could be a problem driving 5uF or with with insensitive electrostatics if the
amplifier was only rated for 10 watts, and if high levels were wanted. Quad ESL57
are equivalent approximately to 1.6 ohms in series with 2uF with 16 ohms across
the 1.6R&2uF also as a load. The response peaking effects caused by 2uF as the
sole load are usually prevented by the series R in front of the C; the other 16 ohms
has little loading effect at all, and in fact ESL57 are very easy loads to drive by
my 300W amp or any other amp. However, Quad's later model, the Quad 63
and following models need more than a 20watt amp.

There is current limiting that operates to prevent the output stage from developing
excessive current which is dangerous to solid state devices which can very rapidly
fail if the excessive current lasts for long enough to heat them up and fuse the small
pn junction size of the active devices which are not much bigger than a 4mm x 4mm
area.
Tubes have large metal areas
of perhaps 12 sq.cms, ( EL34 ) and temporary adverse heat disipation can be
tolerated for longer than in a solid device.
The maximum theoretical peak load current = rail voltage / (  load + Ron + Rs )
where Ron is the minimum 'on' resistance of the mosfet when fully turned on,
Rs = the source resistor. In this case if RL = 3.5 ohms,
load on each mosfet = 10.5 ohms,
Ron = 1 ohm and Rs = 0.5 ohms, I max = 70 / 12 ohms = 5.8 amps peak.
The rating for the mosfets is 7 amps peak.
The voltage gain of the mosfets reduces with RL value so a single mosfet with
a 10.5 ohm load plus 0.5 source R ohms will have a voltage change of 0V to +63.8V,
and the Vg-s change needed is about +8V. There will be 2.9V across the 0.5 ohm Rs,
and the 9V zener diode plus other diode will conduct to prevent any increase in signal
at the gates from over driving the mosfet. To cut a long story short, in effect,
the maximum current able to be produced by the amp = 14 amps peak for 3 mosfets,
and in fact my measurements indicate maximum load current cannot be any higher
than when load = 3.5 ohms or lower, and the output power = 350 Watts. So with a
1 ohm load there is still 14 amps peak produced, and maximum output power is
about 100 watts.
The amplifier would become overheated with a load of 1 ohm and worked hard but
then the fuses in rails or at the output ( not shown ) would blow. The amp is meant
to be used with loads above 4 ohm, but will tolerate normal home use with loads
down to 2 ohms where power is unlikely to average more than 3 Watts per
channel, ever.
The only way to increase low impedance speaker load power handling is to use
more of the same output devices or use devices with higher ratings and more
of them on a bigger heatsink. Exicon flat pack mosfets rated for 16 amps each,
which would allow 40Vrms into 2 ohms giving 800 Watts.
I have never ever needed more than 50 watts for hi-fi at home.

There is an input voltage limiting circuit included which prevents excessive input
voltages from being applied to the input transistors. Diodes D4 and D5 act to
shunt excessive input voltage just before the amp output stage clips.
Below clipping the input limiters have zero action on the input signal because each
diode which is a fast type with low forward turn on voltage <0.25V is biased by
the resistance networks.

The input limiting may be omitted if really wish, but if used it requires very
careful setting up to avoid any input clipping below high power levels.

The amp has been reliable and trouble free and was a great learning exercize to
build. I have surrepticiously substituted it for a 50 watt PP class A tube amp
during a demonstration about 10 years ago and the 3 other guys present did
not hear any noticable change. However, I had much poorer speakers in those
times using cheap asian made drivers. The speakers I have built since then use
far better sounding SEAS drive units and much better enclosures and perhaps I
now would not get away with playing such skullduggerous tricks on
unsuspecting audiophiles.

Maximum output voltage is about 45Vrms. It is possible to use such an amp in
pure class A. The bias current rail to rail may be increased to about 500 mA,
so that about 70 Watts may be comfortably dissipated in the heatsinks.
Using a toroidal output transformer with 3 : 1 step down ratio, there is a possible
15Vrms which will give 34 Watts of pure class A into 6.6 ohms,
and all power into loads above 6.6 ohms will be pure class A.
There will be 75 watts class AB into 3 ohms with 13 watts of pure class A. 
A suitable toroidal is made by the man at www.zeroimpedance.com in the USA.

When I wound the power transformer I used 4 x 25Vrms secondaries all in series
to give 50V-0-50V which produces +/-70Vdc rails. The 25V rails could be
paralleled to allow +/- 35V rails. The six mosfets could still dissipate 70 watts but
be able to put 32 watts of pure classA into 16 ohms, 60 watts class AB into 8 ohms,
including 16 watts of class A, 100 watts of class AB into 4 ohms with 8 watts of
class A.

The measured distortion would become extremely low at audiophile levels of power,
and perhaps the class A would convey some better subjective fidelity.

People have told me why my tube amps and mosfet amps both sound well.
One reason is that they run warm but not stinking hot, they have some class A
working rather than none, and were designed and made by the same man, with
an un-compromising attitude about how things should work.                                   
One friend I had, now deceased, had 4 ohm Duntech Sovereign speakers which
take some driving.
He had a vinyl recording of 20 Africans beating away on drums and percussive
instruments thus generating a very difficult and transient rich signal. We were able
to just get the 300 Watt amp into clipping on the transients but the outcome sounded
exactly like 20 Africans belting away furiously on leather and wood. Of course such
levels could not be sustained lest the street full of people call the police for noise
control.
Some Lieder music from olde europe also sounded detailed and sweet, and methinks
the 'Elephant' treads softly when required.

300w amp power supply.

Underside 2 x 300W amp.

Back to Index page