A Class A Power Amplifier
Back in the eighties Jean Hiraga published the Monster amplifier
among other designs in the french audio magazine l’Audiophile. Despite
the name, it was a class A amplifier only rated for 8 W of
output power, but the huge power supply (with batteries) gave name to
the amplifier. You will find the original article (in French) with an
English translation here.
At this site you will also find a 20 W amplifier with only bipolar
At about that time I modified the Hiraga Le Monstre schematic and built a 25 W amplifier instead. The original Monster schematic is shown below with my modifications. But after having moved around a couple of times, I replaced the amplifier with new ones.
Now I have started to go back to this amplifier, following the attached schematic, but not using the pair C2922 and A1216 (I don't remember which drivers I used in my first version many years back). At that time I used a 2x18 V transformer with CRC-filtering giving plus/minus 24 V (no huge power supply). The schematic I am using now is shown below.
As can be seen, I have replaced the Offset potentiometer with two (RV13 and RV14). This makes it much easier also to set the output bias current (in Q22 and Q23). With only one potentiometer, one must change R7 and R8 to set the bias. Now it is possible to vary the current through the JFETs (Q11 and Q12) with these potentiometers, so both the output offset can be nulled and the bias current can be set to a suitable value; I have chosen about 1.5 Ampere.
It is necessary to say a few words about the transistors. The JFETs used is matched 2SK170GR and 2SJ74GR with an IDSS of 4 mA. These transistors are discontinued of Toshiba, but are produced in small quantities by Linear systems (called LSK170 and LSJ74). The bipolar transistors 2SC2240/2SA970 (Q9/Q10) were handy at the time of building, but these are not critical, so I would think e.g. 2SC1775/2SA872 among others will do the job. The drivers 2SA1209/2SC2911 (Q17/Q18) were chosen because of their low collector-base capacitance (Cob), but not without problems, as pointed out later. The power transistors 2SC5200/2SA1943 (Q22/Q23) were chosen since they are very good transistors easily obtainable for a reasonable price and with a proper package (TO3P). The input FETs are cascoded, reducing the effect of the Reverse transfer capacitance, especially for the P-channel JFETs. This result in a reduced Miller effect, with lower distortion as a result for higher frequencies, high source impedance and higher gain. The gain of the input stage is approximately:
The gain of the output stage is approximately:
forming a Darlington Sziklai
as a linear common emitter amplifiers since the output signal is
the emitters of the power transistors. The
to limit the current variation in Q17
in lower distortion. The
closed loop gain is set by R16/R15,
20x (26 dB). If R28
used, the gain drops to the half: 10x (20 dB). The gain affects the
output impedance, which is about 1.4 ohms for 26 dB gain and about
the half (0.7 ohm) for 20 dB of gain. But,
be aware, not all speakers are fond of the highest
output impedance. With
20 dB gain the distortion is about 0.03 % for 12.5 W (half
with a power supply of plus/minus 24 V. The
open loop bandwidth is very high. With the
drivers 2SA1209/2SC2911 this
led to instability. The best cure was to solder 47 pF capacitors
directly between base and collector on these transistors. The
chosen to be 330 pF when a closed loop gain of 20 dB was used. This
resulted in a closed loop bandwidth exceeding 1 MHz.
A two layer PCB for each channel has been made for the amplifier. Small heatsinks are mounted on the driver transistors, see the picture below. The power transistors are mounted on the larger heatsinks (as are the PCBs). I am using metal film resistors with 1 % tolerance all over. All resistors are rated for 0.6 W maximum effect dissipation with the exception of R20 and R21 (3 W) and R16 and R28 (2 W). A large heatsink rated at 0.3 K/W is used for each channel.
The component side, bottom side and the component placement is shown below. If you are interested, the PCB and/or the Gerber files are available. KiCad was used for the schematic and layout part of the design.
The power supply is a 2x18 V 500 VA transformer common for both channels. I am using separate rectifiers and filters for each channel. The CRC filtering consists of 22000 μF capacitors and 1 ohms resistors. I consider this as a minimum. For efficient loudspeakes larger values probably should be higher.
Finally, I must say, this amplifier sounds remarkably neutral and good sounding despite its simplicity. Yes, it sounds better than amplifiers with many more transistors and higher output power: Quality before quantity, so to say.
Knut Harald Nygaard