ESP Logo
 Elliott Sound Products Project 3B 

25W Class-A Hi-Fi Power Amplifier

© January 2004, Rod Elliott (ESP)

pcb PCBs are available for this project.  Click the image for details.


The P3A amplifier has proven extremely popular, and the DoZ (Death of Zen - see Project 36) continues to provide enthusiasts with a simple, reliable and easy to build Class-A amplifier.  For some, the DoZ is too simple, and I have had many requests for a PCB for Project 10 - a Class-A amp, not too different from the P3A.  Unfortunately, P10 is still sufficiently different that the P3A PCB cannot be used, so after some initial simulations and a trial run, this new version is born.

For a photo of the completed (latest revision) PCB, see Project 3A.

Since PCBs are available (using the PCB for P3A), this makes it that much easier to build, and it uses the latest OnSemi transistors designed specifically for audio applications.  These new transistors have been tested in the P3A and P3B, and give excellent results.

The output devices are MJL4281A (NPN) and MJL4302A (PNP), and feature high bandwidth, excellent SOA (safe operating area), high linearity and high gain.  Driver transistors are MJE15034 (NPN) and MJE15035 (PNP).  All devices are rated at 350V, with the power transistors having a 230W dissipation and the drivers are 50W.

You can also use BD139/140 transistors as drivers, and MJL21193/4 power transistors, with little or no loss in performance.

The amp may also be operated at lower supply voltages for less power, but I do not recommend less than ±18V, which will provide around 15W into 8 ohms.  This supply voltage (approximately) may be obtained by using a 15-0-15V transformer.


The basis for this amplifier has been around now for several years as Project 3A, and requires only an increase in quiescent current to be able to operate in Class-A.  The biggest change is in output power (reduced dramatically from the 60-100W of the Class-AB version), but at 25W is still more than enough for many people.

The power supply is where you will see changes - it has to be able to supply a continuous current of 1.5A, and needs very low ripple and noise.  The design shown below will be seriously expensive to build, but this is the case with any Class-A amplifier, and is to be expected.

The first thing that needs to be examined with a Class-A amplifier is power dissipation of the output transistors, and also the drivers.  At the recommended supply voltage of ±25V DC (nominal) and a quiescent current of 1.5A, each power transistor will dissipate 37.5W, or 75W for the pair in a single channel.  The thermal resistances that need to be considered are listed below, along with typical values ...

RthTypical Value
Junction - Case0.7°C / W
Case - Heatsink1.0°C / W
Heatsink - Ambient0.5°C / W
Total (Junction - Ambient)2.2°C / W
Table 1 - Thermal Resistances (Each Transistor)

The typical derating is a linear curve, starting at 25°C junction temperature and allowing zero dissipation at 150°C.  OnSemi often use a derating factor of 1.43W/°C, starting from 25°C - not an unreasonable value, but this assumes a maximum junction temperature of up to 200°C.  For maximum reliability I will use a figure of 1.6W/°C, which derates a 200W transistor to zero Watts at 150°C - much safer.

Based on Table 1, each transistor will operate at ...

Tj = Rth × PowerSo ...
Tj = 2.2 × 37.5 = 82.5°C Above Ambient!

Based on a typical ambient temperature of 25°C, this means that the transistor junctions will operate at 107.5°C, and applying a derating factor of 1.6W/°C, the transistor must be derated by 87.5 × 1.6 = 140W.  A 200W device is now rated for a maximum dissipation of 60W! This assumes the use of a heatsink of 0.5°C/W for each transistor, or a total of 0.25°C/W - this is a very large heatsink indeed.

There is not a lot of margin, so although it may be possible to operate the transistors slightly hotter than suggested (by using a smaller heatsink), I absolutely do not recommend this.  The best way to reduce thermal resistance is to use the thinnest insulation possible, and make sure that the transistor-heatsink interface is perfect (or as near to perfect as you can make it).

Use of a clamping bar (rather than relying on the transistor mounting holes) will help to reduce thermal resistance to the minimum possible, in conjunction with thin insulators and the exact amount of thermal grease needed.  Consider using a small fan, operated at low speed.  The airflow should be directed towards the heatsink fins, and even a small amount of air will make a surprisingly big difference.

Like nearly all Class-A amplifiers, there is no output short circuit protection, so if speaker leads are shorted while the amp is working with a signal, there is a very real risk of the transistors being destroyed.

The supply voltage should be a maximum of ±25V.  This supply is easily obtained from a 20-0-20V transformer as shown below.

figure 1
Figure 1 - Amplifier Schematic

As can be seen, it is not a complex amp, and in fact is absolutely identical to that for P3A.

For use into 4 ohms (including bridging into 8 ohm loads), do not exceed ±25V, and do not exceed the 1.5A quiescent current.  The amplifier will operate as Class-A up to around 9W into 4 ohms, and will go into Class-AB mode beyond that.

D1 is a standard green LED, and is not optional, nor should it be used as a panel indicator! Don't use a high brightness LED, or change the colour.  This is not for appearance (although the green LED looks pretty neat on the board), but for the voltage drop - different coloured LEDs have a slightly different voltage drop.  The LED sets the current through the differential pair input stage.  The aim is to have a voltage across the LED of around 1.9-2V.  This may seem to be on the low side for typical green LEDs, as they are normally rated at 2-2.2V (although some are much higher and cannot be used).  However, a nominal 2.2V LED will have the right voltage across it at low current - only 1.1mA is provided by R8 with a ±25V supply.

VR1 is used to set the quiescent current, and normally this will be a maximum of 1.5A.  The amp will work happily at lower current, but will not be Class-A.  The Class-A driver (Q4) has a constant current load by virtue of the bootstrap circuit R9, R10 and C5.  Stability is determined by C4, and the value of this cap should not be reduced.  With fast output transistors such as those specified, power bandwidth exceeds 30kHz.

With the suggested and recommended 25V supplies, Q4 will normally not require a heatsink.  The output drivers (Q5 and Q6) will benefit from a heatsink, although it does not need to be large.

Although I have shown MJL4281A and MJL4302A output transistors, these are very recent and may be hard to get for a time.  The recommended alternatives are MJL21193 and MJL21194.

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

Before applying power, make sure that VR1 is set to maximum resistance to get minimum quiescent current.  This is very important, as if set to minimum resistance, the quiescent current will be very high indeed (more than enough to blow the output transistors!).


Since I have boards available for this amp, I obviously suggest that these be used, as it makes construction much easier, and ensures that the performance specifications will be met.  Note that the layout of any power amplifier is quite critical, and great pains were taken to minimise problem areas - if you make your own PCB, it is unlikely that you will be able to match the published specifications.  The P3A PCB is designed to be able to be cut down the middle to make two separate amps, and this is essential for this design.  Don't even consider trying to run a pair of amps on one heatsink!

All resistors should be 1/4W or 1/2W 1% metal film for lowest noise, with the exception of R9, R10 and R15 which should be 1/2W types, and R13, R14 must be 5W wirewound.

The bootstrap capacitor (C5) needs to be rated at at least 25V, but the other electrolytics can be any voltage you have available.  The trimpot (VR1) must be a multiturn type, as the current setting is critical.

Each of these amps will require a 0.25°C/W heatsink (very large).  Consider using a fan or even water cooling to keep temperatures as low as possible.  Remember - there is no such thing as a heatsink that is too big.

Do not use 'Sil-Pads' - even if you have access to the very best (and most expensive) low thermal resistance types, as in my experience they are still not good enough.  Kapton (25um maximum, or 0.001") is recommended.  Unless you can get mica that is 25um or less (down to about 10um), do not use it, as thermal resistance will be too high.

Basic Specifications

The following shows the basic measurement results ...

Input Impedance24k
Input Sensitivity600mV for 25W (8 ohms)
Frequency response [1]15Hz to 30kHz (-1dB) typical
Distortion (THD)0.04% typical at 1W to 25W, 1kHz
Power (25V supplies, 8 ohm load) [2]25W
Power (25V supplies, 4 ohm load) [3]50W
Hum and Noise [4]-73 dBV unweighted
DC Offset< 100mV (< 25mV typical)
Table 2 - Specifications

Notes ...

  1. The frequency response is dependent on the value for the input and feedback capacitors, and the above is typical of that (at full power) when the specified values are used.  The high frequency response is fixed by C4, and this should not be changed.
  2. This figure is typical, and is dependent on the regulation of the power supply (as is 3, below).
  3. Typical.  Only approximately 9W will be produced in Class-A, after which the amp will revert to Class-AB
  4. Hum in particular is highly dependent on layout, power supply and internal wiring.

Powering Up

If you do not have a dual output bench power supply ...

Before power is first applied, temporarily install 22 Ohm 5 W wirewound 'safety' resistors in place of the fuses.  Do not connect the load at this time! When power is applied, check that the DC voltage at the output is less than 1V, and measure each supply rail (at the amplifier, and after the safety resistors).  They may be slightly different, but both should be no less than about 20V.  If widely different from the above, check all transistors for heating - if any device is hot, turn off the power immediately, then correct the mistake.

If you do have a suitable bench supply...

This is much easier! Slowly advance the voltage until you have about ±20V, watching the supply current.  If current suddenly starts to climb rapidly, and voltage stops increasing then something is wrong, otherwise, continue with testing.  (Note: as the supply voltage is increased from zero, the output voltage will decrease - down to about 2V, then quickly return to near 0V.  This is normal.)

Once all appears to be well, connect a speaker load and signal source (still with the safety resistors installed), and check that suitable noises (such as music or tone) issue forth - keep the volume low, or the amp will distort badly with the resistors still there if you try to get too much power out of it.

If the amp has passed these tests, remove the safety resistors and re-install the fuses.  Disconnect the speaker load, and turn the amp back on.  Verify that the DC voltage at the speaker terminal does not exceed 100mV, and perform another "heat test" on all transistors and resistors.

When you are satisfied that all is well, set the bias current.  Connect a multimeter between the collectors of Q7 and Q8 - you are measuring the voltage drop across the two 0.33 ohm resistors.  The correct quiescent current for 'full' Class-A is 1.5A, but I strongly suggest that you use a lower current to start with! The voltage you measure across the resistors should be set to 500mV ±5mV.

If you set the quiescent current to around 1A, the amp will run in Class-A up to about 8W, and will transition to Class-AB at higher power.  This reduces dissipation and still allows Class-A operation at most listening levels.  Class-A amps are not designed for high power, and it's unrealistic to expect output power to match Class-AB amps.  Reducing the current also means that both the amplifiers and the power supply will run cooler.

After the current is set, allow the amp to warm up (which it will - and rather quickly too), and readjust the bias when the temperature stabilises or the current exceeds the rated 1.5A - this will need to be re-checked a couple of times, as the temperature and quiescent current are slightly interdependent.  Under no circumstances should you wander off while the bias is being set! If current keeps increasing, remove power immediately.  If the heatsink is too small or the thermal contact between transistors and heatsink is not good enough, the amplifier will get hotter and hotter until it fails!

NOTE CAREFULLY If the temperature continues to increase, the heatsink is too small.  This condition will (not might - will) lead to the destruction of the amp.  Remove power, and get a bigger heatsink before continuing.  Note also that although the power transistors are mounted to the board, never operate the amp without a heatsink - even for testing, even for a short period.  The output transistors will overheat and will be damaged.

When all tests are complete, turn off the power, and re-connect speaker and music source.

Power Supply

Before describing a power supply, I must issue this ...

WARNING:   Mains wiring must be done using mains rated cable, which should be separated from all DC and signal wiring.  All mains connections must be protected using heatshrink tubing to prevent accidental contact.  Mains wiring must be performed by qualified persons - Do not attempt the power supply unless suitably qualified.  Faulty or incorrect mains wiring may result in death or serious injury.

A simple supply using a 20-0-20 transformer will give a power rating of about 25W into 8 ohms.  This is influenced by a great many things, such as the regulation of the transformer, amount of capacitance, etc.  For each amplifier, a 120VA transformer will be (barely) sufficient, and 150VA is preferable.  To operate a pair of amps from one transformer, the transformer must be not less than 300VA  500VA is preferred to ensure that the voltage doesn't fall too far due to the constant load.  Feel free to increase the capacitance - as shown it is enough, but anything above 50,000µF (per supply rail) for each amp will not achieve significant benefits.  Lower capacitance may also be used at the expense of some additional ripple.  As shown, ripple will be about 20mV P/P at 1.5A loading.

The inductors need to have the lowest DC resistance possible, or significant voltage will be lost as heat.  It is perfectly ok (in fact it is preferable) to use iron cored inductors, but they must have a significant air gap to prevent saturation.  A cored inductor will require fewer turns and have lower resistance than an air-cored coil of the same inductance.

Figure 2
Figure 2 - Recommended Power Supply

For the standard power supply, as noted above I suggest a minimum of a 300VA transformer for one amplifier board (i.e. two amplifiers).  For 115V countries, the fuse should be 6A, and in all cases a slow blow fuse is required because of the inrush current of the transformer and filter capacitors.  C9 is an X2 mains rated capacitor.  When placed in parallel with the transformer secondary it reduces RF interference (conducted emissions) by a useful amount.  It's not essential, but is recommended.

The supply voltage will be dependent on the transformer rating, and the DC resistance of the 10mH inductors.  It will not be possible to obtain the rated power if the transformer is not adequate, or inductor resistance is too great.  Because of the continuous load and poor transformer regulation with capacitor input filters, it's generally advisable to use a transformer with the highest VA rating you can afford.

The bridge rectifier should be a 35A type, and filter capacitors should be rated at a minimum of 35V.  Get capacitors with the highest ripple current rating you can - the ripple current is high and constant, and inadequate capacitors will fail.  All wiring needs to be heavy gauge, and the DC must be taken from the last set of capacitors in the filter.


HomeMain Index ProjectsProjects Index
Copyright Notice. This article, including but not limited to all text and diagrams, is the intellectual property of Rod Elliott, and is © 2000-2003.  Reproduction or re-publication by any means whatsoever, whether electronic, mechanical or electro- mechanical, is strictly prohibited under International Copyright laws.  The author (Rod Elliott) grants the reader the right to use this information for personal use only, and further allows that one (1) copy may be made for reference while constructing the project.  Commercial use is prohibited without express written authorisation from Rod Elliott.
Page Created and Copyright © Rod Elliott 22 Jan 2004