Micro Class-D

    Class-D AMP

   Ribbon Speaker

  ▪  Measurement Mic.

   Mini Subwoofer

  ▪  25 W Power amp

  ▪ NEW  Home Theater


  ▪  NEW  XLS Sub


   Current projects

  ▪  Finished projects


  ▪  Amplifiers

   Ribbon Speakers

  ▪  Subwoofers


  ▪  Main page



Contact: e-mail


Updated 00-08-07


25 W DIY Power Amplifier

DSC00001.jpg (115421 bytes)

Try something simple for a change

When you look at DIY constructions of amplifiers, it's often large expensive designs! Why not try something simple for a change?

I guess that a lot of would be DIYs are held back from realising a dream of making their own amplifier, be cause DIY projects often means investing a lot of money without knowing whether it will work or not!

A lot of online projects or commercial DIY kits, promises better sound for less money than ready build of the shelf amplifiers. It often involves a lot of very large transformers and a lot of bier can size capacitors, lots of exotic components and some black magic ..... which sums up to a lot of money. But how can you be sure it will sound better? For one thing it's quite hard to compete with large companies development departments employing hundreds of full time engineers and designers.

So before you decide to build anything, make up your mind .... you should only do it be cause you think it is fun, not to save money!

The design presented here does not promise state of the art sound, on the other hand it is neither the simplest circuit in the world that is only suitable for kitchen speakers, TVs and getto blasters. The objective was to design an amplifier that gets the best out of a bunch of cheap components.

Design of a 25 W Power Amplifier

When designing this amplifier I put up the following requirements:

Req1:    The amplifier must be cheap

Req2:    The amplifier must be reliable

Req3:    The amplifier must sound fairly good

Req4:    The amplifier must be easy to build

Req1:    The amplifier must be cheap

As this is the first design done entirely by myself, I wanted to get the basics straight before jumping to some mega project. I also think it is to start simple making a good sounding amp out of cheap components, thereby controlling parameters and improvements step by step. This amplifier is not intended to beat the best, but as a first step with a lot of possible improvements.

That the amplifier is cheap also means that it is "available" for more people.

To make it cheap I of course had to keep the output down to a minimum. 25 watt is not much these days, but if the amplifier is well constructed it will be sufficient to satisfy a lot of applications (remember to get the sensation of "double as loud" will require 6 to 10 db, that is the same as 100 to 250 W!).

To keep cost down it should use cheap standard components. This is actually not as bad as it sounds. Today there is a lot of hocus pocus in Hi-Fi. A lot of money is spent on exotic components, claimed to "sound" better without being measurably better (sometimes these components are even worse that the most basic standard components). The used transistors is not the best, but actually they are not at all bad, only the output transistors leaves a lot the be desired when it comes to linearity (but that is easy to fix for a little money).

Together with a small rotary switch and a potentiometer it would do as a good sounding integrated amplifier. Put two or three of them in a loudspeaker together with an electronic crossover, and you'll have an active speaker. Use to for the rear channels in a surround system ..... most integrated surround amplifiers have terrible sounding amplifiers for the rear channels, or use it to power a small active subwoofer.

Req2:    The amplifier must be reliable

The amplifier should be reliable, not going up in smoke when used. This means that the construction should be sound and the components should not be pressed to their limit. It must be able to deliver the promised power in all realistic loads for hours and hours.

Req3:    The amplifier must sound fairly good

This amplifier is not intended to beat the best, but it should provide a sound that makes it worth while to build. By using well known techniques this is not too hard to accomplish.

Req4:    The amplifier must be easy to build

As this is a DIY project that I would like a lot of you handy people out there to try, it must be well documented, well tested and thought through.

The construction is held on a single PCB (printed circuit board) preventing a lot of loose wiring, and thereby preventing a lot of possible faults and burned down circuits.


The construction

The Diagram

The diagram is quite straight forward and basically follows the guide lines recommended by Douglas Self [1]. There is nothing revolutionary about the construction, but it relies on sound and well proven circuits.

diagram.gif(16127 bytes)

The diagram can be divided into three stages: a differential input gain stage, a second gain stage and an emitter follower output stage.

Differential input gain stage

input.gif (4853 bytes)

The differential input stage is based around Q1 and Q2. A differential input stage provides an easy way to handle overall feedback, and is used in 99% of today's operational amplifiers. Feed back has got a very bad reputation, and has ultimately resulted in people now accepting to pay enormous amounts for single ended designs with large distortion figures. If applied with care feed back has some almost magical properties, two of which is lower distortion and lower output impedance.

To livearies the input stage the "normal" collector resistor for Q2 has been removed [1] and local feed back has been applied by means of R4 and R5. The gain of the input stage has been set at 7.5 times.

To livearise the input stage even further a constant current generator provides the bias current for Q1 and Q2. The current generator is formed around Q3 and Q4, and provides 2 mA DC.

Second gain stage

gain2.gif (4675 bytes)

Second gain stage is based round Q7 forming a normal common emitter configuration. The stage has been linearised in three ways. Firstly the bias current is held constant by Q9, and an input buffer has been applied (Q5) [1], and thirdly local feed back is provided by R11  setting the gain in this gain to 140.

P1 is used to adjust the DC level on the output. The adjustment might seem a bit unusual, as it in fact adjusts the bias current through Q7, but it works quite effectively without adding a lot of additional components. With standard components it should be no problem to get near 0 VDC on the output, and still get close to 5 mA through Q7.

Complementary Feedback Pair (CFP) output stage

output.gif (6212 bytes)

The output stage is formed by Q10 - Q13, forming a push pull complementary feedback pair. Q8 forms a constant voltage between basis of Q10 and Q11. The voltage is a bit higher than two times Vbe = 0.6 and determines the idle current in the output devices Q12 and Q13 to drive the output in class AB thereby reducing crossover distortion. D3 limits the voltage over C8 thereby allowing us to use capacitor rated at less that the the double rail voltage (2 * 25,5 V = 51 V) which can occur at switch on/off or if anything should go wrong. This avoids blown capacitors.

The CFP output stage has been chosen because of it's better distortion performance over the most common double emitterfollower output stage. It can be shown that the distortion of the CFP is only around half of what can be obtained by the emitterfollower stage.

Filtering of the rail supply is provided by C9-10 and R17-18. The filtering causes a fall in maximum output voltage (and therefore power) due to the voltage drop on R17-18. The filtering though gives a much better power supply rejection ratio (PSRR) and thereby lower distortion.

Feedback and filtering

The overall feedback is handled by R15 and R7 giving a total gain of:

R15/R7 + 1 = 10k/301 + 1 = 34.2 ~ 30.7 db

This gives an input sensitive of 0.6 V for full output (25W / 8 ohm).

feedback.gif (2301 bytes)

DC gain is set to zero by C3, C20 and C21. C3 and C20 forms a bipolar capacitor allowing both positive and negative voltage at point 12. The voltage at this point is very small (~100mV) so normally s single electrolytic capacitor is used at this place, but even if we're only talking about very small voltages, such a capacitor is not made to be wrongly polarised. The two diodes allows us to use capacitors with low voltage ratings which is both physically smaller and cheaper.

The output is provided through a Zobel network R24, C13 and a resistor damped inductor R23, L1. These are included to secure stability when the load is either highly inductive (Zobel), could be the case when the amp is connected to a wideband speaker without filter, or highly capacitive, as could be the case of some electrostatic speakers.

zobel.gif (1874 bytes)DSC00007.jpg (102479 bytes)

The coil L1 is just 10 windings around R23 with 1 mm cooper wire.

The power supply is straight forward. It "only" uses 2 x 2200 uF capacitors. I know you'll think it's a joke and that it's way too little. But in contrast to what a lot of companies and dealers may want you to believe, it really is quite enough. I tried to parallel connect 2 x 10.000 uF but the sound didn't change at all. The bass didn't get tighter or faster or anything and the sound didn't get any cleaner. So now I'm sure it's enough. Do keep in mind that this is a 25 W amp (and save your money for something else). The PCB is made to allow a bit larger capacitors if you insist.

psu.gif (3392 bytes)DSC00003.jpg (111801 bytes)

The power supply is mounted on a separate board.


Download the diagram in postscript format here.


The amplifier has been thoroughly simulated in PSpice, especially to verify stability but also to adjust component values.

To measure the the stability the amplitude and phase was measured in open loop at the feed back network. That is R15 was disconnected from Q2 and connected to a dummy resistor of 100 ohm (same as R9) which was connected to ground. The amplitude and phase was then measured at this connection point V(40) with 1 mV sinus at the input. The load was set to 4 ohm / 1 uF (8 ohm // 0 uF will give higher stability).

C3 in the spice file (replacing C3, C20 and C21 in the diagram) and C1 was also omitted in the stability simulations.

The gain must be less than 1 then the phase is less than -180 degree, otherwise the amplifier will oscillate when the loop is closed. The trace below shows the frequency response and phase of the amplifier with 1 mV sinus on the input and a load of 4 ohm // 1 uF on the output.

Where V(40) and V(1) crosses each other the gain from the input to the feed back point is 1.


Setting C5 = 560 pF,  the gain is 1 at 518 kHz and the phase is -39 degree. This gives what is called a phase margin of (180-139) 41 degree, which in turn means the amplifier is quite stable.

wpe12.jpg (30464 bytes)

The design criteria is normally a phase margin of 45 degree. This can be obtained with C5 = 680 pF:

wpe13.jpg (30411 bytes)

Whether to use 560 pF or 680 pF for C5 is up to you, using 680 pF you are on the safe side in al conditions, but 560 will give a little more feedback at higher frequencies giving lower distortion.

Download the PSpice file here.

Printed Circuit Board (PCB)

The amplifier is based on two PCBs one for the amplifier circuit and one for the power supply. The power supply comes in two versions, one for mono and one for stereo use. The PCB is single sided to make it easier for DIYs to make, but it of cause results in some jumpers on the board.

wpe3.jpg (26920 bytes)wpe5.jpg (31434 bytes)

The PCB for the AMP

The PCB layout is available in post script and HP Laser Jet compatible printer file.

PCB files including all three PCBs (Amp, Single PSU and Double PSU)

PCB Layout Post Script PRN File
Component Mounting Plan Post Script PRN File
PCB Layout and Component Mounting Plan together   PRN File

When printing the PRN files, just write:

    copy "filename".prn lpt1

in a dos command window. This will copy the file to your printer (if it is connected to lpt1 ..... most likely).

Parts List

R1 301R 0.25 W 1% C1 1uF / 63 V MKT (2 - 4 modul)
R2 10k0 0.25 W 1% C2 1nF / 63 V MKT (2 - 4 modul)
R3 1k96 0.25 W 1% C3 220uF / 10 V E-Lyt. (1 modul)
R4 100R 0.25 W 1% C4 100uF / 35 V E-Lyt. (2 modul)
R5 100R 0.25 W 1% C5 560pF - 680pF / 63 V MKT (2 - 4 modul)
R6 348R 0.25 W 1% C6 100nF Polyester (2 - 4 modul)
R7 301R 0.25 W 1% C7 100uF / 35 V E-Lyt. (2 modul)
R8 5k62 0.25 W 1% C8 100uF / 10 V E-Lyt. (1 modul)
R9 5k62 0.25 W 1% C9 220uF / 35 V E-Lyt. (2 modul)
R10 649R 0.25 W 1% C10 220uF / 35 V 100uF / 25 V E-Lyt. (2 modul) (2 modul)
R11 20k5 0.25 W 1% C11 220uF / 35 V E-Lyt. (2 modul)
R12 Not used. C12 220uF / 35 V E-Lyt. (2 modul)
R13 1K00 0.25 W 1% C13 10nF Polyester(2 - 4 modul)
R14 68R1 0.25 W 1% C14 100nF Polyester (2 - 4 modul)
R15 10k0 0.25 W 1% C15 100nF Polyester (2 - 4 modul)
R16 619R 0.25 W 1% C16 100nF Polyester (2 - 4 modul)
R17 220R 0.25 W 1% C17 100nF Polyester (2 - 4 modul)
R18 220R 220R 0.25 W 1% C18 2200 or 3300 uF / 35 V E-Lyt.

(2 - 3 modul)

R19 34R8 220R 0.25 W 1% C19 2200 or 3300 uF / 35 V E-Lyt.

(2 - 3 modul)

R20 34R8 220R 0.25 W 1% C20 220uF / 10 V E-Lyt. (1 modul)
R21 0R1 5W C21 100nF Polyester (2 - 4 modul)
R22 0R1 5W
R23 2R2 5W
R24 10R 2W D1 1N4148
D2 1N4148
D3 6.2 V Zenner Dioded
P1 200R Multiturn Pot (Bourns 3296Y) D4 5 A DiodeBridge (+ ~ ~ -) (3* 2 module on a line)
P2 1K Multiturn Pot (Bourns 3296Y)
L1 10uH (10 vindings round R17)
Q1 BC550B TR1 2 x 18V, 100 VA or more.
Q2 BC550B
Q3 BC550B F1 5 A Slow blow
Q4 BC550B F2 5 A Slow blow
Q5 BC556B F3 0.5 A Slow blow
Q6 Not used.
Q7 BC556B 3 fuse holders
Q8 BD139
Q9 BC546B
Q10 BD139
Q11 BD140 Heat Sink Just use a pice of aluminium (10 cm * 20 cm * 3 mm)
Q12 BD250 (alt. TIP36 / TIP2955)
Q13 BD249 (alt. TIP35 / TIP3055)

Building the Amplifier

(The construction of cause counts on you being able to make the PCB your self or know someone who can. That is first step.)

Start by soldering all jumpers and resistors. Go on with the smaller capacitors and the small transistors. Q1 and Q2 should be tied together with a plastic strap to get a good thermal contact, as is the case for Q3 and Q4. Take the large components last, ending with mounting the transistors in the output stage.  The best way to do this is to drill the holes in the heat sink first. Place the transistors in the PCB (a little bending of the legs are necessary) and mount them on the heat sink, remember to use thermally conducting insulation washers either mica together with thermally conducting paste (on both sides of the washer) or the newer types based on silicone (these does not need thermally conducting paste and are a lot easier to work with). Also remember to use an insulator sleeve between the output transistors and the screws otherwise the back plate of the transistor which is connected to the collector will be short circuited to the heat sink. The three transistors Q7, Q8, Q9 must be mounted on a small piece of aluminium (all three on the same piece). This is paramount to gain good thermal stability. These transistors don't need much cooling so the size is not important.

Make sure the devices are mounted very tightly and use a ohm meter to make sure there is no connection to the heat sink (measure between the legs of the transistors, one by on, to the heat sink).

Heatsink mounting.GIF (5650 bytes)

After mounting the transistors to the heat sink you can solder them.


Start by testing the power supply. Check that there is +- 25.5 V DC on the output terminals.

Now is the time to try the amp. Before you do anything you must turn P2 all the way anticlockwise, so that no bias current is running in the output when connected. P2 should be set to around 100 ohm.

The best way to try a new amp is to use a vario transformer on the main side so the main voltage can be turned up slowly and the circuit tested initially on a lower voltage. This is very helpful if anything is wrong as this can very well result in a burned amplifier.

The second best way is to connect an additional transformer before the "real" transformer. Take for example a 250V to 125V transformer, connect the primary side to the main (250V) and the secondary side to the primary side of the 2x18V transformer. This will result in a supply of 2x12VDC on the output of the power supply. this is a lot less harmful than the normal 2x25,5VDC!

Connect the output to an old speaker (one you don't really mind loosing!), leave the input unconnected and turn on the amp. If there is any strange sound, lots of hum or noise turn the amp of immediately ..... something is wrong (you need to check all before trying again).

When testing have a mV meter connected over R21,R22. Initially this should read 0, but when the bias current is adjusted with P2 the current in the output stage can be determined by the voltage drop over these resistors.

The best is to have two meters so you can also verify the DC voltage on the output at all times.

If there is only a soft hum or noise in the speaker everything should be ok. Measure the DC voltage on the output and adjust it to near zero with P1. Now you only need to adjust the bias current in the output stage. Measure the DC voltage over R21 and R22 (it should be near zero) and turn P2 slowly until you read 20 mV DC. Now there will be flowing 100mA through the output stage which is enough to eliminate crossover distortion.

Wait some minutes with the amp on for it to get in thermal stability, and adjust P1 and P2 again. Note that when adjusting the DC on P1 you are also affecting the bias current, so some iterations are needed.

Now you can connect the input to your pre amp (turn the amp of before you connect anything .... always!), and enjoy the sound.

Congratulations you have now build your own amplifier.

The Sound

At this point I have only build a mono version for a small sub-woofer, so I can't give you a full report yet. But what I've heard so far is more than promising. .... Stay tuned.


The amp posted here is actually a second version. A proto type was initially build and a lot of improvements have been made in the present version.

All the below stated improvements are implemented in the current version posted on this page, but the last column shows when it was first introduced. This way you have a chance to update an older version.

Improvement Description Implemented in
Emitter resistors in output stage. To reduce distortion, the first version was designed with an output stage without emitter resistors. The emitter resistors does introduce a small amount of distortion, but they have a quite important role in terms of thermal stability. As current flows through the transistors in the output, both DC from the bais current but also from the delivered current to the load, the temperature for the transistors in the output increases. As the temperature rises the basis emitter voltage Vbe decreases (-2mV/ºC). This causes the bias current to rise as long as the bias voltage is held constant. As the bias circuit is constructed around a transistor which is mounted on the cooling fin together with the output stage, this transistors Vbe also decreases. This courses the bias voltage to decrease which in turn makes the bias current decrease. But if the emitter transistors are left out the decrease in bias voltage is not enough to eliminate the decrease in bias current. Remember that the output stage has 4 Vbe in series. This is called thermal run-away, and causes the current in the output stage to rise towards infinity, which in turn causes the output devices to go op in smoke!

If the emitter resistors are to be omitted an improved way of producing the necessary thermal feed back must be developed. Until then two emitter resistors must be included in the output stage.

(The strange thing is that I have seen the emitter resistors omitted in an other quite similar construction [2] ..... supposedly without burning off!)

Better grounding on PCB Noise and hum reduced by better layout, following guidelines proposed by Douglas Self [1]. AMP01PA1
Current mirror in input stage I have tried to implement (in Spice only) a current mirror for the differential input stage instead of using a simple collector resistor ( se [1] page 66). This gave the possibility for a much higher gain, but the interfacing to the next stage is not easy to get to work. The difficulties comes from the fact that the DC level at the output of the mirror is higher than that of the normal collector resistor. I think that the best way of making it work is to implement the VAS stage as a similar differential stage using the same type of current mirror. Not implemented.
Buffer before VAS stage   AMP01PA1


Ref. 1 "Audio Power Amplifier Design Handbook" by Douglas Self (ISBN 0-7506-2788-3)
Ref. 2 High Fidelitty, October + November 84 issue. "Kaskodekoblet effektforstærker" by Steen Michaelsen, Michael Madsen and René Jørgensen (Danish HiFi magazine).