Updated May 20. 2007


    Micro Class-D


       The Shot Out

    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


Micro class-d amplifiers UcD vs. Hysteresis



The construction

As stated above the construction is based on Phillips' application note. There are though several differences.

  • Input resistance increased to 10k Ohm, to make it work wit any normal preamp.

  • Choice of transistors. The goal was to make the amp really small and inexpensive. The use of BC846 for Q12/Q13 will limit the use of supply voltage to some +-30VDC giving a theoretical output of 56 W/8 Ohm. It is though fairly easy to change these to transistors with higher voltage capability to make the amp work for higher output.

  • DC offset is handled by trimmer pot R29. The DC offset seems quite low even without the pot, and this can easily be omitted to lower cost.

  • Most components are different from the application note.

The resistors R15/R19 should be omitted. First of all these where wrongly placed on the wrong base of the input pair. They are supposed to help getting a better PSRR but seems to do more damage instead (picking up noise), this also when placed at the right base input (Q7).

When running in UcD configuration R1, R2 and C1 should be mounted.

When running in Hysteresis mode, R4, R14 and C6 should be mounted.

The Diagram

The Circuit

Click here for a pdf version


The PCB is double sided and measures just 3.3 cm x 7.8 cm.

Six of these amps will fit on one euro card (160 x 100 mm).

You can download the Gerber files needed for the production of the class-d amp PCBs here.

I have had god experience with Olimex, who has a very good offer on fast prototype boards. One double sides euro board is 33 USD, which means that each of the amp PCBs are 5.5 USD. It's easy convenient and cheap. Now I simply wouldnt think of trying to etch my own boards any more! Another unique point is that you can mix diferent designs on one eour card as long as there is space enough. Cutting out the PCBs are free of charge.

A free Gerber viewer can be downloaded from GraphiCode.


  • Power supply 2 x 26 VDC

  • PCB size 3.3 cm x 7.8 cm (only 56% the size of a regular credit card)

  • Good for about 40 W with the current power supply in 8 Ohm

Switching waveform before filter.


Comments when running in Hysteresis mode:

There is overall a bit too much noise at the output. At first the following components where used:

R18  3k3, R9 = 10k, R4 = 270k, R14 = 1M, C6 = 220 pF.

This results in a switching frequency a bit over 250 kHz.

There is a bit of 50 Hz noise from the PSU. This can be lowered by using bigger caps in the PSU, and is overall a result of a rather low PSRR (which is quite normal for a class-d amp).

The noise from on the output (besides the 50 Hz) is quite broad band and is probably coming form the high switching currents/voltages. I think that this can be reduced overall by applying snubbers over the mosfets, and on the output, further more fast diodes could be mounted over the mosfets. Also the PSRR, and general rejection of induced noise could be improved by increasing the open loop gain. But all this was omitted to make the amp as small as possible and to require as few components as possible.

Improvements could also be made to the PCB layout, by placing the feedback components as close to the input as possible. As it is these components are actually placed close to the output!

Changing the following components, noise was though decreased considerably and is now at a level where it is not a concern anymore for normal use.

R18 = 3k3, R9 = 3k3,  R4 = 100k, R14 = 470k, C6 = 1 nF.

R29 was not mounted, but the DC output was always less than 100mV.

R6 gets a bit hot. Must be considered if running from a higher PSU voltage.

Note 13/5 2007: Actually mounting snubbers (10R 1206 + 1nF 100V COG) on the mosfets, did nothing to the output noise, but the ringing on the fets was naturally almost eliminated.

Switching waveform with the above mentioned components. Fs = 245 kHz.

Power supply board

  • Same size PCB as the amp.

  • One PSU will be used to power two amp boards

Power supply board.


On the first assembled unit I used a coil wound on a Micrometals T130-2 core. This is overkill in this small an application. Therefore I will try out the smaller version Micrometals T106-2

As seen from the analysis 40 windings with 1 mm wire (windings counted on outside diameter not including leads) will give 21,5 uH, which should be sufficient (1,7 m wire was used per. coil).

The two coils, wound using banking to minimize capacity [1], [2].


What could be improved? What are the learning points?:

  • The overall open loop gain is to little to give a proper PSRR. This is making the amp sensitive to self generated noise (EMI), ripple on the supply lines etc.

  • The output coil is unnecessary large. For powers less than 100W it is quite over dimensioned.

  • Q1 for the 12 V supply, is getting quite hot. Think it is ok for the +-24 V DC used, but both a heat sink, and another device should be considered if aiming for higher power levels.

  • R6 is there for easing startup. It gets quite hot and must be reconsidered for higher power levels.

  • The feedback network R1, R2, R4, C1, should be placed near the inputs (basis of Q6 and Q7) and not near the output as it is now. This will help reducing noise induction.

  • Ground, should be laid out with more care. The layer is to fragmented, and there is not a good separation between the input ground and the power ground.

  • The of set adjustment seems unnecessary. R29 can therefore be omitted.

  • For higher powers some current limiting should be implemented.

  • To make the construction smaller and easier to handle, it should be considered to place 2 amps and a PSU on a single board. This would eliminate a lot of tiresome and clumsy cabling.

  • For the presented power level, the mosfets does not get more than a little warm .... therefore no heat sink is really necessary. SMD mosfets should be considered as well.