Simple Ripple Eater PSU for the sx and kx-Amplifiers


The latest  PCB’s are gold plated,  through hole plated and silk screened for just $11.99 (excl. shipping) and are available from Jim’s Audio here:

Ripple Eater PSU for kx and sx Amplifiers

Here is an Excel file with the BOM:  Ripple Eater PSU BOM

The BOM is provided for assistance – always carefully check the part numbers and quantities before ordering.


Here is a very simple capacitor multiplier PSU aka ‘ripple eater’ for the sx or kx-Amplifiers.  The LTspice simulated ripple rejection at 100 Hz is about 65dB and at 10 kHz it is 85 dB.  These figures assume you have a 2.5A load and also include the on-board filter capacitors on the amplifier modules themselves (220uF per rail on the sx-Amp and 1000uF per rail on the kx-Amp per amplifier module).

Class A amplifiers draw heavy quiescent current from their power supplies, and this gives rise to significant amounts of mains related ripple that will affect the amplifiers noise performance.  VFA amplifiers generally have better LF PSU noise rejection than CFA’s, but it can still be a problem with a class A VFA.  

A further benefit of this supply is that the radiated mains noise from the PSU to amplifier module wiring is greatly reduced because you will be supplying close to DC into the amplifier module local reservoir capacitors, so the associated mains ripple charging currents are very low. Low frequency (because the HF currents will be provided by the local on board reservoir capacitors – 220 uF for the sx-Amp and 1000 uF for the kx-Amp) signal related currents will still be flowing in the wiring of course, so you have to pay careful attention to layout, how you dress the wiring, loop area minimization and common impedance coupling issues – you can read more about this in ‘How to Wire -up a Power Amplifier’. Nevertheless, this PSU goes a long way to making sure your finished amplifier can be as quiet as theoretically possible. When combined with the 50~60dB mains noise rejection of the actual amplifiers themselves, the noise levels on the amplifier output will be down as much as 100 dB which is an outstanding result by any measure. At HF, the performance will be even better, and that is important for the amplifier sound in the mid and high range.

The series pass transistors (8A MJE15032/33 devices) are designed to be mounted underneath the PCB with their tabs screwed to the amplifier chassis base using the appropriate thermally conductive insulator (same mounting technique as the main rectifier D4 as well but you do not need the insulating washer on the rectifier – only some thermal grease).  Each pass transistor will dissipate about 10W in a stereo set-up delivering full output power into the loudspeaker load.

In normal operation, you will drop ~2  volts across each pass transistor, so if you want an output of say +-27 V, the loaded DC voltage into the PSU should be 29 V.  Note carefully as well, this power supply does NOT regulate the output – it will track the raw filtered DC voltage, but remove the ripple hence the ‘ripple eater’ name.

How Can I use Ripple Eater PSU on class AB Amplifiers with Higher Supply Voltages?

You can use this ripple eater power supply as is up to +-35V.  You can also use it on amplifier with higher supply voltages up to +-50V. However, you must change the filter capacitors to 50 or 63V types (this is C1 and C6 in the schematic).  In a class AB amplifier, you can use values that are about 3x LOWER than you would on a class A amplifier – so about 15 000 uF at 50 or 63V would be correct.  You must also change C5 and C7 to 50 or 63 Volts.  Finally, change R5 and R6 to 10k each. Note that the Ripple Eater PSU is not suitable for amplifiers with power supply rails much above +-50V because the series pass transistors (Q1 and Q3) are rated at 8A continuous and this places an upper limit on the output power.

Can I Increase The Power Handling Capability of the Ripple Eater?

Yes – you can. 

Replace Q1 with an NJW3281 or a MJW3281

Replace Q3 with an NJW1302 or MJW1302

The PCB mounting holes for Q1 and Q3 will NOT accept the larger types above, you  must run short wires from the PCB to the larger suggested power transistors above.  Do not mount the offboard transistors more than 10cm away from the PCB. Make sure the transistors have a good heatsink – again, I recommend you mount them to the chassis if it is constructed of 3mm thick or more aluminium.

Remember, the transistor collectors must be insulated from the chassis – use a suitable silicon heatsink pad



The graphic below gives some indication of the performance (simulated) of the ripple eater supply. I used a ripple eater on the e-Amp to clean-up the supply to the amplifier front end and got better than 30 dB ripple rejection, so the technique works extremely well in practice. In the simulation below, for the first 40 ms, both LF and HF noise are presented to the input, and for the remaining 60 ms, just LF noise is presented. The aqua trace at bottom is the output and as you can see, the LF and HF noise is significantly attenuated.  

A Word of Warning

There is a considerable amount of energy stored in the capacitors and there is NO CURRENT LIMITING on the ripple eater output.  If you accidently short the output of the power supply, the series pass transistors (Q1 and Q3) will blow up in spectacular fashion.  So, exercise caution when wiring up!  

Secondly, remember the series pass transistors are rated at 8A – so they are not designed to supply full load current into  a 4 Ohm load with continuous sine wave testing both channels driven – they will pop if you do this.  If you want to do full load continuous wave testing, short out the collector to emitter on Q1 and Q3.  Once testing is complete, remove the short.

Component Overlays, Assembly and Some Pictures

Here is the component overlay for the PSU.

Below are some pictures of the finished board. Note carefully how the main bridge rectifier (D4) and the series pass transistors (Q1 and Q3) are mounted underneath the board and lay flat on the amplifier (metal) chassis. 



You do not need to insulate D4 from the chassis as the thermal pad on the underside of the device is electrically isolated from the internal diodes – you must use thermal grease however to thermally couple it to the chassis – it will overheat and fail if you don’t.  Remember to check with a DVM to see that there is no continuity between the chassis and the series pass transistor collectors.

Further Notes on Heatsinking/Thermal Management

If your chassis  base plate is made of aluminium of at least 3mm thick, you can mount the Ripple Eater PSU as suggested above. If however your chassis is made of steel, it probably wont be too efficient at dissipating the pass transistor heat, so its best then that you use a separate heatsink and mount the devices to that.  Here is a suitable one from Mouser Aavid 3.7C/W Heatsink

If you do mount the series pass transistors (Q1 and Q3) off-board, remember to twist the base, collector and emitter wires from the PCB to the heatsink mounted transistors and keep them as short as practicable (avoid going much above 10cm if you can).





You must use 8 or 10mm stand-off’s  – not longer. Suitable Mouser Pt# 534-24337 or 534-24433 or 534-24443.  In Europe, you can try RS components 280-9023 (expensive) or Reichelt in Germany DI 10 (best price)


Here are some photos of the ripple eater in action.  In this set-up I am running the kx-Amp off 33V rails with 400 mA per channel standing current.  I used 33 000 uF capacitors (the 47 000uF were not available). The voltage drop across the series pass transistors (Q1 and Q3) with 800 mA total load current   is 1.66V

This is the input voltage and the 100 Hz  ripple is about 1.5V pk~pk


This next shot is the output of the ripple eater, and shows virtually no noise – the scope vertical scale is the same at 500 mV/div.


The final shot below zooms in on the ripple (vertical scale is now 20 mV/div) and we see it is in fact about  14 mV pk~pk,  which is a reduction of 40 dB and there are far fewer HF harmonics in the remaining (very low) ripple.