KEF PSW2000 Subwoofer Repair


Because of the number of issues I have seen with the KEF PSW2000 subwoofer, I am no longer accepting repair jobs on them. The mechanical construction makes them awkward to work on, and they seem to have just too many almost insoluble problems.

However, I leave this page up as it may be of interest or useful to some.


These boxes, made by British manufacturer KEF, are floor standing cubes that are about 35 cm square. They contain a single bass speaker, some preamp electronics and a power amplifier.

In this article I describe some modifications which makes them run cooler. This should help their long term reliability. Two are described : one is a redesign of the low voltage power supply. The other is a simple mechanical change which allows air to flow in and out of the enclosure that houses the electronics. I also give some measurements I made of the internal air temperature, with and without the mods.

SAFETY WARNING : These devices, when powered up with the front panel removed from the speaker cabinet, have exposed connections at mains voltage. Take great care while working on them! As always, be aware that servicing electronics equipment is a skilled task, which should not only be performed by those with appropriate skills and training.


The power supply mod is not that difficult for someone experienced, but it is a bit much for a beginner. You do work with mains connections, and if something goes wrong you need to know how to fault find and fix it. You also need a scope to test your work properly, once it is done.

OK, with all that out of the way – here are the mods I did.

KEF PSW2000 Sub Woofer


Two modifications are suggested:

  1. An small additional mains transformer, bridge rectifier, and two smoothing capacitors are fitted, creating a +/-24V unregulated supply. This powers the 7815/7915 regulators, and improves the original design, as four dropper resistors, which ran hot, can be removed. It has other advatages too : the power dissipated by the regulators is reduced, and the input voltage to them becomes less dependent on preamp current, and independent of the power amp supplies. Two “bleeder” resistors, which also generate heat without appearing to do anything useful, are also removed.
  2. Air is allowed to flow in and out of the enclosure by inserting spacers between the front panel and the box.

Although the first mod sounds like a lot of work and expense, it isn’t that bad. If you are going to remove the power amp board anyway (presumably to fix a fault), you must remove the power transistors from the heatsink and so on – it is quite a job because of the mechanical design. Fitting the transformer requires drilling two 3.5mm holes, and of course, there is a bit of wiring. The whole mod should be done in an hour or two, along with the repair which you had to do anyway. If you are really thorough, you could recap the whole unit, which is another hour or so and perhaps 15€ worth of caps. (The component cost of the PSU mod is only around 5€.)

Note: KEF originally inserted a piece of fibreglass between the mounting bracket which holds the power amp board and the front panel. For all of these tests, this was removed and the bracket thermally bonded to the front panel with a smear of heatsink paste. (Most likely that the purpose of this was to try to reduce the temperature of the front panel. It gets very hot, and a warning sticker was fitted. However, as the only way for heat to get out (in the original design) is via the front panel, it is hard to see how it helps.)


No mods, no output (quiescent)45°C
No mods, 20-25W continuous into 4 ohms85°C and rising(*)
PSU mod only, no output (quiescent)36°C
PSU mod only, 20-25W continuous into 4 ohms85°C and rising(*)
PSU and airgap mod, no output (quiescent)33°C
PSU and airgap mod, 20-25W continuous into 4 ohms80°C
Temperature Measurements

(*) as this was a unit which I was repairing for someone, I didn’t want to risk damaging it. The temperature was still rising, albeit very slightly, when I decided to stop the test.

As the amp heats up, the output level drops slightly, probably due to the temperature compensation altering the voltage gain. Hence the 20-25W – what started out as 10V rms ended up being about 9V.


You could also use black anodised M3 screws (Farnell 1420016), to fix the transformer, if you care about cosmetics. Some 7/0.2 wire and sleeving is also needed.

the transformer prewired, ready to fit


  1. Start by completely removing the power amp board. Disconnect the removable connectors, and remove the four screws that secure the power transistors. (You need a very short phillips head screwdriver for this, as the preamp board is in the way.) Remove four more that secure the PCB to the mounting bracket.
  2. There may be other components (power resistors, transistors) fixed to the heatsink. Remove them as well, so that they hang from the power amp board by their wires.
  3. You may prefer to desolder the four connectors which still connect the power amp board to the preamp. Mark them clearly or draw a sketch so that you will know where to refit them! (Why Kef did not use a single 20 way ribbon cable, or similar, is a mystery.)
  4. Use paper towel and isopropyl alcohol to clean off the heatsink paste from transistors and heatsink bracket. As fitted from the factory, there is usually much too much. Handle the thin, transparent insulating washers carefully, and set aside. If the fibreglass insulator is fitted between the mounting bracket and front panel, clean off the heatsink compound (usually very sloppily applied) and apply a fresh, thin, even smear of compound before fixing the bracket back to the front panel.
  5. Use a pencil to mark the positions for two holes for the transformer, and carefully drill them, 3.5mm diameter. Remove burrs and debris thoroughly.


Solder the rectifier directly to the transformer and fit the wires that will go to the power amp before attaching it to the front panel (Figure 2). The two AC terminals of the bridge go to the outside pins of the transformer secondaries. Note that the bridge is bent upwards so that it will not accidentally short to the front panel. The two inside pins of the secondaries (the centre tap) are joined. Four wires connect to the transformer primaries. These must be sleeved for safety. (In general I do not sleeve low voltage connections, because they can hide intermittent connections. I would rather the wire just falls off. But that is my personal preference. However, lethal voltages should always be insulated.)

prewired and ready to be bolted down

I use red/green/black twisted to take the unregulated DC from the transformer to the power amp. The green (0V) goes to the centre tap of the transformer, the red and black to the + and – terminals of the bridge, respectively. See the next photo. The transformer can now be fixed to the front panel.

New Transformer Primary Wiring – Schematic
Transformer Primary Wiring (back of selector board)


Above are a photo and diagram of the primary connections. The primaries are in parallel with the those of the existing transformer. The connections are most easily made to the back of the mains selector switch, which is on a PCB above the mains switch. You will probably need to disconnect the spade terminals which connect the transformer, and mains switch to get access. Be sure to make a sketch so that you will know how to put the wires back!

Check and double check these connections before going any further. If you have a scope you could power up and check the output of the bridge rectifier, between green and red and then green and black. You should see full rectified sine waves peaking to around +/-30V. Be careful that the secondaries of the main transformer are not touching anything! (Use insulating tape to be sure.)


side view of the modified power amp board. The smaller blue axial capacitors are the new ones.

Next you need to remove the six resistors from the power amp board and fit the two capacitors which smooth the unregulated DC. (It is also good to add points to solder the wires from the transformer to, after the board is refitted. 0.1″PCB pins would be good, but I didn’t have any, so I used some tinned copper wire.)

  1. Remove the six resistors as shown in figure 7. R143 and 144 might be on flying leads, if they were fitted to the heatsink.
  2. Fit the two capacitors on end as shown in figure 6 and figure 8. Take care about polarity.
  3. Fit PCB pins or similar ready, for the +24V and -24V wires, at the points shown in figure 8.

The mod doesn’t look that pretty with the caps on end, but it is solid enough. Their junction is at 0V, so it makes a handy reference point when testing.

power amp board mods 1 – remove resistors
power amp board mods 2 – smoothing caps and connection points for wires from transformer


With the mods done, you can now connect the +/-24V wires from the new transformer and bridge rectifier to the +24V (red), 0V (green, to the junction of the two new caps) and -24V (black) points on the power amp board. You can leave the main (toroid) transformer unconnected for now. Before refitting the PCB to the frame, we should test that the rails are OK.

Make sure that the back of the PCB is not shorting to anything, power the unit up, and test the unregulated DC with a voltmeter. It should be around +/-24V (exact value depends a bit on your local mains voltage). At least 20V, in any case. You can also check the +/-15V at one of the opamps (pins 4 and 8 as usual). Check again with a scope. Unregulated ripple will be about 1V pk-pk. Regulated supplies should be free of signs of oscillation, of course.

I also tested some of the electrolytics – many of them seemed to be going a bit high in ESR. However, the owner didn’t want to go for a recap, so I did not change them. I did replace the two at the output of the 15V regulators though. If I was being really thorough, I would replace them all.


You can now refit the PCB to the frame. (I assume you already know how to thermally bond the power transistors to the heatsink compound, and so on.) Connect everything back up, power up, and check the amp works OK, first unloaded, and then into a dummy load, if you have one.

the power amp with mods done


Even with all these mods done, the unit still has a problem – lack of airflow. Here is a simple way to improve that.

This modification really could not be any simpler. When refitting the panel to the box, use spacers of about 3mm on each of the eight fixing screws. This creates a small air gap around the edge of the panel, which allows hot air to escape the enclosure and cool air to be drawn in. I used Farnell 1466918.


The first KEF subwoofer that I came across was the PSW3500, in which the output transistors are mounted on a large fan cooled heatsink. This is a redesign of the PSW2000 and 2010, and runs a lot cooler.

Every electronic device generates heat, and a power amp needs to get rid of a fair amount of it. Making sure that things do not get too hot is a big deal. When an output transistor or a voltage regulator (for example) cannot lose heat quickly enough, its temperature rises, and eventually it will fail.

How hot is too hot? Every device has, as one its design parameters, a maximum internal temperature. For the 2SA1943, one of the power transistors used in the KEF boxes, it is 150°C, fairly typical. The designer can estimate the maximum allowable air temperature inside the case, which will produce that temperature inside the device. This depends on the thermal resistance of the device and the power being dissipated. Basically, the lower the air temperature, the cooler the inside of the transistor will be (“cool” being a relative, when we are talking about temperatures at 100°C and higher.)

There is another reason to keep the internal air temperature in the case as low as possible. Other components, especially electrolytic capacitors, which need to be slightly wet inside to work, degrade over time, and much more severely at high temperature. In the case of the KEF2000/2010, this is evidenced by a number of common faults such as “chattering” relays, and so on.


corner of the power amp board before modification, showing signs of overheating

Above is the (unmodified) power amp board. The bad capacitor has been removed, not yet replaced – you can see the circle on the board, by the white connector. You can also see that the connector is damaged by heat – it is a bit brown and part of the plastic has flaked away. (Luckily, most electronic hardware is made of flame retardent material.) It is harder to see on the photo, but the green capacitor next to connector, and the back of the board itself, are all brown.

Like most modern audio amps, these boxes have two pairs of internal DC power rails. An unregulated high current supply at +/-50VDC feeds power amp – quite normal. There is also a low voltage regulated supply, +/-15V in the PSW2000/2010, which is for the preamp, and part of the power amp, circuits. All fine so far.

Unfortunately, the two 15V voltage regulators are supplied from the +/-50VDC rails in the original design. This was a bad idea. Here’s why.

The LM7915 and LM7815 regulators can only handle a maximum input voltage of around 32V DC. So, the designer had to use dropper resistors – 270R and 2.2k – to reduce the volts at the input side of the regulators. These are four of the six that we replace with this mod.

Below is a schematic of the droppers, also showing the voltages that I measured. (Capacitors omitted for simplicity.)

part of the Kef PSW2000 power supply design before modification
same schema with power calculation

(Using droppers like this is a bad idea, even without the overheating it causes : because the voltage on the input side of the regulator now depends on the current it is drawing. Of course, as long as the voltage remains high enough at all times, it is not, strictly speaking, a problem. But, a sudden spike of current would easily cause the input voltage to drop too low.

We can now figure out the current through each path in the circuit, and the power dissipated by each component. In fig 13, current is shown in blue and power in red.

The 4 resistors and the two regulators (which have 13 and 17V across them) are burning over 5W between them. That 5W of heat heats the air. (Moving two of them off the board will have helped this a bit.)

That accounts for four of the six resistors. It turned out that the other two are simply connected across the +/-50V rails. They each have a value of 1k8, which means that they each dissipate a further 1.4W or so (using P=V²/R). So that means that the 6 resistors and the regulators dissipate over 8W between them. In a sealed box with no airflow!


What should the designers have done? Well, the moment they decided to use linear regulators to get 15V from 50V, there was a problem. The circuit takes something like 100mA worst case (that might be a bit pessimistic, but as I measured 50-70mA it seems a decent guess at the maximum current required). That means that 35V at 100mA must be burned, about 3.5W for each rail, 7W in total.

The better alternative was to use a separate low voltage unregulated supply at about 20 to 25V to power the +/-15V regulators (exactly what the PSU mod described above does). This means that each regulator only drops about 5-10V, at 100mA that is a dissipation of about 0.5W to 1W per rail, 1W to 2W total. And no need for those big dropper resistors!

As for the resistors directly across the +/-50V : the usual reason for these is as “bleeder” resistors, which make the DC rails discharge towards 0V fairly quickly when the unit is powered down. But, in this case, I suspected they are not needed – normally, in a circuit of this type, the rails discharge quite quickly via the amplifier. (In fact, it appears that in early versions, they are not fitted.) And, a quick test showed this to be true. I measured the time taken for +50V to drop to +5V with and without these resistors – I got 9s with, 12s without. Hardly a big deal, and certainly not worth the extra 2.8W of heat. So I decided to that I would simply remove them.


The above is presented for reference only. My experience has been that these units are quite difficult to work on because of the mechanical construction, and I no longer repair them. The original investigation and mods took quite some time! In the end, I guess I at least saved one or two units from becoming yet more landfill.

KEF make lots of great audio gear, particularly some classic loudspeakers, and I don’t want to make it seem like I have a downer on them. But to be honest I think that whoever designed the electronics of these things for them could have done a bit better.

If the info above helps someone out there to get some use from their unit, I guess the effort was worth something.

Newsletter Updates

Please subscribe to my newsletter!


  1. Boa noite,
    Tenho um problema com este sub,
    Ao ligar por vezes faz um estrondo e não debita som, tenho que ligar e desligar algumas vezes para que funcione.
    Não sei a quem recorrer, nem o que fazer.
    Faço aqui uma tentativa para que me consiga ajudar.
    Desde já obrigado.

    Rui Matias

  2. Thanks for this, have carried out the modifications while changing the main smoothing caps.
    I also changed C52 which blocks DC from the Relay because it was chattering.
    C52 actually sits between the dropping resistors so it is no surprise that it dried out after at least ten years since I bought the device.
    As I had a 6VA PCB mount transformer I made a PCB to hold it & the rectifier & smoothing caps.
    The main dropping resistors were mounted on the main heat-sink with wires going into the PCB so i simply connected my new LV supply direct to a pair of the protruding wires & insulated the connection with Helermann rubbers.
    I added a small fan in the cupboard that my Sub is housed in &used 6mm nylon stand-offs to create a cooling gap which the small fan blows into. all in all a success.

  3. Oh! My PSW2000 didn’t work and I found the burnt point of PCB pattern which you modified.
    I tried to reconstruct the pattern by a wire but the temperature was extremely high.
    And the sub-whooper output was very small with a big noise.
    I wanna share the pics soon.

    Thank you for your information!

  4. In my case, PCB burnt near R143/R144. After fixing with wire and I realized very hot temp of cement register, above 115 degree during 2 min operation.
    Your information is very useful for me. Thank you for your effort!

  5. Thank you for your extremely useful and informative information regarding the PSW2000, I have had one of these sitting in my loft for about 20 years meaning to repair it, knowing there are several posts on the web regarding faults and ‘quick fixes’. I was pleasantly surprised to see that there are quite recent responses to your post, most seem to be older. Your explanation of why ther is a design fault and how to correct it has caused me to take the beast apart and rebuild. Fortunately there seems to be minimal heat damage and C52 is sitting about 15mm above the board suggesting a previous repair. I shall replace C59 and 60 as they look intact but the top is slightly domed. I have in the past built a Vellemann kit which provides a 1A symetrical supply variable between 12 and 15V, it has proved to be a stable kit although now not needed. What I wish to ask is, having removed the 6 ‘heating elements’ aka spurious resistors, is there a problem if I also remove the two regulator chips and tap my + 0 – 15V power supply into the holes left by the output pins and 0V pin, do I need the two capacitors that follow the chips C55 and 56 if not then I could use the holes left by them as a neater fix for my power supply. I am also intrigued by the reason for the tiny C53 and C54, which seem to parallel C55 and C56, are they also redundant in my rebuild. Whilst perfectly capable of repairing what someone else has identified as a problem, I am not too good at identifying the precise reason for a component’s function, if you have a few minutes I would be grateful for an assist.
    Mechanically the heavier aluminium heatsink liberally pasted to the chassis is actually not flat as there are about 1mm humps where it has been drilled and tapped, I doubt this aids good heat conduction so have smoothed them off and there seems to be a much tighter joint and less paste! As you point out, I can see little reason for the piece of copper clad fibreglass board between the bent chassis plate and the main plate.
    Again many thanks, I am now ready to delve into your other posts which I am sure will be as enlightening

    • hello Gareth, first I’m glad that this post was helpful. I have to say that it is a long time ago, and I don’t remember the finer details of the KEF design. As I noted somewhere, I eventually gave up on these boxes as they are a royal pain to work on and not that great even after all the mods listed above. I don’t recommend any changes to the original KEF design unless you are a qualified electronics engineer, and then whatever you do is at your own risk, I’m afraid. Good luck!

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.