psu logo back to main page


AT PSU - Higher Output Voltage Modification

The aim of this was to hack, bodge and generally modify an AT type PC PSU (Power Supply Unit) to give a higher output voltage than originally intended by the original design. This voltage should also be variable over as wide a range as practical.

This bodge will have a use as a high power bench supply, but the primary aim was to provide a power source for a small SSTC. Help was received from the good folks at in this thread which made the process much easier.

This project was based on a 200W AT PSU kindly donated by a colleague from work. The outcome was a variable power supply that could supply between 3V and 48.5V at power levels up to 150W - ideal for what I wanted. Success!

The principles here should also apply to an ATX type PSU (see last section).

Pulsed loading (as in an interrupted SSTC) is unsuitable for this type of power supply as the high current looks like a short circuit and the supply shuts down. It excels as a general lab supply though.

You should not embark on modifying a power supply unless you understand the risks involved. Recommend that you read this safety guide on working on mains powered equipment.


AT Power Supply Architecture

A good place to start is at who have a selection of AT power supply schematics on their site. None exactly resemble the circuitry of the PSU that I was modifying, but they were near enough to give me an idea where to look.

The AT PSU architecture is pretty similar across devices, consisting of a half bridge primary on the transformer. Current sensing is provided by a current transformer in series with the main transformer centre tap, the output of which is added into the feedback node.

The low voltage secondary has a grounded centre tap with half wave rectified outputs. The controller is typically a derivative of a TL494 voltage mode controller with dual switching outputs (in my case a KA7500).

PC power supplies are good bits of engineering: designed for mass production to be as cheap and as simple as possible but with a wide range of safety features and a compact size for the amount of available power. They are also cheap plentiful, which makes them ideal for messing around with.

Basic modified schematics for my power supply are shown below.


Equipment and Materials

Some extra materials are required for this modification namely

  • 4 x 100V to 200V fast rectifiers, suitable for 4A to 8A current
  • One or two 50V electrolytics, low ESR type, approx 1000uF total capacitance
  • Variable resistor, probably about 100k (depends on circuit)
  • Fixed resistor, 1k to 10k (depends on circuit)

In terms of equipment, very little is required. Indeed my initial work was performed with just a multimeter to measure the output voltage. This was in addition to the usual tools lke clippers, wire strippers, screwdriver, de-solder pump and scalpel.

Some kind of load is required for the power supply output. I had the use of a 150W electronic load, but a 60W light bulb would be more than adequate. I've also seen 12V halogen bulbs and car headlight bulbs used.


Modification Principles

The idea is to:

  • Disconnect the grounded centre tap from the transformer
  • Connect the bottom of the transformer to ground
  • Half-wave rectify the other end of the winding relative to ground. This will effectivley double the output voltage
  • Remove all of the other secondary rectification components
  • Disable the Over Voltage Protection (OVP) circuitry to prevent it from shutting the power supply down when we obtain a higher voltage fom it.
  • Insert a variable resistor into the feedback path to allow alteration of the output voltage


Modification Process

psu schematic before  modsThis is a schematic similar to that of the power supply that I modified. There are some differences, especially in the OVP (which is transistorised instead of a comparator) and over current protecton comparator circuit. The one I modified used transistors instead.

The modification procedure I used was as below:

  • Initial test - can get 8A @ 12V into an electronic load without loads on any other rails. It starts up with no load on output.
  • Disconnect 5V single choke to isolate 5V output - power supply now gives out 33V (no 5V feedback, so 12V is increased to compensate and get correct current at feedback node).
  • Insert a 7805 regulator between 12V and +5V output (choke still disconnected). 12V output back down to correct level. Regulator stays cool.
  • Start removing diodes from other rails.
  • Removing -12V rectifiers causes PSU to go into a hiccup mode. 12V traced through a resistor to the cathode of D15. This was removed and the PSU started up again.
  • Removed all secondary rectification components from all the rails.
  • For the replacement rectifiers I used 2x BYW80-200 diodes (200V, 8A) in a TO220FP isolated tab package. These were both mounted together (thanks to the isolated tabs) on the old 5V diode heatsink which had plenty of pre-drilled holes and was exactly the right width for four TO220s. Connected in the same fashion as the eisting rectifiers to half wave rectify the new 12V transformer output.
  • Pulled all windings off the main coupled choke toroid except the +12V. Replaced output capacitor with 2 x 820uF, 50V Nichicon low ESR caps in parallel.
  • Feedback resistor from 12V removed and feedback is now fed through a 100k linear pot with a 1k series resistor.

Now, the power supply schematic looks a bit like this.

The schematic above does not show a 78M05 regulator that provides the 5V rail into the control circuit, the input to which is provided from the KA7500B supply voltage.



Variable output range at 1A load current was from 3V to 48.5V. The upper limit was caused by the controller duty cycle limiting at 50%.

Left running at 30V @ 4A = 120W for an hour over lunch with the fan running from an external 12V supply. No significant heating of the components was observed.



The biggest problem caused by this modification is a lack of a 12V regulated rail to power the fan. I scratched together a SEPIC converter using a National Semiconductor simple switcher running from what would have been the +5V winding.

This provides a regulated 9V rail for the fan and the Tesla coil control electronics.


ATX Modification

An ATX power supply can be modified in much the same way as the main converter architeture is very similar to that of the AT.

However, some differences are evident.

  • Power on signal
  • Minimum output load
  • Standby power supply

The power on signal is a pin on the main 20-pin output connector, which needs to be connected to ground for the power supply's main converter to start up. This can be done with a bit of wire shoved into the connector.

ATX converters have a minimum output load requirement for them to start up, the level of which varies between supplies. Not loading the outputs up enough will cause the power supply to turn on only briefly and then go back into standby. This can especially be seen by observing the fan movement on turn on.

ATX supplies also have a stanby power supply, usually 5V at about 1A or so. This is independent from the main converter, with some supplies shutting it down once the main converter has started up.

It should be possible to modify this supply to give a separate 12V feed by either playing with the feedback and/or squeezing a few more turns onto the transformer.