Changing the output voltage of a commodity type switch-mode power supply (to power tube type equipment and other fun stuff.)

How to make a silk purse out of a sow's ear.

This is not an article about designing switch-mode power supplies. There are several good books and many application notes on the subject. It's intended to describe, in a practical way, how small power supplies can be reworked to power small tube-type radios and for other projects where a power supply with the required output voltage is difficult to acquire.

__Identification of suitable candidates__

First of all, you have to know your ears. The ear you want is a universal input switch mode power supply using the flyback topology. Don't worry, there are many of these available from surplus dealers. They can also be obtained from junked electronics.

Some identifying features are; only one big high frequency power transformer, no output chokes, (except perhaps small noise chokes) lots of electrolytic output capacitors and a single power mosfet switch. The power supply in the photo at the bottom of the page a typical example. Look for an 8-pin pulse width modulator IC with the 384 in the part number UC3842, MC3842. It might also be marked UC3844, 43, 45... or with some other prefix and suffix, we don't care. We're not going to touch any of the primary side circuitry. In these power supplies the control chip is on the primary or AC line side of the transformer. Below is a typical primary side schematic.

It's technically not a transformer but a coupled inductor. But I'll follow the common convention and call it a transformer to avoid confusion. The mosfet switch charges the core when it is on and when the switch turns off the core discharges though the output rectifier. Unlike a transformer where the current flows in the primary and secondary simultaneously.

The nice thing about these flyback power supplies is they regulate a volts-per-turn on the secondary. As long as we maintain that number we won't have to touch the primary.

We only need to rewind the regulated output winding, change the value of one voltage sense divider resistor as well as the output capacitors, output rectifier and possibly the compensation capacitor.

__Gather a little information first__

There should be a label on the power supply somewhere with the ratings. Usually the highest continuous power output is the regulated output. Note, highest power not highest voltage. Often this is the 5V output because logic needs tighter regulation. The stuff that runs off +12 or whatever other voltages may be provided usually can live with less tightly regulated voltage. Sometimes the slave outputs also called tertiary outputs are cross regulated and rated for high peak power for the purpose of starting motors or operating solenoids. At other times it may be additionally regulated with a 3-terminal regulator.

Look for the secondary side reference and error amplifier this part will look like a TO-92 package transistor. It will be marked with the numbers 43, possibly TL431, like the pulse width modulator IC the prefix and suffix may vary. From the reference lead follow the etch to the voltage reference divider resistors and note their values. For the TO-92 part the pins are (looking at the marked face with the pins pointing down) 1. Cathode 2. Anode 3. Reference. Below is a typical secondary side schematic.

__How to disassemble the transformer__

After you determine which output is the regulated output, unsolder the power transformer from the board. Now pulling the core halves apart to have access to the windings can be a problem. What I do is remove any tape from the core then place the transformer with the bobbin pins pointing up on a hot plate. Once the core is hot enough the glue will give up and it can be pulled apart while wearing heavy gloves. Try to heat only the core as much as possible.

Now unwrap the tape from the bobbin until the first winding is exposed. Carefully unsolder the wire and unwrap it. Then unwrap more tape to get to the next winding or layer. Rinse and repeat until you get to the regulated output winding.

If you find a winding that is connected to the primary side before you get down to the main winding, stop. This transformer has an interleaved primary and secondary winding. If you feel like winding a new primary, which includes providing for proper creepages and clearances for safety isolation, you could still use this power supply - but you're on your own. Fairchild published a good application note on designing flyback power supply transformers.

As the main or regulated output winding is removed count the number of turns. Write it down. This will be important later.

Do not damage the insulation beneath the output winding or you might be electrocuted. Examine the way in which the wire is placed on the bobbin, is it sleeved? Is there margin tape to keep it centered on the bobbin? You'll want the new winding to be of similar construction.

It is very important that to maintain primary to secondary isolation in this transformer or it becomes a death trap. It is generally a good idea to ground the output common either at the power supply or at the load. This will hopefully open the fuse if there is an isolation fault.

__How to figure the new secondary winding__

The main output voltage plus about one volt for the rectifier and winding losses, divided by the number of output turns is the volts-per-turn. It's common to see 1 to 2 volts-per-turn. The transformer shown in the schematic has 3 turns for a 5V output so I add the rectifier Vf or forward voltage drop of about 0.4V (because it's a schottkey diode and a big one for a 2A output) then divide 5.4V/3T= 1.8 volts-per-turn

Lets make a new output winding... I want a 140VDC output. So I need 140V/1.8 or 77 turns. Handy, isn't it! I could have chosen a 12V output and wound 7 turns on the bobbin in a single layer, or even two layers if the gauge was large enough.

Try to stay as close to the original volts-per-turn as possible. Do not try factional turns as they couple poorly to the primary and increase the leakage inductance. More on that later. You can also play with the rectifier Vf voltage using larger or smaller rectifier diodes, swapping PN junction diodes for Schottkey junction diodes or slip a small ferrite bead over a lead of the rectifier to increase the apparent Vf.

You do not have to use magnet wire but the wire must be insulated. The more insulation the wire has the less copper you can fit in the transformer. You'll want to have 300-600 circular mils of wire cross section for each ampere of load current. Try to fill the bobbin evenly. If you have too few turns to fill the bobbin spread the winding out. This improves the magnetic coupling between the primary and secondary. If the coupling is lousy the leakage inductance will be high. High leakage inductance makes the power supply, noisy, and inefficient. It may also cause a voltage spike on the mosfet switch that causes it to fail or cause snubber resistors to burn up.

The primary to secondary insulation should not have been disturbed but here is the story on that. Because tape or thin sheet insulation may have imperfections. It's common practice to use three layers of sheet insulation between the primary and secondary, where any two layers of sheet insulation can pass the hi-pot test. Larger transformers may use nomex or mylar strips wrapped between primary and secondary with just the end secured by a piece of tape. Another option is 0.8mm thick solid insulation. This is double insulation as well since it's twice the minimum 0.4mm. I hardly ever see this in switch-mode power transformers though because it increases the leakage inductance by degrading the coupling between the windings due to the increased spacing..

Divide the width of the bobbin available, after subtracting any margin tape, by the diameter of the wire to find the number of turns-per-layer. I put one layer of tape between each layer of wire and three layers of tape over the outside. My 140V winding fit in four layers.

Because of the superior temperature resistance and thinness of the insulation, I almost always use magnet wire, or rarely, Teflon, tex-e, kynar, mylar or mica insulated wire for special applications. MW-35 insulated magnet wire is my favorite but it must be mechanically stripped. Nyleeze or other solder strippable wire is usually good enough. Ordinary PVC insulated wire will work but it won't take much heat and you're wasting valuable winding window by filling it with insulation instead of copper.

For a handy tool to help in selecting the wire look for wtsetup.exe at wiretron.com on teh intarweb!

The tape can be a problem. Transformer manufacturers use specialized tape which is very thin and has excellent temperature, mechanical and electrical insulation properties. Surplus Sales sells some transformer tape but it is expensive. Teflon tape will work well but each layer should be .002" thickness minimum.

__How to figure the new divider__

The TL431 has an internal reference of 2.49V so the control loop will force the point where the output sensing voltage divider is connected to the TL431 reference pin match 2.49V as best it can. The current in the reference divider is 2.49V divided by the bottom resistor value. Subtract 2.49V from the desired output voltage then divide what is left by the resistor current to calculate the resistance you need in the top half of the divider. Use a couple of resistors in series to obtain the desired value.

On the schematic the resistor between the reference pin and ground is 4.99K and the output voltage is to be 140V...

2.49V/4.99K=0.5mA    140V-2.49V=137.5V    137.5V/0.5mA=275K

The value of the resistors making up the top half of the reference divider will be 275K

__Selecting the new output rectifier and capacitor__

Because these power supplies generally operate between 50KHz and 200KHz, you should use parts designed for this service. The diodes and capacitors used to build linear power supplies may work poorly or not at all.

The output rectifier should be a fast or ultra-fast recovery type The voltage rating is decided by the turns ratio of the transformer. Remember the rectifier is blocking when the MOSFET is on. Then you have the input voltage divided by the turns ration plus the output voltage to withstand. The US line voltage is nominally 120V and the bulk storage capacitor may peak charge at light load. So expect 180V or so worse case. The primary turns we don't know. You could measure it after the new secondary is installed by means of a signal generator and oscilloscope. That seems like a lot of work to me. Figure between about 50 turns for the primary (it could be less than half or more that twice that number.) The turns ratio in my example is 45/77 for a turns ratio of about 1:1.7 So divide the primary voltage by the truns ratio then add the output voltage to that (180V/0.558) + 140V = 448V. An 600V PIV rating is a standard value.

The current rating of the rectifier should be about three times higher than the output current. My power supply is 60W so 60W/140V = about 0.5A. Keeping with the x4 rule of thumb I'll use a 2A 600V ultra-fast recovery diode.

The output capacitors should be low ESR type designed for use in switching power supplies with high ripple current ratings. The capacitor voltage should be rated 120-130% of the output voltage. The ripple current rating should be at least 1.2 times the load current. If you cannot find a decent capacitor with the voltage rating needed try paralleling the electrolytic capacitor with a film capacitor rated for the ripple current and voltage.

Suitable rectifiers and capacitors are available from Digikey and Mouser.

__A Few Observations__

The flyback power supply is a constant power converter. If you short circuit the output the voltage will be very low and the current will try to increase to keep the power constant. This usually means burnt up traces and blown out rectifiers. The primary current limit that (hopefully) protects the MOSFET switch may not save the output. If you're lucky the bias power supply, a small 12V supply on the primary side that powers the control IC, will collapse because the secondary volts per turn is now very small. Then the power supply will hiccup trying to restart into the short circuit condition. Without secondary side protection an overload which doesn't cause the supply to hiccup will often lead to failure. A fuse in the output is a good idea.

Flyback supplies are noisy but with some shielding from a metal enclosure and some filtering they are good for operating radios. The Samlex power supply popular for operating ham radio equipment is a flyback power supply.

Heat sinks are usually live, do not touch them or you could be electrocuted. Anything connected to the AC line has some potential to cause death or injury, be careful.

The circuit boards of these power supplies usually have some mounting holes connected to grounding pads. If these are not connected together by a low impedance path, like a sheet metal chassis, the power supply will be noisy.

Using the techniques above I've built several dual output power supplies. One handy version has 105V and 2V outputs to run depression era tube radios with good success. The 105V was regulated and the 2V was derived from a three turn winding of several parallel wires spread across the bobbin. A schottky diode in series with the output was used to drop the 2.3V down to 2V.

Schematics from Micrel Application Note AN-18 MIC38C43 Off-Line Reference Design by Jeff Dixon... worth getting and reading if you wish to know more.

Any modification to these power supplies will void the warranty and the render any safety agency recognition or certification invalid. You could cause a fire or electrocute yourself. Even engineers that work on these things for a living make mistakes so be warned. I will not be responsible for how this information is used. It is only here to document what I have done.

March 23 2007 re-fixed output diode PIV