Simple, regulated PSU. By Lee Davison.

This is a simple, regulated, linear PSU. There are no special components in the basic circuit, they can all be changed or substituted depending on need and availability. All the itterations of this PSU were built with parts from my bits bins.

The idea was to put together a PSU suitable for running CB radios and similar ~12V DC vehicle devices with what I had laying about the laboratory. The most usefull part was the case and PCB from a long dead CB power supply. This may seem like cheating but only the PCB, case and on/off switch were of any use. Everything else had either been removed or was possibly damaged and so not reliable enough to be reused.

Everything was removed from the case and a heatsink for the output transistor was pulled from the big box of assorted heatsinks. The one I found was such a good fit it could have been made for the job.

With this screwed in place a small piece of card was cut as a template for cutting a mounting hole for an IEC mains socket taken from a dead PC PSU. This template was stuck in place with double sided tape over the original entry hole for the fixed mains lead.

A jigsaw was used to cut the hole and the hole was filed until the socket was a snug fit. The socket itself was then used at the template for drilling the securing screw holes.

With the mains socket fitted a 2N3055 power transistor was dug up and fitted to the heatsink with an insulating mounting kit. Leads were soldered on to allow connection to the PCB.

The circuit is really very simple. A centre tapped transformer is rectified by two 1N5401 diodes to give a 24V DC unregulated supply. The transformer does not need to be a centre tapped device, a single output transformer and four individual diodes or a monolithic bridge rectifier could just as easily be used but was not what was to hand at the time.

This rectified voltage is smoothed by a total of 6900uF of 35V capacitors which were also what was to hand. If I had larger values that would have fitted I would have used those instead.

A 15V zener and a conventional silicon diode in series are used as a 15.7V reference voltage. This is fed from the unregulated supply rail through a 1K5 0.5W resistor giving a current of around 8mA. The actual current here isn't critical. As long as there is enough to make the zener work and supply the reference to the rest of the circuit and not so much that the zener cooks all is well. This voltage is bypassed by a 100uF 25V capacitor to keep noise to a minimum.

A 0.5W resistor is specified here because, with the output shorted, it will need to dissipate approximately 22V^2/1500 ohms = 0.32W which is just a litte much for a standard 0.25W resistor.

This reference voltage feeds a TIP120 darlington transistor which itself feeds the base of the 2N3055 output transistor. These two devices together make a high current gain voltage follower that can readily supply a few amps of current.

To stop it all going horribly wrong should to great a demand be placed on the output, a 0.2 ohm resistor in the 2N3055 emitter circuit is used as a current shunt. The voltage developed across this resistor can increase to the point where the 2N2222 transistor begins to conduct which is around 0.6V to 0.7V. Once the 2N2222 transistor starts to conduct it diverts current away from the base of the TIP120 reducing the overall current gain and limiting the output current to about 0.65V/0.2 ohms = 3.25A.

Again the current limiting transistor is not critical. Any small signal transistor that can pass 20mA or more, withstand a Vce of 25V and has an Hfe in double figures should work.

A 2N3055 was chosen as the output transistor because it was the correct shape, TO-3, for the existing case holes and heatsink and with adequate cooling should be able to survive prolonged abuse in this circuit. Any similarly rated transistor should be just as happy to do this job.

An 820 ohm resistor, chosen to light the LED not light the room with the LED, feeds a front panel mounted LED from the regulated output to show that the PSU is working.

With no load on the voltage at the output sockets is about 14V. This is a little more than was anticipated but not so high as to be any worry. A load of only a few tens of mA, about what the a CB will draw in receive, quickly reduces this to 13.5V.

A 21W vehicle bulb drops the voltage to 12.6V. This is to be expected as, even if the voltage source were perfect, the 0.2 ohm current sense resistor will mean a drop of up to 0.6V in normal use. As it isn't perfect this 1V drop at almost 2A current is to be expected.

Adding a 5W vehicle bulb to the load reduces the voltage to 12.4V, not an excessive drop, but replacing that 5W bulb with another 21W bulb causes the output to go into current limit and drop to 10.2V.

Even running in current limit the only parts that get warm are the ones you would expect to. The transformer - eventually, the output transistor and the current sense resistor which actually being a 5W device and not 2.5W as specified, handled the 0.65V^2/0.2 ohms = 2.1W dissipation with ease.

The output regulation suffers in part because of the 0.2 ohm current sense resistor in the output transistor emitter circuit. If this is to be improved then feedback needs to be taken from the output end of this resistor.

This is done by replacing the zener and diode voltage reference with an LM431 or equivalent, these are almost ubiquitous in switch mode power supplies. Feeding the input of the LM431 with a proportion on the output voltage this shunt regulator will begin to conduct when this input voltage gets to typically 2.5V, this will then divert current from the base of the TIP120 keeping this voltage at 2.5V.

The LM431 needs at most 10uA of input current so to make this current negligible a divider current of 1mA was chosen. This will give an input to ground resistance of 2.5V/1mA = 2.5K ohms but as 2.5K ohms is not a prefered value 2.2K ohms is used instead, giving a divider current of 2.5V/2.2K ohms = 1.14mA.

The other resistor is then calculated from the rearranged formula for a potential divider, R1 = Vin/Vout * R2 - R2 which in this case gives ..

 R1 = Vin/Vout * R2 - R2
 R1 = 13.8/2.5 * 2.2K - 2.2K
 R1 = 9.944K

Again this is not a prefered value so a 10K resistor is used instead.

Now when the PSU is tried the no load voltage is 13.79V, already an improvement, and with a 21W vehicle bulb as a load the output is 13.22V. Adding a 5W vehicle bulb to the load reduces the voltage to 12.8V which suggests that current limiting has begun and replacing that 5W bulb with another 21W bulb causes the output to drop to 10.2V as it did before.

Not perfect regulation but an improvement nonetheless.

To improve matters further the 2N2222 transistor can be replaced by another LM431. This will give a much sharper knee as the LM431 has a much higer gain than any single transistor.

The current sense resistor will need to be changed to accomodate the 2.5V input of the LM431. In this case a the 0.2 ohm resistor is replaced by a 1 ohm resistor giving a current limit of 2.5V/1 ohm = 2.5A. If you want to keep the limit at nearer 3A then a 0.82 ohm resistor will give a limit of 2.5V/0.82 ohm = 3.05A or if you want a higher limit, say near 6A, then a 0.43 ohm resistor will give you a limit of 2.5V/0.43 ohm = 5.81A.

This resistor should be a little bigger than the 2.5W resistor previously specified as it will need to dissipate up to 2.5A^2 * 1 ohm = 6.25W in normal use. To accomodate this with room to spare a 10W resistor should really be used here.

Now when the PSU is tried the no load voltage remains at 13.79V and with a 21W vehicle bulb as a load the output is still 13.68V. Adding a 5W vehicle bulb to the load reduces the this by only a small amount to 13.67V which suggests that current limiting has not yet begun. Finally, replacing that 5W bulb with another 21W bulb causes the output to drop significantly to 6.8V, much better current limiting.

The improved current limiting is good but suffers from a couple of drawbacks.

Because the LM431 needs a 2.5V input to work with the sense resistor needs to dissipate quite a few watts. For a 3A current limit a resistor capable of handling at least 2.5V * 3A = 7.5W would be needed which makes the resistor physically large and expensive. It can also be awkward to find exactly the right low value resistor for the desired current limit in such high wattages.

This circuit adds a fixed voltage offset to the current sense resistor voltage which reduces the size of the current sense resistor needed. A bonus is because the current sense circuit is now in the negative leg of the supply there is enough voltage drop for an indicator LED can be added to show when the PSU is current limiting.

A 2.5V reference is provided by yet another LM431, a proportion of which will be added to the sense resistor voltage depending on the position of the 2.2K pot. When this proportion of the voltage plus the voltage across the sense resistor gets to 2.5V the LM431 will begin to conduct, brightly lighting the LED and reducing the voltage at the base of the TIP120, so preventing the current from rising any further.

With this pot set the cutrent limit behaves exactly as it did before but now with a much smaller and cooler sense resistor and an LED which lights when it is active.

This is now essentially complete, just some 'i's to cross and 't's to dot and some matching screws to hold the lid on to find.

This can be used as a bench power supply for vehicle equipment or, with a no load output of 13.8V, even as a float charger for 12V lead acid batteries. It's nothing special but it is more use than the collection of parts that it was up until now.

With the adjustable current limit and smaller current sense resistor the only part that gets more than warm under current limiting now is the 2N3055 pass transistor on its heatsink. This could be fan cooled if the power supply was frequently going to be operated in current limit. A temperature controlled fan, such as those in PC power supplies, would be ideal for this, only comming on fully when needed.

Find four, preferably black, preferably matching screws to hold the cover on.

I found four, matching screws and screwed the cover on nice and tight and it all works .. until that is you load it. Then the stray magnetic field from the transformer bounces the thin steel cover off the top of said transformer at twice the line rate. BZZZZZZZZZ. Now I just need to find a thin foam pad. Two thin slices of pencil eraser stuck to the top of the transformer with double sided tape seem to have quietened all the buzz.

Add an internal mains fuse. As it is it has no mains fuse and so relies on the fuse in the power lead used. This is fine if a lead with a 3A fuse is used but not so fine if a lead with a 10A fuse is used.

Insulate all the exposed mains connections. It's pretty lethal with the cover off as it is and I know that if I leave it like this sooner or later it's going to bite me. Done.

Last page update: 30th June, 2011.

e-mail me