Useful Tips

How to make an adjustable power supply from a computer


I got a little carried away with electroplating (I’ll tell you more about that), and for her I needed a new power supply. The requirements for it are approximately the same - 10A of output current with a maximum voltage of about 5V. Of course, my gaze immediately fell on a bunch of unnecessary computer power supplies.

Of course, the idea of ​​remaking a computer power supply into a laboratory one is not new. I found several designs on the Internet, but decided that another one would not hurt. In the process of remaking, I just made a dofig of mistakes, so if you decide to make yourself such a power supply, consider them, and you will do better!

Attention! Despite the fact that it seems that this project is for beginners, nothing like this - the project is quite complicated! Keep in mind.


The power of the power supply that I pulled out from under the bed is 250W. If I make a 5V / 10A power supply unit, then precious power disappears! No matter! We raise the voltage to 25V, it may work, for example, to charge the batteries - there you need a voltage of about 15V.

For further actions, you must first find the scheme for the source block. In principle, all BP schemes are known and googled. What exactly you need to google is written on the board.

A friend threw me my scheme. Here she is. (Opens in a new window)

Yes, yes, we have to climb in all of these guts. The datasheet on TL494 will help us with this.

So, the first thing we need to do is check what maximum voltage the power supply can give out on the +12 and +5 volt buses. To do this, remove the feedback jumper prudently placed by the manufacturer.

Resistors R49-R51 tighten the positive input of the comparator to the ground. And voila, we have the maximum voltage at the output.

We are trying to start the power supply. Yeah, it doesn't start without a computer. The fact is that you need to enable it by connecting the output of PS_ON to ground. PS_ON is usually signed on the board, and we still need it, so we will not cut it. But we turn off the incomprehensible circuit on Q10, Q9 and Q8 - it uses the output voltage and, after cutting it, will not allow our PSU to start. Soft start will work on resistors R59, R60 and capacitor C28.

So, bp has started. The output maximum voltage appeared.

Attention! The output voltages are greater than those for which the output capacitors are designed, and therefore, capacitors can explode. I wanted to change the capacitors, so I was not sorry for them, but you can’t change your eyes. Neatly!

So, I learned about + 12V - 24V, and + 5V - 9.6V. It seems that the voltage margin is exactly 2 times. Very well! We will limit the output voltage of our PSU at 20V, and the output current at 10A. Thus, we get a maximum of 200W of power.

It seems that they have decided on the parameters.

Now you need to make the control electronics. The tin case of the PSU did not satisfy me (and, as it turned out, in vain) - it strives to scratch something, and it is also connected to the ground (this will interfere with measuring current with cheap opamps).

As the case, I chose the Z-2W, Maszczyk office

I measured the noise emitted by the power supply - it turned out to be quite small, so it’s quite possible to use a plastic case.

After the case, I sat behind Corel Draw and figured out what the front panel should look like:

The sequence of actions for remaking the ATX PSU in an adjustable laboratory.

1. Remove the jumper J13 (you can use pliers)

2. We remove the diode D29 (you can just lift one leg)

3. The PS-ON jumper is already on the ground.

4. Turn on the power supply only for a short time, since the input voltage will be maximum (approximately 20-24V). Actually, we want to see this. Do not forget about the output electrolytes, rated at 16V. Perhaps they will heat up a bit. Given your "bulges", they will still have to be sent to the swamp, not a pity. I repeat: remove all wires, they interfere, but only earth wires will be used and + 12V then solder them back.

5. We remove the 3.3-volt part: R32, Q5, R35, R34, IC2, C22, C21.

6. We remove 5V: the assembly of the Schottky HS2, C17, C18, R28, you can also "type choke" L5.

8. Change the bad ones: replace C11, C12 (preferably with a larger capacitor C11 - 1000uF, C12 - 470uF).

9. Change the inappropriate components: C16 (preferably at 3300uF x 35V like mine, well, at least 2200uF x 35V is necessary!) And the R27 resistor - you don’t have it anymore and that's great. I advise you to replace it with a more powerful one, for example 2W and take a resistance of 360-560 Ohms. We look at my board and repeat:

10. We remove everything from the legs of TL494 1,2,3 for this we remove the resistors: R49-51 (we release the 1st leg), R52-54 (. The 2nd leg), C26, J11 (. The 3rd leg)

11. I don’t know why, but R38 was chopped off by someone :) I recommend that you chop it too. He participates in voltage feedback and stands parallel to R37.

12. Separate the 15th and 16th legs of the microcircuit from "everyone else", for this we make 3 cuts of the existing tracks and to the 14th leg we restore the connection with a jumper, as shown in the photo.

13. Now we solder the cable from the regulator board to the points according to the diagram, I used the holes from the soldered resistors, but by the 14th and 15th I had to peel off the varnish and drill holes, in the photo.

14. The core of loop No. 7 (power supply to the regulator) can be taken from the power supply + 17V TL-ki, in the area of ​​the jumper, more precisely from it J10 / Drill a hole in the track, clear the varnish and there. Drilling is better on the print side.

I would also advise changing the high-voltage capacitors at the input (C1, C2). You have a very small capacity and have probably dried up pretty much. There normally will be 680uF x 200V. Now, we collect a small scarf on which there will be adjustment elements. Support files see here.


I decided to break the electronics into two parts - a bezel and control electronics. The reason for this breakdown is that there was simply not enough space on the front panel to accommodate the control electronics.

As the main power source for my electronics, I chose a standby source. It was noticed that if it is properly loaded, it stops beeping, so 7-segment indicators turned out to be ideal - and the power supply is loaded and the voltage with current is shown.


On it are indicators, potentiometers, LEDs. In order not to drag a bunch of wires to the 7-segment, I used the shift registers 74AC164. Why AC, not HC? HC has a maximum total current of all legs of 50 mA, while AC has 25 mA for each leg. I chose 20mA current indicators, that is, the 74HC164 would definitely not be enough for current.

Control electronics - everything is a little more complicated here.

In the process of compiling the circuit, I specifically adjusted, for which I paid a bunch of jumpers on the board. You are given the corrected scheme.

In short, then - U1A - differential. current amplifier. At maximum current, the output is 2.56V, which coincides with the reference for the ADC controller.

U1B - actual current comparator - if the current exceeds the threshold set by the resistors, tl494 “shuts up”

U2A is an indicator that the PSU is operating in current limit mode.

U2B is a voltage comparator.

U3A, U3B - repeaters from alternators. The fact is that the alternators are relatively high-resistance, and their resistance is changing. This will greatly complicate feedback compensation. But if they lead to one resistance, then everything becomes much simpler.

With the controller, everything is clear - this is a banal atmega8, and even in the dip, which was in the cover. The firmware is relatively simple, and is made between soldering with the left paw. But, no less, working.

The controller works at 8 MHz from the RC generator (you need to put the appropriate fuses)

On good, the current measurement needs to be moved to the “high side”, then it will be possible to measure the voltage directly on the load. In this circuit, at high currents in the measured voltage there will be an error of up to 200mV. I made it and repent. I hope you do not repeat my mistakes.

Alteration of the output part

Throw away all unnecessary. The scheme is as follows (clickable):

I redid the common-mode choke a bit - sequentially connected the winding which is for 12V and two windings for 5v, in the end it turned out about 100 μH, which is dofiga. I also replaced the capacitor with three parallel connected 1000uF / 25V

After modification, the output looks like this:

We start. Awesome of the amount of noise!

300mV! Bundles are like stimulating feedback. We brake the OS to the limit, packs do not disappear. So it's not about the OS

Poking around for a long time, I found that the reason for such noise was the wire! O_o Simple two-wire two-meter wire! If you connect the oscilloscope to it, or turn on the capacitor directly on the oscilloscope probe, the ripple is reduced to 20mV! I really cannot explain this phenomenon. Maybe one of you will share? Now, it is clear what to do - there must be a capacitor in the feeding circuit, and the capacitor must be hung directly on the power supply terminals.

By the way, about Y - capacitors. The Chinese saved on them and did not deliver. So, the output voltage without Y-capacitors

And now with the Y capacitor:

It is better? Undoubtedly! Moreover, after installing Y - capacitors, the current meter immediately stopped glitching!

I also put X2 - a capacitor, so that at least there was less trash on the network. Unfortunately, I do not have a similar common-mode choke, but as soon as I find it, I'll put it right away.


I wrote a separate article about her, read


Here I had to tinker! After several seconds under full load, the question of the need for active cooling was removed. Most of all, the output diode assembly was heated.

In the assembly there are ordinary diodes, I thought to replace them with Schottky diodes. But the reverse voltage on these diodes turned out to be about 100 volts, and as you know, high-voltage Schottky diodes are not much better than ordinary diodes.

Therefore, I had to screw a bunch of additional radiators (how much fit) and organize active cooling.

Where to get power for the fan? So I thought for a long time, but still came up with. tl494 is powered by a 25V source. We take it (from jumper J3 in the diagram) and lower the stabilizer 7812.

For blowing, we had to cut the cover under the 120mm fan, and attach the corresponding grill, and put the fan on 80mm. The only place where this could be done was the top cover, and therefore the design turned out to be very poor - some metal crap could fall from the top and close the internal circuits of the power supply. I put myself 2 points. Do not go away from the power supply housing! Do not repeat my mistakes!

The fan does not mount at all. It just presses the top cover. So it’s good with the sizes I hit.


Total. So, this power supply has been working for a week, and we can say that it is quite reliable. To my surprise, it radiates very weakly, and that’s good!

I tried to describe the pitfalls that I ran into. I hope you do not repeat them! Good luck

Good day. I would like to clarify the values ​​of the resistors R3, R8, R14 and R18, the parameters L1 in the control electronics, the values ​​of the resistors R22 and R25 in the false panel, and also whether it is possible to lay out printed circuit boards. Thank.

The author of course respect for the development! But for repetition, you first need to conjure the BP control scheme, which are in PDF. Pancake! What makes you first encrypt a circuit? And the one for whom it is laid out here, then decrypts this scheme. What a moron so thought up. Really it was impossible to draw both control schemes (pdf) normally on one sheet and without any links like: Vref, AGND ... What kind of mediocrity is this. BSVi - you have a big minus for drawing schemes! You are mediocrity. Never do this again. Ask the experts to do it

The author did a decent job and wrote a useful article.
As for the schemes, excuse me, on the contrary, you show your ignorance 🙂
Take an example of the use of any imported microcircuit (App Note), and you will see there the same style of design of electrical circuits.

This style, by the way, is very convenient in that even a fairly voluminous scheme remains easy to read, and does not turn into a hard to read “vermicelli”.