Another ATX PSU to benchtop supply conversion

[triacontahedron]

Hi,

I am new to the electronics design and need a benchtop power supply so I decided to
convert one of my ATX ones. I have looked around for conversion schematics etc. but
did not find one that fully implemented all my ideas about such conversion kit. Basically
I would like to have:

1. Ability to plug in any ATX PSU without drilling holes or soldering.
2. Outputs for all voltages: 12, 5 , 3.3.
3. Fuse protected outputs so my project wont go up in flames on my table.
4. Switches on all outputs, no need to plug unplug stuff.
5. Build in USB charger for all gadgets that I have.
6. Use PCB as front panel, no need to mess with a bunch of wires.
7. Low cost, variable voltage output, measurement circuit etc. will be placed on
a separate PCB and it is a completely different story.

So after spending a few hours in KiCad I came up with the attached schematic.
I would appreciate any input on it or proposal of features that can be added and
might be useful in a benchtop psu.

Thanks

 

[Reloadron]

Have you tried using the PWR_OK signal to drive a LED yet? If I look at the PSU design specification I see the following:

Signal Type +5 V TTL compatible

Logic level low < 0.4 V while sinking 4 mA

Logic level high Between 2.4 V and 5 V output while sourcing 200 μA

High-state output impedance 1 kΩ from output to common

However, I also found this:

Could this signal be of any use when converting a PC powersupply to desktop usage? Realistically, the answer is YES. A voltage on the PWR_OK line indicates that the PS has completed a successful POST and that the output voltages are stable and within design specification. If you wanted to use an LED (light emitting diode) as an indicator that the PS is on, rather than tie it to one of the +5V or +12V lines, attach the PWR_OK line to the anode (+) side of the LED and place a 220 ohm resistor on the cathode (-) leg before grounding it. The cathode leg is normally shorter on new LEDs -- if the legs have been clipped, the cathode will be on the same side as a flat spot on the LED base. The LED should be bright for normal operation -- it could possibly glow faintly if the PS has withdrawn the signal simply due to bleedover.

Leaving me curious as to if the PWR_OK signal can drive a LED? Never really gave it any thought. Beyond that depending on how much current you may want it may be advantageous to tie the 12 volt, 5 volt and 3.3 volt lines out on the 24 pin connector in parallel configurations for each. I may be wrong but I think (operative word think) is the Molex connector pins are rated at about 5 to 7 amps and a single line from the PSU is AWG18?

<EDIT> The later quote came from here. </EDIT>

Ron

 

[triacontahedron]

Have you tried using the PWR_OK signal to drive a LED yet?

No, but I copied this part from this guy https://www.mbeckler.org/powersupply/ and it seems his mod works.
Maybe I should use 330R instead of 220R like he did.

I don't think you can connect together same voltage lines due to stability or something. Not sure where I read
about it and can not locate the link atm. As far as I understand, making a PCB that can handle > 5A without
increase in temperature would require 10 mm (0.4 in) wide tracks so it is not very practical. It is better to run some wires
between input and output or to make separate high current PSU. The other thing is that most bench top tasks don't require much current nowadays.

PS: there are 4 pin connectors on the right side that don't have fuse and switch, you can solder wires on top of tracks
if you need high current at +12 or +5 V

 

[Reloadron]

Before I forget the value of the series resistor used with the LED will be a function of the LED forward voltage and current. For example if a basic Red LED has a forward current of let's say 20 mA and a forward voltage drop of 1.8 volts and you want to limit the current to 15 mA (so it has a happy life) you would take Vsupply 5 volts - Vled 1.8 volts / Iled .015 amp = 213 Ohms so we use an off the shelf 220 Ohm resistor.

When I mentioned placing voltage rails in parallel I limited myself to the 24 pin connector. For example pins 10 & 11 are both 12 volts and in fact the same 12 volts so placing pins 10 and 11 in parallel is not a problem. Where problems can arise is when rails supplied by different regulators are placed in parallel. Even that depends on variables as to how the PSU was designed. This is an interesting read on the subject of multiple rails.

Something I did notice is you chose pin 13 for your 3.3 volts. Depending on the PSU that pin may be 3.3 volts or it can be used as a 3.3 volt sense line. My thinking is to use another pin for the 3.3 volts and tie pin 13 to your choice at the fuse. This way if the supply relies on 3.3 volt sense you have it covered and are not using pin 13 as a power source.

Anyway, I fully agree if you have no need for higher currents there is no need to parallel lines. It should work fine and last a long time.

Ron

 

[triacontahedron]

Thanks for the input, the link is quite interesting.

220R is indeed optimal but I thought that if you consider PWR_OK signal problems 330R might
be a safe choice (LED will not be that bright but it is ok). Anyway I can always replace a resistor.

But the other reason I did not use pin 11 and also 12, 23, 24 is that you dont have them in old ATX PSU
or you might want to put 2x10 connector instead of 2x12 (one 12V pin i think should be able to provide
enoughth power).

As to about 3.3 sence as far as I undestand in advanced PSU there are 2 wires that go to pin 13: power and sence.
So that PSU can stabilize voltage at the pin, that is exactly why I connected 3.3 output to pin 13.

 

[Reloadron]

Yeah, I didn't bother to get into the earlier 20 pin versions and the thinking as to the LED series resistor also makes sense. Hey, better to start high and go low only if needed.

Several years ago I was much more into PSUs including building load banks for testing the things. Especially during the great PSU wars in marketing. I wasn't aware of the 3.3 volt sense being tied at the pin or actual PSU output. I just figured the tie point was actually on the motherboard for sense.

So all things considered go ahead and build the thing up. It should work fine. Feel free to post some pictures and maybe another member has something else to add.

Ron

 

[tcmtech]

I have taken apart dozens of different computer power supply's over the years and every one I have ever taken apart has all of the main 3.3, 5, and 12 volt leads coming from the same points internally. The only odd ones are the negative voltage and 5 volt standby power outputs.

As far as increasing PCB circuit current capacity thats easy. Just solder strips of stranded wire on over your circuit traces.

 

[Mikebits]

Keep in mind, these supplies are plum full of switcher noise and you may want to add some output filtering, especially with audio.

 

[triacontahedron]

I was going to add separate board with voltage/current regulation features, some displays etc. Though I should probably check
how bad noise is on bare ATX PSU, it might be worthwhile to add some inexpensive filter circuitry. Will do this when I get my hands
on an oscilloscope.

 

[triacontahedron]

Hi, I went ahead and build this thing, see attached pictures. As expected a few issues came up:

1. I wired usb backwards and successfully shorted 5V standby rail. It seems that PSU did
not mind this that much but refused to power on. I was not impressed because 3A fuse did
not blew up, no fun at all.

2. I did not have 220R resistors so powered LEDs through 470R. They a re still too bright.

3. 10R 10W resistor quite warm to the touch, maybe should have installed 20R one.

4. -12V on PSU is actually around 11.5V and all other voltages are quite off as well.

5. It could be a problem with my PSU but it does not turn off some times when I open
main switch.

6. It would be nice to have LED on each output.

All in all, I am quite happy with the thing, now just need to build some kind of case for it.

 

[Reloadron]

Came out pretty nice.

Ron