Bench Power Supply (PSU conversion)

[cprobertson1]

Firstly, sorry if this has been answered - I did have a look around but I can't find an existing thread or at least one that worked

Its a simple converted ATX power supply (Model No. FSP145-61GN)

When I try to power it on it quickly turns off again (I would assume the protection circuitry is kicking in)

It's been constructed as per this tutorial only all the wires coming out the PSU go into an ABS case (containing all the circuitry and terminals) mounted on the outside of the PSU

The wires are as follows:
13x GND == BLACK
3x 3.3V == 12.0A: ORANGE*
7x +5V == 180A: RED
4x +12V == 4.2A: YELLOW

Brown == Sense wire? (bound to the thinnest of the orange wires)
Grey == Power Good Signal
Purple == +5V   0.8A StandBy
Green == Unknown (assumed soft-power on)
White == -5V   0.3A
Blue == -12V   0.5A

Green is tied to ground via a switch (as per the tutorial
Purple and Grey are tied to ground via LEDs and 330 Ohm resistors (for power good and mains voltage signals)

When plugged in 5V standby voltage (purple) line is live: and if turned on (via the switch) the fan starts for a second and then stops again (the other wires jump to their proper voltages for a second and then go dead again) - so I'm assuming the protection circuitry is kicking in for whatever reason.

I have a 10 Ohm load (its not 10W though: which may be the problem) and a 12V fan connected to the 5V lines (red).


Any ideas? Am I just doing something daft? Maybe I'm not wasting enough power to start the powersupply (it is a switchmode PSU)?

sorry to bother you all

[potatogun96]

I would suggest trying to load down the 5 and 12 volt rails some more. I had one that didn't work well until there was a 5 ohm load between 5 volts and ground. If you have some automotive headlight bulbs or something, you could probably load down the supply enough to get it to work.

[cprobertson1]

Good call! loading down the 12V rail worked a charm!

Thanks a bundle - pics coming soon!

[cprobertson1]

Will putting an LED/Resistor parallel to the outputs mess anything up on the output side?

It just occurred to me that adding a "fuse good/connected" LED indicator might be useful in conjunction with the removable fuses. xD

--edit-- fixed typo (LEDs are NOT in parallel to the output fuse! just the output itself (so they go out if the fuse blows or if there is nothing plugged in)

I've just noticed that the LED to go in with the 12V rail would require a 1W resistor at 220 ohms, but I only have 1/4 W resistors: are you able to stack them in serial/parallel to increase their power rating (i.e 4 1/4W resistors with resistance equal to the 220 ohms required) - or would I be better using a transistor as a switch for the LED?

[cprobertson1]

Finally finished! (my power resistor was delayed 



Looks nice enough: not perfect (I'll get the cables wrapped away when I can be bothered - still needs painted too: I'll get the airbrush out later and give it a nice matte coat and cover up those unsightly hot glue splutters

Only question is with the outputs: I shoved them through an oscilloscope and... well are they supposed to look like this?

3.3V
Noise bounces up and down from 0-3.3V: is that just the switching voltage regulator at work? (10 uSecond timebase - cant be bothered calculating the frequency; its passed midnight

5.0V
Couldnt get a good image of it (couldn't get a good trigger on it) - its a damn-near pretty good triangle

12V
Not a lot of periodic noise on the 12V line but a lot of hum (thats at the smallest time range (.5us/div): might be a triggering problem though (that was the best trigger I could get: couldnt focus on anything else apart from 50Hz hum from the mains.


Any suggestions? is this sort of noise to be expected? I don't know much about power supplies nor noise (apart from switching noise from ICs which is rarely a problem in my hobbyist world )

[IanB]

Well...ATX power supplies don't make laboratory power supplies. That is not what they were designed for.

With a switching supply you may find lots of noise on the outputs unless you put a decent load on them. And even then, there will still be noise. The outputs will contain as much noise as digital circuitry can tolerate--that's what keeps the cost down. If you want a power supply with clean outputs for analog use, look elsewhere. If you want a basic power supply for motors, lamps and other non-fussy things, you have the just the thing.

[Rigby]

What if he wanted a clean supply and added some linear regulation?  He'd lose voltage; would he get finely regulated voltage?

[IanB]

What if he wanted a clean supply and added some linear regulation?  He'd lose voltage; would he get finely regulated voltage?

The linear regulator would need to be able to reject the switching noise on its input. That would depend on the specs of the regulator. However, some passive filtering in front of the regulator may help, for example an LC filter or two.

[cprobertson1]

Well...ATX power supplies don't make laboratory power supplies

Aye, I know - it was a fun little project though - and with it cost me about £5 so it wasn't exactly breaking the bank. I'll own a real bench power supply one day, this is just to tide me over until I haz money again

If you want a basic power supply for motors, lamps and other non-fussy things, you have the just the thing.

I think the fussiest thing I'm running are Atmel microcontrollers; I've not had any problems yet (they get programmed with the constant supply from the USBAsp (I should probably check how clean that one is too xD) - not to say I wont have problems in the future: I can always turn to batteries should I require quieter input


Thanks for the feedback btw!

Ps - quick question about linear regulators: am I right in saying they simply waste the excess voltage/current? (i.e if Im drawing from a battery it will be wasteful)?

[cprobertson1]

Your 3.3V rail almost looks like it has no output capacitance on it.

Is there a possibility of a bad cap inside it? It's quite an old supply =/ (And I kinda dont want to go poking round inside it if I can help it: the board looked like a nightmare to detach from the metal frame for one thing; massive high voltage caps for another

Could I maybe get away with putting a cap inside the "control box" I attached to it? What sort of size would I need (or better still: instead tell me how to calculate it or even just the name of the equation - can't beat a challenge!)

Your 5V rail may be normal depends on V/div scale

Im pretty sure it was down at 5mV: so thats about 20-25mV at the peaks.

[rjk5378]

FWIW, I use little device called an ATX Breakout Board. It plugs into the 24 pin connector used on most ATX supplies. Actually, I think it only uses 20 of the 24 pins, but the point is that it is designed to interface with the ATX with no need for modifications. No need to open up the ATX, solder all the correctly coloerd wires, etc. (I did, however, cut off all the other little connectors and associated ribbon cables, so there is just the one cable to the 24 pin connector coming out.) Plus when the ATX supply gives up the ghost--this arrangement frequently involves old ATX supplies from discarded computers--you can just unplug the breakout board and move it to a different ATX. (Also, if the ATX supply is *REALLY* old, you can alos get a -5 volt jack as well.)

As has already been stated, not a substitute for a good lab power supply, especially if noise is a concern, but it is handy. And for certain hobbyists, it may be all that is needed.

The link below is to the board I bought from Dangerous Prototype. This lists the cost at $14. I can't remember what I paid, it may have been closer to $20, but it is cheap.

http://dangerousprototypes.com/docs/ATX_Breakout_Board

[cprobertson1]

@rjk5378: That's a nice little board xD: I might grab one at some point (though I'm hoping to get an actual power supply by august at the very latest: Im hoping the DIY version won't die before then xD)

Well it couldn’t hurt to try.

It already has hurt... it's tasted blood... Will try that out today and let you know how it goes!

--EDIT--
noise on the other rails is 20mV base to peak

Adding at 470uF cap across the 3.3V reduced the noise to about 15mV: and a 1000uF reduced it to about 10mV (am I right in saying the waveform should stay roughly as it is; with caps affecting the amplitude of the noise?)

[daveatol]

But all bets are off when you dont know the topology. I'm just assumuing a buck converter. That and the fact because your output cap is far away I can't  say for certain. Ripple shapes on the output can vary quite a bit based on topology output filtering,board layout etc..

ATX supplies generally use a half-bridge topology.

[diyaudio]

Firstly, sorry if this has been answered - I did have a look around but I can't find an existing thread or at least one that worked

Its a simple converted ATX power supply (Model No. FSP145-61GN)

When I try to power it on it quickly turns off again (I would assume the protection circuitry is kicking in)

It's been constructed as per this tutorial only all the wires coming out the PSU go into an ABS case (containing all the circuitry and terminals) mounted on the outside of the PSU

The wires are as follows:
13x GND == BLACK
3x 3.3V == 12.0A: ORANGE*
7x +5V == 180A: RED
4x +12V == 4.2A: YELLOW

Brown == Sense wire? (bound to the thinnest of the orange wires)
Grey == Power Good Signal
Purple == +5V   0.8A StandBy
Green == Unknown (assumed soft-power on)
White == -5V   0.3A
Blue == -12V   0.5A

Green is tied to ground via a switch (as per the tutorial
Purple and Grey are tied to ground via LEDs and 330 Ohm resistors (for power good and mains voltage signals)

When plugged in 5V standby voltage (purple) line is live: and if turned on (via the switch) the fan starts for a second and then stops again (the other wires jump to their proper voltages for a second and then go dead again) - so I'm assuming the protection circuitry is kicking in for whatever reason.

I have a 10 Ohm load (its not 10W though: which may be the problem) and a 12V fan connected to the 5V lines (red).


Any ideas? Am I just doing something daft? Maybe I'm not wasting enough power to start the powersupply (it is a switchmode PSU)?

sorry to bother you all

I have a similar power supply on my desk modified similar to what you have done, I used it primarily to test and repair car audio amplifiers and stereos 6 years ago, it still works fine up till today.

Did you know, you don't need to load the 5 volt rail to increase voltage ? if your ATX SMPS is similar to the build I have, it should a TL 494 as the control IC, with two loops one for voltage mode and one for current mode optically isolated or transformer isolated. you can as I did, "gently" increase the duty cycle by a factor of 2% to impose a increase in voltage, my current ATX is re-programmed to run at 13.8 VDC as long as the new re-programmed voltage is lower than the output filter caps it should be fine. All parameters like OC trip hiccup works,  I can get 200Watt flat from mine, I don't get the residual noise floor you experiencing I haven't modified any of the filters, consider this the ATX power supply is used to provide stable voltages and current to a CPU core and DMA controllers on the mother board so it got to be stable. (do take into account their are more filters on a mother board to iron out ripple from the high current paths)

[cprobertson1]

Definitely something up with the 3.3V Rail, even when loaded there's some very strong 50Hz noise (80mV base to peak)

I reckon there's a buggered cap somewhere =/ Oh well - I guess I don't use the 3.3V that much, and I can smooth it out a little by shoving some caps downstream of the terminals. Oh well... anybody know a cheap way to check for bad caps? I don't have an ESR meter (and there are no obvious bulges on them)

At least the other rails work alright (the 5V noise is minimal when its under load; same with 12V)

[granz]

Keep in mind as you use your converted ATX PS that the DC output ground will be earth ground.  You don't have a floating isolated supply like a proper bench supply.