Large Ripple on an ATX PC Power Supply?

[techie1234]

Being rather new to electronics and having just bought my first oscilloscope (an HP 1743A), I’ve been trying to measure the things I have on hand with it.  One of my recent forays was an ATX PC power supply.  So I hooked things up and was rather surprised to see quite a bit (+/- 750mV) of ripple on the 5V output when looking at it with a TPD of 20 uS, even without any load.  Reading that PC power supplies often need some minimum level of draw (mine said 1.5A @ 5V) to regulate voltage correctly, I wired up some resistors to create a ~2A flow.  The ripple got much worse and I got a high-frequency whine from the supply.  So I popped a 4700 uF capacitor in and that silenced the whine and reduced the ripple – but even then I’m still left with nearly 1.5V PP of ripple.  It doesn’t look like HF noise from my setup – it’s a nice, solid, steady waveform but just one that dances between 4.x and 5.5V.  The ATX spec says the ripple on 5V should be 60mV or less, so either a) the power supply is bad, b) my equipment is bad, c) my technique is bad, or d) all is fine and this what I should expect.  Any thoughts on which it might be?

[Simon]

what frequency is the ripple.

[techie1234]

what frequency is the ripple.

Since I’ve got TPD set to 20 uS and the deformed waveform repeats about every 2 ½ divisions, it would seem to be about 20 kHz.  See attached pic (hopefully it will display) - I’ve got the VPD on the oscope set to 2.0.

[Simon]

how old is the supply ?

[techie1234]

how old is the supply ?

Good question - could be a decade old as I scavenged it out of an old server.  Let's just say it had a 3/12" floppy installed.   I haven't opened it up yet to see if anything looks amiss - I thought this was going to be an easy exercise...

[techie1234]

I also tried a variety of different caps (size and construction) in various combinations - it didn't seem to make much of a difference as long as at least a few uF was included in circuit.

[Kilrah]

After 10 years it could well have dead caps.

[Simon]

Good question - could be a decade old as I scavenged it out of an old server.  Let's just say it had a 3/12" floppy installed.   I haven't opened it up yet to see if anything looks amiss - I thought this was going to be an easy exercise...

Floppy ? remind me what's one of those ?

[Kilrah]

 

[rrinker]

 You aren't using a 10x probe, are you? That's still too much ripple per spec, even it if really is more like 150mv instead of 1.5v. There always have been crappy power supplies sold for PC use that in some cases did not meet the specs.
 Also, do you have a load on this? Many computer power supplies poorly regulate with no load, especially older ones. I'm always reminded of the one brand of PC (AST I think it was) that came with a large sandbar ballast resistor mounted to the top of the drive cage - if you ran the computer without a hard drive, it needed this resistor connected in order to provide sufficient load to keep the power supply in regulation.
 Or, it's just toast, which is also highly likely given the age.

[techie1234]

After 10 years it could well have dead caps.

Yeah - I was starting to think about that.  I have another scavenged ATX supply sitting at the office - so I'll bring it home on Monday to compare results.  But, wouldn't the 4700uF (and other value) cap I put in circuit smooth things out to such a degree that it wouldn't be a problem?

[techie1234]

You aren't using a 10x probe, are you? That's still too much ripple per spec, even it if really is more like 150mv instead of 1.5v. There always have been crappy power supplies sold for PC use that in some cases did not meet the specs.
 Also, do you have a load on this? Many computer power supplies poorly regulate with no load, especially older ones. I'm always reminded of the one brand of PC (AST I think it was) that came with a large sandbar ballast resistor mounted to the top of the drive cage - if you ran the computer without a hard drive, it needed this resistor connected in order to provide sufficient load to keep the power supply in regulation.
 Or, it's just toast, which is also highly likely given the age.

Nope - it's a 1X probe.  I do have a ~2A load - which I added early on thinking it might be the problem.

[Simon]

 

Slightly too fast but almost authentic tones

[Simon]

After 10 years it could well have dead caps.

Yeah - I was starting to think about that.  I have another scavenged ATX supply sitting at the office - so I'll bring it home on Monday to compare results.  But, wouldn't the 4700uF (and other value) cap I put in circuit smooth things out to such a degree that it wouldn't be a problem?

Maybe the old cap is causing an overload as partly shorted ?

[Kilrah]

But, wouldn't the 4700uF (and other value) cap I put in circuit smooth things out to such a degree that it wouldn't be a problem?

There are many caps in a PSU - the primary could be throwing crap down the line as well...

Reading that PC power supplies often need some minimum level of draw (mine said 1.5A @ 5V) to regulate voltage correctly

Note that while it is very often the 5V rail that is regulated, it is also not necessarily the case.

[tggzzz]

Floppy ? remind me what's one of those ?

The only mass storage device on a computer I bought last week for the sole purpose of (re)running the first freely available Smalltalk implementation. 

Somewhat slow, as expected, but the worst things are the sheer mass of the mouse and thickness of the keyboard. No wonder RSI became a problem fir many people

(Well, free = some money for the 7 distribution floppies).

[techie1234]

As I think Dave would say, "Winner, winner - chicken dinner!"  I don't think those caps are in the same shape as when this left the factory...looks like this one will get de-soldered for parts...

[mariush]

The Fuhj*$^$  capacitors would have to be replaced, the Teapo capacitors may still have some life in them but it wouldn't hurt to replace them.

Besides those, you'd still have a problem, that dark yellow glue which goes brown as it ages and dries out and becomes conductive , risking shorts or changed resistor values. It's best to remove it. 

[techie1234]

The Fuhj*$^$  capacitors would have to be replaced, the Teapo capacitors may still have some life in them but it wouldn't hurt to replace them.

Besides those, you'd still have a problem, that dark yellow glue which goes brown as it ages and dries out and becomes conductive , risking shorts or changed resistor values. It's best to remove it.

Yeah...I think I'm gonna just gonna junk this one and use it for components and connectors. Most importantly: looks like my 1743A was tellin' the truth!

[oldway]

It is a very good oscilloscope and it never lies.

But you would have to replace his time base knob....