Electronics > Beginners

oem atx power suply noise and ripple measurment

(1/2) > >>

kokodin:
Hello
 I am building a custom pc in old oem computer case, so i have to use certain form factor power suply, and it can't be anything more standard than what i could find already. I need to rate if the power suply is up to the task for long term use.
atx standard specs say
Supply (V)    Tolerance       Range, min. to max. (V)    Ripple, p. to p., max. (mV)
+5            ±5% (±0.25 V)    +4.75 V to +5.25       50
−5             ±10% (±0.50 V)    −4.50 V to −5.50       50
+12         ±5% (±0.60 V)    +11.40 V to +12.60    120
−12       ±10% (±1.20 V)    −10.80 V to −13.20    120
+3.3      ±5% (±0.165 V)    +3.135 V to +3.465    50
+5 standby    ±5% (±0.25 V)    +4.75 V to +5.25       50
But i have no idea how to test it in real life.
for instance
My test computer 12V rail have idle voltage of 12,2V, but it sags under load (graphic card benchmark) down to 11,9-11,85V and sensor sweep on pc makes it looks like a very ugly graph, many dips and raises. This is the kind of load power suply will experience, and was measured from inside software on the pc itself, so it might not be very accurate. But i am still concerned that under bigger load it may even fall out of atx specs or die.
Power suply didn't get even a little bit warm under load mind you, and i was benchmarking it for over an hour. And 2 of oem special sauce wires were not even connected (i suspect they are for fan speed controll).
Test system i was using also might be at fault of wrong measurments, because it is an old computer that liked to kill power suplies in the past.

By the way the power suply in question is made by delta and it is 300w fujitsu branded part, I also have 250w model. I have no schematics or specyfication sheat other than the label on them.

Is there even a point hooking up a scope to my test system as it is and looking at the ripple and noise, or should i have linear load like some dumb 300w resistor or bulb? I mean cpu and graphic card are not really a good electronic load candidates, they are very spiky power consumers, very inconsistent. Then again if we consider fluctuations under load as ripple, it would be already out of atx specs by the factor of 2, still in range of min/max voltages though.

I don't really have a way to directly measure power draw on each line . I can measure  power draw of entire system  with a clamp meter, but the power draw on each dc line is different and spreaded, so even scoping cpu 12v socket or atx plug, or molex connector can bring in different data.

What do you think, how should i unpack this problem?

magic:
Current is going to be roughly constant if you apply constant load, such as 1 or 2 or N cores running Prime95 and nothing else. You could probably see the switching ripple by probing in some unused HDD plug. Be sure to connect the ground clip to the exact same plug. Use AC coupling, obviously.

Dave recently did a video on scoping PSU noise, you may watch it for inspiration.

But my first worry would be current capacity, I hope you have checked the label. And close second: cross regulation. As you load 12V, the output on 5V and 3.3V may rise up and possibly go out of spec. Such are the wonders of some ancient PSUs designed for systems utilizing primarily low voltage rails.

kokodin:
Well it isn't that ancient, it is 2007-2009 vintage . Labeled as maximum continuuous output 300W. And A*V all summed up come up exactly as 300W with 10 sec surge overpower of around 315W
3,3V - 8(9)A
5V - 15(17)A
12V -17A
-12V - 0,3A
aux/standby 2(2,5)A
And for i could say, putting it under heaven benchmark made 12V jump around while 3,3 was rock solid in sensor sweep in aida64 set to 50ms probing.
I was using it a lot last year with 98W phenom 2 x4 and the same gt1030 4 sticks of ram and 2 hdd's as a joke pc, and it was keeping up. But it was running paired oem board with special sauce bios to accept phenom 2 processors at all.
i wish to turn it into something a little bit more modern (ryzen 3 1200 on b450 board) and it would be a disaster if it gone in smoke with the broken power suply. I might be overcaucious because it does run a test system of pentium dual core on nforce  board with the same gt1030 with no problems. What shocked me was it being cold to touch while under load and i started checking things up.

I hope analog 10mhz scope would do to visualise ripples and noise, but i did watch daves video and my scope doesn't even have bandwidth selection :] Watching Daves video might be the root of my interest in ti in the first place. I don't really have many engineering tools , only basic ones i use for work.

As i understand i can do 2 things
1 put psu back in phenom pc and put it to crunch prime or other syntetic cpu stress test and load psu to around 120W I wonder if there is similar thing for graphics, because i could het to around 150-170w this way , but still 3 phase cpu power would make more noise than psu itself
2 try to do it on non oem board but then i have even less load capailities

I am not really sure where am i getting with that, i am more interested in how it should behave in real life scenario, than testing it's ripple and stuff in a lab way.
How much of a voltage fluctuations should i exect under wierd loads like that turning on and off constantly. stuff like that.

tunk:
I believe Delta is a reputable brand. That said, the capacitors
may be beyond their sell by dates. If they are, you should be able
to get an indication if the ripple is too high. I've also got the
impression that recent PCs mostly uses the 12V rail - if that's the
case, the PSU can provide around 200W.
 

kokodin:
Well yeah, new one was rated around 200W on 12V rail, that is obvious by a label. But it is quite small unit all things consider , so it have to run in quite high rfrequency to compensate for lack of weight and size. Recaping capacitors might be actually a good idea. There is not many of them there anyway.  I doubt increasing the output caps capacity would help anything, but i could also do that (in reasonable amount). Don't know how would it play out in the long run with the switching frequency.

I am just concerned if it is a good indicator that under random load of a game benchmark, 12V power rail jumps from 12,2 flat line,  to 11,8~12,1 stairs just after launching the benchmark. I might test it with different power suply first , because it might be just a sensor glitch on the board. (i am dataloging voltages from the pc via software on that pc) I have comparable 350W atx psu with only difference being 22A on 3.3V rail and 4A on standby.  I was busy remodeling the inside of that computer case yesterday so i didn't do much testing .

Navigation

[0] Message Index

[#] Next page

There was an error while thanking
Thanking...
Go to full version
Powered by SMFPacks Advanced Attachments Uploader Mod