SOLVED [VIDEO] What is this sound my brand new PSU doing?

[dougfrippon]

I got this CORSAIR Professional Series HX1050 1050W PSU, to power up my new rig.. basically Intel 3770k IVY with dual GTX 670 sli.

Whenever my computer is under heavy GPU load, the PSU starts doing some weird sound. And it's really just under HEAVY GPU load.
Video convertion / Prime 95= no sounds.
Normal games like Diablo3 = no sound.
Uningine and 3Dmark and such = huge sounds...

I was convinced the sound came from my video card (since the sound appears when I use the GPUs intensively) But no, it's really coming from the PSU unit, I guess they video cards must be draining extra wattage under heavy load which explain the link.

So, have you guys heard that before?
I guess I'll have to RMA now... 2-3hours worth of cable management to the trash

edit: solution.. I found out this is coil whine, and that we can't do much about it..

 

[Janus67]

Even if it is coil whine you can still contact the manufacturer to see if they will RMA it.

 

[cyberfish]

Mine does the opposite. It whines only when my computer is idle or very low load.

I think what happens is PSUs nowadays all use variable frequency PWM depending on load, and at some frequency things resonate. I guess we'll just have to find and avoid that range for each PSU... I would still blame the manufacturer, though, if the whine is really that bad (they should specify a maximum noise level, and if the PSU exceeds that while operating in rated load, that should be covered by warranty).

 

[Bobnova]

I seriously doubt that a low end/midrange thing like the CX uses viable PWM frequencies.
I have yet to see a PSU that does, actually. I've scoped a couple dozen now.

Differing amounts of load will maintain different strengths of magnetic field around the inductors as well as different rates of change. That's what your issue is about.

It shouldn't whine. NZXT has taken to wrapping almost all the inductors in heatshrink to prevent it, and gluing (*&@#*% everything.

 

[cyberfish]

That's interesting. I have never scoped PWM on a PSU.

If they do constant frequency, shouldn't it be outside of audible range?

Unless there's some sub-harmonic stuff going on, which is possible if load is not constant.

 

[Bobnova]

There's tons of harmonics going on in there. Actual switching frequency is typically somewhere in the 50-150kHz range, but after a few layers of harmonics (both magnetic and physical) it can hit audible ranges.
Some PSUs will go into electronic oscillation and harmonics as well, the last three units I've tested all did interestingly enough.
Quick example from the most recently posted review. This is the 3.3 V rail outside and inside the oscillation issue zones. Scope is set at 10mV/div in both, 10 microseconds/div on the first pic, 50 microseconds on the second (the other PSUs have had much slower oscillations, to where I could easily hear the whine. Around 1.5kHz and 3kHz if I recall correctly)




You really should scope one if you have access to a scope, they're fascinating.

In that second shot you can see the waveform from the first shot inside the larger sine wave.
It's fuzzy because the sine wave isn't exactly the same frequency every cycle, it shifts enough that it is in a slightly different spot every time the scope draws it. You probably wouldn't see that on a digital scope though, it's one of the pluses (or minuses, depending on what you're up to) of an analog scope.



On the original post's subject, if it annoys you it's RMA time. It shouldn't be making a noise.
Most of the time it's just something that didn't get enough caulk smeared on it, sometimes it is something actively failing.

 

[cyberfish]

That is very cool. On the second one it looks like the frequency is ~5kHz, which would definitely be audible.

Does this happen on higher end PSUs, too? Or just low end stuff?

It looks like there's something wrong with frequency compensation inside the PSU or something. Are those just with constant purely resistive loads?

 

[cyberfish]

By the way, the effect can be reproduced on digital oscilloscopes as well, in infinite persistence mode (not sure if all scopes have that, but the $100K one beside me does ).

 

[Bobnova]

Any unit can have it, if something is wrong inside.
The units with similar looking issues were 430w, 750w and 1000w, all are actually capable of putting out their rated power.

Loads are purely resistive and entirely passive, load resistors with mechanical hand operated switches

 

[cyberfish]

At what load range do you usually see the oscillation?

And I just tried scoping a random PSU I found here in the lab... turned out my expensive active probes cannot do AC coupling .

 

[cyberfish]

I wonder what will adding capacitance do to it. Hopefully not oscillation into explosion.

 

[Bobnova]

You can add all the capacitance you'd like to, don't worry about it.

The issues happened at medium-low to medium-high load after sustained high load and high intake air temps on all three.
Two of them it was the 3v3 rail, one it was all rails (I suspect the APFC on that one).
One of the two 3v3 rails an inductor change between ES and preproduction was the culprit. Same specs, but one oscillates and the other doesn't.
All three made some noise while doing it, in rough proportion to the amount of ripple observed.

Grab a couple $7 passive probes off ebay, they actually work quite well

 

[cyberfish]

Just bought some big capacitors and Molex connectors from Digikey. Going to attach them externally and see what happens. At the low low oscillation frequency, lead inductance from the wires should not be an issue.

Maybe the phase margins were so low that high temp causes the electrolytic caps ESRs to change enough to make the loop unstable.

Grab a couple $7 passive probes off ebay, they actually work quite well

Or just grab a capacitor .

 

[cyberfish]

Finally had time to hack something together to see the rails -
http://www.overclockers.com/forums/showthread.php?t=715001