Overload and or Dust bunnies at work. Don't let this happen to you

I was going to post in geek food but I heard humans don't like eating fried PC hardware so posted in electronics.

So my trusty old Zalman 460w power supply in my PC has failed. I'd noticed a bit of a hot dusty stanky smell the last couple of days but couldn't pinpoint the smell source. I woke up this morning and my PC was off and had that heart sinky to stomach feeling.

I thought crap maybe there's something wrong with my UPS but everything else was still powered up and okay so let's check out the power supply.


Click to view full size!

Click to view full size!

Click to view full size!

See sig specs for PC specs + 4 hard drives to get an idea of the load it was under.

So what was the ultimate cause ? Please speculate I'm curious. I want to know how I can prevent this happening again.
I had an alternative theory. That the power supply was infact not truely rated for 460 watts and commited suicide at the thought of supplying any more than about 300w

I suspect a multiple chain issue, dust causing excessive heat as well as a mild overload perhaps. No idea.

 

Hmm, it looks like there was moisture or oil involved.

What does the fan look like? is their a ring of stuck dust around the ducted region? I noticed this happened on my Noctua fan, which threw grease around the edge of itself.

 

Accumulation of dust and the components most likely heating up to a higher than normal temp, leading to more resistance, overworking the psu and leading to its demise sounds good to me. Looks like some oxidation was happening as well but I'd like to hear the opinions of the electronics gurus here. That's just my best guess.

 

Plain old dust bunnies mate. I shifted all of the dust outside and it came out easily. it was dry.

The oily shit in there I'd say is from the plastic coating on the wire breaking down.

 

Rightio, "clean psu" has been added to my to do list.

 

^ Ditto.

I was going to say - it does look a bit like the insulation on that coil's wiring has melted and been soaked up by fluffy dust doesn't it?

Wild speculation aside, there is a _significant_ amount of crap in there, and your specs seem quite high for a 'not entirely trustworthy, probably something like' 460W PSU (Not knocking Zalman specifically here, just mean - they're not exactly a major OEM like CWT or Seasonic). So, odds are, the heat has just gotten to it over the years.

 

Im an sparky not a electronics expert but looks like overheating causing the dust to smolder which would then result in insulation breakdown.
After that it would become a chain reaction as further coils short out and increase current/heat.
Even see a couple coils have oxidation (green) which would indicate the failure of insulation.

 

So if that's the case. What is the chain of events that my PC would have saw as the power supply began to fail ?

 

From my experience designing and building a switchmode PSU here's a guess:

The coil you photographed has obviously become very hot at some stage, this causes 2 things:

1) Increased resistance in the wires
2) Eventual enamel insulation breakdown

In a transformer point 1) would simply cause the secondary voltage to drop, reducing the current flowing to the load.

However a switchmode PSU is not a simple transformer. They have a feedback circuit which maintains the output voltage. Because of this an increased voltage drop (ie: increased losses) somewhere in the circuit will cause the feedback circuit to push more current through. In a buck regulator (ie: the fundamental circuit the ATX PSU is built off) this results in the main switching coil being turned on for longer.

What happens then is that more heat is dissipated in the coil, which causes more current to be forced through it etc. This is a positive feedback loop which, without some external limit, will result in the coil burning out.

What might have happened is that the increased loss in the coil, or the shorting of coils reducing its inductance, caused the PSU to operate at a higher switchmode frequency. When this happens the switching transistors heat up more due to switching losses, this would further upset the coil due to increased heating.

As the switching frequency increases it eventually reaches a limit due to the maximum frequency that the switching transistors can handle. At this point the output voltage (ie: your +12V, +5V etc supply rails) will start to drop.

In a well designed PSU this will cause a self-shutdown because the feedback circuit can't maintain the output voltage. Simply shorting a rail to ground will almost always cause an overload protection part of the PSU to power off everything, ideally this would have happened.

However, since the PSU is full of dust and probably overheated/old/etc chances are something failed due to it being driven too hard.

If you want to do a detailed autopsy you can check all the schottky diodes and the switching transistors (everything on a heatsink), chances are one of them is dead. Obviously the inductor in the photos is fried too, but shorting coils will only reduce it's inductance, causing something else to blow before the PSU finally dies.

Anyway, that was long, rambling and probably not all that helpful. Thanks for reading it anyway.

 

Actually it was very educational and insightful Thank you.

I'm not very happy with myself for allowing the unit to become this dusty. I often made an effort to blow the power supply out with compressed air but it wasn't enough. I defintely should have cracked it open and cleaned it properally. Despite being against the old dont' play with 240v rule.

I've done all I can with this unit. I removed the fan, Stuck it in the front of my PC. Pruned the cables off. I'm sure I can find a use for them and thrown the rest in the garbage.
Very satisfying!

It's funny you should mention that an overload probably took the power supply out. I only upgraded to the Radeon 4890 from a Geforce 8800GTS about 1 week before the PSU died. I recall maximum load from my power meter was about 280 watts for the system.

I'd say the power supply did exactly what it was meant to do. Failed in the least spectacular way possible. No hardware was damaged and there was no smoke and flames like I see cheapies often blow out the back.

I believe I got a good life out of this one. It's just a shame it was cut short.

 

My opinion:

The main components in a power supply that need cooling are the semiconductors,mainly switching transistors and diodes.

That coil is on the input side of the power supply. I'm not exactly sure of it's function so I can't say why it has overheated. It is possible that another component has failed and put too much current through it.

However there is not enough dust in the PSU to seriously impede the airflow over the heatsinks. The coils themselves should not normally get very hot.

I suspect the PSU died a natural death from other causes and that the dust was irrelevant.

 

That coil is a ferrite choke that is used to filter noise, specifically Electromagnetic Inteference and Radio Frequency Inference. Even if the coating fries on the coils it would not 'short' catastrophically as such as it is still the same voltage and it is not grounding but it would lose inductance and increase the amount of EMI and RFI entering the PSU as you would have less loops through the ferrite core.

Looks like it was getting pretty toasty. Dust bunnies FTW.

Nice explanation LINUX and I agree, there would have been another point of failure for sure.

 

Assuming you mean one of those cheap jaycar style power meters. Since PC power supplies are switch mode the sine wave is modified.
I tested a Jaycar meter about 2years ago and they are not True-RMS thus will read a computers draw up to about %30 low.
Ie my PC measured at 0.9amps but when measured with true-rms meter was really drawing 1.2amps.

Newer power supplies have active power factor correction to reduce this effect so the inaccuracy will vary depending on age of power supply.

Still below the rated max of your PSU.

 

If it makes any difference the system is hooked up through a UPS and the power meter constantly reads a power factor of 100.

 

If it's a switch mode PSU (eg, an inductive load), the the true RMS (apparent power) will be higher than the actual current/actual power due to the reactive current which is drawn by the device, then momentarily returned to the grid out of phase with the voltage.

Too much reactive power is bad for the network, so most PSU's have power factor correction which reduces the gap between apparent power and actual power. A good energy meter should be able to differentiate between reactive power and apparent power.

 

I had a PSU at work where due to dust/heat/age the gap between the bottom of the PCB & solder nibs and the bottom of the PSU case was bridged (either by heat of the board flexing and dust bridging the gap) and it let go with a massive bang and smoke.

 

so its settled your ferrite choked buy a corsair and never have this issue again.
i know ive spilt beer in one

 

oh wow! seeing this just reminded me to clean out my comp of dustt and dead bugs that get caught in the fans

 

That's not a dust problem, this is -

 

That's exactly what mine looks like.....