One reason not to buy cheapie PSUs

[tortillasoup]

https://www.youtube.com/watch?v=Vb9zjxqklJA



the question I have is, what exactly is the failure mechanism that is causing them to spark like that under load? What's typically in the design of a cheap power supply to cause it to fail like that?

 

[Raduque]

Ultra cheap components, possibly even bad components, assembled without a QC and highly overrated.

http://www.jonnyguru.com/

Look at their cheap PSU reviews, especially the Death of A Gutless Wonder series of reviews.

 

[MrTeal]

https://www.youtube.com/watch?v=Vb9zjxqklJA



the question I have is, what exactly is the failure mechanism that is causing them to spark like that under load? What's typically in the design of a cheap power supply to cause it to fail like that?

There likely isn't a single exact failure mechanism that produces the sparking, you'd have to open up each one and see what blew up. In pretty much all cases though, a component failed in such a way that there was a large potential between to conductors, and not enough insulation. Caps are a pretty common failure point, as are underspeced transistors and diodes.

 

[Hugo Drax]

Why I stick with Seasonic. They last forever

 

[Charlie98]

Jeepers!

 

[MongGrel]

PSU melt downs are fun to watch, one of the things I cringe about whenever I see someone trying to shove a cheap one into a new build here.

Sometimes they can damage other things when they go down, of course, so it's not just the PSU you need to worry about.

 

[Insert_Nickname]

PSU melt downs are fun to watch, one of the things I cringe about whenever I see someone trying to shove a cheap one into a new build here.

Sometimes they can damage other things when they go down, of course, so it's not just the PSU you need to worry about.

IMHO, the absolutely worst possible place to save is the PSU. For that exact reason.

 

[piasabird]

I am also reluctant to use power supplies from an old computer in a new one. Just like the CPU fins the cooling fins in a power supply also collect dust and cause it to run hot. The higher the load the faster your Power Supply fans turn the more dust they suck into your power supply.

 

[sm625]

Components heat up. Metal expands. When this expansion causes a local electrical resistance to increase, a runaway temperature/expansion condition occurs very rapidly. Small fractures between where the pad meets the trace or where the component leg (or ball) meets the pad are probably the most likely causes. When the unit heats up, the metal expands differently on one side of the fracture compared to the other, causing the fracture to grow. This doesnt present a problem until the fracture crosses the threshold of how much current can be passed through the remaining unfractured metal. Like I said this last part usually happens very quickly.

Thermal expansion can also cause heatsinks to warp off their components. Sometimes this expansion is not linear, it can be more like two tectonic plates building up pressure and then popping apart in an earthquake-like manner resulting in sudden and rapid failure as soon as the heatsink separates from its component.

Dielectric failure inside of capacitors also can occur due to temperature cycling. This is usually due to microfractures that expand ever so slightly during each temperature cycle. Ceramic capacitors seem to be more prone to this type of failure. Electrolytic capacitors seem to be more prone to buildup of solid particles in the electrolye, which eventually form a bridge between the two plates. Think of a cave with stalactites that slowly grow to reach the floor. When the stalactite reaches the floor, you get an instantaneous overcurrent which heats the plates and warps them, which causes yet more connections between the plates and results in component failure, often catastrophic. Even if the "stalactite" is nonconductive, it can still push the plates apart causing it to in turn push against the plate behind it since the plates are usually two long strips of metal wrapped together and stuffed into a can.

Microfractures in inductors and even resistors can also cause sudden catastrophic failure in much the same way. If a resistor is regulating the voltage or current of a power supply, a sudden increase in its value could cause the output to spike which then blows some other component down the line.

A well designed power supply will limit the temperature fluctuations to only the components that are specifically designed to handle them. It is about more than just the ratings on compoenents. Even the crappiest supplies probably use components that are rated with a reasonable spec. But just because a component is rated up to 85C doesnt mean that you should subject it to that unless absolutely necessary.

 

[Jeff7]

Ahh that takes me back.
I killed a cheap "400W" power supply: I overclocked a good old 1GHz Thunderbird CPU to 1.4GHz, then ran a CPU stress test on the "maximum power/heat" setting.
A power resistor in the power supply promptly blew apart.
Antec vs a cheap-crap supply.
400W cheap and 350W Antec.
"Duro" was the cheapie's brand/manufacturer, I believe. I don't know....anytime a company name tries to use a heavy-handed approach and beats you over the head with "OMG you can't imagine how good our quality is!" I get nervous.

Smaller input caps: 120-240V is the input rating, so those are in series, giving 165uF on the 400W and 410uF on the 350W. I believe a power supply needs to run at least 16.6 milliseconds at full load on capacitor power alone: If there's a power outage, a UPS needs to see a missed cycle to know that the voltage is gone, and that it needs to immediately switch to battery.
So there's smaller caps, less input filtering, less output filtering, smaller transformers, smaller heatsinks (hey, maybe they're using more efficient FETs and don't need better heatsinks:awe, shady vendors for components.....cost cutting everywhere.

Components heat up. Metal expands. When this expansion causes a local electrical resistance to increase, a runaway temperature/expansion condition occurs very rapidly. Small fractures between where the pad meets the trace or where the component leg (or ball) meets the pad are probably the most likely causes. When the unit heats up, the metal expands differently on one side of the fracture compared to the other, causing the fracture to grow. This doesnt present a problem until the fracture crosses the threshold of how much current can be passed through the remaining unfractured metal. Like I said this last part usually happens very quickly.

Good old thermal runaway: A good example of why a system with positive feedback can be more of a problem than it may sound at first.

...
A well designed power supply will limit the temperature fluctuations to only the components that are specifically designed to handle them. It is about more than just the ratings on compoenents. Even the crappiest supplies probably use components that are rated with a reasonable spec. But just because a component is rated up to 85C doesnt mean that you should subject it to that unless absolutely necessary.

My reputation at work has become something of a capacitor snob. I spec capacitors for a minimum of 5000hrs @ 105C, with as much ripple current capacity as I can squeeze into the budget. I've replaced more bulging capacitors than I care to have done, and I have no desire to use crappy capacitors in anything this company ships out.

Yup, you can get a cap that's rated for high temperatures, and it may proudly say "105C" on the sleeve.
I quickly checked the capacitor listing at Digikey: Anything from 1,000hrs @ 85C up to 26,000 hours, just in ones spec'd at 85C. They've also got some that are spec'd at 1,500hrs at 150C.
Capacitors don't like heat. Every degree you can shave off of the ambient temperature will help increase life expectancy. That's the other reason for low-ESR caps: Resistance causes heating, so the cap will heat itself up, which will likely make a bad situation even worse.

I took apart a dying laptop power supply recently: Some cheap crap thing where the manufacturer didn't even want to put their company name or logo on the thing. Two things were hotter than anything else: The transformer and the main input capacitor. Not any of the diodes, not any transistor or FETs....you know, the thing that are usually warmest. Nope, one of them was the input capacitor. No attention paid to ripple current, capacitance requirement, only cost.

 

[MongGrel]

Ahh that takes me back.
I killed a cheap "400W" power supply: I overclocked a good old 1GHz Thunderbird CPU to 1.4GHz, then ran a CPU stress test on the "maximum power/heat" setting.
A power resistor in the power supply promptly blew apart.
Antec vs a cheap-crap supply.
400W cheap and 350W Antec.
"Duro" was the cheapie's brand/manufacturer, I believe. I don't know....anytime a company name tries to use a heavy-handed approach and beats you over the head with "OMG you can't imagine how good our quality is!" I get nervous.

Ah, the old days

I killed probably several cheap PSU's and probably at least one Thunderbird board in the past, just even from lightning strikes in the Tampa Bay area here with low end PSU's once upon a time.

Probably about 5 PSU's, over a couple AMD Thunderbird builds with Gigabyte boards long ago, probably promoted the bleeding green out of the caps that killed one board off at the time.

That was a couple decades ago I guess.

The first "real" PSU I bought was an Antec after the cheapo failures, I tend to be a bit of a fanboi along those lines after that, but have used a few others.