[Mad Vlad]

What transient property causes damage to other components in a PC when the Power Supply castastrophically fails? (i.e. a component failure, like a capacitor bursting, rather than a trip on the protection devices inside the PSU) I assume it's some kind of surge, but I don't know about the electronics enough to pinpoint what's caused the problem.

Any thoughts?

[DeeDub]

I would say that you have pretty much hit the nail on the head. A surge could cause rapid overheating and short circuits. If a capacitor fails there is a chance that the voltage supplied to the componets could jump. Other than that you will need someone who does EEE for more info.

[elsie87]

I might be wrong, but if you get a surge, high voltage and high resistance, you will get overheating which can cause the circuitry to bend and damage the components. That's how the old xbox 360's used to get the red ring of death, the motherboard would bend twisting all the chips out of shape.

[giran]

As power supply capacity increased, the ATX power supply standard was amended to include:

3.2.4. Power Limit / Hazardous Energy Levels

Under normal or overload conditions, no output shall continuously provide more than 240 VA under any conditions of load including output short circuit, per the requirement of UL 1950/ CSA 950/ EN 60950/ IEC 950.
—ATX12V Power Supply Design Guide, version 2.2

This is a safety limit on the amount of power that may pass, in case of a fault, through any one wire. That much power can significantly overheat a wire, and more would be likely to melt the insulation and possibly start a fire.

Taken from http://en.wikipedia.org/wiki/Power_supply_rail

My guess is there was an overload/surge which caused this.

[imhiya]

The pinpoint problem is it has no method of prevention, a cut off point. But its normally caused by high amount of components attatched which gives the circuit a bigger resistance, which draws more voltage and current. Which causes over heating but also makes for unstable rails on 12v, 5v and 3.3v

[Chaz_Hack_Rabbit]

From what I've read in my research into buying a new PSU, when one catastrophically fails the voltage on one (or more) of the rails will go above it's intended level. PC components are designed to handle a pretty steady voltage so aren't able to cope with the spike in the voltage on that particular rail e.g. chips will short if the voltage accross them is too high or if they don't short will most likely overheat. Basically if this happens on the 12V rail with hard drives on it, the control chip etc will get shorted and drive dies. On the 12V rail for motherboard, chips overheat/short and take down certain sub-systems like bios, network, northbridge etc.

[imhiya]

(Original post by Chaz_Hack_Rabbit)
From what I've read in my research into buying a new PSU, when one catastrophically fails the voltage on one (or more) of the rails will go above it's intended level. PC components are designed to handle a pretty steady voltage so aren't able to cope with the spike in the voltage on that particular rail e.g. chips will short if the voltage accross them is too high or if they don't short will most likely overheat. Basically if this happens on the 12V rail with hard drives on it, the control chip etc will get shorted and drive dies. On the 12V rail for motherboard, chips overheat/short and take down certain sub-systems like bios, network, northbridge etc.

(Original post by imhiya)
The pinpoint problem is it has no method of prevention, a cut off point. But its normally caused by high amount of components attatched which gives the circuit a bigger resistance, which draws more voltage and current. Which causes over heating but also makes for unstable rails on 12v, 5v and 3.3v

:P yup.

[Mad Vlad]

Thanks for your thoughts on this, guys Very useful!

[westom]

What transient property causes damage to other components in a PC when the Power Supply castastrophically fails? (i.e. a component failure, like a capacitor bursting, rather than a trip on the protection devices inside the PSU)

The most common reason for power supply failure (even years later) is manufacturing defects. Many will blame the popular myth or what they speculate. Most never do an autopsy.

For example, one power supply that was always powered for months failed after power off. Many assume that power cycling is destructive because that myth is popular. Found a pullup resistor that is always powered with a trivial current. But its only function is to provide the current during power on - bootstrap. While the supply was on for months, constant tiny current caused the resistor to fail - probably months earlier. A manufacturing defect. But those who know only using observation immediately blamed power cycling. Or heat. Or a surge. Blamed only what they understand rather than first collect technical facts - such as which semiconductor or passive component failed.

Also traceable to knowledge from observation or assumption is this myth: power supply failure can cause motherboard or other failure. Completely false as long as the supply contains what all supplies were required to contain even with the original IBM PC. Specs and international standards are quite blunt about this. Power supply failure must never cause damage to other components.

However, electrical ignorance even will purchase supplies only on price and watts. He does not even know that he - not the power supply manufacturer - is responsible for that supply containing required functions. A problem even found with Certified Computer techs because no electrical knowledge is required to pass the test. Because he cannot read specification numbers, a computer assembler selects only what he understands - price and watts. Therefore that supply can be missing essential functions that existed even in 1960 supplies. In this case, failure directly traceable to insufficient knowledge of a computer assembler.

All power supplies (like all appliances) contain significant surge protection. Anything on its power cord that might protect it is already inside the supply. But again, technical ignorance is widespread. Some numbers to demonstrate.

Many 'know' a UPS provided 'cleaner' power. An output from a typical computer grade 120 volt UPS is two 200 volt square waves with a spike of up to 270 volts between those square waves. Double those numbers for a 230 volt UPS. An output harmful to small electric motors and power strip protectors. But perfectly ideal to computer power supplies. Why? Because computer supplies are so robust - make 'dirty' UPS electricity irrelevant. 1) Many ‘assume’ the UPS outputs ‘clean’ power (when even the specs make no such claim). 2) Many do not even know a that all computers must be that robust. Both are reasons why so many have knowledge only from urban myths.

If using a power strip protector, a surge current now has more paths to find earth destructively via the motherboard. A power strip can even bypass protection inside the supply; earth a surge destructively through the motherboard. Few will know this due to knowledge without first learning facts. We even traced damage to a network of powered off computers because two adjacent power strips earthed a surge destructively through the adjacent motherboards and other network computers. A power strip protector bypassed power supply protection. But most never did the autopsy to first learn what is damaged.

All electronics contain significant surge protection. Even in 1970, a 230 electronic appliance was required to withstand transients well above 1000 volts without damage.
Today’s electronics are even more robust. A rare surge (maybe once every seven years) can overwhelm that protection. So that protection inside all appliances is not overwhelmed, earth one 'whole house' protector so that the rare surge does not even enter the building - does not seek earth ground destructively via any appliance.

Moving on to other reasons for failure: static electricity. A static electric discharge to removed components can cause overstress. That component fails weeks or months later. Those who use observation will not blame the irresponsible human who caused that failure. Again, lack of basic electrical knowledge.

Removed components are easiest to damage with static electricity. Once those components become part of a system, its protection increases dramatically. For example, interface semiconductors now must withstand 2000 and 15,000 volts without damage. Protection that exists when the semiconductor is part of a complete system. Once overstressed, that component may fail a month later. Then the naive usually blame ‘assumed’ surges or overheating - only what they understand.

Many see temperature limits on a semiconductor or component. Then assume a higher temperature causes hardware damage. Nonsense. Semiconductors are repeated cycled many hundreds of degrees during manufacturing. Using popular myths, temperatures up to 1000 degrees means complete semiconductor failure. Means a semiconductor fails before it gets out of the factory.

Computers that exceed 40 degrees C suffer timing and threshold changes. Changes that return to normal when the system is cooled. Changes that cause the computer to crash – and no hardware damage. The naïve see the crash – then blame heat as hardware destructive. All computers must work at a perfectly ideal room temperature of 40 degrees C. Failure at this temperature identifies a completely defective internal component (that typically gets worse and fails months later). Many instead add more fans - cure the symptom - as if those extremely low temperatures are hardware defective. Observation and speculation without technical knowledge results in a classic junk science conclusion - heat is hardware destructive.

Remove the heatsink from any Intel processor – then run it. CPU is not damaged. Resulting timing changes will cause the CPU to not function. And when cooled, the CPU is 100% functional. Heat at those trivial temperatures is only hardware destructive in myths.

Yes, excessive temperatures can damage semiconductors. But not at room temperatures. To be damaged, the semiconductor must be so hot as to leave skin when touched. Lower temperatures simply cause signaling errors and computer crashes - no hardware damage.

Heat is a diagnostic to find defective computers long before that defect causes computer failure. Operate any computer with one chassis fan in a 40 degree C room. If the computer crashes, then find the human mistake or manufacturing defect that caused that crash (and may cause complete computer failure later).

Some background on why hardware failure really happens, how to avoid some hardware destructive events, and how to find existing defects before those defects cause failures.

[mfaxford]

The usual component that causes voltage spikes in a circuit is something with inductance where the magnetic field collapses suddenly. Such components can include transformers, motors, coils of wire used for filtering. Usually such parts would have other items around to reduce the impact of the potential spikes they cause (which can be into many kV). If the components providing that protection it's possible those spikes might make it into other parts of the PC and such spikes could cause damage to semiconductors (chips, transistors) etc.

[westom]

(Original post by mfaxford)
The usual component that causes voltage spikes in a circuit is something with inductance where the magnetic field collapses suddenly.

Then why do electronics designs intentionally make that collapse as fast as possible? For example, how does a power supply work? After converting AC mains to DC voltages well above 300 volts, the power supply converts that electricity to radio frequencies - fastest spikes - so that the magnetic field collapses at ten and one hundred thousand times a second. The faster that collapse; the better that design. Causes no damage. And makes a power supply more resilient.

Any damage created by collapsing magnetic fields is directly traceable to a design that was completely defective before entering production. Any spike from that collapsing field is predictable - cannot happen fast enough. If that collapsing field could be faster, the designer would do so. Routine filtering makes spikes from the fastest collapsing fields irrelevant.

Common source of failure is manufacturing defects. For example, the infamous electrolytic capacitor failures because so many capacitor manufacturers used a counterfeit and defective material. Bulging capacitor failures - another manufacturing defect - occurred years later. Failures that many blame only on what they understand - heat, surges, or speculated collapsing magnetic fields. Blame without first identifying then replacing the defective component. An overwhelming majority will blame because they cannot even identify that defective component.