Understanding PSU protections

[Moonzy]

I'm using this page as reference: https://www.tomshardware.com/reviews/power-supplies-101,4193-21.html (thanks @Spotty)

 

Quote

(OCP) Over-Current Protection

Over-current protection (OCP) is a popular protection found in all PSUs with multiple +12V rails, and in most cases, it also protects the minor rails. OCP kicks in when the current in the rails surpasses a certain limit. The ATX 2.2 specification states that if the load at each tested output rail reaches or exceeds 240VA, then OCP should interfere (paragraph 3.4.4). However, the ATX 2.31 specification omits this limit. In order to bypass it, some manufacturers implemented many virtual +12V rails, with each rail rated at 240VA. However, in most cases, the OCP trigger point was set much higher in order to withstand peak currents that some system components (like graphics cards) could draw.

To implement OCP in a PSU, two things are necessary: shunt resistors and a supervisor IC that supports OCP. The shunt resistors are low-resistance, high-precision resistors used to measure the current at the outputs of a PSU, utilizing the voltage drops those currents create across the resistors. By measuring the number of shunts in a PSU in the area where +12V wires are soldered, we are usually able to find the real number of +12V virtual rails. In some cases, when the manufacturer has initially built the PSU as a multiple +12V rail unit and afterward converted it to a single +12V rail unit, the shunt resistors are simply shorted together.

is 240VA = 240W? since P = V I, the unit is VA indeed, or this includes the impedance of a load while W is only resistive load?

meaning 12V rail can only pull ~20A? but it does say it can go higher.

 

Quote

SCP (Short Circuit Protection)
Short circuit protection (SCP) constantly monitors the output rails, and if it finds an impedance of less than 0.1Ω, it immediately shuts down the power supply. In other words, if somehow the output rails are short circuited, then this protection kicks in and shuts down the PSU to prevent damage or fire. According to the ATX 2.31 spec, each +12V rail should have a separate short circuit. This protection is present in almost all contemporary PSUs (at least the branded ones).

0.1ohm at 12v rail = 120A, wouldn't this trip OCP way before that? how does this even work?

technically it should also trip OPP (over power protection) since the power spike is also huge (1440W)

how about 1600W PSU? do they have higher SCP threshold? or there's no single 12v rail high wattage PSU?

 

@Energycorei think we discussed this in DM, so you might be interested

[SpiderMan]

I always thought 240VA means 240Volt-Amphere.

 

I'm really not so sure about the rest though. 

[Sakkura]

9 minutes ago, Moonzy said:

is 240VA = 240W? since P = V I, the unit is VA indeed, or this includes the impedance of a load while W is only resistive load?

meaning 12V rail can only pull ~20A? but it does say it can go higher.

This protection applies to the PSU output, which is DC. As such, impedance isn't really involved.

 

12V rails were originally not supposed to allow more than 20A (240W assuming nominal voltage) per the ATX12V 2.0 specification, but as power demands escalated everyone just started ignoring that part of the specification. It was dropped in later versions.

[Moonzy]

2 minutes ago, Sakkura said:

As such, impedance isn't really involved.

that's what i thought, but then why measure in VA and not W?

[jonnyGURU]

46 minutes ago, Moonzy said:

that's what i thought, but then why measure in VA and not W?

It's never made sense.  In DC circuits, VA is W.

 

But Intel has always used VA in the spec.

 

[jonnyGURU]

2 hours ago, Moonzy said:

0.1ohm at 12v rail = 120A, wouldn't this trip OCP way before that? how does this even work?

technically it should also trip OPP (over power protection) since the power spike is also huge (1440W)

how about 1600W PSU? do they have higher SCP threshold? or there's no single 12v rail high wattage PSU?

You have to take into consideration the resistance, not just of the short itself, but what is added by the circuit itself.

 

When I see SCP working is when there is a direct short between an output and ground.  When I see SCP NOT work, but OCP work instead (if it's there), is when the short is on the far end of a connection, like on the PCB of a failed hard drive or a failed graphics card, etc. where the short is on the device itself.  The device itself typically doesn't have SCP (it might on it's own D2D circuit, but what if the short is before that even?), so the PSU keeps feeding power to the short because it doesn't see it as a short, it sees it as a load.

 

That's how you can know if your PC catching fire was due to the PSU or not.  Are the melted connector on the PSU or are they on the device?  If they're on the device side, that means the device failed and the PSU kept feeding power to it until the connector melted, insulation melted off the wires and this forced a short between a hot and ground (unless the PSU has multiple +12V rails with lower OCP, at which case things would never get so far).

 

And OPP DEFINITELY does not replace OCP or SCP.  Since OPP is on the primary side, you have to take into account the energy stored in all of the components of the PSU before OPP can trip.  This is why some PSUs work better with the Ampere cards than others.  If you have a lot of capacitance storing power, then the graphics card can feed off of that for the transient loads, keeping OCP and OPP completely oblivious to the load. But if the caps drain too quickly, that spike in power hits the OCP or OPP (depending on which see it first).

 

It's very similar to how a stiffener caps work in car audio.  Except the headlights dimming is your OCP or OPP tripping: https://www.crutchfield.com/S-bhH42VRD5iX/learn/car-what-is-a-capacitor-faq.html

 

[Moonzy]

4 minutes ago, jonnyGURU said:

When I see SCP work is when there is a direct short between an output and ground. 

Hmm, how does SCP work?

OCP uses shunt resistor to read the amperage (deduced from voltage drop across it), how about SCP?

 

Will SCP trigger before huge current is passed through if it's a short to ground? Therefore not triggering OCP?

What if the short to ground is happening while operating? Is it OCP or SCP that triggers?

 

9 minutes ago, jonnyGURU said:

You have to take into consideration the resistance, not just of the short itself, but what is added by the circuit itself.

The SCP can only measure the full circuit resistance right? Since it can only refer to what is between the rail and ground.

[jonnyGURU]

11 minutes ago, Moonzy said:

Hmm, how does SCP work?

OCP uses shunt resistor to read the amperage (deduced from voltage drop across it), how about SCP?

It's not simple to explain.  Especially with multiple outputs where +12V to +5VSB or -12V is also considered a "short".

 

I'm not sure if/when Aris will put it into his PSU testing/reviewing, but he could tell you that MOST PSUs have SCP that doesn't function properly 100% of the time.  Fortunately, most of the time the PSU just self destructs (no damage to PC components), though some damage is possible in some corner cases.

 

14 minutes ago, Moonzy said:

Will SCP trigger before huge current is passed through if it's a short to ground? Therefore not triggering OCP?

Usually, yes.

 

18 minutes ago, Moonzy said:

What if the short to ground is happening while operating? Is it OCP or SCP that triggers?

I don't understand why you mean by this.  If SCP triggers, the PSU latches off and stays off until the short is removed.

 

18 minutes ago, Moonzy said:

The SCP can only measure the full circuit resistance right? Since it can only refer to what is between the rail and ground.

Correct.

 

[Moonzy]

9 minutes ago, jonnyGURU said:

I don't understand why you mean by this.  If SCP triggers, the PSU latches off and stays off until the short is removed.

say, i have a desktop that is running furmark

 

i then proceed to short the 12v to GND (stupidly), which one will trigger?

im thinking OCP, but idk how SCP functions so...

im guessing the answer is "depends" since as you mentioned, many PSU have SCP that doesnt function properly

 

10 minutes ago, jonnyGURU said:

It's not simple to explain.

is it possible to point me in the right direction to read?

maybe a certain circuitry that i can study about

 

thanks for your insight, by the way, it's much appreciated

im just a curious person

[Spotty]

40 minutes ago, jonnyGURU said:

When I see SCP working is when there is a direct short between an output and ground.  When I see SCP NOT work, but OCP work instead (if it's there), is when the short is on the far end of a connection, like on the PCB of a failed hard drive or a failed graphics card, etc. where the short is on the device itself. 

The device itself typically doesn't have SCP (it might on it's own D2D circuit, but what if the short is before that even?), so the PSU keeps feeding power to the short because it doesn't see it as a short, it sees it as a load.

 

That's how you can know if your PC catching fire was due to the PSU or not.  Are the melted connector on the PSU or are they on the device?  If they're on the device side, that means the device failed and the PSU kept feeding power to it until the connector melted, insulation melted off the wires and this forced a short between a hot and ground (unless the PSU has multiple +12V rails with lower OCP, at which case things would never get so far).

What sparked these questions from @Moonzy was the recent video from gamers Nexus testing the nzxt H1 case catching fire.

 

They concluded SCP didn't trip because there was enough resistance from the motherboard tray and chassis between the short and the power supply. Also supported by when they connected the faulty PCIe riser directly to the enclosure of the PSU grounding it, the SCP did kick in to shut it down.

 

That would maybe be a good example of when SCP doesn't trip and you would want OCP. I have no idea what the current draw was so no idea if having a multi rail PSU would have made a difference in this scenario.

The PSU they used was the NZXT S650, rebranded Seasonic focus SGX 650. 

I do wonder if a PSU with a lower configured OCP (ie multi rail PSU) would have tripped before it ignited? 

[jonnyGURU]

5 minutes ago, Moonzy said:

say, i have a desktop that is running furmark

 

i then proceed to short the 12v to GND (stupidly), which one will trigger?

im thinking OCP, but idk how SCP functions so...

 

Like I said in my first post, depends on where the short is.  If you just take a +12V wire and a ground and short them against each other, that's SCP tripping the PSU.

 

If you damage a PCB and the short is there, and the PSU has OCP, then OCP would shut the PC off before too much power is delivered to the short.

 

Here's a prime example of SCP not seeing the short, and the PSU not having OCP:  

 

9 minutes ago, Moonzy said:

is it possible to point me in the right direction to read?

maybe a certain circuitry that i can study about

Here's a simple project:  https://circuitdigest.com/electronic-circuits/short-circuit-protection-circuit-diagram

 

Keep in mind this is a SINGLE output.  When you have multiple outputs, you have to do this in duplicity and throw some Zener diodes into the mix.  

[akio123008]

4 hours ago, Moonzy said:

i then proceed to short the 12v to GND (stupidly), which one will trigger?

The thing is, it could be either one. Depending on just how bad the short is and how the circuitry works (not everyone uses the same design) it may be either one of them that trips first.

 

For instance, a real bad short may trigger SCP, as SCP typically has a faster response time than OCP, but if the short is imperfect, there may not be sufficient current to trigger SCP, and so in that case OCP would trip first.

 

In general, not just in computer power supplies, SCP systems tend to be extremely fast, but only respond to insane currents (shorts), whereas OCP systems detect smaller faults, but have a slightly slower response time (sometimes even deliberately slow to allow for startup of machinery)

 

5 hours ago, jonnyGURU said:

It's never made sense.  In DC circuits, VA is W.

Well there's some ripple on that DC so technically there is reactive power, but I suspect it's negligible.