Understanding Over Current Protection

[Merman]

It was my understanding that since OCP is required by ATX12V that it was used on the real 12V rail in addition to the individual OCP on multiple virtual rail units.

Using the ASUS U-75HA 750 W Power Supply Review
with its 4 virtual rails with a 24 amp limit and a 56 amp total limit on the actual single 12V rail:

I could not understand why the OCP 56 amp limit could be higher if more current was loaded to the first two rails???

It was my understanding that if there was OCP that there was one total limit on the real rail no matter how balanced or unbalance the loads on the virtual rails. ???

How does a real rail OCP work on a unit that has multiple virtual rails???

[Gabriel Torres]

Hello Merman,

I understand your confusion and I hope to clarify.

OCP, over current protection, is a current sensor installed on each power supply output. The presence of these sensors is what defines a power supply being "single rail" (only one or no sensor at all for all +12 V outputs) or "multiple rail" (one sensor per each group of +12 V wires).

Most manufacturers do not announce at what level the OCP is triggered ("trigger point"). This trigger point has nothing to do with what is printed on the power supply label. I.e. what is printed on the label IS NOT the OCP trigger point.

During our reviews we always try to check the trigger point from the OCP circuit. I like to see OCP configured at a level close to what is printed on the label, but this is me talking, ok?

Let's say, as you mention, that a power supply has four +12 V rails each one with OCP configured at 24 A. Does this mean that we can pull 96 A (24 A x 4) from the +12 V rails? Absolutely not. This only means that if you try to pull 24 A from any +12 V rail the power supply will shut down, and only that.

As you know, the manufacturer prints a maximum current for each individual +12 V rail and then a maximum combined current for the +12 V rails. OCP only monitors individual rails (of course on single-rail design OCP will monitor "all" rails, as the unit has only one).

If you try to pull more power than the PSU can deliver, one of its other protections (OPP, OVP or UVP) will probably kick in and shut down the power supply, assuming that it implements one of these protections, of course. If not, the unit will burn/explode. The unit can also burn/explode if these protections are configured to trigger at a value that is too high (or too low, in the case of UVP), e.g. OPP configured to shut down the PSU at 600 W and the PSU can only "resist" up to 550 W without burning. If the values of these protections are too "loose" the unit will obviously work as if no protection existed.

I hope I have helped you to understand how this protection works.

Cheers,
Gabriel.

[Merman]

Thank you for such a concise answer. It was so good you should save it for any up date the protection section of Everything You Need to Know About Power Supplies. http://www.hardwaresecrets.com/article/181

But there is still one point you haven't explained.

Originally Posted by Gabriel Torres

As you know, the manufacturer prints a maximum current for each individual +12 V rail and then a maximum combined current for the +12 V rails. OCP only monitors individual rails (of course on single-rail design OCP will monitor "all" rails, as the unit has only one).

How could more than the total 56 amp OCP limit be exceeded by being able to put a higher load than 33 amps of the loadtester's first rail, which was attached to two virtual rails of the power supply unit???

[Olle P]

Originally Posted by Gabriel Torres

This trigger point has nothing to do with what is printed on the power supply label.

The nominal (printed) maximum current should be on or below the trigger point. (As was not the case for one of the PSUs tested not too long ago...)
Too bad if the trigger point is half of the nominal value...

Originally Posted by Gabriel Torres

I like to see OCP configured at a level close to what is printed on the label, but this is me talking, ok?

As do I. But as noted above, better have the trigger close above than close below the nominal value.

Originally Posted by Merman

But there is still one point you haven't explained.
How could more than the total 56 amp OCP limit be exceeded by being able to put a higher load than 33 amps of the loadtester's first rail, ...

In that particular test the loadtester was attached to two rails of the PSU, each of which could supply ~23A before the OCP kicked in. Therefore it should be possible to pull some 46A combined output from those lines before triggering any OCP (and instead possibly trigger some other protection at a level below 46A), but the loadtester can't pull more than 33A on each 12V input line.
One solution would be to combine the PSU's 12V2 and 12V4 to the tester's second input, thus being able to feed up to 66A 12V through the tester without triggering any OCP.

Cheers
Olle

[Merman]

Originally Posted by Olle P

In that particular test the loadtester was attached to two rails of the PSU, each of which could supply ~23A before the OCP kicked in. Therefore it should be possible to pull some 46A combined output from those lines before triggering any OCP (and instead possibly trigger some other protection at a level below 46A), but the loadtester can't pull more than 33A on each 12V input line.

The idea that 46 amps could be loaded on only two lines was never questioned since the OCP was set at 24 amps for each rail. The idea that some other protection would kick in at less than 46 amps when the unit should put out 50 or more amps doesn't seem plausible. If the unit could put out 56 amps which seems to be the case, 46 amps on two virtual rails should not have been a problem.

Originally Posted by Olle P

One solution would be to combine the PSU's 12V2 and 12V4 to the tester's second input, thus being able to feed up to 66A 12V through the tester without triggering any OCP.

I will quote what was said from the review again.

Overload Tests
When we tried to pull more than 23 A from the +12V2 rail from this power supply the over current protection (OCP) entered in action shutting down the power supply.

Since our load tester has only two inputs (and each one limited to 33 A), we were limited by our instrument, since we could “only” pull up to 33 A from the +12V1 input (which was connected to the +12V1 and +12V3 rails) and 23 A from the +12V2 input (more than that OCP would kick in, as explained). Thus it is possible that ASUS U-75HA can deliver even more than we were able to pull.

This power supply has four rails, distributed like this:
+12V1 (solid yellow wire): Main motherboard, SATA and peripheral power connectors.
+12V2 (yellow wire with black stripe): ATX12V/EPS12V connectors.
+12V3 (solid yellow wire): One of the video card power cables.
+12V4 (solid yellow wire): The other video card power cable.

The way I understood the second paragraph of the Overload Tests was the 12V4 was inconsequential or tried since only 23 more amps kicked in the total OCP but you are right Ollie it could just be the 12V2 virtual rail OCP that shut the unit down.


So if OCP was being tested why wasn't 12V4 connected to the second loadtester input???

This question prompted me to read about the loadtester a SunMoon SM268.

Here it is important to explain something before people get confused on our reviews. This equipment has two separated +12 V inputs, labeled +12V1 and +12V2, which does not necessarily relate with the power supply multiple rails (+12V1, +12V2, +12V3, etc). All plugs that provide +12 V (main motherboard power connector, peripheral power connectors, video card power connector and EPS12V/ATX12V connector) are connected to the machine +12V1 input. The second input is connected only to a second EPS12V/ATX12V connector that is available.
http://www.hardwaresecrets.com/article/522/2

Now I understand that 12V4, which is another video card power cable, would have to be attached to the loadtester's first input and this input is limited to 33 amps. Connecting 12V4 was inconsequential as 12V1 and 12V3 already maxed out the 33 amps of the first input and thus why if the loadtester had more capacity, more amps could have been pulled from this unit.

Sorry for the confusion and thanks for the help.

[Gabriel Torres]

Originally Posted by Merman

Thank you for such a concise answer. It was so good you should save it for any up date the protection section of Everything You Need to Know About Power Supplies. http://www.hardwaresecrets.com/article/181

It will be. I am working on an update for this tutorial, but it is becoming so long that I may have to break it down into several parts or subjects. Maybe I will write a tutorial only on the protections and how they are physically implemented inside the power supply.

Originally Posted by Merman

How could more than the total 56 amp OCP limit be exceeded by being able to put a higher load than 33 amps of the loadtester's first rail, which was attached to two virtual rails of the power supply unit???

I think you finally got this from the above post, it is because we were mixing rails that you got confused. I will try to draw it:

Power Supply Load Tester
+12V1 -------------------->+12V1 (33 A Limit)
+12V3 -------------------->+12V1
+12V2--------------------->+12V2 (33 A Limit)
+12V4-- Not Connected

So, at +12V2 is simple to view: we can pull up to 33 A on our load tester, but OCP kicked in when we tried to pull more than 23 A.

At +12V1 (load tester) we are limited to 33 A, but this this we connected this to +12V1 and +12V3 at the same time. Pulling 33 A from this input was making +12V1 to work at 16.5 A and +12V3 at 16.5 A, thus below the 24 A OCP limit.

Cheers,
Gabriel.

[Merman]

Originally Posted by Gabriel Torres

As you know, the manufacturer prints a maximum current for each individual +12 V rail and then a maximum combined current for the +12 V rails. OCP only monitors individual rails (of course on single-rail design OCP will monitor "all" rails, as the unit has only one).

If you try to pull more power than the PSU can deliver, one of its other protections (OPP, OVP or UVP) will probably kick in and shut down the power supply, assuming that it implements one of these protections, of course. If not, the unit will burn/explode. The unit can also burn/explode if these protections are configured to trigger at a value that is too high (or too low, in the case of UVP), e.g. OPP configured to shut down the PSU at 600 W and the PSU can only "resist" up to 550 W without burning. If the values of these protections are too "loose" the unit will obviously work as if no protection existed.

Thanks again for all your instruction. I have re-read the quote and my understanding may not be correct.

A unit with only one real 12V rail that has virtual rails is confusing. This type of unit isn't usually called a single rail design. So does this type of power supply have OCP only on the virtual rails and total amps are limited by another protection???

I thought this type unit also had total OCP for the one real rail but am not sure now. Or could it depend on how the OCP is implemented???

[Olle P]

Originally Posted by Merman

A unit with only one real 12V rail that has virtual rails is confusing. This type of unit isn't usually called a single rail design. So does this type of power supply have OCP only on the virtual rails and total amps are limited by another protection?

You've got it just about correct.
It's just that today a "rail" is defined simply as "all outputs covered by the same OCP".

There are very few modern PSUs that have multiple 12V transformers, as was more common before using "switched" designs.

Cheers
Olle