Power Supplies 101: A comprehensive guide

[Doc Overclock]

Yes Jon did a wonderful job on that article and for those who want to split hairs i say let-Em,,, The article can stand on its own merit....

I am proud of Jon for this and thank him immensly for his efforts///

Maybe he can write a how to guide on how to accurately test a PSU...

Hey Johnny?? You Game??

[jonnyGURU]

Dunno...

But thanks for the good words, Doc.

But I do think we need to fix that efficiency equation. People are right. I typo'd. It should be (DC Output)/(AC Input) = Efficiency.

[Doc Overclock]

Please send an email to Evan Jon-----------

[jonnyGURU]

Please send an email to Evan Jon-----------

Evan must be on vacation. I emailed him two days ago and haven't heard back.

[Twisty]

On the whole a good article but there are bits that niggle me.

I appreciate that the article is aimed for all but there are technical untruths, especially in the PFC section.

For example the third diagram is the same as the previous two 'sinusoidal' waveforms, yet it is supposed to be illustrating a non-sinusoidal current! Actually none of the waveforms look sinusoidal, especially the dashed line, and there are no labels.

A switching inductive DC load will have a lagging current and hence VA not equal to Watts. VA=Watts only for ‘real’ currents.

The poor power factor of PSU is primarily generated by the 3rd harmonic distortion caused by the high switching frequency of the PSU. (Hence the name SMPS => Switched Mode Power Supply)

Poor power factor is bad because: -
Noise can affect other equipment on same supply
High load on neutral conductor due to phase current mismatch (especially with 3rd harmonic)
Overloading/heating/losses in conductors due to reactive (imaginary) currents..

Basically if you have an office full of computers without PFC you are likely to run into problems with supply noise.

Discussion on resistance of connectors is interesting but means nothing to me without any figures. What’s the resistance range of a typical Molex connection? What is the typical volt drop across that connection at, say, 10A? Without this information one cannot gauge if resistance in connectors is of any significance or not.

[jonnyGURU]

On the whole a good article but there are bits that niggle me.

I appreciate that the article is aimed for all but there are technical untruths, especially in the PFC section.

For example the third diagram is the same as the previous two 'sinusoidal' waveforms, yet it is supposed to be illustrating a non-sinusoidal current! Actually none of the waveforms look sinusoidal, especially the dashed line, and there are no labels.

A switching inductive DC load will have a lagging current and hence VA not equal to Watts. VA=Watts only for ‘real’ currents.

The poor power factor of PSU is primarily generated by the 3rd harmonic distortion caused by the high switching frequency of the PSU. (Hence the name SMPS => Switched Mode Power Supply)

Poor power factor is bad because: -
Noise can affect other equipment on same supply
High load on neutral conductor due to phase current mismatch (especially with 3rd harmonic)
Overloading/heating/losses in conductors due to reactive (imaginary) currents..

Basically if you have an office full of computers without PFC you are likely to run into problems with supply noise.

Correct, correct, correct and correct. But the article is edited down to keep things simple and cover as much as possible over a short period. But you are correct on all of your points. I should probably do an edit and have Evan add them at a later date.

But furthermore, I'm coming from this from a 115V US user standpoint where you're coming from a 230V UK standpoint. Ever wonder why some power supplies in the US are 115/230V non-PFC and some of those PSU's in the EU are 230V only?

For example, over here, a Topower P6 600W (like the OCZ Powerstream) or the new 650W the Mushkin is based on doesn't have full range APFC. That's because at 115V, the amperage required to move 600W+ through the PFC circuitry generates so much heat, and in some cases the components used simply can not handle the amperage.

You know in many cases, 800W APFC power supplies simply can't do 800W @ 115V? More likely, such a power supply can only sustain 650W @ 115V. But the thinking is "nobody's ever going to find out, so who cares." Label it 800W and be done with it.

And so many "reviewers" mistake power factor with efficiency it blows me away, so I wanted to devert so much positive energy away from PFC. Not that it's a bad thing, of course. But because it's been so misunderstood on so many levels.

But again, all of your points are correct. Including the "room full of non-PFC equipment" analogy. Yes. I should do some sort of edit.

Discussion on resistance of connectors is interesting but means nothing to me without any figures. What’s the resistance range of a typical Molex connection? What is the typical volt drop across that connection at, say, 10A? Without this information one cannot gauge if resistance in connectors is of any significance or not.

Such a discussion would also be hard to have and still keep the article concise. You ask, what's the voltage drop at the connection at 10A? What gage wire? What kind of connector? How long is the wire leading up to the connector? It's a valid concern, but the article is written more from a standpoint of "be mindful of the resistance the modular connector creates" and not "this is how much resistance a modular connector creates."

Personally, I use a modular power supply. I am "mindful" of it. I do not try to persuade people away from it, but I don't pretend that there isn't a potential for problems.

[Twisty]

Thanks for the reply. I apologise my post was a bit bitchy, it was a 2am 'I cant get to sleep post'.

I feel sorry for you guys on 115V supplies, it must take minutes to boil a kettle

The resistance of connectors issue is a matter of some interest to me, and I dont have the equipment to measure the contact resistance with any accuracy. When I am making my own connectors I usually solder the wire onto the connector , opposed to the dry compression joints commonly found on PSU connectors. I wonder if this has a significant effect, as if one has a modular PSU there are effectively 4 or 5 dry contact points in the conduction path.

[jonnyGURU]

I would say the solder is definitely the way to go, but then you still have the resistance of the male to female coupling.

One time I did an experiment with an ATE to see what resistance was added due to a "modular" connection.

If I remember correctly, I went from a 0 to 20A load and only lost .1A. But pair that .1A with the decline you typically get with an average power supply and you could end up with a 3% or more drop.

The thing that REALLY matters is when people throw a modular PSU in their rig is to double check the connectors on the PSU side before putting the side panel back on. Too often I see people plug everything in, slap the side panel on and fire things up without even double checking that everything is secure on the PSU.

[Twisty]

One time I did an experiment with an ATE to see what resistance was added due to a "modular" connection.

If I remember correctly, I went from a 0 to 20A load and only lost .1A. But pair that .1A with the decline you typically get with an average power supply and you could end up with a 3% or more drop.

I assume you mean a .1V drop across the connector, as you cant lose current (Kirchoffs current law) unless your connector is radioactive! Still 0.1V at 20A will result in a 2W power loss to heat.

The thing that REALLY matters is when people throw a modular PSU in their rig is to double check the connectors on the PSU side before putting the side panel back on. Too often I see people plug everything in, slap the side panel on and fire things up without even double checking that everything is secure on the PSU.

I have seen fried lappy power connectors due to only being half plugged in. It surprises me a little that no modular PSU designs have used a more updated socket at the PSU end with a more secure engagement.

[jonnyGURU]

One time I did an experiment with an ATE to see what resistance was added due to a "modular" connection.

If I remember correctly, I went from a 0 to 20A load and only lost .1A. But pair that .1A with the decline you typically get with an average power supply and you could end up with a 3% or more drop.

I assume you mean a .1V drop across the connector, as you cant lose current (Kirchoffs current law) unless your connector is radioactive! Still 0.1V at 20A will result in a 2W power loss to heat.

LOL! Yeah. I meant .1V. Not .1A.

The thing that REALLY matters is when people throw a modular PSU in their rig is to double check the connectors on the PSU side before putting the side panel back on. Too often I see people plug everything in, slap the side panel on and fire things up without even double checking that everything is secure on the PSU.

I have seen fried lappy power connectors due to only being half plugged in. It surprises me a little that no modular PSU designs have used a more updated socket at the PSU end with a more secure engagement.

Yeah... Hiper had a good idea with the BNC connectors, but then screwed that up by reducing two ground leads to one wire, one connector. Super Flower seems to have a similar design and I hope (for their sake) they implemented it correctly, but their power supplies can be a little "weak."