Modular Power Supplies: The Reality of the Resistance

For discussions related to the main web site or the forum itself are now combined into one to clean things up a bit..

Modular Power Supplies: The Reality of the Resistance

Postby ollE » Fri Aug 11, 2006 3:29 pm

ollE
Anti-Static Strap
Anti-Static Strap
 
Posts: 404
Joined: Mon Oct 31, 2005 1:22 pm
Location: Deep Space

Postby Twisty » Sun Aug 13, 2006 3:30 am

Thanks a lot for putting that work in, it answers a lot of questions for me.
So it appears at 11A there was a varience of about 0.1V between supplies, which equates to 1.1W of loss to heat.

The only question left in my mind is what effect wetting the compression joints with some solder would have.

Was the length of cable run roughly equal for all the supplies? Is the load purely resistive? As if not a significant portion of the losses across the wire could be inductive.
- Webster - :tongue8:
Twisty
Black Belt 1st Degree
Black Belt 1st Degree
 
Posts: 1916
Joined: Fri Mar 02, 2001 1:01 am
Location: UK

Postby jonnyGURU » Sun Aug 13, 2006 4:11 am

The length of cables were:

Ultra X2 550W: 24"

Corsair HX620W: 21.5"

Antec NeoHE 430W: 23.5"

I need to measure the fixed cable units. I don't recall what the cable length was on those.

One of the problems of using an ATE is that you can not introduce inductive loads. Why is that a problem? Well, if I'm measuring ripple, I know that my results are going to be best case scenario because the inductive resistance, EMI and RF create by an array of four or more drives can wreak havoc on a power supply. But I can't reproduce that... not "technically." I could always plug in a bunch of hard drives, or small motors, but I'm not sure that would really prove anything.
jonnyGURU
Green Belt
Green Belt
 
Posts: 145
Joined: Sat Mar 25, 2006 5:45 am

Postby tedybear » Sun Aug 13, 2006 6:44 am

Webster :P wrote:The only question left in my mind is what effect wetting the compression joints with some solder would have.


It's been my experience it does not make much/if any difference. Provided the compression joints are crimped correctly that is.

This is also why I recommend useing a contact cleaner (electronic solvent) when I see people ask about lower then expected voltages reported in their bios. Over time each connector will generate "heat" which does increase resistance causeing a mild voltage drop. If not addressed and allowed to continue? That heat will bake the connector and create even more resistance. The result is a burnt contact. Keeping the physical connector free of any oxidation and crap will ensure a long lifespan.

The ATX connector is a classic example of this problem. Over the course of a few years you might see the edges of the retaining plastic turning "yellowish" or brown in color. That happens when the contacts start turning into resistors and it gives off enough heat to discolor, and even in more extream cases it can deform the connector to the point you'll not be able to unplug it. The most extream case I've run across is having several pins that broke free after burning the trace/solder pad on the motherboard.

Basic maintence will prevent this from happening, but you'll seldom see it addressed-it's not something we tend to think about.

S-
tedybear
Black Belt 5th Degree
Black Belt 5th Degree
 
Posts: 7251
Joined: Sat Feb 14, 2004 12:55 am
Location: Fulton, New York, on Earth

Postby jonnyGURU » Sun Aug 13, 2006 7:10 am

Exactly.

All people need to think about is their old Nintendo. Remember cleaning the cartridge with a pencil eraser? ;)

Or an older stereo. Turn the volume knob up and hear "pop pop pop?" Time for some contact cleaner!

Remember how your slot car race track came with the little piece of Scotch-Brite? ;)
jonnyGURU
Green Belt
Green Belt
 
Posts: 145
Joined: Sat Mar 25, 2006 5:45 am

Postby tedybear » Sun Aug 13, 2006 7:21 am

jonnyGURU wrote:Exactly.

All people need to think about is their old Nintendo. Remember cleaning the cartridge with a pencil eraser? ;)

Or an older stereo. Turn the volume knob up and hear "pop pop pop?" Time for some contact cleaner!

Remember how your slot car race track came with the little piece of Scotch-Brite? ;)


That pencil eraser trick is prehaps the best item in my electronics toolbox for work. Or a dollar bill (rough fiber...rub up and down on the jamma harness contacts...Cleans 'em up real good, but turns the bill almost "black")

It's all about resistance :)

S-
tedybear
Black Belt 5th Degree
Black Belt 5th Degree
 
Posts: 7251
Joined: Sat Feb 14, 2004 12:55 am
Location: Fulton, New York, on Earth

Postby zepper » Sun Aug 13, 2006 9:00 am

What contacts at the ATX connector have most often overheated and caused melted wires and shells? 12V - no, 3.3V - no, 5V - YES!!! So why are we testing at 12V? The reason for using 12V for so many things now is to reduce the necessary current load per wire/contact. And almost nothing in the PC uses 12V directly anyway - it is always reregulated and filtered down to a much lower voltage. Only drives and fans use the 12V directly. So the actual 12V number is almost irrelevant. When CPU power is taken from 5V, the Amps go way above 11 - but not all on one wire, of course.

And I'll mention here that most switching PSUs require a minimum load for accurate regulation. That explains the anomalies JG saw at 0 load. Anyone that knows SwPSUs at all should know that... ;) The minimum load on +12 and +5 should be 2 Amps for most. Some older SwPSUs can actually be damaged if run with 0 load - most newer designs have zero load protection. Better case/PSU combos and barebone systems were (back in the day) delivered with dummy loads connected to the PSUs to prevent just such an occurrence.

What about the gauge (not gage ;) ) of the wires on the PSUs? Some PSU makers have gone to 20 or 22 gauge for a lot of the wires. Better made PSUs will have 18 ga. for most high-current circuits. Even so, in the one case where the wire seemed to drop voltage, it would more likely be the crimp connection (or faulty stripping - I've actually found wires totally broken inside the insulation from poorly adjusted strippers) than the wire itself. At the lengths mentioned here, even 22 ga. shouldn't drop much at all as long as none of the strands were lost or broken in the stripping/crimping.

The problem with modular is not when new, but after many cycles of heating cooling the wires and contacts as well as multiple more opportunities for bad crimps/strips. The cheap contacts used on PSUs (seldom genuine Molex or Amp - generally some oriental knock-offs) these days will loosen more with heat/cool cycles and allow corrosion to start - starting the cascade on the way to a melted connector and shutdown or other corrosion effects. In 12V circuits, since current drop isn't really the issue there, the loose/corroded contacts could rectify a local high powered radio station and thus modulate the DC with the audio frequency signal. I'm sure that could cause some interesting system effects...

Granted, those ends will not occur on many modular PSUs, but the potential is there.

Make sure ALL the modular connectors at the PSU end have latched connectors - that will go a long way to helping prevent problems. Treating your contacts when new with Caig Labs Deoxit Gold or similar will also reduce or eliminate contact-related problems. But as for myself, I'll stick with hard-wired PSUs, thank you.

.bh.

8)
zepper
Pilgrim
Pilgrim
 
Posts: 6
Joined: Tue Sep 09, 2003 6:39 pm
Location: Finger Lakes area of NY

Postby jonnyGURU » Sun Aug 13, 2006 10:34 am

Hey Zepper.

Glad to see you over here at the Motherboard homeworld. I trust your experience as invaluable, but if you don't fully read the article your commenting on, then your comments are nothing more than trolling. ;)

I know you're not a troll, though. I have my share of those over at AT. :D So I will address your concerns.... and actually agree with some of them! :D

zepper wrote:What contacts at the ATX connector have most often overheated and caused melted wires and shells? 12V - no, 3.3V - no, 5V - YES!!! So why are we testing at 12V? The reason for using 12V for so many things now is to reduce the necessary current load per wire/contact. And almost nothing in the PC uses 12V directly anyway - it is always reregulated and filtered down to a much lower voltage. Only drives and fans use the 12V directly. So the actual 12V number is almost irrelevant. When CPU power is taken from 5V, the Amps go way above 11 - but not all on one wire, of course.


It is stated in the conclusion that connector should always be double checked. But that's true of modular and non-modular connectors. Whenever I've seen burnt contacts, it's actually at the 3.3V lead mroe often than not, but the burnt contacts happen to fixed cable PSU's just as much as modular.

zepper wrote:And I'll mention here that most switching PSUs require a minimum load for accurate regulation. That explains the anomalies JG saw at 0 load. Anyone that knows SwPSUs at all should know that... ;) The minimum load on +12 and +5 should be 2 Amps for most. Some older SwPSUs can actually be damaged if run with 0 load - most newer designs have zero load protection. Better case/PSU combos and barebone systems were (back in the day) delivered with dummy loads connected to the PSUs to prevent just such an occurrence.


So you skipped over page three ENTIRELY?!?! Man... :(

Allow me to quote myself:

The main ATX connector and 8-pin EPS connector are hooked up and there is already an 11A load on the +12V leads of these connectors. There is also a 12A load on the +3.3V rail, 15A load on the +5V rail, 2A load on the +5VSB and 0.5A load on the -12V. In other words, the power supply is running with a 250W load on it as if it were plugged into a computer.

What I do from here is gradually increase the load on the PCI-e connector from 0A to 5A to 7A to 9A to, finally, 11A, record the voltages with the DMM and log them in Excel.


So when I said "Zero Load" what I meant was "Zero load on the PCI-e connector." All of the other loads remained the same; a total of 250W.

zepper wrote:What about the gauge (not gage ;) )


It can be spelled either way. Pick up a dictionary. Of course, the first definition of "gage" is a variety of plum. ;)

zepper wrote:...of the wires on the PSUs? Some PSU makers have gone to 20 or 22 gauge for a lot of the wires. Better made PSUs will have 18 ga. for most high-current circuits. Even so, in the one case where the wire seemed to drop voltage, it would more likely be the crimp connection (or faulty stripping - I've actually found wires totally broken inside the insulation from poorly adjusted strippers) than the wire itself. At the lengths mentioned here, even 22 ga. shouldn't drop much at all as long as none of the strands were lost or broken in the stripping/crimping.


Agreed. Could be the wire. Could be the crimp. That's why I said I couldn't explain it. Since the wires were brand new with no breaks in them and the gage of wire the same on all five PSU's, then I think problem would be how the wire is attached to the connector pins. And the drop we saw in the Silverstone? Probably the pins themselves. Doesn't change what voltages were seen. THAT is the point. Here we have five different power supplies with voltages taken from no less than three different points along the line and we see different voltages.

zepper wrote:The problem with modular is not when new, but after many cycles of heating cooling the wires and contacts as well as multiple more opportunities for bad crimps/strips. The cheap contacts used on PSUs (seldom genuine Molex or Amp - generally some oriental knock-offs) these days will loosen more with heat/cool cycles and allow corrosion to start - starting the cascade on the way to a melted connector and shutdown or other corrosion effects. In 12V circuits, since current drop isn't really the issue there, the loose/corroded contacts could rectify a local high powered radio station and thus modulate the DC with the audio frequency signal. I'm sure that could cause some interesting system effects...

Granted, those ends will not occur on many modular PSUs, but the potential is there.


Of course, that's true of ALL modular connections. Incuding those connectors on power supplies that ARE NOT modular, where you plug in the peripherals, drives, video card, etc. The problem is about as common as tin contact DIMM's corroding in gold plated DIMM slots.

Make sure ALL the modular connectors at the PSU end have latched connectors - that will go a long way to helping prevent problems.


The latched connectors are good because you can't accidently yank a connector out of a power supply. But the clip does not hold the connector up against the interface tight enough to prevent the connection from coming loose enough to cause potential resistance.

Treating your contacts when new with Caig Labs Deoxit Gold or similar will also reduce or eliminate contact-related problems.


I totally agree with you there. Then again, I used to own a bike shop and was one of the few guys that greased every mating surface of two metal surfaces. You always knew when a bike I had assembled had come back a year later and we could actually get the stem and seatpost back out of the frame. ;)

But as for myself, I'll stick with hard-wired PSUs, thank you.


And no one is trying to change your mind. ;) There will always be manufacturers that will never do modulars and there will always be users that will never use modulars. The article was trying to point out facts with numbers as opposed to regurgatatng fud.

Do modular connectors cause resistance? Yes.

Can modular connectors come loose? Yes.

Do modular connectors give a PSU one more potential point of corrosion? Absolutely.

But do modular power supply connectors have as much resistance as five feet of wire? Absolutely not. Even if the connector is three years old and full of oxidation. There's more fud than fact out there and that's the point of the article.
jonnyGURU
Green Belt
Green Belt
 
Posts: 145
Joined: Sat Mar 25, 2006 5:45 am

Postby Insight Driver » Thu Nov 23, 2006 3:00 pm

For me, I don't see the point of modular power supply wiring. I understand the concept, but it's a non-issue with me. In the builds I've done, I use the connectors I need and tuck the unused ones out of the way, on top of a 5 1/2 in drive. Since power supplies are commodities and have a swiss army knife's worth of connectors on them, there are always going to be unused connectors. A modular supply would have to ship with all the connectors an ATX 2.2 supply has anyway, the only difference is unused ones just get put in a box somewhere or lay loose in the bottom of the case.
MSI K7N2 Delta ILSR
Barton 2500+ AMD
Kingston value ram 512Mb PC3200
Western Digital 200Gb 8Mb
Western Digital 60Gb 2Mb
Gigabyte N57128D 8X AGP
Plextor 8X DVD +/-R/RW, CD-R/RW
Acer 16X DVD ROM
Antec SLK3700AMB case
Insight Driver
Initiate
Initiate
 
Posts: 62
Joined: Mon Apr 26, 2004 6:29 am
Location: Sacramento

Postby Insight Driver » Thu Jan 04, 2007 5:26 pm

I decided to re-visit this discussion as I ran into a difference of opinion on another board. Basically it's a power supply 101 article and one of the links in the article is to the PC Power and Cooling site. It's their site, with their, "power supply myths," that states that a modular connector may have as much drop as 2 feet of wire. Of course that is not true, the proof is based on the excellent testing done by the author. Having the urge to debunk myths I took it on to defend the usefulness of modular power supplies.

Ironically I come back to this discussion over a month later finding no further replies and even more ironic, I found I myself was not defending modular power supplies! At least, to my credit I simply stated it was a non-issue as I did not have a clutter problem with my wiring.
MSI K7N2 Delta ILSR
Barton 2500+ AMD
Kingston value ram 512Mb PC3200
Western Digital 200Gb 8Mb
Western Digital 60Gb 2Mb
Gigabyte N57128D 8X AGP
Plextor 8X DVD +/-R/RW, CD-R/RW
Acer 16X DVD ROM
Antec SLK3700AMB case
Insight Driver
Initiate
Initiate
 
Posts: 62
Joined: Mon Apr 26, 2004 6:29 am
Location: Sacramento

Next

Return to Motherboards.org Web Site

Who is online

Users browsing this forum: No registered users and 0 guests