Are minimum load resistors actually necessary??

[Th3_uN1Qu3]

Got on my hands a couple wounded deers (or LC power maybe, but they use "2005" controllers so they're probably deer), they had exactly THREE main caps, all blown on one of them (the other one has those resistors farther away from the caps...). The insulation on some of the cables was also damaged. Guess where all the molten plastic was? On the minimum load resistors of course.

Also the labels indicated different distribution on the two "450W" supplies, but internally they are almost the same. But we already know that so i won't get into it again. I'll be making my own labels to stick on 'em when i sell 'em. Right now i'm waiting to get some MOVs into them so i don't blow anyone's computer up.

So, back on the wounded and burnt deer. All rails were low due to the bad caps. I was generous with it and added all missing caps, and one pi filter inductor for the 5v rail. The pi filters for the other rails were silkscreened on the top, but the PCB tells another story... to save on wire links, they redesigned the PCB to short the placeholders for the inductors from the get go. I've seen cheap but this is ridiculous.

After completing the output filtering (well, as much as the PCB allowed for), all rails are in spec. The darn load resistors however, the ones on 3.3v and 12v, threatened my pretty Nichicons. The body of the resistors was at 100C!!! I noticed that the one on the 5v rail had actually unsoldered itself from the board, yet that rail was in spec too.

So i pulled 'em all out and whatcha think. The 3.3v rail is exactly the same at 3.35v, and voltage on 12v increased by only 30mv, from 12.27v to 12.30v. 5v is at 5.18v.

That obviously brings the question: Are those resistors actually necessary, or are they simply there because they keep using the same obsolete designs? I know that modern controllers generally do not need a minimum load, so why are they there?

[momaka]

Well, you know my opinion on that one
While I'm no expert on this, I did read an article a while ago about PSU ripple bypassing and filtering and from what I understood, one of the methods for removing ringing from the LRC filters in the PSU output was by putting a resistor in parallel with the capacitor/load. Here's the article (page 7, figure 8 part D, specifically):
http://www.designers-guide.org/Design/bypassing.pdf

My guess is that cheap PSU manufacturers put these resistors there because they maybe also help with ripple suppression in some way.

Originally Posted by Th3_uN1Qu3

The pi filters for the other rails were silk-screened on the top, but the PCB tells another story... to save on wire links, they redesigned the PCB to short the placeholders for the inductors from the get go. I've seen cheap but this is ridiculous.

I have an L&C like that, too. What's even more ridiculous about mine is that L&C actually installed the inductors, even though the silkscreen on the PCB underneath it goes straight trough! Needless to say, that PSU is sitting in a drawer, ready to be used for parts. I got it for free from a shop. Apparently something arced on the primary to the case (no plastic sheet underneath). POS.

[Th3_uN1Qu3]

Mmm, ringing, yes... Well with just 3 caps they probably need those resistors. With a proper cap job, methinks not. I've finished recapping the one that still worked too, this one had a spot for a pi filter inductor on the 12v rail only as opposed to 5v on the other one. I'll get some MOVs on Monday, install those, then away they go. On a more pleasant note, they can actually do 200W on the 12v rail, which isn't bad at all.

I'll throw 'em on the scope and check for any ringing. If they turn up fine, i won't bother about those resistors anymore.

[goodpsusearch]

Originally Posted by Th3_uN1Qu3

The pi filters for the other rails were silkscreened on the top, but the PCB tells another story... to save on wire links, they redesigned the PCB to short the placeholders for the inductors from the get go. I've seen cheap but this is ridiculous.

Originally Posted by momaka

I have an L&C like that, too. What's even more ridiculous about mine is that L&C actually installed the inductors, even though the silkscreen on the PCB underneath it goes straight trough! Needless to say, that PSU is sitting in a drawer, ready to be used for parts. I got it for free from a shop. Apparently something arced on the primary to the case (no plastic sheet underneath). POS.

Please post some photos of the pcb!

[momaka]

Originally Posted by Th3_uN1Qu3

I'll throw 'em on the scope and check for any ringing. If they turn up fine, i won't bother about those resistors anymore.

Good idea. Actually, could you check the ripple both when the resistors are installed and when they're not (if possible, of course)? That way perhaps we may be able to see if they really do make a difference in ripple suppression.

I don't usually completely remove them, but just put higher values (I think I said it right this time ).
33 Ohms min. for 3.3v rail, 100 ohms min. for 5v rail, and 470 ohms min. for 12v rail. 1 or 2 Watt rated resistors, of course.

Originally Posted by goodpsusearch

Please post some photos of the pcb!

Here you go... picture attached below. Just noticed that the only place where the PI coil was actually functional was for the 5v rail. All other rails were just bypassed with a trace on the PCB.
Also note the PCB discoloration due to the load resistors. Typical of L&C, Deer, and Allied units.

[Th3_uN1Qu3]

Originally Posted by momaka

Good idea. Actually, could you check the ripple both when the resistors are installed and when they're not (if possible, of course)? That way perhaps we may be able to see if they really do make a difference in ripple suppression.

They don't. I'll post pics when i have a chance, but i just tested the two PSUs i fixed that have had their load resistors removed, against one which still has them (because they weren't getting hot). Ripple is fine on all three.

I also load tested the two supplies today. For now the "load tester" is a messy setup with a bunch of watercooled resistors (fancy term for saying grab a bunch of resistors, wire them together, stick'em to a large can and fill can with water).

I was surprised to see both of them doing over 200W on +12v! Especially since one of them has a 16A rectifier. I stressed the one with the 20A rectifier more, and IT SHUT DOWN!!! Incredible, they actually have working protections! Makes you think whether it's simply the high ripple and lack of MOVs that actually kills computer parts (i'm not talking about the 5vsb problem on mobos, but about the other parts that cheap PSUs like to take out, like HDDs and memory).

Anyway i think i know why the resistors are there. Crossloading protection. When loading the 12v rail exclusively and excessively (getting it down to 10 volts or so) 3.3v goes to 3.8v, and when loading 5v or 3.3v exclusively 12v goes over 13v. The 3.3v thing happens only on one of them, which simply has an extra mosfet for 3.3v, no magamp regulator. However the 12v rail going WAY high when load is applied only to 3.3 and 5v is common to both (well, to group regulated PSUs in general). When a load as low as 1A is placed on the 5v rail, 3.3v returns to spec. But in practice you'll never have a motherboard which uses nothing on 3.3v and 5v, so this is a non-issue.

The only systems where crossloading is a problem are AXPs and PIIIs which use basically no 12v at all. I put the PSUs up for sale and explicitly said they will not work in an Athlon XP or PIII, and i have a 200W PSU for those systems, with a really anemic 12v rail of 6A.

In practice i've seen AXPs running on crap power supplies where the 12v rail was at 13.something volts. They still ran just fine. OVP on the ATX2005 controller is set to 14.4v on the 12v rail, maybe this very high value wasn't pointlessly chosen. Come to think, i know some years back people were running CD-ROMs in cars with just a 5v regulator, and the 12v was wired directly to the battery, which as we know goes up to 14.4v when the engine is running.

Originally Posted by momaka

Also note the PCB discoloration due to the load resistors. Typical of L&C, Deer, and Allied units.

Same thing on those two of mine.

[PCBONEZ]

The load resistors have a dual purpose in that they are also a discharge path that isn't through the mobo when you you shut it off.

.

[goodpsusearch]

Originally Posted by momaka

Here you go... picture attached below. Just noticed that the only place where the PI coil was actually functional was for the 5v rail. All other rails were just bypassed with a trace on the PCB.
Also note the PCB discoloration due to the load resistors. Typical of L&C, Deer, and Allied units.

Thanks for the pic. I have to confess though that even now I can't believe it!

This is completely idiot and cheap even for LC.

Originally Posted by Th3_uN1Qu3

I stressed the one with the 20A rectifier more, and IT SHUT DOWN!!! Incredible, they actually have working protections! Makes you think whether it's simply the high ripple and lack of MOVs that actually kills computer parts (i'm not talking about the 5vsb problem on mobos, but about the other parts that cheap PSUs like to take out, like HDDs and memory).

The problem with those psus is that their quality can vary a lot. Consider also that they have very tolerant overvoltage or undervoltage protection, as you mention, that may help to damage the components when the caps fail..

All your LC psus have the ATX2005 controller?

[momaka]

Originally Posted by Th3_uN1Qu3

They don't. I'll post pics when i have a chance, but i just tested the two PSUs i fixed that have had their load resistors removed, against one which still has them (because they weren't getting hot). Ripple is fine on all three.

Cool, thanks for verifying that.
I'll still keep some higher-ohm load resistors in my PSUs though, just so there's a discharge path like PCBONEZ mentioned.

Originally Posted by Th3_uN1Qu3

For now the "load tester" is a messy setup with a bunch of watercooled resistors (fancy term for saying grab a bunch of resistors, wire them together, stick'em to a large can and fill can with water).

Hey, if it works, it works .

Originally Posted by Th3_uN1Qu3

I was surprised to see both of them doing over 200W on +12v! Especially since one of them has a 16A rectifier. I stressed the one with the 20A rectifier more, and IT SHUT DOWN!!! Incredible, they actually have working protections!

Interesting. I also have an L&C LC-B300ATX that burned its wires randomly one day when I tried to boot it up. Good thing I unplugged it quickly, though. I still haven't figured out why that happened. Obviously the protections didn't work. Controller is "2003" in that one.
Basically, both 5v and 12v wires on one of the peripheral connectors burned (the connector was connected to a floppy and a CDROM drive - both of which tested good after this incident). Surprisingly, PSU worked too after this. WTF!

Originally Posted by Th3_uN1Qu3

Makes you think whether it's simply the high ripple and lack of MOVs that actually kills computer parts (i'm not talking about the 5vsb problem on mobos, but about the other parts that cheap PSUs like to take out, like HDDs and memory).

I think MOVs only help if a high-voltage spike appears in the power lines. They absorb it so the primary caps don't get damaged. So more or less they are there to protect the PSU rather than the attached hardware.

Originally Posted by Th3_uN1Qu3

Anyway i think i know why the resistors are there. Crossloading protection.

I doubt they'd help much, really. Take for example a load resistor on the 12v rail of a typical L&C or Deer which is normally 100 Ohms. This gives you 0.12A of current draw, and 1.44 Watts of dissipated power in the PSU. A cheap 80mm fan typically draws about 0.14A. Perhaps they help if there's absolutely no load on any of the rails.

[Th3_uN1Qu3]

Originally Posted by goodpsusearch

All your LC psus have the ATX2005 controller?

Yes. One has ATX2005Z, the other one ATX2005AZ.

Originally Posted by momaka

Interesting. I also have an L&C LC-B300ATX that burned its wires randomly one day when I tried to boot it up. Good thing I unplugged it quickly, though. I still haven't figured out why that happened. Obviously the protections didn't work. Controller is "2003" in that one.
Basically, both 5v and 12v wires on one of the peripheral connectors burned (the connector was connected to a floppy and a CDROM drive - both of which tested good after this incident). Surprisingly, PSU worked too after this. WTF!

A PSU with "2003" fried two of my mobos a few years back. Not pretty, and yes it still worked afterwards... It's common for 5v wires to burn up if shorted even on brand PSUs, the current is high enough that one wire presents enough resistance to melt, and isn't low enough to trigger OCP.

Originally Posted by momaka

I think MOVs only help if a high-voltage spike appears in the power lines. They absorb it so the primary caps don't get damaged. So more or less they are there to protect the PSU rather than the attached hardware.

Not really, a high voltage transient WILL get thru to the secondary. The primary caps cannot absorb it, and the controller is too slow to compensate for it.

Originally Posted by momaka

Perhaps they help if there's absolutely no load on any of the rails.

Well, in L&C/Deer/Solytech and whatever it may be called now engineer's minds, they probably do. And btw... it's more like 56 ohms on 12v and 10 ohms on 3.3 and 5v... yes that IS 2.5W, no wonder they burn up the PCB.