A few Technical PSU Questions

Post by **Rusty075** » Sun Oct 20, 2002 9:51 pm

I'm planning a sequel to my PC-in-a-desk project and I've run into some technical questions I could use some help with. It looks like this project is going to require some PSU mods, and that is probably the one part of the system that I feel the least comfortable tinkering with.

So here goes:

Does the housing of the PSU serve any purpose besides keeping people from frying themselves? I could see how the PSU may produce interference (RF?, EMI?) that the rest of the computer would have to be protected from, but I'm not sure.

The two heatsinks in the PSU, they're live, right? I accidentally touched one in an old AT power supply years ago and it sure felt live to me.

The bits that connect directly to them (MOSFETS?) are the bits that get really hot, correct? Hence the big heatsinks screwed to them? Could you use a shim of ceramic to isolate them from the sink? Or does it have to be live for a functional reason?

Those bits, I'm going to call them MOSFETS for now since that's what they resemble to me, could they be relocated? Could I cut them off the PCB and attach extension wires to them? Would this have any "real" effect on the operation of the PSU? I realize there would by some hypothetical performance changes due to the increased resistance of the wire lengths, but would a few inches of wire make a real difference?

I realize that's alot of questions. Any help with any of them would be much appreciated. This sort of info is hard to find on the web. It all seems to be either way too technical for my poor brain, or its from dubious sources that can't be trusted. I don't want to fry my components, or myself for that matter!

TerryW, let me thank you in advance. I'm sure you'll have some opinion of these, and I'm looking forward to reading your thoughts.

quokked · Post by **quokked** » Mon Oct 21, 2002 12:50 am

ummm rusty the
- heatsinks are live in the PSU treat with respect

- the extenstion wires works in one case I've seen on the web
- EMI, EFI, I was running my 250W ATX PSU without the top on so I could get the fan airflow pumping more air out (long story)

good article to check out for u would be http://www.zerofanzone.co.uk/content.as ... 50w&page=1 this guy liquid cools his enermax PSU and looks at a lot of issues u look at here

Gekkani · Post by **Gekkani** » Mon Oct 21, 2002 1:07 am

The housing of the PSU works as 'Faraday Jail' (ermm.. not sure if its a good translation, I'm spanish) Its used to keep away interferences with rest of the puter (avoid Hard disk corruption for example) and to avoid interences with other electricals devices like monitors, audio equipment, etc...

The bits that u said are MOSFETS.. u r right, and gets really hot...

I'm not sure about the other questions .. so I prefer not answer

But hey... please man take care if your going to 'mod' your PSU.. its very dangerous.

gekkani

quokked · Post by **quokked** » Mon Oct 21, 2002 2:01 am

Faraday Shield

ChiefWeasel · Post by **ChiefWeasel** » Mon Oct 21, 2002 3:25 am

Ive heard it called 'Faraday Cage'... Although Faraday Jail is much better

With regards to the extension wires, if you use as thick a wire as possible, i cant imagine the extra resistance would cause any problems. Just been reading the article quokked linked to and he said '...maybe I could lengthen the mosfet pins with wire so they could be easily mounted to a single large block. After confirming there would be no trace length issues in doing this...', so looks ok

What kind of heatsink are you planning on using for the MOSFETS?

Gekkani · Post by **Gekkani** » Mon Oct 21, 2002 6:28 am

lol!

So the correct term is Faraday Cage/Shield... heheh... Good to know it

gekkani

TerryW · Post by **TerryW** » Mon Oct 21, 2002 7:03 pm

Hi all, TerryW here -

I find it hard to pass up reviewing such articles without providing my own comments.

Ya Ya, I know what you are going to say: Another long winded tecno junkie reply from TerryW. Sorry, I just can't help myself.....

As I have indicated in other forum articles, the PSU is a switcher design. The high voltages & magnetics can generate & radiate transient energy within the PSU and without a proper grounded case, into the computer electronics.

The ground electrode on the PSU power cord has two main objectives:

1. First and foremost: electrical safety, in the event of a failure inside the PSU. Not all PSU are created equal and not all have protective fuses, etc.
2. Provides an electrical path to ground for radiated noise within or outside the PSU. Remember EMI / RFI?

Years ago, older computers were sensitive to close proximity of cordless phones, cell phones, etc. These devices radiate a lot of transient noise that computers can be affected by. The computer & PSU cases provide reasonable protection against externally radiated noise.

In my house & at work, I am very strict when people bring their cell or cordless phones near the computers. Hey, you never know when disaster will strike your expensive computer "toys".

The PSU casing is indeed a Farady Sheild / Cage or Jail (I like the jail dea too). The case is also there to provide a sheild to keep the John Q. Public out of dangerous areas. Without a case sheild, I doubt the PSU manufacturers could get CSA, UL and/or VDE electrical approvals.

MOSFETS versus transistors. In the PSU that I have serviced, I have seen either NPN and/or PNP transistors (cheap - cheap). MOSFETS can be expensive and are usually reserved for higher-end PSU designs.

The heatsinks could be live for several reasons:

- Heatsink is physically connected to transistors and/or MOSFET devices.
- Heatsinks electrically tied to other parts of the PSU switcher circuitry.
- At the switching frquencies of the PSU, high voltage can build up on the heatsinks by the transistors and/or MOSFET devices. This is done by capacitive coupling.

Now for the extension of the MOSFET terminations. Yes, the extra wire will add resistance into the circuit, albeit very low, but could be enough to unbalance the PSU switching circuitry. Another thing could be is now we are adding inductance into the circuitry. The extra inductance of the wires will unbalance the PSU switcher.

I agree with Quokked: Respect the PSU. Leave PSU mods to the experts. VERY dangerous to the uninitiated newcomer. High voltage discharges into bare skin can be hazardous to your health. Some other components could explode and tear into your eyes. Having been there and done that, I don't want others to go there.

Whew, enough of this techo-babble. Proceed with caution!

Yours in computing,

TerryW

gmat · Post by **gmat** » Tue Oct 22, 2002 1:32 am

To provide balance against TerryW post, i'd said that factual evidence proves that open case PSUs work very well. For instance, many current PSUs have huge openings (think Enermax or 3-fan PSUs...) which let EMI get through. For a Faraday cage to be effective it must be totally 'closed', or at least openings must be sized accordingly to the EMI frequency. I doubt that 80mm / 92mm round openings provide any EMI shielding.
Moreover, many people are running their PSU in an open manner with no problem.

For the 'live' part BladeRunner from ZeroFanZone (zerofanzone.co.uk) answered it. It is a side-effect only, because they are in contact with live parts. And having seen what our modder here (Rusty) has done i'm quite confident he knows what he's doing.

Voltage regulators in our switched PSUs are MOSFETs indeed. And yes they are the components that get cooled. Extending the wires will not harm anything provided:
- the wires are thick enough to sustain high wattages
- PCB traces in most PSUs are crappy enough and certainly more resistive than thick wire...
Just make sure the solder points are *very good*, this is usually the weak link.

Post by **Rusty075** » Tue Oct 22, 2002 6:03 am

Thanks all.

I had been thinking in similar lines to Gmat in terms of the PSU shell as a shield. The Enermax's come factory equipped with a 92mm hole directly between the PCB of the PSU and the motherboard. If they can get it certified with a huge hole like that in, I can't imagine that removing some more of the material would make a difference.

This thread illustrates a common and amusing theme in modding PC's;

The Engineer says, "You can't do that! It won't work! You'll destroy everything!"

And the Modder says, "Did it already, and it works just fine, thank you."

gmat · Post by **gmat** » Tue Oct 22, 2002 6:54 am

Lucky me (or not), i am both an engineer and a modder.
Anyway, good luck with your mod. Could you tell us exactly what you plan to do ?

Additional note on MOSFETs: you can insulate them from the heatsinks with mica shims. See BladeRunner's excellent article on this topic. Extending them is mainly a test of your soldering skills... thats all.

ChiefWeasel · Post by **ChiefWeasel** » Tue Oct 22, 2002 8:21 am

Regarding fan holes, remember though that they always have a metal fan guard, or grill, which means that the holes are infact much smaller than the full 80 or 92mm. Not sure what your planing on doing precisely but maybe bare that in mind?

That article by bladerunner, or rather the work he did to the PSU, is very impressive!

Post by **Rusty075** » Tue Oct 22, 2002 10:42 am

But the fan guards have openings that are more than a centimeter apart. Given the wavelengths we're dealing with that's not going to stop anything.

My plan, so far, for the psu mod is to spread the heat production from the Mosfets out over a larger area, so I can use less ventialtion.

TerryW · Post by **TerryW** » Tue Oct 22, 2002 11:55 am

Hi all,

I found on the 'net a document "ATX Power Supply Design Guide (Ver 1.1)". Copywrited by Intel Corporation @ 2000. This Power Supply Design Guide makes for some interesting (techno-jargon) reading.

Each PSU manufacturer has to go through rigorous testing to meet the requirements of this Design Guide.

Specifically, ANSI C63.4-1992 is for the radio noise emmissions from low-voltage electrical & electronic equipment in the 9kHz to 40 gHz for EMI radiation testing.

1) Are these manufacturers designing to the ATX Power Supply Design Guide? Or do they care?
2) Are these manufacturers meeting the requirements for CSA, UL and/or VDE? Or do they care?

I quote some details in these specifications:

Fan size / speed:

"An 80 mm or larger fan is recommended ... to provide enough cooling. Exact CFM varies .... but 25-35 CFM is typical ...."

"For ... noise-sensitive applications, it is recommended that a thermally-sensitive fan speed control circuit be used to balance system-level thermal & acoustic performance."

Venting:

"In general, more ventin in a PSU yields reduced airflow impedance and improved cooling performance. Intake & exhaust vents should be as large, open & unobstructed as possible not to impede airflow or generate excessive acoustic noise..... A flush-mount wire fan grill can be used instead of a stamped mmetal vent for improved airflow & reduced acoustic noise."

"There are 3 caveats (neat techo word) to the venting guidlines:"

1) "Openings must be ... designed to meet safety requirements.. "
2) "Larger openings yeild decreased EMI sheilding performance."
3) "venting in appropriate locations can detrimentally allow airflow to bypass those areas wher its is needed."

Whew, some interesting details, eh?

BTW - Why is it Canadians say 'eh' after everything? Its the way we spell our country: C-A(eh)-N-A(eh)-D-A(eh)! eh? Snicker snicker.

For those that might be interested in reading the ATX Power Supply Design Guide, I have the link @ home (MikeC has too) and I can provide later.

Cheerio!

TerryW

Post by **MikeC** » Tue Oct 22, 2002 12:38 pm

Um... This is not really news, Terry; the form factors site has long been one of the information links recommended here, and many of its documents (including the one you quote) has been referred to directly in some of our articles.

With regard to your questions, who do you mean by "these manufacturers?"

In an attempt to answer,

1) PSU makers are well aware of the document; how closely they follow it depends on what their goals are, I expect
2) All the PSUs I have tested or examined have had umpteen approval stickers & marks, including the ones you mention. Some more than others. So whether they REALLY meet the standards, they have the certification marks, anyway.

Post by **MikeC** » Tue Oct 22, 2002 1:05 pm

My plan, so far, for the psu mod is to spread the heat production from the Mosfets out over a larger area, so I can use less ventialtion.

Hmmm, but the mods you're considering have the potential to make your PSU -- well, dead -- and it's not clear that you'd get what you're looking for. Here's another idea: (assuming you're thinking about modding your desk PC further)

1) You're not limited by any form factors, yet your PSU, which makes heat, is in the same space as your other gear. What's to stop you from moving it outside the drawer with some method to pull fresh air directly to the PSU? This takes the PSU heat out of the main PC case (the drawer). It also stops the PSU from getting heated by other components.

2) mount the PSU on the outside of the drawer back panel so that its main fan hole is pointing up. This means air is drawn from the bottom & passes much more usefully up through the heatsinks before exiting out the top -- see the "Run a PSU without a fan?" article on the main site.

3) increase the surface area of the heatsinks in the PSU by screwing/bolting an appropriately sized piece of aluminum HS material extendng across/between the existing ones. I've done this, but haven't built anything around it. For your situation, you could even get a huge pice of extruded aluminim heatsink matrial (say 10" x 6" with 1" tall fins running lengthwise), screw/bolt the flat side on the outside drawer back panel above the PCI slots, and then bolt the PSU HS to the big aluminum piece (without the cover on) so that the whole PSU is hanging of the al. piece. At that pont, you could discard the PSU case all together, as most of it is on a single PCB, and find other places for the power switch & AC cable. This whole assembly could be covered with a tube (cardboard? PVC?) cut in half & so it become a heater in a chimney.

You could probably run this without a fan but there would have to be some way for the rising heat to escape.

Just some wild thoughts. If what I have written is unclear (or if my assumption that you're working more on your desk is wrong), let me know -- if the former, I'll see if I can show rather than tell.[/quote]

TerryW · Post by **TerryW** » Tue Oct 22, 2002 2:15 pm

MikeC,

Gee, I thought I was helping and not hindering the issues. However, I do tend to go on & on.

There is no question whether PSU manufacturers meet or exceed the Design Guide. All I really wanted to point out to prospective modders is the design criteria the PSU manufacturers need to abide by when they design their products, and to provide a level of awareness so that modders will modify for the right reasons.

As you can appreciate, anyone that mods an electrical product does so at their own risk.

Any mods to an approved product naturally voids the original PSU maufacturers warranty, voids the computer vendors warranty AND any electrical approval / certification that may be on the PSU and/or the computer system. Also, if a fire was to happen and the fire marshall was to inspect and the cause was a modded PSU, your house insurance could also be voided.

I have heard of similar situations where people mod house wiring and/or applicances, a fire starts & there is no insurance to cover damages. That can be a huge downside to modding electrical products.

TerryW

Post by **MikeC** » Tue Oct 22, 2002 5:54 pm

Terry, no intention to ruffle feathers; I wasn't sure why you were quoting all that and what you were really trying to say, because no one mentioned anything about manufacturers before. Anyway, your last post makes clear what you meant. I would guess that Russ has got to be aware of the warranty voiding aspect of his planned activity.

I recall a long discussion somewhere about whether PSUs can actually catch fire, and the answer by several techs were, no they can't, there are so many retardent materials used in PSUs that the worst that could happen is a bang and a breakdown. I guess if your house wiring / mains breakers are bad, something could happen elsewhere.

I've done lots of dumb things inside PSUs -- zapping myself with DC and AC more times than I care to remember -- um maybe I keep doing it because the zaps make me forget the last time.

I have destroyed 4 PSUs over the last 12 years or so, only the last one in a big BANG! The rest kind of got beaten into submission after umpteen shorts and other abuses and died quietly. But I've never had one catch fire.

siliconacoustics · Post by **siliconacoustics** » Tue Oct 22, 2002 10:56 pm

Maybe I am just unlucky with computerz (I have always believed this to be true) but let me assure you that PSUs can in fact catch fire. I do not want to injure the maker of this small production open frame PSU by revealing its name (it is nothing we sell), but it did burn. At some point of great electrical resistance there was enough heat generated to inflame proximate board level componentry. The good news was that there is not a lot of fuel in a computer case. The bad news is that what fuel does exist burns with noxious fumes that made me think that the kitchen trash compactor motor had gone berserk and caught fire. The heat melted cables, and the smoke damage was so severe that I can’t remove it from aluminum case even with a solvent. All damage was confined, except for the smell, to the inside of a substantial aluminum case.

The ATX specification does describe that a PSU should not catch fire under any catastrophic condition, but that is just a specification after all.

Jonathan

Post by **MikeC** » Tue Oct 22, 2002 11:08 pm

Yikes! I stand dramatically corrected!

Unlucky with computers so that's why you're in the business of selling computer components?

But that was an open frame PSU not necessarily intended for use within a PC -- probably intended to be used as a part engineers would integrate into whatever, right? How about an ATX/PS2 PSU? Have you had one of those burn, Jon?

siliconacoustics · Post by **siliconacoustics** » Tue Oct 22, 2002 11:24 pm

It was a 150-watt ATX PSU to power a computer system. It was open frame because of a special requirement to attach its heat producing bits to the frame of the case. I've not had a fire in a normal enclosed PSU (yet). The case of a normal PSU would have been enough to contain the fire spread. But not the smell

Jonathan

gmat · Post by **gmat** » Wed Oct 23, 2002 3:03 am

TerryW, i suspect Rusty has already voided the warranty of a lot of things...
And come on, anyone that can handle a soldering iron (and takes electrical precautions, such as discharging capacitors...) can mod a PSU. I am an electrical engineer and during my training (that was when i was young...) a 1st year exercise was to *build* a PSU ! Trust me they were quite ghetto compared to a neat mod in an ATX PSU.
As for respecting standards, Rusty did *not* respect ATX recommendations :p So what, his house is gonna take fire, his wife quit him, his car taken away, and insurances paying nothing *just* because he didnt conform to ATX standards ? What country do you live in, i wouldnt want to live there... Crying "it's gonna take fire!!" is not constructive or even informative. It's like telling a modder he's risking severe cuts and injuries handling a hacksaw and a dremel. Or telling an overclocker that he risks burning his CPU...

Let's stay realistic. This site aint intended for Mr Joe Shmuck and his Auntie. And this thread was started by a guy who built a neat in-desk PC, i'm sure he knows what he's doing...
As far as switching PSU go, this is low-tech enough for anyone to hack in it. I repeat, extending the MOSFETs is harmless, there is absolutely *no* reason it could harm the PSU (apart from a fumble while soldering...). Putting huge heatsinks in place of the original ones is a great idea and could lead to a fanless solution.

note on Terry's first post: imabalance the switcher ?? The MOSFETs are here to regulate the voltage, not to provide switching timing. The switching, IIRC, is triggered by a diode bridge.

Post by **Rusty075** » Wed Oct 23, 2002 2:28 pm

Mike, in reference to your comment, no this PSU mod for for a completely new system. I've decided that any further silencing of the Desk Machine would require some seriously radical changes (Watercooling, phase-change cooling, etc.)

Although the plans for this new system are still sketchy I know a few characteristics of it for sure. It's going to be Powerful, Portable, Silent, and oh yeah, completely fanless.

Gekkani · Post by **Gekkani** » Wed Oct 23, 2002 2:32 pm

Fanless? So u'll use VIA C3 right?

WaterCooling usually needs at least 1 fan.

Post by **Rusty075** » Wed Oct 23, 2002 3:06 pm

I don't think Gekkani.

This is going to have a "real" processor, probably 2ghz or better

crisspy · Post by **crisspy** » Thu Oct 24, 2002 1:25 am

In response to gmat:

"I repeat, extending the MOSFETs is harmless, there is absolutely *no* reason it could harm the PSU..."

I mostly agree. The mosfets are a capacitive load to the driver IC, and are probably only compensated/driven as such. Very long leads, like across a whole tower case to the other end, might add enough inductance, that combined with the mosfet's capacitance, at the high switching frequency, would risk 'dulling' the sharp square wave switching signal. I can imagine those fets spending a long time in transition from off to on to off. Those mosfets only handle the power because they switch too fast to spend much time as a transitional load. An uninformed but compitent and adventurous electrician/hobbiest (wires own house etc.) could inoccently make the mistake of saying "Hey, I have heatshrink, I have #12 lead wire, I have insulating shims, and I have a neat home made copper water block with space on it for those mosfets, so I'll just put some 2 foot wires on and roll up the excess later." Poof? I'm sure that 6" leads wouldn't do much harm, but I would want to do some extensive bench testing with an oscilloscope before I went out past a foot or two long.

I think that as is often the case, staying within reason is what is really required here. Maybe as a rule of thumb stay within the psu case dimensions, and make the mosfet leads just long enough to do the job. That should be short enough to not cause any big trouble.

And keep a fire extinguisher handy LOL

As for ATX specifications, compliance, etc... The issue here is Electrical Codes. In Canada the laws say "You can't sell things without CSA approval." The insurance companies follow along and say, "We won't pay your insurance claim if an electrical fire is caused by a non-CSA approved device." Most other countries are probably similar. If you modify an electrical device, you void it's CSA approval, and having a CSA'd PSU burn after modding it, burning your property, could put $$ $$ in the eyes of an insurance claims lawyer. Maybe it would be best to lie and say "bought it that way" or try to call attention elsewhere somehow??? Anyways, who wants a fire. Insurance is not the answer to fires. Prevention is. Maybe a small automatic fire extinguisher is available and inexpensive, to go inside a PC case to limit the potential damage. Strikes me as a good idea with any wooden enclosure like Rusty's desk, or plastic like Mike's breadbox. I don't know how well it would work, but I'll be looking into it. Rusty may have beat me to the punch on the desk idea, but I'll be catching up with my own furniture creation eventually, and I would like to play it safe. I'll post my findings.

It's nice to see the diverse opinions and attitudes co-exist so peacefiully and constructively on these forums, with no flame wars etc..

Cheers to all!

gmat · Post by **gmat** » Fri Oct 25, 2002 3:25 am

Gekkani wrote:
Fanless? So u'll use VIA C3 right?

WaterCooling usually needs at least 1 fan.

Nope. With a radiator big enough or appropriately designed, you can run it passively. You can also use geothermal cooling like BladeRunner here:
http://www.zerofanzone.co.uk

Jolt · Post by **Jolt** » Fri Oct 25, 2002 6:19 am

BladeRunner's system is so cool, must be the holy grail of silent yet high performace computing.

Too bad it requires an excavator and a water tank of a few hundred kilos

BladeRunner · Post by **BladeRunner** » Mon Nov 04, 2002 2:17 pm

Just found this topic through a search

Rusty075

I just love this quote:-

The Engineer says, "You can't do that! It won't work! You'll destroy everything!"
And the Modder says, "Did it already, and it works just fine, thank you."

So....So true, think I'll pinch it for my sig

Talking of which my water-cooled PSU build is coming up for a year old now and other than a mod to help cool a coil that was getting too hot during extended gaming, and some thining out of the plug wiring clutter. it's coped with a high of 38C summer ambient without a fan just fine.

I've started to modify the chasis for better cooling by cutting out windows that will have a fine mesh to aid airflow, it will also be polished up to a shiny chrome finish and lacqured. Other things I've done are a Pump flow related auto shutdown system, and prevention of PC start up if the pump is not connected to the PC.

I haven't noticed any problems with the extened mosfet wiring but the wires probably could do with being thicker if more load was taken from the PSU like a peltier.

here is a current image.

It's currently running this set-up with no instability issues, and for the past 6 months or so it's been run without a lid.

My Current Spec,

Althlon XP 2000 (1666mhz) @ XP2200 (1800mhz) 9 x 200fsb
Abit KX7 Raid
1GB Corsair XMS 3500 at cas2
Radeon 9700 Pro (currently being water-cooled)
2x Seagate Barracuda IV 40gb (single platter)
Enermax 550 watt PSU

Water cooling

Home made or modified blocks for everything
Pump, Eheim 1060 (remote)
Radiator, "bomb" 100 litre propane tank underground
external header tank above PC level for filling, bleeding and circulation

TerryW · Post by **TerryW** » Mon Nov 04, 2002 2:26 pm

BladeRunner,

Very cool (pun intended). Very nice job.

TerryW

BladeRunner · Post by **BladeRunner** » Thu Nov 07, 2002 6:18 am

I finally found some mesh, it will be finshed off and fitted properly inside the PC case soon hopefully. on an R300 Project atm