You are here

Fanless Power Supply: Marko's Homebrew

April 13, 2003 by Marko Djokic (quix)
with
Mike Chin

Marko's
how-to article is not for the faint of heart or for the soldering-challenged.
He describes the steps he took to replace the heatsinks in a conventional ATX
power supply with a massive heatsink mounted on the back of the PC case, allowing
fanless operation. It assumes a certain level of electronics and PC hardware
savvy. It is, as with many SPCR reader submissions, a testament to DIY ingenuity,
sheer will power and courage -- to try, to risk failure.

Marko's willingness to write an article in a language not his mother tongue
is another measure of his courage! (Some might say his willingness to expose
a socked foot on the web is one too. :D) I spent extra time carefully going over edits and proofs with Marko to ensure that the correct information is conveyed, and to smooth out issues with language. I also added more photos to better illustrate some of Marko's text.
If there is any confusion, by all means,
email the author (or me) with questions. Modding a power
supply is intrinsically much more dangerous than something like wiring up a
fan for 5V operation or rigging up a hard drive suspension.

Marko is not an electronics technician, and this article does not delve
deeply into the technical background of the tasks performed. Obviously, if you
try something similar, it is at your own risk. My main caveat: It has
not been that long since Marko finished this project and got his fanless PSU
up and running, so while I have no specific reason to question its longevity,
it has not withstood the test of time. Enjoy, and be careful! - Mike Chin,
Editor.

Introduction

Hello! I'll briefly introduce myself here. My name's Marko Djokic, but you'll
often find me as quix on the net. I live in Switzerland, I'm 20 years old and
studying computer science. I'm mostly coding on the PC for fun, 3D engines and
such.

The PC I'm modding was my main computer until 6 months ago when I got a laptop
(Dell Inspiron 8200, P4-Mobile 1.7 GHz, 768 MB DDRRam, 60 gig HD, ATI Radeon
9000 Mobile with 64 MB, UXGA 15" Display). Now, it's just a file-server
/ leech-machine, and silencing it is mostly a challenge. I'm trying to keep
my mother from noticing that it's on 24/7, she would kill me if she knew that
it has been using up electricity continuously for over a year now ;-). But I
would really like to have it in my room and be able to sleep next to it without
hearing it!

Components

  • Asus P2B Motherboard
  • Katmai-core P3 500 MHz
  • 384 MB (256 + 128) PC100 SDRAM
  • 2x Maxtor 120 GB 7200 rpm Plus9 hard drives (FDB Motors :-D)
  • 1x Maxtor 80 GB 7200 rpm 740DX Hard drive
  • 1x Western Digital 30 GB 5400 rpm Hard drive
  • Elsa Erazor III (NVidia Riva TNT 2) with 32 MB RAM
  • SB Live! Value
  • Realtek-NIC
  • 5-Volted Coolermaster TF8-25IM for HD's
  • 5-Volted Coolermaster TLF-R82 for the CPU
  • 1 CDRW drive which currently isn't connected (until I get a controller)

The PC is installed in an old (now heavily modded) Viper Case, with the original
PSU. I started the heavy modding after I got the laptop, because, if I blew
this system up, I had a backup now! Here is the label from the PSU:

(Editor's Note: As you can see, it is a standard 250W rated for only
8A on the 12V line. It is perfectly suitable for Marko's system, assuming honest
ratings, but not for XP or P4 systems.)

I changed the PSU fan, and 5-volted it. This airflow was also enough (when
idle or downloading at least) to passively cool the CPU. The other Coolermaster
fan is cooling the two 120G HDDs. The hard drives turn off after 10 minutes,
so usually, only the system hard drive (one of the 120G HDDs) is running, and
it's very quiet! Also it didn't need too much cooling as it's running only one
hard drive and the CPU is idle most of the time. The 5-volted fan of the PSU
was the loudest, then the fan for the HDDs, then the hard drives themselves,
and finally the coil buzzing from the PSU.

My Fanless Idea

Since the power MOSFETs (Metal Oxide Semiconductor Field Effect Transistors)
are the only parts that are attached to heatsinks, I figured they are the only
components that need to be cooled. My idea was to extend the leads of the MOSFETs
so that they could be attached to a very large heatsink mounted outside the
case. The heatsink would have to be big and efficient allow the MOSFETs to be
cooled with just passive convection airflow. Then the loud fan in my PSU would
be completely unnecessary.

I had this idea for quite some time, but didn’t have the courage to try it.
I don’t know much about electronics and whether it would really work. The problem
is that the wires would introduce resistance, inductance, impedance, and who
knows what else that might interfere with proper operation of the PSU.

Inspiration

Then, one day, as I went through the links on this great site (SPCR), I stumbled
across Zero Fan Zone
by BladeRunner. There, he wrote on how he extended the leads on the MOSFETs
to mount them on a waterblock. Since it worked very well for him, I decided
to do it too! I had a spare PSU lying around, so it would be no problem if I
blew it up.


WARNING!

If you also want to do this mod, be very careful! Some experience
with electronics work and soldering is helpful. It can be dangerous, the voltages
can kill you. I’m only summing up how I did it, I’m not responsible if you
blow up something or hurt yourself! Also, I
read that a modded PSU could invalidate your home fire insurance
, so if
your PSU catches fire, and your home burns up, if the insurance company finds
out that your modded PSU was the cause, they may not pay anything.


Getting Started

You should start this by first planning how and where you want to mount the
printed circuit board (PCB) and the new heatsink(s), so you don’t use wires
that are too long, like I did. I wasn't sure if the PSU would work at all, so
I started with the soldering!

The PCB should be vertically mounted to have some convection airflow, because
other components still have to be cooled, not only the MOSFETs! You should leave
some place above/next to the PCB, so you can mount additional heatsinks if needed.

The main heatsink for the MOSFETs should be mounted with the fins running vertically
for good convection airflow. I mounted the heatsink on the back because looked
nicest and the least obtrusive. I've also chosen to mount it where the PSU usually
would be, so I can use its mounting holes, and also have a place inside where
I can mount the PCB.

To the PSU

Before doing anything inside the PSU, you should wait some time after unplugging
it, so the capacitors can unload and you won't get a shock when working on it.
I’m not sure how much is enough, some PSU have resistors that unload the caps,
so an hour may be enough, but a few days should be enough on any PSU.
(Editor's note: A thread in the SPCR PSU Forum called PSU
modding dangerous?
discusses this issue. If unsure, take the
safe route and wait at least a day.)

The first task is to examine:

  • the heatsinks,
  • the way the heatsinks are attached to the PCB,
  • the way the MOSFETs are mounted on the HS, and
  • whether there is any electrical connection between the HS and the MOSFETS
    or between the HS and the PCB.

LIVE?!

You must check whether the heatsinks are live -- this means carry any
voltage that could cause a short when touching something else in the PC or cause
a shock if you touch it. Normally, there are at least 2 lugs or anchors that
secure the heatsink to the PCB. The photo below (of another PSU) shows a HS
anchor.

The anchors are usually soldered to a heavy trace (metal contact path bonded
to the board) on the bottom side of the PCB. If other components are soldered
to that same trace, the heatsink is probably live!

If the other anchor also is on a trace with components soldered to it, then
the HS is also acting as a signal path between those two traces. This is the
case with anchors #1 and #2 in the photo above. You should check to see if there's
some other trace that directly connects the two anchor points. If not, you'll
have to solder a wire between them when the MOSFETs and heatsink are moved off
the PCB.

The anchor #5 on the PSU PCB above is an example of an anchor that is NOT connected
to any components. The trace is separate, sort of like an island on the PCB.
It is acting purely as a mechanical support. Note that this HS has 3 anchors,
which means you have to examine any trace contacts among the 3 anchor points
and record that info in case you need to wire in one or more jumper wires.

Mica-shims should be used to isolate the MOSFETs when they are mounted to the
new heatsink. Otherwise, the heatsinks and the case would
be live -- a very dangerous thing for components as well as for anyone touching
the case. You should check if the MOSFETs have a direct electrical connection
to the heatsink. This happens when there isn't a shim, and the MOSFET has a
metal back. Usually all the MOSFETs in a PSU have a mica-shim.

If you don't understand what I've explained so far, you shouldn't do the
mod. Make sure you understand fully all of the above or get someone more knowledgeable
to help you before starting to do any modification in the PSU.

In my PSU, the smaller heatsink was soldered to a live trace on one anchor,
but not on the other anchor, so the heatsink was not acting as a signal path.
The two MOSFETs on this HS had a completely plastic case and didn't make a connection
through their case. The MOSFETs could thus be removed and mounted on a new HS
without worrying about the role of the HS in the circuit. Mica shims were not
even needed because of the insulated plastic casing of the MOSFETs.

Unsoldering the MOSFETs

It would be easiest to unbolt the MOSFETs, and unsolder them separately, but
often you won't be able to do that, because the transformers and capacitors
are in the way. They were for me. By unsoldering the two big capacitors, I could
easily unbolt 4 of the 6 MOSFETs, which really helped, as I didn’t have a de-soldering
"sucker" tool when I did this. Don’t forget to note down where each
MOSFET was!

The other heatsink shown in the 2 pictures above is a bit more interesting.
Two MOSFETs were screwed to the heatsink with mica shims in-between, so they
could be transferred to the new HS and mounted the same way.

But the other two MOSFETs were directly screwed to the HS without insulating
shims. The metal part of the MOSFET with the mounting hole is connected to its
middle pin. So the middle pins of these MOSFETs are connected to the heatsink.
One anchor of this heatsink was also soldered to a live trace. I checked the
traces on the PCB, and saw that the middle pins of these two MOSFETs and the
heatsink are all soldered to the same trace. In other words, the heatsink is
not the only electrical path for the connection between the middle pins and
the anchor point. Even after removal of the HS, the electrical circuit would
be unchanged. Whew!

Extension Wires

The next task was to solder extension wires to the PCB mounting points for
the MOSFETs. I used wires from a old blown-up PSU that was kicking around. They
were the cables to the Molex connectors, which explains the colors. They seemed
to be thick enough so the electrical resistance is not too high. (Editor's
note:
Usually 18 AWG -- gauge)
They shouldn't be any bigger, otherwise
it will be difficult to solder cleanly to the PCB. Now you should plan how you
want to mount the HS, so you don’t use wires that are too long, like I did!
Double-check that you didn’t short any of the wires when soldering on the PCB.

Before soldering the wires to the MOSFETs, you should maybe first
shorten the pins a bit. Check the right MOSFET on the photo above.

Make sure you wires are soldered to the correct pins. Messing this up could
be very dangerous! Color coding the wires for the 3 pin types might be a good
idea to help you with this. Don’t forget to insert the heat shrink tubing over
the wires before soldering, and solder as cleanly as possible!!!
Apply heat from a strong hot hair dryer (or heat gun) to ensure the heatsink
tubing provides insulation protection for the MOSFET pin solder joints.

Testing

Before going any further, I wanted to test my wiring to make sure the PSU was
still working. I placed the PSU on the wood floor and connected some hard drives
I had lying around plus the original fan (just in case it needed to be cooled).

Before plugging the PSU into the wall, I hid myself behind something, so if
it exploded or something, I wouldn’t get hurt! I did the same thing when I powered
it on. To do this, connect wires from a switch to the green wire contact and
any black wire contact on the 20-pin plug that goes to your motherboard. (Editor's
Note:
If the wires are long enough, you can run to your underground bunker
before flipping the switch.)

The first time I turned the power on, it worked!!!! That was a great moment!
Then I checked the 5 and 12 volts rails, and they were OK.

Mounting to Case

After being assured that the electronics were still working, I went on to figure
out the new heatsinks. (Should have done this at the beginning.) I calculated
that the heatsink should not cover the cutout of the motherboard tray, and at
least cover the hole for the PSU. I went to a electronics shop nearby with my
measurements. I wanted a large heatsink that had all fins going into one direction,
but they only had such heatsinks with really short fins, 2 cm, so their cooling
performance wasn't that good.

I got two heatsinks that had perfect width and the right combine height when
stacked on top of each other. They were rated for 1 C/W each. That means the
temperature of the heatsink will rise 1C per 1 Watt of heat applied. Obviously
a lower number is better. Because the heat of the PSU would be split between
the 2 heatsinks, it would be like having one HS with a 0.5 C/W.

(Editor's Note: If 50C is the target maximum safe temperature of the
HS, for an ambient temperature of 25C, the heatsinks would be able to dissipate
50W before hitting 50C. Assuming an AC/DC conversion efficiency of 65%, the
PSU would be able to deliver ~93W in DC voltage to the system at that temperature.
That is probably adequate for Marko's Katmai-core P3-500 MHz system, although
if all the drives were working at the same time with a heavy CPU load, it
would be close.

Mind you, it is not that the PSU would not deliver higher power. It would
simply try to deliver whatever current was being demanded, and the temperature
of the PSU would continue to rise. But because the HS is exposed, there is
more of a fire hazard in case flammable material with a low ignition temperature
comes in contact. So it is important not to load the PSU to the point where
the HS gets much hotter than 50C.

It is a known fact that PSU efficiency tends to drop as thermal limits are
reached, often dramatically. This means that as the PSU gets hotter, it becomes
less efficient and produces more heat, which makes it work harder, and so
on, in a heat-inducing spiral that can become something called thermal
runaway
. Like Chernobyl. Not a nice thing to have happen in your PSU or
heatsink.)

At home, I held the heatsinks on the back panel and marked on them the PSU-mounting
holes. I hammered slightly with a nail on these markings on the heatsinks so
the drill won't slip, drilled the holes and tapped them. The heatsinks now could
be screwed from the inside. Then, I drilled 4 holes on each heatsink, on a place
where I could easily screw the MOSFETs to once mounted on the case. I drilled
and tapped 4 holes on each heatsink for a total of 8. I only needed to mount
6 MOSFETs, but wanted to be able to move the MOSFETs around in case one of the
heatsinks got much hotter than the other. This didn't happen in the end, but
it was good insurance planning.

Find a place in the case where you can mount the PSU and have good convection
airflow, because the other components still have to be cooled, not only the
MOSFETs! In my case, this was no problem, I just put it where the PSU was before,
and there was also metal behind it, so I could directly mount it there.

To mount the PCB directly to the case, I marked on the sheet metal through
the holes on the PCB, and drilled the holes, a bit larger than the screw hole
of the spacer I would place between that metal and the PCB. Screwed everything
together, finished! (Editor's note: To be on the safe side, you may
want to place a piece of that stiff insulating plastic sheeting between the
PCB and the case.
)

Additional cooling

I connected everything, and it worked! But, one of the coils was getting really
hot, and I couldn’t keep the CPU passively cooled as before. It was cooled well
enough by the 5-volt fan in the PSU before. So, as a temporary solution before
running off to have the PC work as the file-server for a LAN-party, I mounted
a fan between the PSU and the CPU. It is a clear fan, shown in the photo directly
below, underneath the 2 parallel white wires.

After the LAN party, I attached a heatsink directly to this coil.
I drilled two holes to the side of the heatsink, and attached it with a cable-tie.
I put a mica-shim in-between and some thermal goop. It's kinda strange that
now I don't even notice that this coil/heatsink is getting warm... there must
have been some really bad airflow.

Next problem was the big transformer, it wasn’t getting hot, I
could hold my finger on it for a longer time, but it was too hot to leave it
like that. The same with the main rectifier. So I hot-glued two heatsinks on
these parts, and it worked! I was really surprised how well hot-glue transferred
that heat! The PSU is now finished, I might change some little things, but it
works very well as it is now!!

Case Cooling Mods

I've also done some work on the case. I drilled a bunch of 6mm
holes on the top of the case (above the PSU and CPU) so the hot air can get
out. 667 holes, in fact! It's great, you can feel the airflow, and see the dust
rising in the CCFL lights, even when the PSU and CPU are only around 40C!

It is quite simple to drill one hole but it takes ages for 667. I took a A4
4mm squared paper, and marked every second crossing so it made a nice pattern.
Every second line was also moved by one crossing so it didn't look completely
squared. You can, of course, do something on the PC and print it out. Then,
I tightly taped the paper to the case, and on each marking I hammered lightly
with a nail to create a starting point.

Now you can start with a smaller drill to be more exact. I started doing this
after I slightly messed up 40 or so holes. I used a 6mm bit as it would leave
at least 2mm between each hole! I thought this would be enough, but I wasn't
careful enough. Try being that careful when drilling 1334 holes! In some places,
there's nearly no metal between two holes, but you really have to look hard
to spot them. While drilling and hammering, put something behind that piece
of your case so it won't vibrate and bend too much (it'll bend back, mostly,
after drilling). At the end take a much larger drill (twice or so), and go very
lightly into each hole (don't make new ones!!) to deburr the edges on both sides.
This also needs painting, as you can now see the bare metal in and a bit around
the holes. I'll do this when I'm finished modding the case.

Additional and
Future Mods

The next thing I’m going to do is to cut out some of the
metal behind and under the PSU to improve the airflow. And href="http://www.cocoon-culture.com/lib/noise-report/psu-2-investigation.htm" target="_blank">hot-glue
some components of the PSU, because it’s buzzing quite loud! Well, relatively
speaking, it seems loud. :-)

I have planned some more mods for it. I already cut out that design
on the left side cover (got a Dremel 2 days ago :-D) so cold air can get in
there. I need to cut out some more holes on the bottom, and remove the fan for
the hard drives, so there would be only one 5-volt fan left on the CPU. I’m
thinking about how to remove this one too, undervolting, larger heatsink, newer
CPU... This one is a Katmai-core P3-500MHz, with 28 watts thermal design power;
more recent P3’s are much better, 10 watts at same speed!

I hope this article helped some people who want a fanless PSU and can’t afford
the commercially made fanless PSUs. If you have some questions, you can email
me, or better (as it would help other people, too) ask in the forum here on
SPCR (which I’m usually checking daily).

April 22, 2003 by Marko Djokic
(quix)
with Mike
Chin

* * *

Discuss
this article in the SPCR Forums

Google

www SPCR