Make a Fanless PSU with (Almost) No Soldering, for Less than $12
By Soldermizer (July 17, 2011)
I found Silent PC Review (http://www.silentpcreview.com
) while searching for fanless PSU reviews. I was looking for solutions that would help to quiet my modest desktop PC. It is my office/bedroom/music PC. The original unit produced too much fan noise. I formatted this post as a possible "article" but later discovered this huge "power supply" forum, and this seem the most appropriate place.
After reading some of the reviews and DIY articles, I concluded that (1) a truly fanless PSU was too expensive for my modest system and (2) that most of the DIY projects were too complex for my modest skills! This article, I hope, covers new ground in the sense of showing that sometimes a "low tech" solution will be adequate.
This is my desktop PC. It is certainly a modest system, and probably representative of many older PCs:
PowerSpec V102 (a clearance system at Micro Center in Dec. 2008; only the case remains; all other internals have been replaced due to failures -- believe it or not, not due to my incompetence, just bad luck!).
Motherboard: Biostar G31M+ (The cheapest that Comp USA had -- do you see a pattern developing?); I re-used 2 GB of RAM and the CPU from the previous motherboard (Intel Celeron D 2.8 GHz)
Hard Disk Drive = WD 320 GB (factory replaced the original 160GB that failed);
Boot drive = OCZ Vertex II 50 GB (that refuses to suspend or hibernate in my laptop, so here it is...);
Wireless network card;
The PSU I modded here is a Diablotek DA 250 watt PSU ($11.76 at Comp USA; not surprisingly also the cheapest unit.) Since the internals of each PSU will vary, the exact mod might be different. But I don't see why my simple method would not work with any PSU.
Let's face it, if you are like me, you scarcely have the mechanical aptitude to pick your own nose, much less perform sophisticated modifications such as found on these DIY sites. Lacking access to a machine shop, and being somewhat lazy also, I want a simpler method. Even if you are handy, sometimes an easy method is all you need.
Disclaimer: if you open the PSU case, you are exposed to potentially deadly voltages (and you void the warranty!) Please be careful. Also plaster of paris while technically not used in this mod, was something I used in my research, generates heat when curing and it is possible to get burns from it. A teenager once got third degree burns and had to have some fingers amputated from having her hand in wet plaster.
I did not want to do a complex modification, such as get a huge heatsink and mount the heat-producing components (e.g. MOSFET) on it. As you can partly see in the photo, my PSU has two heatsinks that are conveniently parallel and close to each other. Brief testing with the volt meter (VOM) showed that one of them was "live", having a few volts AC. So I could not use a metal heatsink. How, then, to conduct heat but not electricity? Prior experiments had shown that one thing that apparently did NOT work is using plaster (more on that later). So what you read here is attempt #2 and apparently successful! I thought that a small piece of rock, ceramic, or brick would conduct the heat away. True enough, but what to use? The most elegant solution would probably to have found a piece of glazed ceramic perhaps the size of a bar of soap. Perhaps a small brick or "paver" would work. I had some spare ceramic floor tiles, but they were far too big. While a larger mass would store more heat (a plus here), it would weigh more and the strength of the epoxy bond and also the weight supported by the hardware would be factors to consider.
Psycho-Ceramics, or A Crackpot approach to Heatsinks
While searching the kitchen, I found a ceramic tea cup that was already conveniently broken. I took a shard that was about the right size and saw that it would fit, joining the two heat sinks along and still leaving a contact area with the metal PSU case. I did chip the shard a bit smaller, to make sure it would fit inside the ATX case. Using regular Epoxy cement, the shard is cemented so the cylinder edge fully joins one entire side of each heat sink (not visible), as well as being joined to the rear of the PSU (visible).
The PSU has a fan that (of course) you should remove (otherwise it won't be fanless!) but there is a problem: the PSU has a fail-safe for fan failure. I was able to "fool" this protective circuit by simply soldering 110 Ohm load across the fan's power line, giving a dummy load. The fan's load was 0.1 A @ 12 V, so 120 ohm would have been ideal. This as showing in the next photo (top right), was the only soldering needed for the whole project. Of course, it could be done more elegantly, perhaps leave the wires longer when you cut them, or put them on the circuit board. Lazy, as I said ...
Once the Epoxy is dry, it is time to test the PSU. Put a test load (spare HDD) on one of the power connectors. Short the ATA pins 14&15 (I use paper clip); have VOM on a 5V or 12 V line, connect AC power, turn on the switch, and hope for the best!
Of course, this should give good voltages even with a normal PSU. The important question now is to see how much heat the PSU will generate, and will it conduct to the case? Lacking a thermometer except my fingers, I checked the shard and the outside case over a period of many minutes and it clearly was working.
The 2nd test is to install the PSU -- without its cover, for best airflow -- into the ATX case. I do not have the tools or patience to install the PSU the way others have that may well be better (e.g. heatsink up, or mounted on another surface, etc.) So I used the factory mounting position. Unfortunately, this is not optimal: As mounted, it is "upside down" and the heat sinks are directly below the PC board which of course heats up more than if it would be inverted. Still, the shard is going to conduct heat (enough of it?) to the PSU case, and via that, to the ATX case it is bolted to.
Now came the extended testing: I reconnected the power lines to the internal connectors. It is very important to monitor the temperatures, and the best tool I have is the hardware health read-outs in the BIOS. The first runs were done with the side panel off, which should provide extra cooling. The first boot up worked ok, but I was still getting a lot of noise. The immediate culprit is the 2nd case cooling fan. Disconnected at motherboard. Next test, the HDD is too noisy. So it gets unplugged (not needed, as the SSD is the boot drive). Now boot up. Amazingly quiet!
The only remaining fan is one I must leave in place: the CPU cooling fan. However, I tested even that assumption. My CPU has a maximum allowable temp of (about) 67°C. I disconnected the CPU fan and booted to the BIOS "hardware health" screen to monitor the temps. Of course, the temp in a matter of minutes climbs to the unsafe levels. I set the auto-temp shutdown to about 60°C for future insurance (which also works, perhaps worth testing it). For now, at least, I will have to keep this one fan. There are passive CPU cooling mods but they are too exotic for me, at least right now. I mean really, do you think I could afford a heat sink like this one, well it won't fit in my case judging from the photo:http://www.silentpcreview.com/article932-page2.html
Note that house temperatures during this testing were around 75°F, or about 25°C.
Initial testing: I left the side panel off. Unit was run all night, about ten hours, and never had a trouble. Temperatures did not climb above 41°C/41°C [CPU/SYSTEM]. No unhappy smells of burning plastic.
Further testing, with the side panel on, show that the temperatures stay well within the safe range. I have not seen a temperature above 50°C, and after many hours, it seems to be no higher than 45°C /45°C. With the PowerSpec case, there is a little duct on the side so the CPU fan draws in air. It is noticeably cooler there. As expected, the warmest part of the case is the top rear, directly above the PSU. Never was it too hot to touch. The many existing slots on the rear of the case (including the PSU fan opening) provide a "chimney" that naturally provides cooling air.
Speculation: as installed, the PSU's heat may lead to shortened life. Who cares? For $12, I'm willing to experiment.
Suspending the HDD: using ideas elsewhere from SPCR, I mounted ("precariously hung" would be more accurate) the HDD using rubber bands (4), paper clip (1) and pipe cleaners (2) to suspend the HDD so that it doesn't contact any surfaces of the ATX case.
Summary: the desktop PC was made substantially quieter by improvising a completely fanless PSU for under $12, the disconnection of the 2nd case fan, and suspending the HDD. Long term reliability is unknown. It is reasonable to assume the PSU's life is shortened because it runs hotter. While both the CPU and system clearly are warmer, the maximum observed temperatures are still well within the safe range (as far as I know).
Getting Plastered: Epic fail , or some methods that didn't work
Earlier I had tried literally "bricking a PSU." In my search for a heat but not electric conducting heat sink material, I had considered regular plaster of Paris. I took a working PSU, disconnected the fan, and filled the case with plaster. At this time, I did not know that the PSU have the fan failsafe. Also, when I powered up this PSU, it hissed and fizzled ... perhaps indicative of a short circuit. I have not tested this method further.
Test #2: the CPU gets plastered
I had a prehistoric (1998 vintage) Pentium-MMX PC for next test victim. Based on my above noted success, I desire to be rid of the CPU fan without having to resort to an enormous heatsink. Perhaps covering the CPU and its heatsink with a plaster mass would be an effective heatsink? Based on my single test, the answer is "no." The system booted but froze during the memory test. It would not boot at all following this. Case of failure is unknown, possibly due to electricity conducting where it shouldn't? Based on the PSU and this test, I pronounce plaster a failure, possibly because it's not an electrical insulator.
Ideas for future testing:
To use ceramic, rock, or other non-electricity-conducting materials would seem to hold much promise as a heatsink strategy. An interesting (and easy) research or investigation would be such ideas as:
1. Research common plaster-like materials for their thermal and electrical conductivity.
2. If a material looks promising (another kind of plaster? Cement? Concrete?) set up simple experiments to see if it be a good electrical insulator.
3. Use throw-away PC hardware for actual tests, like I did.
4. Could an entire PC be encased (say) in ordinary concrete or cement and still operate? This would be an excellent heatsink (if it conducts enough that the CPU doesn't die) and would literally be an "embedded system". The only downsides I see is that servicing the unit would be impossible and lifting it would be nearly so.
About the author
"Soldermizer" is a 49-year-old male who has dabbled in electronics and computing since his pre-teens. Professionally he has worked in telecom, computer networking and support operations. He has long had a love of home audio systems and has applied his -- uh -- talents to fixing up old gear for his own enjoyment or to re-sell. He is the guy who can buy a working 1960s vintage Arvin tabletop tube radio (FM stereo!) at an estate sale for $10, clean it up, and turn around and sell it for $2.50 on Ebay. His interests are highly eclectic, which is not to say talented. In just the past few months, he has tried his hand at such diverse hobbies as making custom ear pads for his headphones, home soap making and the present topic, silencing his noisy PC.