Doug realized that to follow SPCR’s advice of starting with quiet components to make a quiet computer, he’d have to replace most of the noisy components in his computer. He decided this would be too costly, and opted for a different approach: Build a case using pine boards and a design to contain the noise, with a little advice from SPCR forum members. Doug’s successful DIY quiet system should garner a lot of admirers.
April 28, 2005 by Douglas Klassen
SPCR advises that the ideal way to achieve PC silence is to start with quiet components. What to do if you already have a PC with lots of expensive noisy components? Douglas Klassen faced this challenge and opted for an unusual solution: Build a case using pine boards and a design to contain the noise, with a little advice from SPCR forum members. His DIY project is nicely detailed here. Doug lives in Saskatoon, SK, and works for a small company involved in pro audio and acoustic consulting. He is currently looking into ways to silence his fridge.
– Mike Chin, Editor
I bought my current computer about two years ago. It’s an Athlon XP 2100+ on an nForce2 motherboard with integrated graphics. The old one was a PII 266, to give you an idea of where I was coming from. The first time I turned the new one on my first thought was "Wow that’s loud!" I then proceeded to ignore the noise for most of a year. Eventually, though, it started to drive me nuts, and I went on an initial round of quieting which brought the noise levels down to near those of the old PII. Unfortunately, replacing the PSU and the stock HSF unmasked an extremely annoying high pitched chattery hissing sound that came off the motherboard whenever the CPU was under load. In addition to the motherboard, the primary sources of noise were the Thermaltake PurePower 360 PSU, the Arctic Cooling Copper Silent 2 (non-TC, constant 2200 RPM) HSF, and the WD Caviar 80GB HDD.
Recently, I decided that the time had come to do something to bring the volume down further. However, the question of what to do was rather puzzling. Part way through my first round of silencing I’d discovered SPCR, and my reading of the resources here indicated that the ideal way to achieve silence is to start with quiet components, as case damping and the like tend not to attenuate noise levels as much as one would like. However, replacing all the noisy components in my case would have meant buying a whole new system, and the budget for that just wasn’t there.
So, I adopted a different plan: Building a wooden case. Wood, I thought, should block much more sound than the thin sheetmetal of an ordinary case, and by designing from scratch, good ventilation could be achieved without giving sound any easy escape routes to my ears. This, I thought, should cut dramatically into the volume especially at high frequencies, which would mitigate the worst of the hard drive whine and motherboard coil hiss. With better airflow, I could slow down the CPU fan, and since my design called for a differently configured PSU anyway, I would upgrade to a Seasonic Super Tornado.
Here is a complete list of the components in the system:
CPU: AMD Athlon XP 2100+
HSF: Arctic Cooling Copper Silent 2, stock fan later replaced with Panaflo 80L
Motherboard: MSI K7N2G-L
HDD: 80GB Western Digital Caviar
GPU: Onboard Geforce MX
CD-RW: LG 52X (GCE-8520B)
DVD-R: LG 16X (GDR-8163B)
PSU: Thermaltake PurePower 360, later replaced by Seasonic Super Tornado 400
The design revolves around airflow. The intakes are situated in the bottom of the case, with the exhaust pushed out the back. The sides, front, and top are sealed up against noise leaking. To escape via the bottom intakes sound has to bounce off the carpet, which should absorb most of it. Sound escaping with the exhaust is traveling away from the user and less worrisome. The design also calls for separate airflow for the PSU, which necessitates it being placed at the bottom rear of the case. The exhaust for the rest of the case is provided by a single 120mm fan. This isn’t really a revolutionary layout, but the options are fairly restricted if one wants all the I/O jacks at the back. Anyway, given its placement in the room, a case with a roughly conventional tower shape was desirable.
One issue that came up during the design phase of the project was the resonant nature of wood. A few experiments running fans on my desktop convinced me that this could be a significant problem, but one without any really definitive solution. The best option seemed to be to plan to isolate all moving parts from the case, using various sorts of foam and elastic to hold things in place.
The final feature of the design is an acoustic absorbing panel made from 1" compressed fiberglass to be fitted to the inside of the left side door. A panel of this sort effectively absorbs sound down to around 1000 Hz, and should help to muffle the high frequencies coming off the motherboard and hard drive. Even though the panel only covers one side of the interior, sound will have a tendency to bounce around inside the case until it encounters the panel, at which point it will be converted to heat. Or so the theory goes.
I decided to build the case out of pine boards, principally because I had already made a few pieces of furniture from pine, including the computer desk. To make things easier, I opted to cannibalize the motherboard tray and 5.25" drive rack from my old case.
The first few pieces tacked together.
Here we see the first few pieces tacked together in order to establish where the holes in the back needed to be cut. It was a bit tricky – the case needed to be partially assembled to accurately tell where the rear I/O port would be, but I needed to cut the holes before putting anything together.
Here the basics have been assembled – intake vents on the bottom, exhaust and I/O port on the back. The top hole in the front is for optical drives, and the bottom for a switch panel. The top half of the bottom hole also lines up with the bottom slot in drive rack, so it could someday house a fan controller or whatever. I used fairly simple joinery techniques, mostly butt joints, screwed and glued. Most of the strength comes from the glue; the screws act as clamps while the glue sets, but only contributes a bit to the strength of the joints. Where screws would be visible on the outside, I used cleats instead of screwing into the ends of the joints. You can see some cleats in the above picture at the top of the motherboard backing, ready to be fastened to the top of the case.
Primary assembly complete.
All that remained was to mount the doors, and that had to wait till the finish was applied. Note the feet holding the bottom up off the ground with long cutouts for unrestricted air intake. I wanted the side doors to close snugly, so I put in backing strips along the top and bottom of the opening, to which I later glued strips of felt.
I used the same wood stain as I had on my computer desk, covered with multiple coats of clear Varathane. As you can see, I had taken a router to the edges of the case.
Time to get some finish on.
The side door is constructed of boards glued edge to edge. They’re actually a bit of a design weakness, since pine has a bit of a tendency to twist itself into odd shapes and there was a significant chance they wouldn’t end up really flat. They came out fairly well, though I did end up pulling one of them off after the whole thing was together because the very front edge was twisting out and not fitting right. Some planing and sanding put an end to the pine’s insubordination. To do it right, I probably should have run the doors through a thickness planer, but I don’t have access to one. I could also have put some sort of ribs on their backs to hold them straight, but I didn’t really have space to do that.
Here I’ve got the right side door on.
In the center of the picture you can how it is held closed at the front by wingnuts. That proved to be a mistake, as it was extremely difficult to get at the top wingnut after the 5.25" drive rack was in place. I eventually replaced the top wingnut with a catch of the same sort I used for the left door, which can be seen in the extreme upper right of this picture. On the right the switch panel can be seen. It’s a bit of a hack job, with just a pair of single pole momentary contact switches, and a couple holes for the power and drive LEDs I stole from the old case.
I taped the LEDs against the back of the holes with an adhesive strip designed especially for the purpose. Oddly, the adhesive was marketed under the name 3M All-Weather Duct Tape. Go figure. The final empty hole visible is for a multi-position rotary switch wired up with diodes for variable fan speed control. I have vague plans to build some more complex custom fan control circuitry, at which point I’ll replace this switch panel with something nicer, but for now this will do.
MOUNTING THE COMPONENTS
The drive cage is screwed into the top of the case, with some foam strips between the cage and the top.
This doesn’t really isolate the cage from the case, as vibrations can be transmitted through the screws. The foam does make it easier to align the drive faces with the hole in the front of the case, though. You can also see a strip of felt tacked between the drive faces and the case to prevent the drives from vibrating directly against the case. So far the lack of complete isolation between the optical drives and the case hasn’t proven to be a problem. At full speed they are noisy, but there hasn’t been excessive resonance, and at lower speeds they are very quiet.
The Super Tornado 400 is strapped into place, resting on foam strips.
400 watts is severe overkill for this system, but I’m hoping that the PSU will survive through a couple upgrade cycles so I went with extra headroom. The hard drive was suspended on elastic cords strung on a framework built for the purpose. The old 4 GB Seagate seen here was just a placeholder.
The motherboard tray is screwed directly to the case, and the motherboard is mounted.
My 120mm exhaust fan was slow in arriving, so I wedged a 92mm fan into the exhaust port as a stopgap measure. You can see the strips of felt glued onto the backing for the door. All that remained was to build and mount the absorbing panel.
DAMPING & FINAL ASSEMBLY
The fiberglass was cut to size, and then covered with an open weave fabric using a spray adhesive.
The covering was important, as it would be rather unpleasant to have fiberglass dust everywhere on the inside of the case. The completed panel was attached to the inside of the door with wood screws.
The 120mm Nexus finally arrived. The foam block was glued to the case, and the fan held in place against it with an elastic cord.
The finished product. I’ll explain the panel sitting at the back in a bit.
Behind the front door.
The zip drive is just a placeholder again. I’m thinking it will look really sharp if I put in a pair of black optical drives and do up a nicer switch panel. As you can see, the front door actually hinges on the edge of the right side door, so if you open the right side up the front door has to go with it.
A view of the back.
Care has to be taken plugging the cord into the PSU, as it’s not held down very solidly. Obviously that would be an issue in transport, too, so I’ll have to tie the PSU down more securely if I plan on moving it about.
Yes, you say, it looks very nice. But how well does it work?
Upon firing it up the first time, I immediately noticed that I hadn’t been successful in eliminating resonance. The case was emitting quite a noticeable hum. A little experimentation (consisting of pressing my ear against the case and stopping the CPU fan with my finger) showed the vibrations were coming from the CPU fan. This wasn’t terribly surprising, since it wasn’t isolated from the case. There had been no good way to mount the motherboard tray besides screwing it in solidly. So, I pulled the Arctic Cooling fan off the heatsink, and replaced it with a Panaflo 80L held on with elastic cords and isolated from the heatsink with foam standoffs.
CPU HS fan mechanically decoupled
with foam and elastic cord.
As an aside, the little black cube seen at the bottom left between the CPU and the two small heatsinks is the source of the motherboard’s hissing. I imagine it’s an induction coil of some sort involved in powering the CPU. In any event, decoupling the CPU fan improved the resonance issue immediately.
Somewhat oddly, the Panaflo was keeping the CPU substantially cooler than the Arctic Cooling fan had been. I say oddly because the Panaflo’s rated airflow is lower than that of the stock fan. I’ve concluded that the difference was likely due to a combination of a better thermal interface (i.e., the right amount of thermal goop) and washing the dust off the heatsink while I had it off. In any event, I was able to undervolt both the Panaflo and the Nexus. Since I didn’t have my fan switch finished, I used the 7V difference trick on both fans, and recorded some temperatures. Case temperature was taken by means of placing a thermometer inside near the top. I think it shows a little high, since it always says my apartment is 24-25 degrees but it never feels that warm, but anyway. System temperature is that reported by the motherboard. I have no idea where that probe is.
Wood Case System Temps
Ambient temp was 24°C
As you can see, the temps are quite acceptable. I tried dropping the Panaflo 80L to 5V, which resulted in CPU temps of 52°C idle and 64°C under load, but since I couldn’t hear the fan at either speed, I moved it back up to 7V.
With the exhaust and CPU fans at 7V, the new case is dramatically quieter than the old case. From my chair, the fans are essentially inaudible, and I can only just make out the chatter of HDD seeks and the motherboard hiss when I load the CPU. In the old case, simple flash banner ads would be enough to drive me bonkers, as they’d load the CPU enough to induce the dreaded hissing. Ironically, this used to mean that I’d rarely keep the SPCR front page open since a few of the frequent advertisers love their flash. Now it’s just not an issue. The case appears to work exactly as intended. It’s virtually inaudible from the sides and front. If you stick your ear right down at the bottom, you can just make out a bit of sound leaking through the intake ports. There is a fair bit of sound coming off the back of the case, though. The motherboard (when it’s hissing) and the hard drive are predominant, but the gentle whirring of the Nexus can also be heard. This sound can be heard from my chair, as it tends to bounce off the walls and come back to me.
But enough of these subjective descriptions, let’s see some objective data. My employer was kind enough to allow me to borrow his SPL meter, a B&K 2237. Unfortunately (or fortunately, depending on how you look at it), it wasn’t sufficiently sensitive to register a reading at the standard 1m distance, having a minimum sensitivity of 30 dBA. So, we’ll have to do some extrapolation of the data I did manage to gather.
The difference between idle and load is strictly the addition of the motherboard hiss. The fans were turning at a constant rate the whole time. I couldn’t get a reading until I was right up against the case, except for at the back or with the side door open. I couldn’t get any reading from the front at all.
A weighted SPL Readings w/ 30 dBA limit Sound Level Meter
There’s a fair bit of guesswork involved in estimating what the SPL at 1m at the front and sides might be. We could just duplicate the difference between the 1m and 0.01m readings seen in the rear and open measurements, but this is problematic in that the close up readings were highly variable depending on where precisely the meter was held. Still, a difference of around 10 dBA seems reasonable, which would put the 1m figure at ~21 dBA for the side and a couple dBA lower for the front. I think that this is quite remarkable, since as the readings with the open side show, that WD hard drive is a noisy little thing, and yet it simply can’t be heard most of the time. [Editor’s Note: Any measurement device used at the limit of its capability is usually subject to higher than normal error. This SLM was definitely at the limits trying to read below, say, 33 dBA.]
I tried readings on a few noteworthy variations. Removing the absorbing panel from the side door resulted in levels 1-2 dBA higher all around. The subjective difference is rather greater than this, and I’m sure a graph showing sound level by frequency would show a much greater difference in high frequencies.
Spinning the fans up to 12V results in a substantial hum, substantial enough that readings at 1m became possible: 31 dBA front, 33 dBA side, and 35 dBA rear. This is due to the fact that the right side door begins resonating as the Nexus is brought near to full speed. I had thought my foam block/elastic strap mounting technique provided good isolation, but apparently it doesn’t, since the resonance drops off if the fan is pulled away from the foam. Since my temperatures are fine with the fans spinning slower, this isn’t much of an issue, but it’s clearly something I’ll have to deal with if I should ever need a lot more airflow for some upgraded components in the future.
Cost in CA$
The total includes a set of sanding drums for use with a drill that are nowhere near used up, and some sandpaper I didn’t use. However, that’s counterbalanced by the fact that I didn’t have to buy screws or any finish, as I had both left over from previous projects. If the switch category looks a bit high, it’s because two thirds of it is for the rotary switch and diodes for my rudimentary fan control.
My computer is now below ambient noise levels most of the time. Most of what little sound it was producing I wiped out by propping that acoustic panel I promised to tell you about behind the case. It’s far enough back to allow the exhaust to escape freely, but absorbs the great majority of sound coming through the rear ports.
During the design process, I encountered all kinds of worries about building a case from wood. Questions about EMI/RFI, resonance, and cooling came up. Incidentally, some playing with an AM radio demonstrates that the computer is indeed radiating some RF noise, but my LCD monitor generates more. I tried sticking my cordless phone and my cell both right next to the case, but didn’t get any static on either. I consider that case closed. Cooling, as I think those who frequent SPCR know, isn’t really dependent on case material. As the debunking of the aluminum case cooling myth shows, what matters for cooling is airflow, and this case has very adequate airflow.
The resonance of wood is an issue, but it’s manageable. Even with fans and hard drives decoupled from the case, some resonant frequencies can be amplified. I have not found it difficult to work around that so far, though, and the advantages of wood in sound blocking relative to sheetmetal more than compensate. This last is important, as it has allowed me to effectively silence my machine even though it still contains a couple of loud components.
Would I do anything differently if I had it to do over again? Well, yes, I would. I’d make the power led visible from the outside without opening the front door, because I can’t tell when the computer is on. Other than that, I’m pretty happy.
Many thanks to those who contributed to my design thread in the forums, even if I did ignore some of the advice.
* * *