Cocoon Culture No.47




[ index ]


Measurements performed using YF-22 Sound Level Meter (range 30-80 dB) Fast response. All noise measurements are performed by using minimum level hold for 5 to 10 minutes.

Sound clip recordings done with a SHURE 16A condenser microphone through a sound-track 97 pci card.

Abbreviations used in tables etc.:

WA Work Area - Approximate 1 meter from the left side of the computer tower. Meter positioned on desk, immediately to the left of the keyboard.
NC Near Cabinet (reference point) - 1 cm from the case cover, in the center of the left side.
NK Noise Killer
Rb 250 Ohm
Rmin 1.5k Ohm ( in combination with noisekiller 60 Ohm)


Measurements part 1

Measurement of noise levels with initial setup and optimal conditions.

PSU Fan trimmed down to minimum rotation using pot.meter (1.5 k Ohm) , 12 Volts. Factory Noisekiller bypassed.
CPU Fan trimmed down to minimum rotation using resistor (250 Ohm), and voltage reduced from +12v to +5v
Cabinet covered in woolen-blanket, blister packaging inside and some foam on one side.
Hard-drive spinning

Table 1:

Scenario PSU CPU HD WA (dBA) WA (dBC) NC (dBA) NC (dBC) Sound-clip Comment
M-1 Rmin 5V+Rb On 30,2 51,2 n/a n/a n/a  
M-1.2 Rmin 5V+Rb On 29,9 35,4 40 40 n/a *(2)
M-2 Rmin 5V+Rb On 32,4 48,6 49,1 58,6 n/a *(3)

*(2) Measurement M-1.2 performed later on at 18:00 same day, gives some idea about the inaccuracy of the measurements. Average noise-level immediately around cabinet, ca. 40 dBA

*(3) Measurements done with open cabinet, no blanket and no padding. Measurement inaccurate, to much background noise.

Observations so far :

The main noise-source is the CPU Fan. Which appear to be unbalanced. Secondly the hard-drive with an marrow-piercing high pitch whining noise.

The most annoying sounds are the high pitched ones. The low-frequents are easier to live with.

Aim is to reduce fan-rotation to an un-noticeable level. And use the noise killer as a safety measure. Eliminate the need for a fan, by improving on natural ventilation. Open cabinet.

The above ideas where not possible, due to the poor function of the noise-killer.


Measurements - Part 2

Note! All sound-clips have first been amplified by +20dB's by mic.amplifier, and then +48dB's by cool edit 96. These clips are heavily distorted and inaccurate, but serves to give a impression as to the change in noise perception. All recordings in this part (2) are also comparable to each other.

Measurement under optimal conditions with Noise Killer installed.

Scenario PSU CPU HD WA (dBA) WA (dBC) NC (dBA) NC (dBC) Sound-clip Comment
OPTIMAL 1 NK+Rmin 5V+Rb On 31,7 39,9 50 50 clip-1 *(1)

*(1) Without case-covering, Rb=CPU fan resistor (on old fan), Rmin is PSU pot.meter set to just keep the fan turning (approx. 60 Ohm) Fan used is old 12 V fan, working on 5 volts, and with resistor. Noise-Killer operating on +12v

Measurement of worst case :

Resistance on PSU fan removed. Resistor on CPU fan also removed and voltage changed from 5+ to 12+ volts. Open chassis.

Scenario PSU CPU HD WA (dBA) WA (dBC) NC (dBA) NC (dBC) Sound-clip Comment
WC-1 12v 5v+Rb On 41,3 45,3 56,4 59,6 clip-2 *
WC-2 12v 5v On 42,0 54,6 58,1 62,0 clip-3  
WC-3 12v 12v On 40,0 54,2 55,8 61,1 clip-4  

* Direct Sound of PSU fan operating at full speed is in the range 62dBA/65dBA +/- 2dB (Measured approx. 5 cm from the side. out of the airstream)

* Drop in noise level on CPU cooler fan, when operated on 12V. Or at least change in sound shape/character.

With original case cover on

Scenario PSU CPU HD WA (dBA) WA (dBC) NC (dBA) NC (dBC) Sound-clip Comment
WC-4 12v 12v On 40,9 46,5 52,3 56,5 clip-5 *

* Measurements reads lower. But noise have changed character and appears higher, more concentrated, coming from one direction. The fan house.

Replace old fan with new (two voltage fan 5v or 12v) and inserting new noise killer.

Scenario PSU CPU HD WA (dBA) WA (dBC) NC (dBA) NC (dBC) Sound-clip Comment
NEW-1 NK 5v On 31,5 54,0 48,1 59,3 clip-6 *(1)
NEW-2 NK+Rmin 5v On 31,3 35,2 47,7 46,7 clip-7 *(2)
NEW-2.1 Off Off On 30,6 50,4 45,0 57,4 clip-8 *(3)
NEW-2.2 NK+Rmin 5v Off 28,9 34,3 36,4 40,5 n/a ** *(4)

* Factor affecting the measurements prior to and including NEW-1 - A storm outside. Redo measurement and recording for NEW-1 another day, when the storm has passed. (As noted from the measurements below, particularly the dBC recordings are biased in the above measurements.

*(1) NEW-1 No resistor on PSU Fan in measurement. Operated on 12 volts.

*(2) NEW-2 Measurement same as NEW-1 but with resistor on PSU fan. Open cabinet. ( Also note that the dBC measurements are more realistic than in NEW-1)

*(3) NEW-2.1 Just the sound generated by the hard drive. All fans off.

*(4) NEW-2.2 Just the sound generated by the fans(CPU+PSU). ** Sound clip unavailable as HD starts up on record.

Summary so far :

The main noise source has now become the hard-drive. When the hard-drive is powered down, it's livable. Fan noise isn't that bad anymore.

Note. A. Measurement accuracy. It appears that particularly the dBC measurements are   sensitive to background noise. And since they measurements are done in a real environment. Accuracy outside the time-frame of the actual measurement is low. It appears that the dBA measurements are more consistent.

Note. B. New sound observed : Now where does that annoying resonating sound come from. Period ca. 0.8 s, varying intensity. Unplug fans. Source most likely hard-drive. Examine the noise level of just the drive. (When the storm is over).


Now. attempting to insulate and relocating the drive.

First attempt :

There is a high whining noise coming from the drive. There is also the low-vibrations in the chassis. Removing drive from its 3.5 HD slot and see if any noticeable change appears.

Machine froze for some reason during the measurement. Windows protection error. Noticed that processor temperature was higher than normal.

Scenario PSU CPU HD WA (dBA) WA (dBC) NC (dBA) NC (dBC) Sound-clip Comment
HD-1 x x Floor 29,6 32,9 42,0 45,1 clip-9  

Second attempt :

Now. try to mount the drive in 5.1/4" chassis, with sound insulating cotton padded around it. Another approach might be to insert some kind of rubber suspension to try to stop the drive vibrations from spreading into the chassis. It appears to function like an old mechanical gramophone player. Where the drive is the needle, and the chassis is the speaker.

Scenario PSU CPU HD WA (dBA) WA (dBC) NC (dBA) NC (dBC) Sound-clip Comment
HD-2 x x Floor 29,5 39,3 40,7 43,4 clip-10 *

The drive mounted in second-most bay from the top. Partially padded with cotton. top.bottom.sides. But not completely, left breather hole open, as well as 1/3 of back half of the drive. Some improvement in noise-level, but not much. The subjective noise level appears higher than when the drive was on the floor. Chassis vibrations is also still present.

The noise-padding and relocation of the hard-drive did improve some as compared to the original placement. But there was less noise when it was located on the floor.

As the drive noise has been reduced, this marrow piercing whining noise that comes from no moving parts becomes more apparent. (around 8k Hz or above)


Now Locating the high-pitched sound. Using the microphone as a probe.

Noise immeasurable - but located to the Power-supply. Will detach it and try to locate the component causing the noise.

The noise is located to the cooling-ribs, or the transformers just below. It was not possible to determine exactly but most likely one of the three transformers/coils. An interesting observation is that this noise disappear, as the hard-drive surge power for start up. And returns as soon as the drive is up and running. Sample of noise recorded in the file x-noise-probe.wav (44.1 kHz sample).

Not much that can be done about it. Change power supply, change housing, change everything.


Frequency-spectrum analysis of the noise.

Well a lot of hassle setting up the microphone for the computer. and the noise measures through the microphone are very unreliable. But maybe there some information in them. Try to normalize and amplify the recordings to an audible level.

Amplified the original recordings by 48 dB. From the frequency analysis some peaks can be identified, but not any useful information derived from it.



  • Relocation of the HD from the lower 3.5" position to the middle 5 1/4 " reduced noise.
  • Changing the CPU fan, for a 5V specific, reduced the noise level and characteristics. Of all the changes done this was the single most effective. The old fan was worn out.
  • The Noise Killer did not meet the expectations. However, it made the whole setup more reliable and tidy as compared to the use of resistors and potentiometers.
  • Removing the PSU fan shield made the noise characteristics more annoying.
  • As levels are reduced on the major noise sources, previously masked sources emerge.
  • PSU electronics generates high-pitched noise.

The high frequency sounds are not easily noticeable at first. But they are really annoying when you become aware of them. Like TV-tube sounds. They burn an impression into your head that last for some time afterwards. Like strong light. You see it even after the source has been removed. Causing sub-conscious stress.

On installing the noise-killer over the old noise-killer, it was found that the old actually worked some. Reducing the fan voltage to 10.2 Volts.

The noise generated inside the chassis is reflected and amplified inside. The inside surfaces are reflective and noise generated internally is not absorbed and contained within the cabinet, but transmitted outside. It's like having rats running around in a ventilation shaft. If the inside was covered by wall to wall carpets for instance, you wouldn't notice their presence so well. Also metal-pipes of this type, where used in the old days to communicate from the bridge to the engine room of boats. It's a tragic material to use for sound insulation. It' amplifies the noise problem.

A low frequency vibration, such as that from transformers and drives, can generate a high-pitched resonating sound at higher frequencies.



At the end of the whole experiment my video-card exploded. Maybe cotton isn't such a great insulation material. I now suspect that it builds up static electricity quite easy. Anyway cotton as damping material has been abandoned.

So final measurement with everything boxed and sealed. (And a new video card)

Scenario PSU CPU HD WA (dBA) WA (dBC) NC (dBA) NC (dBC) Sound-clip Comment
Final NK 5v Off 29,6 34,5 32,6 40,8 n/a  
Final-2 NK 5v On 30,1 42,8 33,0 41,2 clip-11  

At the end the noise level is as before the experiment started. Possibly a little higher, since the NK operates the PSU fan at a slightly higher rotation than when resistors where used to trim down the speed.

Although there isn't much of a change in the measured noise level, the noise characteristics have changed to a more comfortable one. I also know more on the subject now. And will explore other further means of eliminating the noise.

The most significant improvement is the change in noise characteristics of the new fan. The fan was constructed for operation on two voltage levels. It operated more quietly than the old fan.

The relocation of the hard-drive from within the chassis also improved the noise-level, but did not eliminate it.

Concludes now by wrapping the whole thing up. Chassis is condemned. PSU generates noise. Chassis amplifies the noise. The original noise is reduced to a acceptable level.

Further exploration :

Elaborate on the exhaustpipe noise damping for the PSU fan. Find a way to eliminate the resonance in the PSU. Using foam. (I notice that quite a lot of silicon have been used to glue the whole thing together. Could it been that it is an attempt to solving the transformer generated noise ?). Rubber suspensions for print-card ?

Also to build a noise cage-around the whole chassis as a last-resort. Try to relocate the drive in a separate box. Problems with heat dissipation ?

Another alternative is to use ear-plugs. Or a good headset that shuts out the noise.

Improve on the cache usage. I now have 160 MB of memory, should be possible to totally eliminate disk access once the programs have been loaded. How to keep everything in memory if possible. Programs. Search for. Set-up/configuration of vcache. How to ?


[ index ]

Copyright 1999-2002 Cocoon culture no.47 . All rights reserved.
[ ? ] [ Top ] [ Back ]
Hit Counter