Cocoon Culture No.47



PSU (2) - Noise removal testing ( successful )

Based on the info found about capacitors and coils/transformers being the cause of PSU noise, the PSU (2) where dismantled to verify this new information. And from there on things evolved into an attempt at eliminating it.

The intention was to checkout some stuff on this PSU, before I moved on to the more precious PSU no.3. Find out where the problems are, what can be done, how successful etc.

At the beginning : measurements.

The intention was to use the mic. with the plastic tube to locate the sound. It was however quickly discovered that this didn't work as planned. The noise came from everywhere, in much the same way as with the HDD probing. (see elsewhere).

A long hollow plastic tube approx. 1m in length where used to locate the sounds instead. One great tip that I received was to use the shaft of a screwdriver or similar to push down the suspected component and see if the noise disappeared, or if there was any change. This worked great.

Although my super-tool wasn't very successful in locating the noise, it was useful for getting a representative recording of the noise characteristics. (see descriptions of recording elsewhere).

Opening the PSU : status

Inspecting the board and components, I found that glue had already been used in some areas. Particularly on capacitors. If you look real close on the picture Area A you can see the white stuff in the middle of the clusters of capacitors. Some glued together, some glued to the board. There where no glue on any of the transformers or coils however.

It was difficult do access all areas of the PSU. It was assembled in such a way that some components would have to be soldered of to gain access. The state of the world is also easily observed here. It is not created with love.

[ View detailed image of PSU ]

Locating the noise sources

As mentioned, it was difficult to locate the as it is composed of many different sources blending together varying with time and power load. The capacitors where quickly ruled out. And the main sources where found to be the coils in area A, and the transformers in Area B. The coils made a EEEE like noise, while the transformer made very annoying supersonic IIIIII like noise.

PSU-2-Area-B.jpg (3221 bytes)                   PSU-2-Area-A.jpg (6120 bytes)

Attempting to eliminate the noise

I first tried with some glue on the coils in Area A, this was successful. The noise was reduced. I went from there to pumping 3 sticks of glue at strategic points in the PSU, fixing it for good. I also added some marbles for mass, most noticeable effect on the high pitched frequency of the one transformer that I could reach.

When this was done I was quite happy with the result. The magnitude of the noise was reduced subjectively by 3 times. And the supersonic frequencies where close to eliminated. And what remained was a more bassy EEEE sound.

I also added a vibration-damping mat underneath the print-board. This was intended to improve stability and to dampen possible vibrations generated by components.

On disabling the Windows 98 Advanced power management, and the Intel controller something in the control panel, the noise level was reduced even more (3 times subjectively).


The glue-gun massacre worked very well. The noise problem is alleviated, but not eliminated. Since the steps that have been taken are irreversible and the PSU will never ever be disassembled again cause it's so full of glue, I might try some even more drastic steps like encapsulating entire sections in some material/substance.

One Idea at the moment is using plaster for this. Or maybe concrete is better ?

Update - 1

I've recently  learnt that plaster has poor temperature transferring abilities. And actually works as a heat isolator. Commonly used in walls, and also as fire-barriers. However it is a dead material, so I'll try this on a limited area such as the coils and capacitors. As the heat generated here is modest. I will not however use it on the areas with transformers. Need to find something with a better heat transferring qualities. Sand ? No. Maybe Wax ?

- The rest of area A has been covered with plaster, waiting 24hrs for it to dry before testing if it has had any effect. Oh and I also put some strange sticky tape like rubbery smelly stuff around the noisy transformer, wonder if it has any effect. It is some old stuff that nobody knew what was for, I intercepted it on the way to the trashcan ... think it has been used for in plumbing to wrap around pipes for some reason. But it was heavy, it was rubbery and it was sticky so I'll try it.

- Testing the plaster insulation after 24 hrs, made it clear that it had an noticeable effect. However : moist patches appeared on the PSU walls, needed 48 hrs to dry up completely. The rubbery sticky stuff also worked as a dampener on the transformer in area B. Subjective magnitude improvement approx. one time. Not much in the way of characteristics.

Update - 2

Taking the next step, to further try to dampen the noise. I thought wax would be a good idea. I melted some remains of candle-lights that I had lying about. Poured this on at different locations to try to dampen the noise-levels even more. Particularly the transformer in area B was coated. The results where great on the Higher-Frequency components of the sound, which where transformed from III into some more muffled lower EEE type noise. The noise-level in area A was also further dampened. Improvement by a magnitude of 2 times.

Note ! It was very difficult to apply the wax material accurately around. I had imagined that it would dry more quickly. Furthermore - candles are not made purely of wax, but "stearin" some "paraffin" material that is flammable. On second thought this wasn't such a great idea, as I caught the thought that gas might build up in the computer cabinet or PSU box and explode. But what is done is done.Pure wax would have been a better choice, but I didn't have any. In addition to this, one of the candles were of the type odor. Which means that  "perfume" is coming out of my PSU when it gets hot. Hopefully a passing thing, as I get sick of the smell. But to look at it from the bright side; I could now say that my PSU has a built in air-freshener.

After all this the noise is still not eliminated. I'm running out of Ideas. Last thing I can think of is to seal the whole-thing up with concrete. An increasingly attractive option know that I realize that I've made the PSU into a potential fire-bomb.

Update - 3

The PSU was encapsuled in a special concrete-like compound. Covering all vital areas of the PSU.
After 4 days of drying, the power-plug was inserted and the power-switch turned on. Some ugly uueeehhh sounds came from somewhere. They were of passing duration, and assumed to be the noise of some capacitors charging up or something. Still this was disappointing, as it indicated that the PSU is still not noise-less. Well after some hesitation around whether to move on or not, it was decided to try to connect MB power-cable. This resulted in some strange brumming noises coming from the audio output. ( PSU Power switch was in on position) . Decided to halt the testing of PSU, till some other time. Leave a couple of more days for drying at room temperature.

Update - 4 - Testing OK.

After a total of 11 days of drying (7 days at room temperature) a test system was put together to check out the results.

In brief, a total success with regards to the noise-angle. As to a fan-less operation of this device; there are some problems with the buildup of heat. Assumed operating range [0 .. 70 C], but tested up to a maximum of 55 C. With the waterfall cooling utility, a approximate stable temperature of 53 C where possible. Without testing stopped at 55 C ( still rising at a rate of 0.2 C/minute.). All this was done with a totally open system. no grounding, no shielding, no covers.
The standby temperature was measured to be 27.1 C, after 9 hours of testing. This temperature was reached 3 hours after mains power on.

Further experimentation with this device will be related to getting rid of the heat in a silent fashion. The outcome of from second custom-made cooler for the Celeron processor will provide new knowledge as to how heat can be removed in a silent way.

This document is closed, as the noise aspect of the PSU has been resolved. What remains now is the thermal/cooling issues; which will be discussed elsewhere; possibly in conjunction with the creation of the second custom CPU cooler. Or maybe in a separate document.

To be added [ Thermal measurements and test-results : First session ]

Note ! On opening the electronics in conjunction with the improvement of the cooling surface, it was noted that some of the wax in the region of a resistor close to the coil had melted and floated away. At what temperature does this happen ? What's the melting point for candles ? 60 C ? Also some of the wax around the coil have perspired/soaked into the surrounding coating material.

Thermal issues : ( this content will be relocated )

Planned modifications of  PSU(2) :

  1. Perforation of casing. Motivation for this is not clear, but maybe to improve the natural ventilation and airflow. And oh yeah ... to reduce the PSU fan noise. The difference is expected to be that of talking into a bucket with or without holes in it.
    F.U: This was very time consuming. There were a lot of holes to be drilled. But it t looks cool.
    F.U: From measurement it is also a important part.
  2. Increase cooling surface. Particularly around the coil.
  3. Experimentation with heat export. ( The CuCo Tube )
  4. Customization of thermal fan regulator. ( In conjunction with rest of cabinet ) Setting of operating rage.
    F.U. Based on the exploration of the noise-killer(1), hopes are low as to getting this thing to fly. Cancelled.
  5. Mounting of fan. Relocation to the outside of case. Mounting of noise-trap on this. Modified type given change noise-characteristics.
    F.U. Skipped. Uncertain if PSU would fit into cabinet with FAN mounted externally.
  6. Possible adding of 2nd fan to get a push-pull effect. Take head for this in perforation of case-cover.

Further testing (of thermal behavior) :

  1. With cover-on.  Externally. [ third session ]
  2. Mounted in cabinet. Temperatures with active or passive cooling.
  3. Evaluation of the different modifications.[ summary ]
    Problems with the materials used when overheating. Vaporizing/melting point unknown. When this occurs the HF noise dampening abilities are reduced.


PSU (2) - Measurement & Testing : First session.

Setup :

Using components as decried for system 1. Win98, ISA Video Card, HD, CD-ROM, Floppy, P233-MMX

Software cooler : Waterfall 1.23

Sensor accuracy : Lies in a hot-spot, so that the readings are representative for the PSU core temperature. +/- . Obtained using the digital thermometer with external probe.

Measurements done with the protective grid off. Upside down from normal position.

Measurement of PSU standby temperature :

Reaches a peak of 27 C @ 21.3 C ambient temperature after 2 hours. 5 hours later the temperature is 27.1 C @ 21.1 C(amb). Measurement started from ambient temperature. Reaches 26 C @ 21.6 C (amb) after about an hour.

Measurement of PSU idle temperature :

Running windows 98 with the waterfall 1.23 cooler software. All drives powered on, and running idle.

Note : measurement takes a flying start at the PSU temperature at the time. Pre-heated. Ambient temperature approximately 21.6 C during measurements.

Absolute time Relative time Probe temperature Rate of change (C/min) Comment
00:50 00 33.4 Power Up
01:03 39.8
01:10 42.1
01:19 44.6 0.27
01:20 44.8
01:50 49.0 0.1
02:00 49.8 0.06
02:20 50.9 0.04
02:30 51.1 0.02
02:45 51.7 0.04
03:10 52.2 0.03
03:19 52.3 0.01
03:30 52.4 0.01

Conclusion : Fairly stable below 53 C @ 21.6 C(amb) when system is running idle with cooler software.

Measurement of PSU load temperature :

Disabling the software cooler to find temperatures without. The limit has been set to 55 C, as I'm not sure at this point as to what operating temperature specs. the PSU have.

Absolute time Relative time Probe temperature Rate of change (C/min) Comment
03:33 51.4 Measurement continues in time from table above
03:34 52.6
03:36 52.9
03:38 53.2
03:40 53.6
03:42 54.1
03:44 54.4
03:46 54.8
03:47 54.9 Limit reached, enables s.w.cooler again. Test stopped.
03:52 54.6 Temperature drops
03:55 54.5
03:57 54.4

Conclusion : Effect of the SW cooler is noticeable, and helps keep a lower PSU temperature.

+++ add more detail.


PSU (2) - Measurement & Testing : Second session.

Setup :

As in first session. Improvements have been made in the heat transfer capabilities. Testing were stopped in the previous session because the temperature exceeded the self-imposed limit of 55 C.

Testing starts up with no drives connected. No software cooler (as from now referred to as SWC). Power on PSU and system set in standby mode.

Absolute time Relative time Probe Temperature (C) Rate of change (C/min) Ambient Temperature (C) Comment
18:50 0:00 20.0 20.0 Test started. System in standby mode.
21:20 2:30 25.5 21.2 Representative PSU standby temperature reached. But still rising slowly. Accurate enough.
21:24 0:00 26.0     Boot up. No drives or keyboard connected. System stuck in BIOS.
:26 26.9
:27 27.6
:28 28.1
:29 28.6
:30 29.4
21:40 34.2
21:50 37.9
22:00 40.8
22:12 43.4 21.9
22:15 43.7 0.2
22:20 44.5
22:40 46.6 0.1 21.7
23:20 48.2
23:45 48.9 0.03
00:10 49.2 0.015 approx. rate value
00:11 49.3
00:19 49.3 0.0125
03:43 50.6 22.4
03:48   50.7 0.0055   Stable temp.
04:30 Connected power to all drives.
04:35 50.4* 22.6 * Temp. drop due to power down up when connecting drives.
04:40 50.7
04:50 51.0 0.03
05:00 51.3
:10 51.5 22.7
:30 51.8 22.9
:50 52.1 0.015
06:00 52.2 23.3
:20 52.3 0.005 23.4
:40   52.1*   23.0 * Seem to have connected to the universe.
:50 52.3 23.1
07:10 52.3 23.3 Connecting keyboard. Booting up windows with SWC disabled.. Running defrag on HD.
:30   52.3   24.1 Stable temp
:35 52.3 Activating SWC.
:40 51.9
:45 51.3 0.1
:50 51.0 0.06
08:10 50.0 0.05 24.3
:12 49.9
:14 49.8
:17 49.7
:20 49.6 Testing stopped. Keeps falling at a rate of 0.05 C/min. Estimated stable temp 48 C. Verify sometime.


Key data :

PSU Idle temperature : Estimated to be around 48 C @ 23.4 C (ambient). But only verified down to 49.6 declining at a rate of 0.05 C/min.

PSU Load temperature : 52.3 C @ 24.1 C (ambient). This is running windows without SWC. All drives powered up (HD/CD/Floppy).

Conclusions :

After a testing for 13.5 hours, there visibility in the room is still clear and everything seems to be working fine. The modifications made to the PSU have been effective. The load temperature is now below 53 C, where as it exceeded 55 C before. How much I don't know. By comparing the results for the idle measurements (with SWC) and assuming that the temperature estimate is correct. The effect have been to lower the PSU core temperature by 4 to 5 degrees C.

The accuracy's of the measurement are as with all fanless cooling, very susceptible to   humidity levels. These where made on a failure dry winters day in January.


PSU (2) - Measurement & Testing : Third Session.

(Continuing of from session 2, another day with a different climate)

Setup :

The customize PSU cover has know been put on. Mounted with a blower fan (not active unless specified).

Measurements :

[ Table data have yet not been entered, add some other time ]

Key data :

PSU Idle temperature : 50.2 C @ 22.4 C (ambient). Running windows idle, with SWC.

PSU Load temperature : 57.3 C @ 23.0 C (ambient). This is running windows without SWC. All drives powered up (HD/CD/Floppy).

PSU Load temperature /w blower fan : 36.4 C @ 22.1 C (ambient). Fan operated on 6V static.

PSU Standby temperature : 27.3 C @ 21.8 C

Conclusion :

The PSU Load temperature with cover on was 5 C higher, so there is a major difference. It is possible that further modifications are necessary. The magnitude was not excepted.. But this proves the importance of a good airflow through the shield.


PSU (2) - Reality testing with system (1) in cabinet. : Fourth and Final Session.

After passing the external testing, the PSU where mounted in the original unmodified cabinet(1) for reality testing. There where no shield on the PSU. And the setup was altered slightly from normal. Use of different drives etc. After the initial testing, modifications where made to the cabinet as well.

A short story about the sparks encountered during testing. Add later.

Testing in the Original Cabinet :

Environment (Initial) : Cab. cover on. System running idle in windows98 with no ACPI but with waterfall software cooling. All drives always on. Video always on.

Testing stability of CPU, MB and internal cabinet temperatures. Stable at 32 C @ 22 C ambient.

Testing stability of PSU temperature at idle : 59.0 C @ 22.8 C : Cab cover on.

Load testing of PSU temperature stability. The "test" mode of the utility "CPU Stability Test V 6.0" where used for this purpose.
Test Failed. A temperature of 67.9 C @ 21.8 C ambient where reached within 2 hrs 30 minutes from boot up.This is with the unmodified cabinet. The criteria for failure is that it did not reach stability below 70 degrees.

Testing of stability system (1) under normal operation : (snap-shot of peak measurements recorded) :

  • Ambient temperature : 23.3 C
  • PSU : 61.7 C (observed range 58 .. 62 C)
  • Cabinet : 35.1 C (measure at the top drive bay)
  • CPU : Unknown but stable.

Specification of the equipment used during the testing :

Standard for system(1), but some drives where different :

  • HD (2) : Conner [type no] 540 MB mode 0
  • HD (3) : Conner [type no] 120 MB mode 0
  • CD-ROM (2): Brand : TUV, 8x, Model No. E2850UA, Made in Japan, January 1997, Serial No. B70676011
  • Floppy drive (2)  3.5 " : Sony

Measurement of system(1) Power Consumption :

Measurements during system boot :

  • Peak measurement : 0.3 A
  • In BIOS : 0.22 A
  • In Windows : 0.22 .. 0.24 A
  • In windows98 with SWC on : 150 mA

Measurements in the different modes of CPU stability test :

  • CPU Warming : 250 - 260 mA
  • Typical Test mode reading : 235 mA (210 - 300 mA range)
  • Fibonacci test : 245 mA
  • Prime test : 215 mA

Other measurements :

  • Current wall socket AC voltage : 235 V
  • Windows idle w/o SWC : 225 mA
  • System in standby mode : 27 - 30 mA
  • System Power switch off : 6 - 7 mA (consumed by net filter).

Summary of the system power consumption in its different states (+15% estimate error):

  1. Standby, 6 W
  2. Idle, 30 W
  3. Testing, 47 W
  4. Peak, 60 W

Missing the measurements for the individual devices. Do this another time. Entries here or listed in the specification of equipment used.

Average idle power-consumption : 150 mA, Well socket voltage 235 V.

Testing in the Modified Cabinet :

Some info about the modifications :
Basically just drilled some more strategically located holes to improve natural ventilation. Opened top-bay drive door. Removed the cabinet noise insulation here. It took me 3 working days to complete this. The original estimate was 4 hours :)

Testing stability of PSU temperature at idle : 52.3 C @ 19.6 C . Represents a 3 C degrees temperature drop (normalized) when compared to Original Cab.

Load testing of PSU temperature stability. The "test" mode of the utility "CPU Stability Test V 6.0" where used for this purpose.
Please note that one HD drive has been added in this test (120 MB Conner).
Stable at 63 C @ 21.3 C ambient after 3 hours of testing.

Conclusion : System (1) silenced and stable :) my-star.gif (1965 bytes)


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