Fan noise measurements

NotSilentYet · Post by **NotSilentYet** » Tue Jan 14, 2003 5:46 pm

Today I measured the noise produced by different fans. The noise was measured using a Bruel&Kjaer 2231 SLM. This is an instrument in the $6000 range so I'm thinking it should be fairly accurate. What is not so accurate is the environment in which the measurements were performed ( my apartment). I redid some of the measurements to assure myself that the measuring error caused by switching fans didn't affect the values too much. The remeasured values were within 0.5-1 dB so they were fairly accurate.

I placed the fans ontop of some foam standing 10 inches tall. The SLM was placed at an equal height. I did the measurements with the microphone on the SLM pointed at the side of the fans to avoid having the microphone in the air stream. The instrument was configured to measure using an A-weighting and to measure LEQ which is basically an average. Measurements were done at three different voltages 12 V, 10 V and 7 V. The distance between the fan and the instrument was 0.125 m. This short distance was chosen to get more reliable measurements as the fans are all fairly quiet. To get an approximate equivalent value at 1 m subtract 18 dB from the measured values. All values are compensated for the background noise in my apartment ( 21 dB). All values below 30 dB is a little less accurate. The fans I measured are Papst 8412NGL (12 dB), Papst 8412NGML (19 dB), Panaflo L1A (21 dB), Papst 3412NGL ( 92 mm 23 dB) and for fun a Coolermaster 92 mm.

There are two measurements for the Panaflo. This was because my Panaflo fans are 1 year old and my Papst fans are brand new. I thought the Panaflo was a bit too noisy so I lubed it by dripping some teflon oil onto the bearing. I'll let the results speak for them self. I even measured all the values for the lubed Panaflo twice measuring on one of the Papst fans in between to see that those values were still the same but the results for the Panaflo came back the same. Don't take these measurements too seriously because there is probably alot of things i haven't taken into consideration.

Post by **MikeC** » Tue Jan 14, 2003 6:26 pm

NotSilentYet --

Very intersting! Good to have someone else with another SLM -- and that seems like a good one. A bunch of questions --

1) 21 dBA ambient noise in your apt sounds REALLY low. I have not had access to an SLM that goes down that low in the house, but I doubt my house gets that low until maybe the wee hours of the morning. What is the lower limit of the B&K?

2) To get an approximate equivalent value at 1 m subtract 18 dB from the measured values. I don't know why you say this? It doesn't sound like you calibrated for the difference between the 0.125m and 1m readings...

3) If the ambient noise really is 21 dBA, you would probably have been ok to measure the 12V noise at 1/2 meter or 1/4 meter. By using several louder noise sources (Coolermaster fan and others like it) as reference, calibrate the dBA difference in you setup between the near field measurement (1/2 or 1/4 meter) against the 1 meter measurment. This difference should be fairly consistent. Then you could say "subtract __ dBA to obtain the approx. 1M reading".

Does the above make sense?

BTW -- I presume these were non-clicking Papst?

NotSilentYet · Post by **NotSilentYet** » Tue Jan 14, 2003 6:58 pm

1. I live in sweden so the measurements were taken between 00:00-02:00. I don't know the lower limit of the B&K and it doesn't say so in the manual either. The manual states that it has a dynamic range of 60 dB and then you configure the maximum value to be measured. In this case i chose 60 dB as the maximum. It did come with a 20 dB attenuator that is placed between the microphone and the instrument. It is designed to increase the range of the instrument from 130 dB to 150 dB. I tried putting it on when i was measuring ambient noise and it read 8 dB instead of 21. So perhaps this is the lower limit.

2. This number was taken straight from the theory in the book that was included with the SLM.

3. Measuring the 12 V noise at 1/4 meter was no problem 1/2 meter is probably ok. I just thought it would be interesting to be able to compare the values with each other. Going further away than this would compromise the accuracy of the measurements. Doing these measurements is extremely tedious as I have to turn everything in my apartment off. Another problem is that I live pretty close to the railway and the trains completely screws up the measurements. So I measure every value several times to see that they are consistent.

Calibrating my measurements by measuring at different distances sounds like a good idea. I'll try that next time I'm up late and feel like doing some measureing.

Post by **MikeC** » Tue Jan 14, 2003 7:25 pm

Doing these measurements is extremely tedious as I have to turn everything in my apartment off.

Yeah, tell me about it!

Well it sounds like maybe 21 dBA was an accurate reading. My house is very quiet most of the time, and super quiet at night, but I have been told it is still unlikely to be below 20 dBA. One day I will have to try a "real" SLM in here.

Thanks for your efforts, I am sure others appreciate it too.

The reason I suggested the 1/4m : 1m calibration is because I am always interested to test the claim of the manufacturer. They usually measure at 1M, with the fan suspended in free air, but I think many are incorrect, sloppy or simply exaggerating. One or two samples does not prove anything, but it does give us an idea.

PS -- interesting how the Panaflo quieted down with the lube job.

hyum · Post by **hyum** » Tue Jan 14, 2003 7:54 pm

now if someone had access to a high-end SLM and an anemometer (with a duct to smooth out airflow to measure CFM at different voltages in free air).

NotSilentYet · Post by **NotSilentYet** » Tue Jan 14, 2003 8:13 pm

I wonder how the manufacturers measure the flow. Because looking at their datasheets even the slightest increase in pressure on the positive side reduces the flow. By placing a duct on either side of the fan you would introduce a pressure drop that will make the measurements less accurate. I know we have an instrument at work that uses a hot wire that you place inside the air flow. This will cool the wire down and by measuring the temperature drop it will give a reading of the flow. We use this to measure flows at around 4000 cfm so I'm not sure if its capable of measuring these tiny flows that we are talking about here. I'll check with the guy who usually do these types of measurements and see if he knows of a way to measure small flows.

Post by **MikeC** » Tue Jan 14, 2003 8:34 pm

Fan makers invariably cite airflow against free-air impedance. If you look into any curves provided, you'll see more info, but many don't provide the pressure / airflow curves. To test for airflow, a standard test "plenum" ot jig is required. This is clearly specified by several fan measurement standards. The construction and testing of this jig is part 1 of the UBC Fan Research project. (Which I need to chase up on)

loply · Post by **loply** » Wed Jan 15, 2003 3:29 pm

One thing I was wondering about recently, is it not worth testing the noise of fans inside a PC, rather than hanging independently?

Surely there could be certain fans which vibrate like mad when attatched to a chasis but are quiet otherwise.

Id like to see a conclusive, be-all-and-end-all analysis of the most common fans which makes real conclusions about their noise levels.

NotSilentYet · Post by **NotSilentYet** » Wed Jan 15, 2003 5:41 pm

loply wrote:One thing I was wondering about recently, is it not worth testing the noise of fans inside a PC, rather than hanging independently?

Surely there could be certain fans which vibrate like mad when attatched to a chasis but are quiet otherwise.

Id like to see a conclusive, be-all-and-end-all analysis of the most common fans which makes real conclusions about their noise levels.

Well it is possible. But it would make a big difference depending on the case, from which direction measurements were taken and so on. So it would still not be the end of it. Then I figured it seems most people use some sort of damper between the fan and chassis so vibration has less impact on the absolute sound level. What I learned from these measurements is how difficult it is to measure sound level in an accurate way. Moving the instrument or placing it at a different angle makes a big difference in measured value. This is probably why the manufacturers use a controlled environment in which to do the measurements. It's the only way you're going to be able to compare the levels with each other.

For example I tried to measure the sound level from the same fan at different voltages and distances. It turned out that with a higher noise level the value was 11 dB lower at 1 meter than at 0.125 meters. Lowering the speed of the fan resulted in a 15 db decrease in sound level at 1 meter compared to that of 0.125 meters.

hyum · Post by **hyum** » Wed Jan 15, 2003 6:02 pm

NotSilentYet: if you're going to run a test like this again anytime soon, as a favor to ask, can you also include 5V skews? that would be terrific. i typcially run all fans that are not on dynamic fan control at 5V.

Post by **MikeC** » Wed Jan 15, 2003 6:13 pm

But it would make a big difference depending on the case, from which direction measurements were taken and so on. So it would still not be the end of it.

YUP! One solution is a test jig designed specifically to test for fan vibrations using an accelerometer. Standards for such exist, of course, but no DC fan makers (for PCs) use them, it seems. The UBC fan project will built one.

What I learned from these measurements is how difficult it is to measure sound level in an accurate way.

It's so nice to hear someone else say this!

It is true. And you have to be meticulous to get repeatable results.

with a higher noise level the value was 11 dB lower at 1 meter than at 0.125 meters. Lowering the speed of the fan resulted in a 15 db decrease in sound level at 1 meter compared to that of 0.125 meters.

This sounds a bit odd. Implies SLM non-linearity. Was this repeatable with more than one fan? Several? How about dfferent mic positions? (not distance)

NotSilentYet · Post by **NotSilentYet** » Thu Jan 16, 2003 3:24 am

This sounds a bit odd. Implies SLM non-linearity. Was this repeatable with more than one fan? Several? How about dfferent mic positions? (not distance)

I should have added that I used the Coolermaster fan for these last measurements and its sound level was anything but consistent at any distance. I think the reason for the different values is that the spectral content is different at different RPMs. Depending on the room characteristics different frequencies will be attenuated more or less I think. At least this is my conclusion so far. Perhaps I should repeat the above mesaurement with a sine wave source instead and see if the result is the same or if it is frequency dependent.

NotSilentYet: if you're going to run a test like this again anytime soon, as a favor to ask, can you also include 5V skews? that would be terrific. i typcially run all fans that are not on dynamic fan control at 5V.

I'll give it a try next time but I think the noise levels at 5 V will be immeasureable with the equipment i have except for the loud fans.