Shaker Noise 1

[kaiserman]

I need to design a small sound enclosure to reduce background noise including shaker noise. Could anyone direct me to a site, text, journal etc. providing tips on enclosure design?

Project details:
(1) automotive application
(2) Test part is (l)14in. x (w)6in. x (h)8in.
(3) Application determined to be random. Field PSD's have been collected
(4) Test samples will mount to a fixture which will mount to a UD ST560 single axis shaker and then be excited.
(5) Proposed noise enclosure will surround the test sample and fixture only.
(6) Test samples are mainly composed of structural plastic.
(7) The UD shaker is air cooled.
(8) GOAL: To make noise comparisons, based on sones (N10), of various test samples with differnt materials and geometric changes to numerous parts. Although the shaker is a single axis shaker, it is desired to make measurements in all axes by rotating the test sample/fixture and/or shaker.
(9) PROBLEM:The background noise in the room is too loud. (Doesn't follow the old rule of thumb that the background noise must be 10dB below the entity you are trying to measure for all frequencies of interest.
(10) We have considered running the shaker in a "no-blower" operation for short periods of time until the armature heats up to a certain temperature.

Any help would be appreciated!
Thanks,
Kaiserman

 

[MikeyP]

That's quite a problem you have posed there!

You may consider this an "obvious" response, for which I apologise, but the enclosure would have to be mounted independently of the test structure and the shaker (ie not vibrated directly by the shaker) otherwise the enclosure itself will be also be radiating noise.

Is the shaker/blower the dominant source of background noise? If it is, then maybe you can simply shield the test article with a baffle to bring down the noise floor. Making an effective seal around the skaer armature could be tricky, but it cound still have a significant effect.

Can the blower be positioned remote from the shaker and attached with a long hose?

You might want to consider relaxing your criteria for acceptable background noise margin to say 6 or 7 dB. In this case you cannot simply ignore the background noise (BGN), but you can subtract out its effects by applying some simple statistics, particularly if, as you suggest, you are using a random excitation. The procedure is as follows:

1: measure the background noise in each band averaged over 16 seconds or so.

2: switch on your shaker and perform a noise measurement again averaged over 16 seconds or so.

3: compare the measured band levels with the BGN levels. If in any particular band the difference is less than 6 or 7 dB(depending on which you chose), then discard the data for that band (but keep the rest)

4: in each of the remaining bands, subtract the BGN levels (in dB) from the measured levels (in dB) to obtain your first estimate of the true noise level.

5: keep a tally of the which bands were rejected.

Repeat 1-5 say 6 times (if the background noise level does not change much then you can just use the original BGN measurement)

Now, in some bands you will have valid data from all 6 measurements, and in some you will have less. Call the number of valid measurements in a particular band (250Hz, say) N. Call each level (in dB) for that band (obtained from step 4) x1 x2 ... xN.

You can assume that the levels in one band follow a normal distribution when expressed in dB. You can now estimate the statistical variance (sigma^2) for that band from the formula

sigma^2 = ((x1^2 + x2^2 + .. + xN^2) - ((x1 + x2 + .. + xN)^2)/N) / (N-1)

With this information you can now estimate a confidence interval for that band using the formula

95%CI = sqrt(sigma^2/N)*t

This is a 95% confidence interval (95% of all measurements in that band can be expected to be within this margin of the mean level). t is a value obtained from a standard statistial table known as a "two-tailed Student's t-distribution". Find a copy of this table (there should be plenty of books with it in and it is probably on the net somewhere) make sure it is the "two-tailed" version (not "one-tailed&quot. These tables will give you the values of t for a given number of "statistical degrees of freedom" (in this case this will be N-1) and for a particular level of probability (For a 95% CI this will be 0.975, for 99% CI it will be 0.995. For p% CI it will be 1-((100-p)/200). )

So you can now calculate the variance, confidence interval and mean measured level (the average of the N step 4 levels) in each band.

You then need to decide what is an acceptable level for the 95% (or whatever %age you chose) confidence interval (1 or 2 dB perhaps?). If after your original 6 measurements, some of the bands still have a CI greater than the threshhold you selected, then you can keep doing more tests and repeating the statistical analysis until you get what you want. You can then quote your results as the mean value in each band plus or minus the CI in each band.

Hope that made some sort of sense!

M

 

[kaiserman]

MikeyP
Thanks for the reply. What is obvious to some, is not so to others. Your suggestion of making the enclosure separate from the shaker/slip table was very insightful. I failed to realize how the "enclosure" could effect the measurements. I also didn't consider how difficult it will be to have a fixed enclosure with a hole in it to allow a moving armature in.

Since the shaker tends to be the largest contributer to the BGN your other suggestions seem helpful as well. I just discussed the possibility of moving the blower to a remote location with our maintenance crew.

Also, thanks for the suggestion regarding statistically removing the effects of the BGN. The techniques you have described will ultimately simplify the enclosure design. I hadn't thought of a statistical approach for comparisons. I'll let you know how it goes!

Thanks!
Kaiserman

 

[GregLocock]

Your initial suggestion of testing with the blower off is also valid. With a little bit of work you could build an interface box between the measuring system and the blower so that the measurement only runs if the blower was off, using a mercury switch and a flap in the exhaust from the shaker, and the trigger channel.

Don't burn the coils out! Cheers

Greg Locock