Attached are some plots of sound data from a 2120-2.

What do you see in it? Need some ideas here, I think.

These_are_3_noise_spectrums.doc (302 Kb, 53 downloads)

The beat-frequency-variation behaviour you are showing is as expected. You can check this out in any basic vibration text.
A single discrete tone is more offensive to the ear than the same sound level with more than one frequency, or broadband. I recall, just from memory, that in environmental acoustics and related standards, there's a 5 or 6 dB penalty for a single-tone-dominant noise environment.

Buddy,

Human perception (annoyance) of beating tones is limited to a specific delta-F range. Small delta-F yields a long beat period that is not noticeable. A large delta-F yields a short beat period that is also not noticeable. I have seen this described in a reference a long time ago, so you could search on beat frequency audibility.

I have worked on many power plant ID fan noise projects. It is better to keep two fans running near same speed by keeping same load on each fan. Your sound pressure spectra should be in decibels referenced to 20 micro-Pascals.

Walt

quote:

Your sound pressure spectra should be in decibels referenced to 20 micro-Pascals.

Walt - How do you set that up ?

Mark

Buddy, do you see the same sort of behavior (beat duration) in the vibration data? Is this a vibration-induced "noise" problem, or a flow-induced problem? Do the fans have a common suction and/or discharge duct or plenum?

Mark,

My comments are based on CSI analyzer and MasterTrend software. The simplist procedure is to select Microphone as a sensor type and then input the sensitivity in volt/Pascal. The waveform would then be scaled in Pascals (Pa) Vs time. The spectrum wold be sound pressure level (dB) Vs frequency (Hz).
In MasterTrend Spectrum display Select Axis Options --> Type of Amplitude Axis --> Log. From the Spectrum display select Tools --> Global Options and check to be sure Sound Pressure Unit is dB and dB Reference is 0.00002 Pascals or 20 micro-Pa. A word of caution: if you are using a microphone with more than one gain/attenuation setting, then each setting would have a different sensor/microphone sensitivity. An acoustic calibrator that produces a standard sound pressure level makes the setup easy. Suppose the calibrator produces 114 dB at 1000 Hz. Solve this equation for Pa:

dB = 20*LOG(Pa/Pref) where LOG is logarithm to base-10 and Pref is pressure reference of 0.00002 Pascals

So: Pa = 0.00002 * 10EXP(dB/20) or 10.023 Pascals
If the AC voltage was 1.0 volts, then the microphone sensitivity is 1.0/10.023 or 0.09976 V/Pa

If you want A-weighted sound spectra, then either use the sound meter's A-weight filter or configure a Parameter Set for your measurement point where you have the option of A-weighting. A Parameter Set can also be used for Third Octave and Octave Band sound pressure spectrum that are common for acoustical measurements.

Most any type/brand of analyzer should allow you to make a sound pressure level measurement.

Walt

Rusty,

Sorry so long in getting back, but what with the holiday approaching, been real busy.

I do not have a waveform showing the beat frequency pattern but here is some data showing the spectrums of the worst amplitude of vanepss on a normal bearing position. Also the noise in Db units.

velocity_and_noise_data_compared.doc (95 Kb, 10 downloads)