top of page
loud time.jpg
PSD_all (1).jpg



Class: Mechanical Vibrations (ME 301)

Year: Spring 2019

Goal: To design and conduct an experimental setup that uses concepts from Mechanical Vibrations.


  • Custom-programs for Digital Signal Processing

  • Unknown source in room identified and verified, with both mechanical and acoustic vibrations

  • Experimental values converge, for both loud noise data and low noise data

  • Easily repeatable, for other rooms and for further research

Technical Details

Often times, there are instances of unwanted noise, in classrooms, laboratories, or offices. As a method to identify this unknown, unwanted noise, this project uses Digital Signal Processing (DSP)methods of vibroacoustic measurements.

Room LL210 of 41 Cooper Square was chosen, as there is a persistent noise present. Several hypothesis of the source of the noise was made, including that of the water pipe belonging to the HVAC system. In order to verify this hypothesis, several measurements were taken. The overall room noise/sound profile was taken using a conference, omnidirectional microphone in the center of the room, and a reference microphone and an accelerometer were placed on the exposed water pipe, the hypothesized source of the noise. The noise data was taken both during times of loud noise and low noise.

In order to gain useful information, discrete Fourier Transform was taken of the time series data and was analyzed in the frequency domain. The sampling rates were manipulated so that all three signals have the same length and sampling frequency. Once the Fourier Transform was completed, the Power Spectral Density of each signal was found. They all had a peak at around 148 Hz.

The Cross Spectral Density of each pair of signals (room microphone and noise microphone, room microphone and accelerometer) was found to compare the frequencies, and to ultimately find the Coherence, or the frequency relation, of the signals. The coherence of all data, for both loud and low noise signals, had peaks around 148 Hz, further verifying the hypothesis that the noise source is the pipe.

To ensure that the DSP results were accurate, the recorded noise of the room was played alongside a generated tone of 148 Hz. The two matched, further suggesting that the noise source of the room was indeed the water pipe.

All DSP program was custom-created, to ensure that the values are scaled correctly, as MATLAB's built-in functions fail to do so. The technical report can be found here.

bottom of page