Mach Wave Elimination
A new method for quieting high-speed jets
(U.S. Patents 5590520 and 5916127)

THE CONCEPT

EXPERIMENTAL SETUP

ACOUSTIC DATA

APPLICATION TO AEROENGINES

TECHNOLOGY EXCHANGE

FUTURE PLANS

PUBLICATIONS

PUBLICITY & AWARDS

# NOISE DEFINITIONS

This is a brief overview of the various measures of noise and annoyance used in our study.   The microphone output is a time trace of voltage, V(t).  The voltage is prortional to the pressure fluctuation p'(t) via a constant provided by the manufacturer.   The pressure flutuation p'(t) is normalized by a commonly-used reference pressure, pa=20 mPa.  The spectrum of the normalized pressure fluctuation, S(f),  is computed using a fast Fourier transform (FFT) algorithm.   The spectrum is corrected for the response of the microphone and for atmospheric absorption.   In our experiments, the resolved frequency f ranges from 500 Hz   to 150 kHz.The following noise metrics are then computed:

Sound Pressure Level (SPL) Spectrum
This is simply a logarithmic representation of S(f):
SPL(f) = 10 log10 S(f)    [units: dB/Hz]

Overall Sound Pressure Level (OASPL)
The total energy contained in the spectrum is

E = ò S(f) df

where we integrate over all resolved frequencies.  The OASPL is

OASPL = 10 log10 (E)   [units: dB]

A-weighted (dBA) spectrum

 This metric is used widely to assess the human perception of sound.  Humans are most sensitive to the frequency range 1000-10000 Hz.  The spectra S(f) are scaled-up to the size of a typical engine exhaust (this is done by dividing the frequencies by the scale-up factor), discretized to 1/3-octave bands, and corrected with the A-weighting displayed on the right. The peak of the dBA spectrum gives an overall assessment of perceived noise A-weight correction

PNL

PNL (Perceived Noise Level) and its variants are  metrics used for aircraft noise certification.  For an overview of the PNL calculation the reader is referred to this web page.  To compute the PNL, the spectra are extrapolated to higher frequencies using a drop of 30 dB per frequency decade.  The extrapolation allows resolution of scaled-up frequencies up to 10,000 Hz necessary for computation of PNL. For the scale-up factors used here (50- 100) the calculated PNL is very insensitive on this extrapolation.