enbw

Equivalent noise bandwidth

Syntax

bw = enbw(window)

bw = enbw(window,fs)

Description

bw = enbw(window) returns the two-sided equivalent noise bandwidth, bw, for a uniformly sampled window, window. The equivalent noise bandwidth is normalized by the noise power per frequency bin.

bw = enbw(window,fs) returns the two-sided equivalent noise bandwidth, bw, in Hz.

Examples

Equivalent Noise Bandwidth of Hamming Window

Determine the equivalent noise bandwidth of a Hamming window 1000 samples in length.

bw = enbw(hamming(1000))

bw = 1.3638

Equivalent Noise Bandwidth of Flat Top Window

Determine the equivalent noise bandwidth in Hz of a flat top window 10000 samples in length. The sample rate is 44.1 kHz.

bw = enbw(flattopwin(10000),44.1e3)

bw = 16.6285

Equivalent Rectangular Noise Bandwidth

Obtain the equivalent rectangular noise bandwidth of a 128-sample Hann window.

Generate the window and compute its discrete Fourier transform over 2048 frequencies. Shift the transform so it is centered at zero frequency and compute its squared magnitude.

lw = 128;
win = hann(lw);
lt = 2048;
windft = fftshift(fft(win,lt));

ad = abs(windft).^2;
mg = max(ad);

Specify a sample rate of 1 kHz. Use enbw to compute the equivalent noise bandwidth of the window and verify that the value coincides with the definition.

fs = 1000;

bw = enbw(win,fs)

bw = 11.8110

bdef = sum((win).^2)/sum(win)^2*fs

bdef = 11.8110

Plot the squared magnitude of the window. Overlay a rectangle with height equal to the peak of the squared magnitude and width equal to the equivalent noise bandwidth.

freq = -fs/2:fs/lt:fs/2-fs/lt;

plot(freq,ad, ...
    bw/2*[-1 -1 1 1],mg*[0 1 1 0],'--')
xlim(bw*[-1 1])

Verify that the area of the rectangle contains the same total power as the window.

Adiff = trapz(freq,ad)-bw*mg

Adiff = 7.2760e-12

Repeat the computation using normalized frequencies. Find the equivalent noise bandwidth of the window. Verify that enbw gives the same value as the definition.

bw = enbw(win)

bw = 1.5118

bdef = sum((win).^2)/sum(win)^2*lw

bdef = 1.5118

Plot the squared magnitude of the window. Overlay a rectangle with height equal to the peak of the squared magnitude and width equal to the equivalent noise bandwidth.

freqn = -1/2:1/lt:1/2-1/lt;

plot(freqn,ad, ...
    bw/2*[-1 -1 1 1]/lw,mg*[0 1 1 0],'--')
xlim(bw*[-1 1]/lw)

Verify that the area of the rectangle contains the same total power as the window.

Adiff = trapz(freqn,ad)-bw*mg/lw

Adiff = 0

Input Arguments

window — Window vector

real-valued row or column vector

Uniformly sampled window vector, specified as a row or column vector with real-valued elements.

Example: hamming(1000)

Data Types: double | single

fs — Sampling frequency

positive scalar

Sampling frequency, specified as a positive scalar.

Output Arguments

bw — Equivalent noise bandwidth

positive scalar

Equivalent noise bandwidth, specified as a positive scalar.

Data Types: double | single

More About

Equivalent Noise Bandwidth

The total power contained in a window is the area under the curve produced by the squared magnitude of the Fourier transform of the window. The equivalent noise bandwidth of the window is the width of a rectangle with height equal to the peak of the squared magnitude of the Fourier transform and area equal to the total power.

Parseval's theorem states that the total energy for the N-sample window w(n) equals the sum of the squared magnitudes of the time-domain window samples or the integral over frequency of the squared magnitude of the Fourier transform of the window:

$${\displaystyle {\int }_{-1/2}^{1/2}{\left|W(f)\right|}^2}df={\displaystyle \sum _n{\left|w(n)\right|}^2}.$$

The power spectrum of the window has peak magnitude at f = 0:

$${\left|W(0)\right|}^2={\left|{\displaystyle \sum _{n}w}(n)\right|}^2.$$

To find the width of the equivalent rectangle, and thus the equivalent noise bandwidth, divide the area by the height:

$$\frac{{\displaystyle {\int }_{-1/2}^{1/2}{\left|W(f)\right|}^2}df}{{\left|W(0)\right|}^2}=\frac{{\displaystyle \sum _n{\left|w(n)\right|}^2}}{{\left|{\displaystyle \sum _{n}w}(n)\right|}^2}.$$

When a sample rate f_s is specified, enbw returns the previous expression multiplied by f_s:

$$\text{ENBW}=f_s\frac{{\displaystyle \sum _n{\left|w(n)\right|}^2}}{{\left|{\displaystyle \sum _{n}w}(n)\right|}^2}.$$

For normalized frequencies, the function divides by the bin width, f_s/N:

$$\text{ENBW}=N\frac{{\displaystyle \sum _n{\left|w(n)\right|}^2}}{{\left|{\displaystyle \sum _{n}w}(n)\right|}^2}.$$

See Equivalent Rectangular Noise Bandwidth for an example that, for a given window:

• Compares the result of enbw to the definition for a specified sample rate and for normalized frequencies.

• Plots the equivalent rectangular bandwidth over the magnitude spectrum of the window.

• Verifies that the total power contained in the window is the same as the power contained in a rectangle with height equal to the peak squared magnitude of the Fourier transform of the window and width equal to the equivalent noise bandwidth.

References

[1] harris, fredric j. “On the Use of Windows for Harmonic Analysis with the Discrete Fourier Transform.” Proceedings of the IEEE^®. Vol. 66, January 1978, pp. 51–83.

Extended Capabilities

C/C++ Code Generation
Generate C and C++ code using MATLAB® Coder™.

Version History

Introduced in R2013a