lowpass filter function doesn't filter out data

I have an ultrasonic signal (sample rate 199.2 MHz) with random noise which I want to filter out using the lowpass() built-in function in matlab. I want my cut-off frequency to be 9 MHz. In my code this looks like this:

FilteredSignal = lowpass(NoisySignal,9*1e6,1e6*199.2,'ImpulseResponse','iir');

The resulting signal I get after this seems to include even more frequencies instead of filtering out the noise.

How can I fix this?

Any help is much appreciated.

[Paul]

Hi Maria,

Is the plot on the right really the output of that lowpass command? Can't recereate that plot ...

load('NoisySignal')
Fs = 199.2e6;                                          % Sampling Frequency (Hz)
N = numel(NoisySignal);
FilteredSignal = lowpass(NoisySignal,9*1e6,1e6*199.2,'ImpulseResponse','iir');
t = (0:N-1)/Fs;
figure
plot(subplot(121),t*1e6, NoisySignal),grid,xlim([0 9]);
plot(subplot(122),t*1e6, FilteredSignal),grid,xlim([0 9])

Answer 1

[Star Strider]

The ‘NoisySignal’ has broadband noise, so a frequency-selective filter can eliminate some of the noise, however not all of it. One option to deal with it is wavelet denoisiung (the wdenoise function), and the other is Savitzky-Golay filtering (the sgolayfilt function).

Using both the lowpass and sgolayfilt gives a reasonable result —

load('NoisySignal')
Fs = 199.2E+6;                                          % Sampling Frequency (Hz)
Fn = Fs/2;
L = numel(NoisySignal)

L = 16384

t = linspace(0, L-1, L)/Fs;

figure
plot(t, NoisySignal)
grid
xlabel('Time')
ylabel('Amplitude')

FT_NS = fft(NoisySignal.'.*hann(L));
Fv = linspace(0, 1, L/2+1)*Fn;
Iv = 1:numel(Fv);

figure
plot(Fv, abs(FT_NS(Iv))*2)
grid
xlabel('Frequency (Hz)')
ylabel('Magnitude')
xline(9E+6, '-r', 'Filter Cutoff Frequency')

FilteredSignal = lowpass(NoisySignal,9*1e6,1e6*199.2,'ImpulseResponse','iir');

FT_FS = fft(FilteredSignal.'.*hann(L));

figure
plot(t, FilteredSignal)
grid
xlabel('Time')
ylabel('Amplitude')

figure
plot(Fv, abs(FT_FS(Iv))*2)
grid
xlabel('Frequency (Hz)')
ylabel('Magnitude')
xline(9E+6, '-r', 'Filter Cutoff Frequency')

SGorder = 3;
SGframelen = 301;
SGFilteredSignal = sgolayfilt(FilteredSignal, SGorder, SGframelen);

FT_SGFS = fft(SGFilteredSignal.'.*hann(L));

figure
plot(t, SGFilteredSignal)
grid
xlabel('Time')
ylabel('Amplitude')

figure
plot(Fv, abs(FT_SGFS(Iv))*2)
grid
xlabel('Frequency (Hz)')
ylabel('Magnitude')
xline(9E+6, '-r', 'Filter Cutoff Frequency')

Experiment to get the result you want. (The signal length is quite nicely already a power-of-2, making the code a bit more efficient without having to zero-pad it to a power-of-2 for the fft call.)

.