Zero Pad My FFT Signal and Window

Hi Everyone I want to know how I can zero pad my FFT signal to make it longer so that I can get better frequency resolution. In my line of code below that shows (FTSignal = fft(Signal-meanSignal, 15000000)/N), I thought this number would define the zero padding, but I am not confident it is doing what I believe it to do.

I have also applied my hanning window in my FFT domain, I have been told it is better to do this in the time domain before FFT, but based on my code below I am not too sure what to do. I did try Data=hann(length(Data)) but this is incorrect.

Any pointers for the zero padding and windowing would be appreciated

%% Read in Data
FName = 'Tit_10MHz_110F.flxhst';
MyData = read_history(FName);

Time = MyData.TimeRecB.Time;
Data= MyData.DataRecB(1).Data;

%% Take the signal and normalize

Data = Data./max(abs(Data)); % This is the normalized signal
% Data=hann(length(Data))

%% Plot the time domain signal

figure(1)
plot(Time*1e6, Data);                                              
grid
xlabel('Time \mus')
ylabel('Amplitude')

%% Set up the FFT Parameteres

Signal= Data;
Ts=Time(2)- Time(1); %Time Step

Fs = 1/Ts;             % Sampling Frequency (Hz)
Fn = Fs/2;             % Nyquist Frequency (Hz) half of the sampling rate of a discrete signal processing system
N = length(Signal);    % Lengh of signal
meanSignal = mean(Signal);    % ‘Signal’ Mean
FTSignal = fft(Signal-meanSignal, 15000000)/N;   % Normalised Fourier Transform Of Baseline-Corrected &Signale&  zero Padded
FTSignal = fft((Signal(:)-meanSignal) .* hann(length(Signal)) ) / N; %Apply hanning window to the data
%% Set up FFT Verctors

Fv = linspace(0, 1, fix(numel(FTSignal)/2)+1)*Fn;           % Frequency Vector
Iv = 1:numel(Fv);                                           % Index Vector

%% Plot FFT

figure(2)
plot(Fv/1e6, abs(FTSignal(Iv))*2)
xlim([0 20])
grid
xlabel('Frequency(MHz)')
ylabel('Amplitude')

Answer 1

[Star Strider]

I recognise my code!

I see nothing wrong with that (although I do not understand the two fft calls), however I would change it to:

NFFT = 2^nextpow2(15000000);

FTSignal = fft(Signal-meanSignal, NFFT)/N; % Normalised Fourier Transform Of Baseline-Corrected &Signale& zero Padded

for efficiency.

The ‘Fv’ assignment is correct. (It is taken from the R2015b documentation.)

Similarly, this would likely work:

FTSignal = fft((Signal(:)-meanSignal, NFFT) .* hann(length(Signal)) ) / N; %Apply hanning window to the data

although it wouold be necessary to experiment with it in the event that:

hann(NFFT)

produces a better result.

EDIT —

I cannot import the data:

Uz = unzip('https://www.mathworks.com/matlabcentral/answers/uploaded_files/952214/Matlab.zip')

Uz = 1×3 cell array
    {'Matlab/New_FFT.m'}    {'Matlab/read_history.m'}    {'Matlab/Tit_10MHz_110F.flxhst'}

T1 = readtable(Uz{3})

Error using readtable
'.flxhst' is not a recognized file extension. Unable to detect file type. To read the file as a specific file type, regardless of file extension, use the 'FileType' name-value pair.

.

Answer 2

[Matt J]

The zero-padding is fine. The way you set up your frequency axis, I'm a little unsure about.

N=numel(FTSignal);

NormalizedAxis= (0:N-1)-ceil((N-1)/2);

Fv=NormalizedAxis/N/Ts;

plot(Fv,fftshift(abs(FTSignal)))

[David Harra]

Thank you so much Matt, this method is much much nicer and it does give me the same plot as mine. I just dind't have a great deal of confidence as this is all new to me. However, when I change the number in my line of coding, there doesn't seem to be any difference in the plot, which I was expecting to see because as I increase this number the resolution should be finer. Any thoughts?

FTSignal = fft(Signal-meanSignal, 15000000)/N;

[Matt J]

If you are already over-sampled, you will not see a change. There's no way to know more without attaching Data for us.

[David Harra]

Hi Matt, I have attached my data and files which should just run for you no problem. Really appreciated if you could have a look at this.

[Matt J]

Could you just attach a .mat file containing Data. That's all we need to run your code.