Why does my uivhat produce table completely full of NaNs?

Just started with MATLAB, so absolutely no clue what I'm doing. Any help will be appreciated. I think the main issue stems from uivhat giving NaNs in the resulting table after transforming back into real.

%Express currents in terms of complex velocity
uiv = kadcp.Evel + 1i*kadcp.Nvel;
H = find(sum(isfinite(uiv),2)>10000 & sum(isfinite(uiv))<size(uiv,2));

for n=1:length(H)
    A = find(isfinite(uiv(:,H(n)))); 
    B = find(isnan(uiv(:,H(n))));
    uiv(B,H(n)) = interp1(A, uiv(A, H(n)),B); % interpolate across missing data depth by depth
end

%Define sample frequency
tlength = 17824;
samplefreq = 2;
deltafreq = samplefreq/tlength;
freq = deltafreq :deltafreq :samplefreq/2;


%Compute Fourier coefficient
uivhat = real(fft(uiv).*conj(fft(uiv)));
%Retrieve spectra coefficient for neative and positive frequencies
CW = uivhat(2:(length(freq))+1,:); % clockwise components
CCW = flipud(uivhat([2+length(freq):end, 1],:));% counterclockwise components
%Band-average to reduce noise
CW = runmean(CW, 2);
CCW = runmean(CCW, 2);
%Plotting the spectra for CW and CCW
figure(4);
loglog(freq,CW,'r',freq,CCW,'b');

[Jan]

Without having your input arguments, we cannot run the code and cannot check, where the NaN values are coming from.

You can step through the code line by line using the debugger: Set a breakpoint in the left bar of the editor. Now check line by line, which calculations produce the NaNs.

[Di]

Hi Jan,

Thanks for your reply. I've isolated the issue to fft(uiv) where uiv is a table filled with numbers and some NaNs. When I run fft(uiv) I get the table full of NaNs... Attached is the uiv file if you want to take a look. Stuck on this for many days.

[Jan]

I do not see any table objects here. Do you mean "vector"?

Did you check, where the first NaN occurs?

dbstop if naninf

I boldly guess it is interp1 outside the defined range.

Answer 1

[Hornett]

Hi,

I understand that you have a dataset called "uiv". However, when you apply the "fft" function to it, the resulting matrix is entirely composed of NaN values.

The “fft” function will return all values as NaN even if a single value among them is NaN, so you will have to remove all the NaN values from your dataset.

Here is the example code which you can use to remove NaNs.

matrix = uiv;
matrixWithoutNaNs = matrix;
matrixWithoutNaNs(isnan(matrixWithoutNaNs)) = 0;
% Display the matrix without NaNs
disp(matrixWithoutNaNs);

You can also use “fillmissing” function to fill any NaN value with constant.

Refer to the documentation link of “fft” and “fillmissing” functions

• fft: https://www.mathworks.com/help/matlab/ref/fft.html
• fillmissing: https://www.mathworks.com/help/matlab/ref/fillmissing.html