Loop generation doubt.

Hi,

I need to create a loop. I have to do a numerical analysis of thousands of data points.

My equation is,

E=A*cos(k-(w*t))

I have A,k and w three of them have same length 1x10. In the case of t which i have 1x50.

what i have to do is i need to take the first value of A,k and w and vary t (means run with range of values) . I need to do the same processs for the rest of the rows of A,k and w and vary t.

To be more precise.

A k w T

1 1 1 1

2 2 2 2

3 3 3 3

4 4 4 4

N N N 5

6

N

the equation will be

First calculation need to be

E=1*cos(1-(1*(1....N)))

second

E=2*cos(2-(2*(1....N)))

.......

continue upto length of A,k,w

Answer 1

[the cyclist]

% DO NOT USE THESE. USE YOUR DATA. THESE ARE FOR AN EXAMPLE ONLY
T = rand(160547,1);
A = rand(44880,1);
k = rand(size(A));
w = rand(size(A));

%%%%%% Only use the code below %%%%%

% Preallocate memory for E
E = zeros(1,160547);

% Calculate E
for ii = 1:size(E2,2)
    E(ii) = sum(A.*cos(k-(w.*T(ii))));
end

[the cyclist]

FYI, if your data weren't so large (which you did not mention in your original question), then

E = sum(A.*cos(k-(w.*T')));

which is just the sum of my original answer, would give the same result.

This code work even though the vector T is a different size, because of the transpose and the implicit expansion I mentioned. It would calculate the 160547x44880 array, as you state. I suggest you try it on a smaller dataset

Your data are too big, so that huge intermediate array cannot be stored. But here is proof that on a smaller dataset, it gives the same result:

% DO NOT USE THESE. USE YOUR DATA. THESE ARE FOR AN EXAMPLE ONLY
T = rand(1605,1);
A = rand(44880,1);
k = rand(size(A));
w = rand(size(A));

% Preallocate memory for E
E = zeros(1,1605);

% Calculate E
for ii = 1:size(E,2)
    E(ii) = sum(A.*cos(k-(w.*T(ii))));
end

% Calculate using implicit expansion instead
E2 = sum(A.*cos(k-(w.*T')));

% Show that the two results are equal
isequal(E,E2)

[the cyclist]

You can attach your data file here, using the INSERT icon that looks like a paperclip.

[Martin Thomas]

please find the attachment for cyclist.m is the code

[Martin Thomas]

hope you understand that.

[the cyclist]

For my solutions above, it was important that your vectors were all row vectors, as you stated in your original question. In your actual code, they were not.

I believe the following aligns all the pieces correctly:

% Preallocate memory for eta
eta = zeros(160547,1);

% Precalculate K*X.
KX = K.*X;

% Calculate eta
for ii = 1:numel(eta)
    eta(ii)=sum(amp'.*cos(KX-(f.*T_a(ii))));
end

The result I get is all NaNs, but I have not investigated why.

[Martin Thomas]

same for me as well, im not good at coding, is any chance to sort it out. I realy need it that important for me

[the cyclist]

eta is always NaN because K has a NaN in it (so the sum is always NaN).

The first element of K is NaN because it comes from 1/f, and the first element of f is zero.

The first element of f is zero because you defined it that way:

f = (Fs/2*linspace(0,1,nfft/2+1)).*(2*pi);

So, you need to fix that in a way that makes sense for your application.

[Martin Thomas]

Let metry with

KX(isnan(KX))=0;

[Martin Thomas]

hey got the result but i need to consult with my professor is that correct or wrong. Thank you so much. If anything i will ask you.

[Martin Thomas]

hey one more doubt, i think you know the fft, in the fft amplitude plot from the above code you can see lot of peaks, moreover if i do the fft on any signal it will give me some more additional values is that possible to remove those values from that.

[the cyclist]

I don't really use fft. I would open a new question, and upload an image of the plot using INSERT again, along with the code that generates it.

Try to be very clear about your exact question.

Answer 2

[the cyclist]

You don't need to use a loop. MATLAB will do vectorized calculations.

If you have a fairly recent version of MATLAB, you can do this using implicit expansion of the vectors:

% Make up some data
A = rand(1,10);
k = rand(1,10);
w = rand(1,10);
t = rand(1,50);

% Calculate E
E = A.*cos(k-(w.*t'));

The output E will be 50x10.

Note that I took the transpose of t, and also used element-wise multiplication and not matrix multiplication. You can read about the difference here.

[Martin Thomas]

Thanks for the reply,

is that possible to put the A=rand(FIRST ROW TO LAST ROW) without giving the actual value becauseA might have multiple same values.

[Martin Thomas]

one more thing i have calculated A,k,w, t values already. i need to take that values itself.

[the cyclist]

The only code you need is the last line I wrote:

% Calculate E
E = A.*cos(k-(w.*t'));

The other lines of code are just making up some pretend input values, which I used to make sure that last line worked. You should just use your values for A, k, and w.

I didn't really understand your other comments, so if that doesn't help, maybe try to explain it again?

[Martin Thomas]

I have a time vector of

T=1:(1/128):1200 160547x1 double

A= fft results (amplitude) 44880x1 double

k=(same length of A) 44880x1 double

w= frequency (same lenth of A) 44880x1 double

The problem is T does not have equal length so i can't do the E. the only way i need to do is calculating first row of A,k and w with varying T will get a result same i need to be don in the second row, thrid until the end of the A,k and w value. then i wll sum it up

E = SUM(A.*cos(k-(w.*t')));

[Martin Thomas]

So there willbe 160547x44880 or 44880x160547 matrix after summing up the values ther will be 160547x1 double matrix