How to centre a graph about 0 on the y axis?

Trying to perform an FFT and need points centered around 0 (i.e. all their values would average out to be 0) about the y axis. Originally I was taking mean of all the points in a column and minusing these from their respective columns. I am now however trying to specifically use the 20001-66000 values in the matrix and the average is not changing for some reason, and the graph of the shortened matrix is no longer centred due to this. Any help greatly appreciated. The full code is as below.

    A = LongTestArray(20001:66000, 1:4);
    M = mean(A)
    A(:,3) = A(:,3) - M(:,3);
    A(:,4) = A(:,4) - M(:,4);

%%Attempting to average down the 46000 data points for signal processing
  B = A;
  Ar = reshape(B, [], 4, 920); % Reshape 

%%to draw the graph, changing it from number of samples to a time domain
  X=Ar(:,4);
  f = 21.7 ; 
  samples = 0:size(X, 1)-1 ; 
  t = samples/f ; 
  %plot column 1 
  plot(t, X ) 
  grid on ; 
  hold on ; 
  title('Low Frequency Accelerometer Test 1')
  xlabel('Time (s)') % x-axis label
  ylabel('Voltage (V)')

Answer 1

[Walter Roberson]

set(gca, 'YAxisLocation', 'origin')

for placing the graphics axis at y = 0

YL = get(gca, 'YLimit');
maxlim = max(abs(YL));
set(gca, 'YLimit', [-maxlim maxlim]);

to force the plot to have the same extent to the top and bottom.

[Lara Severne]

Do you know where I would enter this into my code? As currently it is returning an error code of

Error using matlab.graphics.axis.Axes/get There is no YLimit property on the Axes class.

Error in Untitled7 (line 17) YL = get(gca, 'YLimit');

[Walter Roberson]

Sorry, please change YLimit to YLim in both places.

[Lara Severne]

Okay so code runs through now but wherever I place it it doesn't centre about 0, tried both above and below the plot graph

[Walter Roberson]

I need your data to test with.

[Lara Severne]

here is the data I was using for that code. If you copy and paste the code over you should see a graph like the one attached occurring

[Walter Roberson]

The A(:,3) and A(:,4) you calculate do have individual column means of 0.

But then you take

B=A;
Ar = reshape(B, [], 4, 920); % Reshape

which reshapes the 46000 x 4 array to 50 x 4 x 920. This is followed by

 X=Ar(:,4);

which is the same as X=Ar(:,4,1)

Now, Ar(:,4,1) is extracted entirely from the first 50 x 4 of the 50 x 4 x 920, and so is extracted entirely from the first 200 datapoints of A. But with A being 46000 x 4, the first 200 datapoints of A are A(1:200, 1), and in particular X = Ar(151:200,1) in effect . Which you never subtracted out the mean of. And if you had subtracted out the mean of column 1, then that would have been the mean overall of A(:,1) which would not necessarily be anywhere close to the mean of the 151 to 200 point subset of the first column, since we have no reason to expect that the entries of the first column are statistically independent of each other.

[Lara Severne]

Ah right I see, I have asked previous questions on how to correctly average the data points down so that I can have roughly 100 points per second for example, and not the 20000 that I currently have. I did not realise that this function was not allowing me to centre through correctly as was selecting that data. Do you know a simpler way to average the results down that wouldn't have this negative impact and would allow the data to be centred? Sorry am not very good with MatLab but this function is essential to my data processing. I also cannot thank you enough for working on this with me and trying to help me figure stuff out, and am sorry I am not very knowledgeable

[Walter Roberson]

To confirm: you have about 20000 x 4 points per second? And you want to average batches of 200 x 4 consecutive points, so that you get down to 100 x 4 averaged points per second?

And once the points are averaged that way, you want to center what results for each column, over all of the seconds, so that even though each batch of 100 x 4 points per second might not have a center of 0 per column, that the overall average for each column is 0 ?

To do this we will need a hard variable about the number of points per second (is it exactly 20000 ?), and we will need rules about what should happen if the size of the data is not an exact multiple of points_per_second/100 .

[Lara Severne]

Yes due to having 4 different sensors per each experiment, and the sample rate for each is 20000 per second. Yeah that would be ideal, but this value may also have to alter depending on whether 100 points per second smoothes the curves enough for a successful FFT. I would like the total results for each of the averaged down columns to be centred around 0 for the FFT. So the total length of cropped matrix is 2.3 seconds roughly so across the sample of 230 averaged results, I need this centred around 0 on y axis. It is exactly 20000 per second, and a total of 66000 per channel (I am trying to remove first second of data due to it not being interesting) hence the initial bit of code)

A = LongTestArray(20001:66000, 1:4)

If any of this is unclear then just let me know, thanks again