surface plot

Hello all,

I have the following code:

a = 0:.1:1;
b = 0:.01:.1;
for k = 1:length(a)
  c = a(k) + b(k);
end

I would like to plot 'c' versus 'a' and 'b' using 'surfl'; Is there a way to do it? thanks.

Answer 1

[Grzegorz Knor]

[a b] = meshgrid(0:.1:1,0:.01:.1);
c = a + b;
surfl(a,b,c)

[Walter Roberson]

Bahareh's previous question was specifically asking about combining a(K) and b(K) instead of producing a grid, so it seems your code (fine in most circumstances) is not what is desired _here_

Answer 2

[Walter Roberson]

a = 0:.1:1;
b = 0:.01:.1;
c = a + b;
t = nan(length(a),length(a));
z = diag(c) + tril(t,-1) + triu(t,1);  %defined on diag, else nan
surfl(a,b,z);

I would not expect the plot to have any useful content, though. surface plots do interpolation from vertices to determine what to draw, but you are only defining the result along a line.

I will not suggest that you want the entire grid of a(J) + b(K) values as that would not be in accordance with your previous question.

Answer 3

[Albert J. G.]

This question is about how to plot a surface diagram.

The data are in an excel file, as you can find.

My question is how to plot a surface diagram using data from A2 to C72? (if you think data is no enough, you can use C491)

Maybe x=kp; y = ki; z = delay. What I need is like surf(x,y,z), and it will give a surface plot with NO ERROR (this is the most important, I think).

This is the question borthering me for a long time.

BTW, it not hard to plot a 3D scattor diagram, using:

[num, txt]= xlsread('rank.xlsx', 'A2:C72');
x = num(:,1);
y = num(:,2);
z = num(:,3);

% the point on 3d diagram:
plot3(x,y,z,'.','color','black','markersize',12)
xlabel('kp')
ylabel('ki')
zlabel('delay')
grid on

But I really donnot how to plot a surface since Z do not have a obvious relation with X and Y.

[Walter Roberson]

See scatteredInterpolant() or griddata()

[Albert J. G.]

Thanks for replying.

I tried this but it seems doesn't work.

The data I have is more like scatter so it's a soild one.

The question is how can i draw a perfect surface plot with MATLAB.

p.s. 1. the updated excel file is attached; 2. For the surface plot we have x-axis: kp, y-axis: ki, z-axis: delay; 3. If you can show the source code and the diagram, i will be grateful!

[Walter Roberson]

num = xlsread('rank.xlsx', 'A2:C491');
x = num(:,1);
y = num(:,2);
z = num(:,3);
xv = linspace(min(x), max(x));
yv = linspace(min(y), max(y));
[X, Y] = meshgrid(xv, yv);
Z = griddata(x, y, z, X, Y);
surf(X, Y, Z, 'edgecolor', 'none');

[Albert J. G.]

Thanks for answering!

However, there is a big mistake in your answer: the program didn't show the graph when delay (z-axis) is bigger than 0.11 sec. (The graph should show a full data till 0.21 sec (z-axis).)

A guess is that the program average the data of z-axis.

Is there any way to solve it?

[Walter Roberson]

You have a lot of duplicated data points. You only have 20 different x, and only 10 different y. I do not know what a "surface plot with NO ERROR" means to you.

[Albert J. G.]

It should be like a three-dimensional trapezoid, as shown in the attached file.

I just cannot find a way get it using MATLAB.

[Walter Roberson]

Your data is not compatible with a plot that looks like that.

num = xlsread('rank.xlsx', 'A2:C491');
x = num(:,1);
y = num(:,2);
z = num(:,3);
xv = linspace(min(x), max(x));
yv = linspace(min(y), max(y));
[X, Y] = meshgrid(xv, yv);

[ux, ~, xidx] = unique(x);
[uy, ~, yidx] = unique(y);
zz = accumarray([xidx,yidx],z(:),[], @(v) max(v), nan);

mask = ~isnan(zz);
ind = find(~isnan(zz));
[xind,yind] = ind2sub(size(zz),ind);
xux = ux(xind); yuy = uy(yind); zuz = zz(mask);
Z = griddata(xux,yuy,zuz,X,Y);

surf(X,Y,Z,'edgecolor','none')

[Albert J. G.]

I have to mention that I have updated my excel file (named rank.xlsx below).

num = xlsread('rank.xlsx', 'A2:C10745');

The graph is more and more perfect, but it seems there are a lot of sawtooths, which the attached file shows, around kp (x-axis) smaller than 50 when using the following codes you give:

num = xlsread('rank.xlsx', 'A2:C10745');
x = num(:,1);
y = num(:,2);
z = num(:,3);
xv = linspace(min(x), max(x));
yv = linspace(min(y), max(y));
[X, Y] = meshgrid(xv, yv);
[ux, ~, xidx] = unique(x);
[uy, ~, yidx] = unique(y);
zz = accumarray([xidx,yidx],z(:),[], @(v) max(v), nan);
mask = ~isnan(zz);
ind = find(~isnan(zz));
[xind,yind] = ind2sub(size(zz),ind);
xux = ux(xind); yuy = uy(yind); zuz = zz(mask);
Z = griddata(xux,yuy,zuz,X,Y);
surf(X,Y,Z,'edgecolor','none')

[Walter Roberson]

griddata has to do interpolation. Your data has some location with a single z of 0 and beside it locations with multiple z at higher values. So you end up with adjacent high and low locations when you do the maximum value projection. Interpolation for that turns out sawtooth.

[Walter Roberson]

Perhaps you would prefer a different approach:

num = xlsread('rank.xlsx', 'A2:C10745');
x = num(:,1);
y = num(:,2);
z = num(:,3);
K = boundary(x, y, z, 1);
trisurf(K, x, y, z);

[Albert J. G.]

Wow! That works!! Thx!!!