My arrows are too frequent and too small in quiver-CodePudding

I have a 100x100 unit vectorfield created on a meshgrid as follows. I evaluate some function on every point combination in the X and Y meshgrids.

import numpy as np
import matplotlib.pyplot as plt
N=100 # of points
#set arrays of n, m and td_50 to try
n_array=np.linspace(0.01,1,N)
td50_array=np.linspace(1,100,N)

X,Y=np.meshgrid(n_array,td50_array)

Z=X*Y

I am now trying to plot a vectorfield of the gradient of Z, so I do

#compute gradient,normalized gradient x and y components, as well as gradient raw magnitude
gradx,grady=np.gradient(Z)
gradxnorm=gradx/np.sqrt(gradx**2 grady**2)
gradynorm=grady/np.sqrt(gradx**2 grady**2)
gradmag=np.sqrt(gradx**2 grady**2)
#plot
plt.pcolor(X, Y, gradmag,cmap='rainbow')
plt.colorbar()
plt.quiver(X, Y, gradxnorm , gradynorm,units='xy')
plt.show()

And I want to make a plot of the gradient magnitude overlay with the direction of the gradient (with uniform values)

If I set N=10, then I get a nice clean plot that looks like this. This is exactly what I want

However, I want to evaluate at more than just 10 points in each direction, but if I set N=100, I get a plot like this instead where the arrows are far too small and far too frequent to be useful overlayed with the gradient.

I can see that this is just because I have many more points, but is it possible to make the arrows on this second plot look more like the arrows on the first plot? I don't need an arrow for every single point in the grid, so I would like fewer arrows that are actually larger instead of one tiny arrow at every single pixel.

CodePudding user response：

When you use plt.quiver you might want to plot the arrows every N datapoints instead of every datapoints. This requires indexing your X, Y,gradxnorm and gradynorm accordingly. Below is an example where I only plot points with x and y coordinates multiples of 10:

import numpy as np
import matplotlib.pyplot as plt
N=100 # of points
#set arrays of n, m and td_50 to try
n_array=np.linspace(0.01,1,N)
td50_array=np.linspace(1,100,N)

X,Y=np.meshgrid(n_array,td50_array)

Z=X*Y

#compute gradient,normalized gradient x and y components, as well as gradient raw magnitude
gradx,grady=np.gradient(Z)
gradxnorm=gradx/np.sqrt(gradx**2 grady**2)
gradynorm=grady/np.sqrt(gradx**2 grady**2)
gradmag=np.sqrt(gradx**2 grady**2)
step=10
plt.pcolor(X, Y, gradmag,cmap='rainbow')
plt.colorbar()

plt.quiver(X[::step,::step], Y[::step,::step],gradxnorm[::step,::step] , gradynorm[::step,::step],units='xy')
plt.show()

The output gives:

You can modify the variable step to increase or decrease the number of arrows.