Ask Your Question

Plotting in 3D in spherical coordinates

asked 2011-04-19 08:28:25 +0200

this post is marked as community wiki

This post is a wiki. Anyone with karma >750 is welcome to improve it.

My problem is that I am trying to plot (in full 3D spherical coordinates) a set of values stored in a 2D lookup table or LUT. The LUT is actually stored as a numpy 1801*3601 2D array indexed by theta and phi respectively in 0.1 degree steps. The LUT in fact represents an antenna radiation pattern (i.e. antenna gain/ radiation intensity as a function of theta and phi). However, my problem generalises to any one of plotting a function in spherical coordinates.

My first attempt at plotting this in Sage was to use the 'spherical_plot3d()' function. First I defined a function called:

getGain(phiInRadians, thetaInRadians)

which returned a value from the lookup table (LUT) representing antenna gain (a positive number in decibels). Then I tried plotting this as follows:

sage: spherical_plot3d(getGain,(-3.142,3.142),(0,3.142)).show(aspect_ratio=(1,1,1))

Now this almost does what I want, but not quite. My LUT has a high resolution with 0.1 degree intervals. However, the 3D plot which the above command delivers (via Jmol) seems to smooth the pattern where I don't want it to be smoothed (because it has abrupt edges), and is too 'blocky' where I would like the pattern to be smooth. Is there any way I can have fine control of the step-size in phi and theta (u and v in Sage-speak), or must I leave it to Sage to control these?

I also tried a different approach, which is to use list_plot3d, but to transform the coordinates from spherical to rectangular when building up my list to plot. To discuss this case we can simplify the problem to say that we wish to plot the radiation pattern of an isotropic antenna, i.e. one which has equal gain in all directions. Thus what we are simply trying to do is to plot a sphere in 3 dimensions from a list of 3-tuples, where each tuple represents an (x,y,z) coordinate in Cartesian space. However, when generating the points to plot I transform from spherical to cartesian coordinates when setting up the list of points to plot, thus:

sage: import numpy as np

sage: r=1 # Representing the gain of an isotropic antenna

sage: listOfPointsOnSurfaceOfSphere = [ (r* sin(theta) * cos(phi), r* sin(theta)* sin(phi), r* getGain(phi,theta)* cos(theta)) for theta in np.arange(0.1,pi,0.1) for phi in np.arange(-pi,pi,0.1) ]

sage: myPlot1 = list_plot3d(listOfPointsOnSurfaceOfSphere).show()

Now what I find is that my plot is all 'spiky', whereas I was hoping to see a smooth sphere. Of course, what is happening is that I have multiple points with the same (or similar) (x,y) coordinates, but very different z coordinates, since every point on the sphere 'above the equator' (i.e. above the x,y plane) has effective neighbours as mirror images below. This seems to be screwing up the interpolation routine, which isn't able to identify that it ... (more)

edit retag flag offensive close merge delete

2 Answers

Sort by ยป oldest newest most voted

answered 2011-04-21 17:08:37 +0200

this post is marked as community wiki

This post is a wiki. Anyone with karma >750 is welcome to improve it.

Can you use the plot_points option in spherical_plot3d to get a finer u,v grid?

edit flag offensive delete link more


To follow up: you can do something like spherical_plot3d(getGain,(-3.142,3.142),(0,3.142), plot_points=[200,100]).show(aspect_ratio=(1,1,1)) to plot 200 grid points in the azimuth angle and 100 points in the inclination angle. You can also just pass in a plot_points=<number> to set both numbers of points to the same value.

Jason Grout gravatar imageJason Grout ( 2011-04-21 17:13:05 +0200 )edit

Another great answer! Yes, this worked a treat, and avoids the residual 'shaded plane' problem I found with the 'list_plot3d()' approach described above. So this solves my problem entirely. Thanks.

deebs67 gravatar imagedeebs67 ( 2011-05-03 12:54:29 +0200 )edit

answered 2011-04-19 09:22:40 +0200

this post is marked as community wiki

This post is a wiki. Anyone with karma >750 is welcome to improve it.

There may be an easy fix, using two copies of your solution (b): If you add two plots, and show the result, you will see both plotted on the same axes. If I understand correctly, this will resolve your issues with (b), no?

sage: N = parametric_plot3d((lambda u,v: cos(u)*sin(v), lambda u,v: sin(u)*sin(v), lambda u,v: cos(v)),(0,pi),(0,pi))
sage: S = parametric_plot3d((lambda u,v: cos(u)*sin(v), lambda u,v: -sin(u)*sin(v), lambda u,v: cos(v)),(0,pi),(0,pi), color='red')
sage: (N+S).show()
edit flag offensive delete link more


Great answer! The only slight issue is that using parametric_plot3d() as you did above works fine for the sphere, but I still get the 'blocky/smoothed' problems as per a) when I try to plot my full (complicated) antenna radiation pattern in this way. This is presumably because Sage takes care of controlling the resolution, without giving me the option (??). But I can still use your 'North'+'South' trick with list_plot3d(). However, this does have the drawback that the plots also shade the x,y plane. So when I plot my full antenna radiation pattern in this way (or the sphere described above) it is bisected by a coloured plane at z=0. But other than that the problem is solved. Thanks.

deebs67 gravatar imagedeebs67 ( 2011-04-19 10:34:04 +0200 )edit

yes, I just didn't have a bunch of data to give an example with list_plot3d(). Sorry it's still not ideal -- maybe there's a keyword which will remove the shaded plane? For this, you might have to trace through the source code of list_plot3d() to see what keywords are accepted by the other functions it calls.

niles gravatar imageniles ( 2011-04-19 11:17:34 +0200 )edit

Your Answer

Please start posting anonymously - your entry will be published after you log in or create a new account.

Add Answer

Question Tools


Asked: 2011-04-19 08:28:25 +0200

Seen: 3,108 times

Last updated: Apr 21 '11