Torch Lens Maker

Appearance

Hello! I'm Victor, the author of this project.
I'm looking for funding to be able to keep working on it. If you can, please consider donating or sponsoring.
Thank you! ❤️

Spot diagram ​

python
from IPython.display import clear_output
clear_output()
python
import torch
import torch.nn as nn
import torch.optim as optim
import torchlensmaker as tlm
import numpy as np

Let's define a simple optical model that we'll use to demonstrate various ways of producing spot diagrams.

python
surface = tlm.Sphere(diameter=15, R=tlm.parameter(32.6656))
lens = tlm.BiLens(surface, material="BK7", outer_thickness=1.5)

optics = nn.Sequential(
    tlm.ObjectAtInfinity(beam_diameter=10, angular_size=5.),
    tlm.Wavelength(400, 800),
    tlm.Gap(1),
    lens,
    tlm.Gap(30),
    tlm.ImagePlane(15),
)

Spot diagram with rows and cols ​

The row and col arguments control the sampling dimension thats use to layout the rows and columns of the diagram. row='object' and col='wavelength' is a typical choice.

python
clear_output()
tlm.show2d(optics, sampling={"base": tlm.sampling.dense(10), "object": tlm.sampling.dense(3), "wavelength": 3})
tlm.show3d(optics, sampling={"base": 100, "object": [[np.deg2rad(0), 0.], [np.deg2rad(5), 0.]], "wavelength": 3}, end=100)
python
sampling = {"object": [[np.deg2rad(0), 0.], [np.deg2rad(5), 0.]], "wavelength": 3}
f, _ = tlm.spot_diagram(optics, sampling | {"base":1000}, row="object", col="wavelength", figsize=(12, 12))

png

Spot diagram with color_dim ​

If either row or col is None, then there will be only one row or column containing every ray sample. Still, the color_dim argument can be used to color points by one dimension.

python
sampling = {"object": [[np.deg2rad(0), 0.], [np.deg2rad(5), 0.]], "wavelength": 3}
f, _ = tlm.spot_diagram(optics, sampling | {"base":1000}, row="object", col=None, color_dim="wavelength", figsize=(12, 12))

png

python
# TODO object coordinates are given as radian here
# but as degrees in light source init / tlm.Rotate

# if you give exact values in a sampling dict
# it becomes dimension specific