2_custom_parameters.py

import badcrossbar
import numpy as np

# Plotting  module produces vector images (as PDF files) that can then be
# edited in any vector graphics manipulation program. However, it also provides
# option to modify some of the features of the diagram that might be difficult
# to change once the image is produced.

# This example demonstrates how to use some of these options. In particular,
# we consider how one might visually represent node voltages in a large crossbar
# array where using default parameters produces a diagram with nodes that are
# hard to see without zooming in.

# Shape of the crossbar array.
num_word_lines = 128
num_bit_lines = 64

# Number of different sets of inputs applied.
num_examples = 100

# Interconnect resistance in ohms.
r_i = 0.25

# Randomly generated resistances and voltages
# (in ohms and volts, respectively).
resistances = 100 * np.random.rand(num_word_lines, num_bit_lines)
voltages = np.random.rand(num_word_lines, num_examples)

solution = badcrossbar.compute(voltages, resistances, r_i, verbose=0)

# We are going to make the interconnects and crossbar devices 4 times thinner.
my_wire_scaling_factor = 1 / 4
my_device_scaling_factor = 1 / 4
# We are also going to make them lighter in color.
my_default_color = (0.9, 0.9, 0.9)

# When we reduced the width of the devices, we also made the diameter of the
# nodes smaller (the two are controlled by the same variable
# `device_scaling_factor`). However, we want to make the nodes larger; for
# that we can use a separate variable `node_scaling_factor` which controls
# the size of the nodes without affecting the device width.
# The following line will result in net scaling factor of 8 for the nodes.
my_node_scaling_factor = 1 / my_device_scaling_factor * 8

# Additionally, we are going to change the color bar colors, so that the
# zero would associated with black and that the maximum positive value would be
# associated with red.
my_zero_rgb = (0, 0, 0)
my_high_rgb = (1, 0, 0)

badcrossbar.plot.nodes(
    voltages=solution.voltages,
    wire_scaling_factor=my_wire_scaling_factor,
    device_scaling_factor=my_device_scaling_factor,
    default_color=my_default_color,
    node_scaling_factor=my_node_scaling_factor,
    zero_rgb=my_zero_rgb,
    high_rgb=my_high_rgb,
    axis_label="Average voltage (V)",
    filename="Ex-2",
    allow_overwrite=True,
)

# The produced diagram hopefully helps to visualize the voltage decreases
# across the crossbar array more clearly. In this specific example, the
# new diagram has nodes that are larger than the crossbar devices (visible when
# zoomed in). However, this might be appropriate design decision if one
# wanted to have a better global view of how the voltages are distributed.