"""IQ plane comparison plot."""
import json
from pathlib import Path
import matplotlib.pyplot as plt
import numpy as np
from .common import load_exported_dataframe
[docs]
def _plot_iq(exp_0: str, exp_1: str, save: bool = False) -> None:
"""Plot the IQ plane of two experiments and the rotated CDF analysis.
:param exp_0: directory of the first experiment.
:type exp_0: str
:param exp_1: directory of the second experiment.
:type exp_1: str
:param save: also save the figures.
:type save: bool
"""
exp_path_0 = Path(exp_0)
exp_path_1 = Path(exp_1)
df_0 = load_exported_dataframe(exp_path_0)
df_1 = load_exported_dataframe(exp_path_1)
if df_0 is None or df_1 is None:
return
# Load experiment configuration
with open(exp_path_0 / "config.json") as f:
config_0 = json.load(f)
with open(exp_path_1 / "config.json") as f:
config_1 = json.load(f)
for config in [config_0, config_1]:
if config["variables"] or config["sys_config"]["/axisAcquisitionIP_0"]["$output_type"] != "accumulated":
print("Experiment config contains variables or the value is not accumulated")
return
iq_0: np.ndarray = df_0["value"].to_numpy()
iq_1: np.ndarray = df_1["value"].to_numpy()
plt.scatter(iq_0.real, iq_0.imag, s=5)
plt.scatter(iq_1.real, iq_1.imag, s=5)
if save:
plt.savefig(exp_path_0 / "a_figure.png")
plt.show()
# define the mean value for 0 and 1
mean_0 = iq_0.mean()
mean_1 = iq_1.mean()
# find the middle point between the two means
mid_point = (mean_0 + mean_1) / 2
# rotate around the middle point and move back
angle = np.angle(mean_1 - mean_0)
# Rotate the data so that the line joining the two means is parallel to the X axis
iq_0_rot: np.ndarray = (iq_0 - mid_point) * np.exp(-1j * angle) + mid_point
iq_1_rot: np.ndarray = (iq_1 - mid_point) * np.exp(-1j * angle) + mid_point
mean_0_rot = (mean_0 - mid_point) * np.exp(-1j * angle) + mid_point
mean_1_rot = (mean_1 - mid_point) * np.exp(-1j * angle) + mid_point
# plot a scatter to show the data, show the two means and a line connecting them
plt.figure(figsize=(10, 5))
plt.subplot(1, 2, 1)
plt.scatter(iq_0_rot.real, iq_0_rot.imag, s=5, alpha=0.5, label="Rotated 0")
plt.scatter(iq_1_rot.real, iq_1_rot.imag, s=5, alpha=0.5, label="Rotated 1")
plt.plot(
[mean_0_rot.real, mean_1_rot.real],
[mean_0_rot.imag, mean_1_rot.imag],
color="red",
marker="o",
label="Means Line",
)
plt.title("Rotated IQ Plane")
plt.xlabel("I (Real)")
plt.ylabel("Q (Imag)")
plt.legend()
plt.grid(True)
# consider only the values on the x axis, create cumulative distribution
# for both, plot the two cumulative distributions
plt.subplot(1, 2, 2)
# Take only the X (real) component after the rotation
x_0 = iq_0_rot.real
x_1 = iq_1_rot.real
# get the limits of the plot
x_max = max(x_0.max(), x_1.max())
x_min = min(x_0.min(), x_1.min())
# create a linspace for the x axis of the CDF
# 1. Sort your input arrays first (required for searchsorted)
x_0_sorted = np.sort(x_0)
x_1_sorted = np.sort(x_1)
# 2. Create your grid
x_cdf = np.linspace(x_min, x_max, endpoint=True, num=(x_0.shape[0] + x_1.shape[0]))
# 3. Use searchsorted to find how many elements are strictly less than each point in x_cdf
# 'left' ensures we find the index where x_0_sorted < x_cdf[i]
x_0_cdf = np.searchsorted(x_0_sorted, x_cdf, side="left")
x_1_cdf = np.searchsorted(x_1_sorted, x_cdf, side="left")
# normalize to 1
x_0_cdf = x_0_cdf / x_0.shape[0]
x_1_cdf = x_1_cdf / x_1.shape[0]
# find where the two distances are maximized
x_cdf_abs_diff = np.abs(x_0_cdf - x_1_cdf)
threshold = np.argmax(x_cdf_abs_diff)
fidelity = 100 * x_cdf_abs_diff[threshold]
print(f"maximum fidelity with threshold: {threshold} is equal to: {fidelity:.2f}%")
# Plot the two cumulative distributions (CDF)
plt.plot(x_cdf, x_0_cdf, label="CDF 0", linewidth=1.5)
plt.plot(x_cdf, x_1_cdf, label="CDF 1", linewidth=1.5)
plt.plot(
[x_cdf[threshold]] * 2,
[x_0_cdf[threshold], x_1_cdf[threshold]],
color="red",
marker="o",
label=f"Point of maximum fidelity: {fidelity:.2f}%",
)
plt.title("Cumulative Distribution (X-axis)")
plt.xlabel("Projected X value")
plt.ylabel("Probability")
plt.legend()
plt.grid(True)
plt.tight_layout()
if save:
plt.savefig(exp_path_0 / "rotated_analysis.png")
plt.show()
