mirror of
https://gitlab.science.ru.nl/mthesis-edeboone/m-thesis-introduction.git
synced 2024-11-13 10:03:32 +01:00
117 lines
3.7 KiB
Python
117 lines
3.7 KiB
Python
"""
|
|
Routines to assist in plotting
|
|
"""
|
|
|
|
import matplotlib.pyplot as plt
|
|
import numpy as np
|
|
|
|
def annotate_width(ax, name, x1, x2, y, text_kw={}, arrow_kw={}):
|
|
default_arrow_kw = dict(
|
|
xy = (x1, y),
|
|
xytext = (x2,y),
|
|
arrowprops = dict(
|
|
arrowstyle="<->",
|
|
shrinkA=False,
|
|
shrinkB=False
|
|
),
|
|
)
|
|
|
|
default_text_kw = dict(
|
|
va='bottom',
|
|
ha='center',
|
|
xy=((x1+x2)/2, y)
|
|
)
|
|
|
|
an1 = ax.annotate("", **{**default_arrow_kw, **arrow_kw})
|
|
an2 = ax.annotate(name, **{**default_text_kw, **text_kw})
|
|
|
|
return [an1, an2]
|
|
|
|
|
|
def beacon_sync_figure(
|
|
time, impulses, beacons,
|
|
delta_t=0,
|
|
beacon_offsets=[],
|
|
impulse_offsets=[],
|
|
f_beacon=1,
|
|
colors=['y','g'],
|
|
show_annotations=False,
|
|
multiplier_name = ['m','n'],
|
|
fig_kwargs = {'figsize': (12,4)},
|
|
ns=1e-3
|
|
):
|
|
if not hasattr(delta_t, "__len__"):
|
|
delta_t = np.array([0, delta_t])
|
|
|
|
if not hasattr(impulse_offsets, "__len__"):
|
|
impulse_offsets = np.repeat(impulse_offsets, 2)
|
|
|
|
if not hasattr(beacon_offsets, "__len__"):
|
|
beacon_offsets = np.repeat(beacon_offsets, 2)
|
|
|
|
N_axes = 2
|
|
if show_annotations:
|
|
N_axes += 1
|
|
|
|
fig, axes = plt.subplots(N_axes,1, sharex=True, **fig_kwargs)
|
|
axes[-1].set_xlabel("Time [ns]")
|
|
for i in range(0, 2):
|
|
axes[i].set_yticks([],[])
|
|
axes[i].set_ylabel("Antenna {:d}".format(i+1))
|
|
axes[i].plot((time-delta_t[i])/ns, impulses[i])
|
|
axes[i].plot((time-delta_t[i])/ns, beacons[i], marker='.')
|
|
|
|
|
|
# indicate timing of pulses
|
|
for i, impulse_offset in enumerate(impulse_offsets):
|
|
kwargs = dict(color=colors[i])
|
|
|
|
axes_list = [axes[i]]
|
|
if show_annotations:
|
|
axes_list.append(axes[-1])
|
|
|
|
[ax.axvline((impulse_offset-delta_t[i])/ns, **kwargs) for ax in axes_list]
|
|
|
|
|
|
# indicate timing of beacon
|
|
for i, beacon_offset in enumerate(beacon_offsets):
|
|
kwargs = dict(color=colors[i], ls=(0, (3,2)))
|
|
tick_kwargs = dict(color='k', alpha=0.2)
|
|
|
|
axes_list = [axes[i]]
|
|
if show_annotations:
|
|
axes_list.append(axes[-1])
|
|
|
|
|
|
# indicate every period of the beacon
|
|
beacon_ticks = beacon_offset + [(n)*1/f_beacon for n in range(1+int((time[-1] - time[0]) * f_beacon))]
|
|
|
|
[axes[i].axvline((tick-delta_t[i])/ns, **{**kwargs, **tick_kwargs}) for tick in beacon_ticks]
|
|
|
|
# reference period in beacon
|
|
# is the first tick > 0
|
|
ref_tick = beacon_ticks[0]
|
|
[ax.axvline((ref_tick-delta_t[i])/ns, **kwargs) for ax in axes_list]
|
|
|
|
if show_annotations:
|
|
# annotate width between impulse and closest beacon tick
|
|
# and closest beacon tick and reference tick
|
|
closest_beacon_tick_id = np.argmin(np.abs(beacon_ticks-impulse_offsets[i]))
|
|
if closest_beacon_tick_id != 0 and beacon_ticks[closest_beacon_tick_id] > impulse_offsets[i]:
|
|
closest_beacon_tick_id -= 1
|
|
closest_beacon_tick = beacon_ticks[closest_beacon_tick_id]
|
|
|
|
annotate_width(axes[i], f"$A_{i+1}$", (closest_beacon_tick - delta_t[i])/ns, (impulse_offsets[i]-delta_t[i])/ns, 0.7)
|
|
annotate_width(axes[i], f"$B_{i+1}={multiplier_name[i]}T$", (closest_beacon_tick-delta_t[i])/ns, (ref_tick-delta_t[i])/ns, 0.4)
|
|
|
|
|
|
if show_annotations:
|
|
axes[-1].set_yticks([],[])
|
|
|
|
# annotate width between beacon reference periods
|
|
annotate_width(axes[-1], "$t_\phi$", (beacon_offsets[0]-delta_t[0])/ns, (beacon_offsets[-1]-delta_t[-1])/ns, 0.4)
|
|
|
|
# annotate width between pulses
|
|
annotate_width(axes[-1], "$\Delta t$", (impulse_offsets[0]-delta_t[0])/ns, (impulse_offsets[-1]-delta_t[-1])/ns, 0.4)
|
|
|
|
return fig, axes
|