mirror of
https://gitlab.science.ru.nl/mthesis-edeboone/m-thesis-introduction.git
synced 2024-11-13 10:03:32 +01:00
170 lines
6 KiB
Python
Executable file
170 lines
6 KiB
Python
Executable file
#!/usr/bin/env python3
|
|
# vim: fdm=indent ts=4
|
|
|
|
import h5py
|
|
from itertools import combinations, zip_longest
|
|
import matplotlib.pyplot as plt
|
|
import numpy as np
|
|
|
|
import aa_generate_beacon as beacon
|
|
import lib
|
|
|
|
|
|
if __name__ == "__main__":
|
|
from os import path
|
|
import sys
|
|
|
|
import os
|
|
import matplotlib
|
|
if os.name == 'posix' and "DISPLAY" not in os.environ:
|
|
matplotlib.use('Agg')
|
|
|
|
fname = "ZH_airshower/mysim.sry"
|
|
|
|
show_plots = True
|
|
ref_ant_id = None # leave None for all baselines
|
|
|
|
####
|
|
fname_dir = path.dirname(fname)
|
|
antennas_fname = path.join(fname_dir, beacon.antennas_fname)
|
|
time_diffs_fname = 'time_diffs.hdf5' if False else antennas_fname
|
|
fig_dir = "./figures" # set None to disable saving
|
|
|
|
basenames, time_diffs, f_beacons, true_phase_diffs, k_periods = beacon.read_baseline_time_diffs_hdf5(time_diffs_fname)
|
|
|
|
f_beacon, tx, antennas = beacon.read_beacon_hdf5(antennas_fname)
|
|
|
|
# TODO: allow multiple frequencies
|
|
if (f_beacon != f_beacons).any():
|
|
raise NotImplementedError
|
|
|
|
N_base = len(basenames)
|
|
N_ant = len(antennas)
|
|
|
|
# reshape time_diffs into N_ant x N_ant array
|
|
sigma_phase_matrix = np.full( (N_ant, N_ant), np.nan, dtype=float)
|
|
#sigma_phase_matrix = np.arange(0, N_ant**2, dtype=float).reshape( (N_ant,N_ant))
|
|
|
|
name2idx = lambda name: int(name)-1
|
|
for i, b in enumerate(basenames):
|
|
idx = (name2idx(b[0]), name2idx(b[1]))
|
|
|
|
if idx[0] == idx[1]:
|
|
pass
|
|
|
|
sigma_phase_matrix[(idx[0], idx[1])] = true_phase_diffs[i]
|
|
sigma_phase_matrix[(idx[1], idx[0])] = true_phase_diffs[i]
|
|
|
|
# for each row j subtract the 0,j element from the whole row
|
|
# and apply phase_mod
|
|
first_row = sigma_phase_matrix[0]
|
|
|
|
sigma_phase_matrix = sigma_phase_matrix - first_row[:,np.newaxis]
|
|
sigma_phase_matrix = lib.phase_mod(sigma_phase_matrix)
|
|
|
|
|
|
# Except for the first row, these are all separate measurements
|
|
# Condense into phase offset per antenna
|
|
if True: # do not use the first row
|
|
my_mask = np.arange(1, len(sigma_phase_matrix), dtype=int)
|
|
else:
|
|
my_mask = np.arange(0, len(sigma_phase_matrix), dtype=int)
|
|
|
|
mean_sigma_phase = np.nanmean(sigma_phase_matrix[my_mask], axis=0)
|
|
std_sigma_phase = np.nanstd( sigma_phase_matrix[my_mask], axis=0)
|
|
|
|
# write into antenna hdf5
|
|
with h5py.File(antennas_fname, 'a') as fp:
|
|
group = fp['antennas']
|
|
|
|
freq_name = None
|
|
for i, ant in enumerate(antennas):
|
|
h5ant = group[ant.name]
|
|
|
|
h5beacon_info = h5ant['beacon_info']
|
|
|
|
# find out freq_name
|
|
if freq_name is None:
|
|
freq_name = [ k for k in h5beacon_info.keys() if np.isclose(h5beacon_info[k].attrs['freq'], f_beacon)][0]
|
|
|
|
h5attrs = h5beacon_info[freq_name].attrs
|
|
|
|
idx = name2idx(ant.name)
|
|
h5attrs['sigma_phase_mean'] = mean_sigma_phase[idx]
|
|
h5attrs['sigma_phase_std'] = std_sigma_phase[idx]
|
|
|
|
|
|
##############################
|
|
# Compare actual time shifts #
|
|
##############################
|
|
antenna_time_shifts = { a.name: a.attrs['clock_offset'] for a in antennas }
|
|
|
|
if True:
|
|
# show means and std
|
|
fig, axs = plt.subplots(1,2, sharey=True)
|
|
fig.suptitle("Comparison Measured and Actual Phases")
|
|
axs[0].set_xlabel("Antenna no.")
|
|
axs[0].set_ylabel("Antenna Phase $\\Delta_\\varphi$")
|
|
axs[1].set_xlabel("#")
|
|
if True:
|
|
forward = lambda x: x/(2*np.pi*f_beacon)
|
|
inverse = lambda x: 2*np.pi*x*f_beacon
|
|
secax = axs[1].secondary_yaxis('right', functions=(forward, inverse))
|
|
secax.set_ylabel('Time $\\Delta\\varphi/(2\\pi f_{beac})$ [ns]')
|
|
|
|
l = axs[0].errorbar(np.arange(N_ant), mean_sigma_phase, yerr=std_sigma_phase, marker='4', alpha=0.7, ls='none')
|
|
|
|
axs[1].hist(mean_sigma_phase, bins='sqrt', density=False, orientation='horizontal', color=l[0].get_color(), histtype='step')
|
|
|
|
# Actual time shifts
|
|
if True:
|
|
actual_phase_shifts = [ -1*lib.phase_mod(2*np.pi*f_beacon*v) for k,v in antenna_time_shifts.items() ]
|
|
antenna_names = [int(k)-1 for k,v in antenna_time_shifts.items() ]
|
|
|
|
# make sure to keep the same offset
|
|
if False:
|
|
phase_offset = mean_sigma_phase[0] - actual_phase_shifts[0]
|
|
|
|
actual_phase_shifts += phase_offset
|
|
|
|
l = axs[0].plot(antenna_names, actual_phase_shifts, ls='none', marker='3', alpha=0.8)
|
|
|
|
if True:
|
|
axs[1].hist(actual_phase_shifts, bins='sqrt', density=False, orientation='horizontal', ls='dashed', color=l[0].get_color(), histtype='step')
|
|
|
|
fig.tight_layout()
|
|
if fig_dir:
|
|
fig.savefig(path.join(fig_dir, __file__ + f".residuals.pdf"))
|
|
|
|
##########################
|
|
##########################
|
|
##########################
|
|
|
|
actual_time_shifts = []
|
|
for i,b in enumerate(basenames):
|
|
actual_time_shift = lib.phase_mod(lib.phase_mod(antenna_time_shifts[b[0]]*2*np.pi*f_beacon) - lib.phase_mod(antenna_time_shifts[b[1]]*2*np.pi*f_beacon))
|
|
|
|
actual_time_shifts.append(actual_time_shift)
|
|
|
|
# unpack mean_sigma_phase back into a list of time diffs
|
|
measured_time_diffs = []
|
|
for i,b in enumerate(basenames):
|
|
time0, time1 = mean_sigma_phase[name2idx(b[0])], mean_sigma_phase[name2idx(b[1])]
|
|
measured_time_diffs.append(time1 - time0)
|
|
|
|
# Make a plot
|
|
if True:
|
|
fig, ax = plt.subplots()
|
|
ax.set_xlabel("Baseline no.")
|
|
ax.set_ylabel("$\\Delta t$[ns]")
|
|
if True: # indicate single beacon period span
|
|
ax.plot((-1, -1), (0, 1/f_beacon), marker='3', ms=10, label='1/f_beacon')
|
|
ax.plot(np.arange(N_base), actual_time_shifts, marker='+', label='actual time shifts')
|
|
ax.plot(np.arange(N_base), measured_time_diffs, marker='x', label='calculated')
|
|
|
|
ax.legend()
|
|
if fig_dir:
|
|
fig.savefig(path.join(fig_dir, __file__ + f".calculated_shifts.pdf"))
|
|
|
|
if show_plots:
|
|
plt.show()
|