m-thesis-introduction/airshower_beacon_simulation/dc_grid_power_time_fixes.py

#!/usr/bin/env python3
# vim: fdm=indent ts=4

"""
Show how the Power changes when incorporating the
various clock offsets by plotting on a grid.
"""

import matplotlib.pyplot as plt
from mpl_toolkits.mplot3d import Axes3D # required for projection='3d' on old matplotliblib versions
import numpy as np
from os import path
import joblib

from earsim import REvent
from atmocal import AtmoCal
import aa_generate_beacon as beacon
import lib
from lib import rit

try:
    from tqdm import tqdm
except:
    tqdm = lambda x: x

def save_overlapping_traces_figure(test_location, ev, N_plot = 30, wx=200, title_extra=None, fname_distinguish='', fig_dir=None, location_in_plot_text_loc=(0.02, 0.95), location_text=None, **fig_kwargs):
    P, t_, a_, a_sum, t_sum = rit.pow_and_time(test_location, ev, dt=1)

    fig, axs = plt.subplots(**fig_kwargs)
    #axs.set_title(
            #"Antenna traces" + ("\n" if title_extra is not None else '')  + 
    #        title_extra )
    axs.set_xlabel("Time [ns]")
    axs.set_ylabel("Amplitude [$\\mu V/m$]")
    if False and location_in_plot_text_loc:
        text_loc = location_in_plot_text_loc

        if not location_text:
            shower_plane_location = rit.location_to_shower_plane(test_location, ev=ev)
            location_text = '(' + ', '.join(['{:.1g}'.format(x) for x in shower_plane_location]) + ')'

        axs.text(*text_loc, location_text, ha='left', transform=axs.transAxes)

    a_max = [ np.amax(ant.E_AxB) for ant in ev.antennas ]
    power_sort_idx = np.argsort(a_max)

    if N_plot is None:
        N_plot = len(a_max)

    for i, idx in enumerate(reversed(power_sort_idx)):
        if i >= N_plot:
            break

        alpha = max(0.4, 1/N_plot)

        axs.plot(t_[idx], a_[idx], color='r', alpha=alpha, lw=2)

    axs.set_ylim([None, max(a_max)*1.2])
    if True:
        axs.set_xlim([-28000, -27300])

    if fig_dir:
        if fname_distinguish:
            fname_distinguish = "." + fname_distinguish

        fig.tight_layout()

        fig.savefig(path.join(fig_dir, path.basename(__file__) + f'{fname_distinguish}.trace_overlap.{case}.pdf'))
        #fig.savefig(path.join(fig_dir, path.basename(__file__) + f'{fname_distinguish}.trace_overlap.{case}.png'), transparent=True)

        # Take center between t_low and t_high
        if True:
            orig_xlims = axs.get_xlim()

            if not True: # t_high and t_low from strongest signal
                t_low = np.min(t_[power_sort_idx[-1]])
                t_high = np.max(t_[power_sort_idx[-1]])
            else: # take t_high and t_low from plotted signals
                a = [np.min(t_[idx]) for idx in power_sort_idx[-N_plot:]]
                t_low  = np.nanmin(a)
                b = [np.max(t_[idx]) for idx in power_sort_idx[-N_plot:]]
                t_high = np.nanmax(b)

                if False:
                    axs.plot(a, [0]*N_plot, 'gx', ms=10)
                    axs.plot(b, [0]*N_plot, 'b+', ms=10)

            center_x = (t_high - t_low)/2 + t_low

            low_xlim  = max(orig_xlims[0], center_x - wx)
            high_xlim = min(orig_xlims[1], center_x + wx)

            axs.set_xlim(low_xlim, high_xlim)
            fig.savefig(path.join(fig_dir, path.basename(__file__) + f'{fname_distinguish}.trace_overlap.zoomed.{case}.pdf'))
            #fig.savefig(path.join(fig_dir, path.basename(__file__) + f'{fname_distinguish}.trace_overlap.zoomed.{case}.png'), transparent=True)

    return fig

if __name__ == "__main__":
    valid_cases = ['no_offset', 'repair_none', 'repair_phases', 'repair_full']

    import sys
    import os
    import matplotlib
    if os.name == 'posix' and "DISPLAY" not in os.environ:
        matplotlib.use('Agg')

    atm = AtmoCal()

    from scriptlib import MyArgumentParser
    parser = MyArgumentParser()
    parser.add_argument('--input-fname', type=str, default=None, help='Path to mysim.sry, either directory or path. If empty it takes DATA_DIR and appends mysim.sry. (Default: %(default)s)')
    parser.add_argument('-X', type=int, default=400, help='Atmospheric depth to start the initial grid.')

    group = parser.add_argument_group('figures')
    for case in valid_cases:
        group.add_argument('--'+case.replace('_','-'), dest='figures', action='append_const', const=case)

    args = parser.parse_args()

    if not args.input_fname:
        args.input_fname = args.data_dir

    if path.isdir(args.input_fname):
        args.input_fname = path.join(args.input_fname, "mysim.sry")

    wanted_cases = args.figures

    if not wanted_cases or 'all' in wanted_cases:
        wanted_cases = valid_cases

    figsize = (6, 4.5)
    figsize_map =  (6, 4.8)
    if True:
        from matplotlib import rcParams
        #rcParams["text.usetex"] = True
        rcParams["font.family"] = "serif"
        rcParams["font.size"] = "15"
        rcParams["grid.linestyle"] = 'dotted'
        rcParams["figure.figsize"] = figsize
        figsize = rcParams['figure.figsize']

    fig_dir = args.fig_dir
    show_plots = args.show_plots

    remove_beacon_from_traces = True
    apply_signal_window_from_max = True

    ####
    fname_dir = args.data_dir
    antennas_fname = path.join(fname_dir, beacon.antennas_fname)
    tx_fname = path.join(fname_dir, beacon.tx_fname)
    beacon_snr_fname = path.join(fname_dir, beacon.beacon_snr_fname)

    # create fig_dir
    if fig_dir:
        os.makedirs(fig_dir, exist_ok=True)

    # Read in antennas from file
    _, tx, antennas = beacon.read_beacon_hdf5(antennas_fname)
    _, __, txdata = beacon.read_tx_file(tx_fname)
    # Read original REvent
    ev = REvent(args.input_fname)
    bak_ants = ev.antennas
    # .. patch in our antennas
    ev.antennas = antennas
    # Read in snr info
    beacon_snrs = beacon.read_snr_file(beacon_snr_fname)
    snr_str = f"$\\langle SNR \\rangle$ = {beacon_snrs['mean']: .1g}"

    ##
    ## Setup grid
    ##
    X = args.X
    zgr = 0 #not exact
    dXref = atm.distance_to_slant_depth(np.deg2rad(ev.zenith),X,zgr+ev.core[2])
    scale2d = dXref*np.tan(np.deg2rad(2.))
    scale4d = dXref*np.tan(np.deg2rad(4.))
    scale02d = dXref*np.tan(np.deg2rad(0.2))

    Nx, Ny = 31, 31
    #Nx, Ny = 15, 15
    scales = {
            'scale2d': scale2d,
            'scale4d': scale4d,
            'scale02d': scale02d,
            }

    N_plot = None
    trace_zoom_wx = 100

    plot_titling = {
            'no_offset': "no clock offsets",
            'repair_none': "randomised clock offsets",
            'repair_phases': "phase clock offsets",
            'repair_full': "phase + period clock offsets"
            }

    power_on_grid_sc_kwargs = dict(
            s=90, # 250 for (10,8)
            cmap='inferno'
            )

    # For now only implement using one freq_name
    freq_names = ev.antennas[0].beacon_info.keys()
    if len(freq_names) > 1:
        raise NotImplementedError

    freq_name = next(iter(freq_names))

    # Pre remove the beacon from the traces
    #
    # We need to remove it here, so we do not shoot ourselves in
    # the foot when changing to the various clock offsets.
    #
    # Note that the bandpass filter is applied only after E_AxB is
    # reconstructed so we have to manipulate the original traces.
    if remove_beacon_from_traces:
        tx_amps = txdata['amplitudes']
        tx_amps_sum = np.sum(tx_amps)

        for i, ant in enumerate(ev.antennas):
            beacon_phase = ant.beacon_info[freq_name]['beacon_phase']
            f = ant.beacon_info[freq_name]['freq']
            ampl_AxB = ant.beacon_info[freq_name]['amplitude']
            calc_beacon = lib.sine_beacon(f, ev.antennas[i].t, amplitude=ampl_AxB, phase=beacon_phase)

            # Split up contribution to the various polarisations
            for j, amp in enumerate(tx_amps):
                if j == 0:
                    ev.antennas[i].Ex -= amp*(1/tx_amps_sum)*calc_beacon
                elif j == 1:
                    ev.antennas[i].Ey -= amp*(1/tx_amps_sum)*calc_beacon
                elif j == 2:
                    ev.antennas[i].Ez -= amp*(1/tx_amps_sum)*calc_beacon

            # Subtract the beacon from E_AxB
            ev.antennas[i].E_AxB -= calc_beacon

    # Slice the traces to a small part around the peak
    if apply_signal_window_from_max:
        N_pre, N_post = 250, 250 # TODO: make this configurable

        for i, ant in enumerate(ev.antennas):
            # Get max idx from all the traces
            # and select the strongest
            max_idx = []
            maxs = []

            for trace in [ant.Ex, ant.Ey, ant.Ez]:
                idx = np.argmax(np.abs(trace))
                max_idx.append(idx)
                maxs.append( np.abs(trace[idx]) )

            idx = np.argmax(maxs)
            max_idx = max_idx[idx]


            low_idx = max(0, max_idx-N_pre)
            high_idx = min(len(ant.t), max_idx+N_post)

            ev.antennas[i].t = ant.t[low_idx:high_idx]
            ev.antennas[i].t_AxB = ant.t_AxB[low_idx:high_idx]

            ev.antennas[i].Ex = ant.Ex[low_idx:high_idx]
            ev.antennas[i].Ey = ant.Ey[low_idx:high_idx]
            ev.antennas[i].Ez = ant.Ez[low_idx:high_idx]
            ev.antennas[i].E_AxB = ant.E_AxB[low_idx:high_idx]

    ## Apply polarisation and bandpass filter
    rit.set_pol_and_bp(ev)

    # backup antenna times
    backup_antenna_t = [ ant.t for ant in ev.antennas ]
    backup_antenna_t_AxB = [ ant.t_AxB for ant in ev.antennas ]

    fig = save_overlapping_traces_figure([0,0,0], ev, N_plot=1, wx=trace_zoom_wx, title_extra = plot_titling[case], fname_distinguish=f'single', fig_dir=fig_dir, figsize=figsize)
    plt.close(fig)

    no_offset_maximum_power = None # set in below loop
    with joblib.parallel_backend("loky"):
        # always force no_offset first to determine the reference vmax in the colorbar
        for case in ['no_offset'] + [c for c in wanted_cases if c != 'no_offset']:
            print(f"Starting {case} figure")
            # Repair clock offsets with the measured offsets
            transl_modes = {'no_offset':'orig', 'repair_phases':'phases', 'repair_full':'full'}

            if case in transl_modes:
                transl_mode = transl_modes[case]
                measured_offsets = beacon.read_antenna_clock_repair_offsets(antennas, mode=transl_mode, freq_name=freq_name)
            else:
                print(f"Warning: unknown repair case requested '{case}', defaulting to none.")
                measured_offsets = [0]*len(ev.antennas)

            for i, ant in enumerate(ev.antennas):
                total_clock_offset = measured_offsets[i]

                ev.antennas[i].t =     backup_antenna_t[i] + total_clock_offset
                ev.antennas[i].t_AxB = backup_antenna_t_AxB[i] + total_clock_offset

                if i == 0:
                    # Specifically compare the times
                    print("backup time, time with measured_offset, true clock offset, measured clock offset")
                    print(bak_ants[i].t[0], ev.antennas[i].t[0], ev.antennas[i].attrs['clock_offset'], measured_offsets[i])

            #
            # determine the maximum power for the true timing
            #
            if case == 'no_offset':
                xx, yy, p, maxp_loc = rit.shower_plane_slice(ev, X=X, Nx=1, Ny=1, wx=0, wy=0, zgr=zgr)

                no_offset_maximum_power = p
                if True and np.isclose(X, 400):
                    no_offset_maximum_power = 200


            # we forced no_offset for determing the reference vmax
            # simply continue with the rest
            if case not in wanted_cases:
                continue

            #
            # Plot overlapping traces at 0,0,0
            #
            loc = (0,0,0)
            fig = save_overlapping_traces_figure(loc, ev, N_plot=N_plot, wx=trace_zoom_wx, title_extra = plot_titling[case], fname_distinguish=f'{case}.0', fig_dir=fig_dir, figsize=figsize)
            plt.close(fig)

            #
            # Plot overlapping traces at simulation shower axis
            #
            dX = atm.distance_to_slant_depth(np.deg2rad(ev.zenith), X, zgr)
            loc = (0)*ev.uAxB + (0)*ev.uAxAxB + dX * ev.uA
            loc[0:1] = 0

            fig = save_overlapping_traces_figure(loc, ev, N_plot=N_plot, wx=trace_zoom_wx, title_extra = plot_titling[case], fname_distinguish=f'{case}.axis.X{X}', location_text=f"on simulation axis, $X={X}$", fig_dir=fig_dir, figsize=figsize)
            plt.close(fig)

            if not True:
                continue;

            #
            # Measure power on grid
            # and plot overlapping traces at position with highest power
            #
            for scalename, scale in tqdm(scales.items()):
                wx, wy = scale, scale
                print(f"Starting grid measurement for figure {case} with {scalename}")
                xx, yy, p, maxp_loc = rit.shower_plane_slice(ev, X=X, Nx=Nx, Ny=Nx, wx=wx, wy=wy, zgr=zgr)
                fig, axs = rit.slice_figure(ev, X, xx, yy, p, mode='sp', scatter_kwargs={**dict(
                    vmax=no_offset_maximum_power,
                    vmin=0,
                    ), **power_on_grid_sc_kwargs},
                    figsize=figsize_map)

                suptitle = fig._suptitle.get_text()
                fig.suptitle("")
                #axs.set_title("Shower plane slice\n" + plot_titling[case] + "\n" + suptitle)
                axs.set_aspect('equal', 'datalim')
                #axs.legend(title=snr_str,loc='upper right')

                axs.set_xlim(1.1*min(xx)/1e3, 1.1*max(xx)/1e3)
                axs.set_ylim(1.1*min(yy)/1e3, 1.1*max(yy)/1e3)

                if fig_dir:
                    fig.tight_layout()
                    fig.savefig(path.join(fig_dir, path.basename(__file__) + f'.X{X}.{case}.{scalename}.pdf'))
                plt.close(fig)

                #
                # Plot overlapping traces at highest power of each scale
                #
                fig = save_overlapping_traces_figure(maxp_loc, ev, N_plot=N_plot, wx=trace_zoom_wx, title_extra = plot_titling[case] + ', ' + scalename + ' best', fname_distinguish=scalename+f'.best.X{X}', fig_dir=fig_dir, figsize=figsize)

                #
                # and plot overlapping traces at two other locations
                #
                if False:
                    for dist in [ 0.5, 5, 10, 50, 100]:
                        # only add distance horizontally
                        location = maxp_loc + np.sqrt(dist*1e3)*np.array([1,1,0])

                        fig = save_overlapping_traces_figure(location, ev, N_plot=N_plot, wx=wx, title_extra = plot_titling[case] + ', ' + scalename + f', x + {dist}km', fname_distinguish=f'{scalename}.x{dist}.X{X}', fig_dir=fig_dir, figsize=figsize)
                        plt.close(fig)


    if args.show_plots:
        plt.show()