#!/usr/bin/env python3

#################################################
######## CoreXY BELTS CALIBRATION SCRIPT ########
#################################################
# Written by Frix_x#0161 #

# Be sure to make this script executable using SSH: type 'chmod +x ./graph_belts.py' when in the folder!

#####################################################################
################ !!! DO NOT EDIT BELOW THIS LINE !!! ################
#####################################################################

import optparse, matplotlib, sys, importlib, os, pathlib
from collections import namedtuple
import numpy as np
import matplotlib.pyplot, matplotlib.dates, matplotlib.font_manager
import matplotlib.ticker, matplotlib.gridspec, matplotlib.colors
import matplotlib.patches
import locale
import time
import glob
import shaper_calibrate
from datetime import datetime

matplotlib.use('Agg')


ALPHABET = "ABCDEFGHIJKLMNOPQRSTUVWXYZ" # For paired peaks names

PEAKS_DETECTION_THRESHOLD = 0.20
CURVE_SIMILARITY_SIGMOID_K = 0.6
DC_GRAIN_OF_SALT_FACTOR = 0.75
DC_THRESHOLD_METRIC = 1.5e9
DC_MAX_UNPAIRED_PEAKS_ALLOWED = 4

# Define the SignalData namedtuple
SignalData = namedtuple('CalibrationData', ['freqs', 'psd', 'peaks', 'paired_peaks', 'unpaired_peaks'])

KLIPPAIN_COLORS = {
    "purple": "#70088C",
    "orange": "#FF8D32",
    "dark_purple": "#150140",
    "dark_orange": "#F24130",
    "red_pink": "#F2055C"
}


# Set the best locale for time and date formating (generation of the titles)
try:
    locale.setlocale(locale.LC_TIME, locale.getdefaultlocale())
except locale.Error:
    locale.setlocale(locale.LC_TIME, 'C')

# Override the built-in print function to avoid problem in Klipper due to locale settings
original_print = print
def print_with_c_locale(*args, **kwargs):
    original_locale = locale.setlocale(locale.LC_ALL, None)
    locale.setlocale(locale.LC_ALL, 'C')
    original_print(*args, **kwargs)
    locale.setlocale(locale.LC_ALL, original_locale)
print = print_with_c_locale


def is_file_open(filepath):
    for proc in os.listdir('/proc'):
        if proc.isdigit():
            for fd in glob.glob(f'/proc/{proc}/fd/*'):
                try:
                    if os.path.samefile(fd, filepath):
                        return True
                except FileNotFoundError:
                    # Klipper has already released the CSV file
                    pass
                except PermissionError:
                    # Unable to check for this particular process due to permissions
                    pass
    return False


######################################################################
# Computation of the PSD graph
######################################################################

# Calculate estimated "power spectral density" using existing Klipper tools
def calc_freq_response(data):
    helper = shaper_calibrate.ShaperCalibrate(printer=None)
    return helper.process_accelerometer_data(data)


# Calculate or estimate a "similarity" factor between two PSD curves and scale it to a percentage. This is
# used here to quantify how close the two belts path behavior and responses are close together.
def compute_curve_similarity_factor(signal1, signal2):
    freqs1 = signal1.freqs
    psd1 = signal1.psd
    freqs2 = signal2.freqs
    psd2 = signal2.psd
    
    # Interpolate PSDs to match the same frequency bins and do a cross-correlation
    psd2_interp = np.interp(freqs1, freqs2, psd2)
    cross_corr = np.correlate(psd1, psd2_interp, mode='full')
    
    # Find the peak of the cross-correlation and compute a similarity normalized by the energy of the signals
    peak_value = np.max(cross_corr)
    similarity = peak_value / (np.sqrt(np.sum(psd1**2) * np.sum(psd2_interp**2)))

    # Apply sigmoid scaling to get better numbers and get a final percentage value
    scaled_similarity = sigmoid_scale(-np.log(1 - similarity), CURVE_SIMILARITY_SIGMOID_K)
    
    return scaled_similarity


# This find all the peaks in a curve by looking at when the derivative term goes from positive to negative
# Then only the peaks found above a threshold are kept to avoid capturing peaks in the low amplitude noise of a signal
def detect_peaks(psd, freqs, window_size=5, vicinity=3):
    # Smooth the curve using a moving average to avoid catching peaks everywhere in noisy signals
    kernel = np.ones(window_size) / window_size
    smoothed_psd = np.convolve(psd, kernel, mode='valid')
    mean_pad = [np.mean(psd[:window_size])] * (window_size // 2)
    smoothed_psd = np.concatenate((mean_pad, smoothed_psd))
    
    # Find peaks on the smoothed curve
    smoothed_peaks = np.where((smoothed_psd[:-2] < smoothed_psd[1:-1]) & (smoothed_psd[1:-1] > smoothed_psd[2:]))[0] + 1
    detection_threshold = PEAKS_DETECTION_THRESHOLD * psd.max()
    smoothed_peaks = smoothed_peaks[smoothed_psd[smoothed_peaks] > detection_threshold]
    
    # Refine peak positions on the original curve
    refined_peaks = []
    for peak in smoothed_peaks:
        local_max = peak + np.argmax(psd[max(0, peak-vicinity):min(len(psd), peak+vicinity+1)]) - vicinity
        refined_peaks.append(local_max)
    
    return np.array(refined_peaks), freqs[refined_peaks]


# This function create pairs of peaks that are close in frequency on two curves (that are known
# to be resonances points and must be similar on both belts on a CoreXY kinematic)
def pair_peaks(peaks1, freqs1, psd1, peaks2, freqs2, psd2):
    # Compute a dynamic detection threshold to filter and pair peaks efficiently
    # even if the signal is very noisy (this get clipped to a maximum of 10Hz diff)
    distances = []
    for p1 in peaks1:
        for p2 in peaks2:
            distances.append(abs(freqs1[p1] - freqs2[p2]))
    distances = np.array(distances)
    
    median_distance = np.median(distances)
    iqr = np.percentile(distances, 75) - np.percentile(distances, 25)
    
    threshold = median_distance + 1.5 * iqr
    threshold = min(threshold, 10)
    
    # Pair the peaks using the dynamic thresold
    paired_peaks = []
    unpaired_peaks1 = list(peaks1)
    unpaired_peaks2 = list(peaks2)
    
    while unpaired_peaks1 and unpaired_peaks2:
        min_distance = threshold + 1
        pair = None
        
        for p1 in unpaired_peaks1:
            for p2 in unpaired_peaks2:
                distance = abs(freqs1[p1] - freqs2[p2])
                if distance < min_distance:
                    min_distance = distance
                    pair = (p1, p2)
        
        if pair is None: # No more pairs below the threshold
            break
        
        p1, p2 = pair
        paired_peaks.append(((p1, freqs1[p1], psd1[p1]), (p2, freqs2[p2], psd2[p2])))
        unpaired_peaks1.remove(p1)
        unpaired_peaks2.remove(p2)
    
    return paired_peaks, unpaired_peaks1, unpaired_peaks2


######################################################################
# Computation of a basic signal spectrogram
######################################################################

def compute_spectrogram(data):
    import scipy

    N = data.shape[0]
    Fs = N / (data[-1, 0] - data[0, 0])
    # Round up to a power of 2 for faster FFT
    M = 1 << int(.5 * Fs - 1).bit_length()
    window = np.kaiser(M, 6.)

    def _specgram(x):
        x_detrended = x - np.mean(x)  # Detrending by subtracting the mean value
        return scipy.signal.spectrogram(
            x_detrended, fs=Fs, window=window, nperseg=M, noverlap=M//2,
            detrend='constant', scaling='density', mode='psd')

    d = {'x': data[:, 1], 'y': data[:, 2], 'z': data[:, 3]}
    f, t, pdata = _specgram(d['x'])
    for axis in 'yz':
        pdata += _specgram(d[axis])[2]
    return pdata, t, f


######################################################################
# Computation of the differential spectrogram
######################################################################

# Interpolate source_data (2D) to match target_x and target_y in order to
# get similar time and frequency dimensions for the differential spectrogram
def interpolate_2d(target_x, target_y, source_x, source_y, source_data):
    import scipy

    # Create a grid of points in the source and target space
    source_points = np.array([(x, y) for y in source_y for x in source_x])
    target_points = np.array([(x, y) for y in target_y for x in target_x])

    # Flatten the source data to match the flattened source points
    source_values = source_data.flatten()

    # Interpolate and reshape the interpolated data to match the target grid shape and replace NaN with zeros
    interpolated_data = scipy.interpolate.griddata(source_points, source_values, target_points, method='nearest')
    interpolated_data = interpolated_data.reshape((len(target_y), len(target_x)))
    interpolated_data = np.nan_to_num(interpolated_data)

    return interpolated_data


# Main logic function to combine two similar spectrogram - ie. from both belts paths - by substracting signals in order to create
# a new composite spectrogram. This result of a divergent but mostly centered new spectrogram (center will be white) with some colored zones
# highlighting differences in the belts paths. The summative spectrogram is used for the MHI calculation.
def combined_spectrogram(data1, data2):
    pdata1, bins1, t1 = compute_spectrogram(data1)
    pdata2, bins2, t2 = compute_spectrogram(data2)

    # Interpolate the spectrograms
    pdata2_interpolated = interpolate_2d(bins1, t1, bins2, t2, pdata2)
    
    # Cobine them in two form: a summed diff for the MHI computation and a diverging diff for the spectrogram colors
    combined_sum = np.abs(pdata1 - pdata2_interpolated)
    combined_divergent = pdata1 - pdata2_interpolated

    return combined_sum, combined_divergent, bins1, t1


# Compute a composite and highly subjective value indicating the "mechanical health of the printer (0 to 100%)" that represent the
# likelihood of mechanical issues on the printer. It is based on the differential spectrogram sum of gradient, salted with a bit
# of the estimated similarity cross-correlation from compute_curve_similarity_factor() and with a bit of the number of unpaired peaks.
# This result in a percentage value quantifying the machine behavior around the main resonances that give an hint if only touching belt tension
# will give good graphs or if there is a chance of mechanical issues in the background (above 50% should be considered as probably problematic)
def compute_mhi(combined_data, similarity_coefficient, num_unpaired_peaks):
    # filtered_data = combined_data[combined_data > 100]
    filtered_data = np.abs(combined_data)

    # First compute a "total variability metric" based on the sum of the gradient that sum the magnitude of will emphasize regions of the
    # spectrogram where there are rapid changes in magnitude (like the edges of resonance peaks).
    total_variability_metric = np.sum(np.abs(np.gradient(filtered_data)))
    # Scale the metric to a percentage using the threshold (found empirically on a large number of user data shared to me)
    base_percentage = (np.log1p(total_variability_metric) / np.log1p(DC_THRESHOLD_METRIC)) * 100
    
    # Adjust the percentage based on the similarity_coefficient to add a grain of salt
    adjusted_percentage = base_percentage * (1 - DC_GRAIN_OF_SALT_FACTOR * (similarity_coefficient / 100))

    # Adjust the percentage again based on the number of unpaired peaks to add a second grain of salt
    peak_confidence = num_unpaired_peaks / DC_MAX_UNPAIRED_PEAKS_ALLOWED
    final_percentage = (1 - peak_confidence) * adjusted_percentage + peak_confidence * 100
    
    # Ensure the result lies between 0 and 100 by clipping the computed value
    final_percentage = np.clip(final_percentage, 0, 100)
    
    return final_percentage, mhi_lut(final_percentage)


# LUT to transform the MHI into a textual value easy to understand for the users of the script
def mhi_lut(mhi):
    if 0 <= mhi <= 30:
        return "Excellent mechanical health"
    elif 30 < mhi <= 45:
        return "Good mechanical health"
    elif 45 < mhi <= 55:
        return "Acceptable mechanical health"
    elif 55 < mhi <= 70:
        return "Potential signs of a mechanical issue"
    elif 70 < mhi <= 85:
        return "Likely a mechanical issue"
    elif 85 < mhi <= 100:
        return "Mechanical issue detected"


######################################################################
# Graphing
######################################################################

def plot_compare_frequency(ax, lognames, signal1, signal2, max_freq):
    # Get the belt name for the legend to avoid putting the full file name
    signal1_belt = (lognames[0].split('/')[-1]).split('_')[-1][0]
    signal2_belt = (lognames[1].split('/')[-1]).split('_')[-1][0]

    if signal1_belt == 'A' and signal2_belt == 'B':
        signal1_belt += " (axis 1,-1)"
        signal2_belt += " (axis 1, 1)"
    elif signal1_belt == 'B' and signal2_belt == 'A':
        signal1_belt += " (axis 1, 1)"
        signal2_belt += " (axis 1,-1)"
    else:
        print("Warning: belts doesn't seem to have the correct name A and B (extracted from the filename.csv)")

    # Plot the two belts PSD signals
    ax.plot(signal1.freqs, signal1.psd, label="Belt " + signal1_belt, color=KLIPPAIN_COLORS['purple'])
    ax.plot(signal2.freqs, signal2.psd, label="Belt " + signal2_belt, color=KLIPPAIN_COLORS['orange'])

    # Trace the "relax region" (also used as a threshold to filter and detect the peaks)
    psd_lowest_max = min(signal1.psd.max(), signal2.psd.max())
    peaks_warning_threshold = PEAKS_DETECTION_THRESHOLD * psd_lowest_max
    ax.axhline(y=peaks_warning_threshold, color='black', linestyle='--', linewidth=0.5)
    ax.fill_between(signal1.freqs, 0, peaks_warning_threshold, color='green', alpha=0.15, label='Relax Region')

    # Trace and annotate the peaks on the graph
    paired_peak_count = 0
    unpaired_peak_count = 0
    offsets_table_data = []

    for _, (peak1, peak2) in enumerate(signal1.paired_peaks):
        label = ALPHABET[paired_peak_count]
        amplitude_offset = abs(((signal2.psd[peak2[0]] - signal1.psd[peak1[0]]) / max(signal1.psd[peak1[0]], signal2.psd[peak2[0]])) * 100)
        frequency_offset = abs(signal2.freqs[peak2[0]] - signal1.freqs[peak1[0]])
        offsets_table_data.append([f"Peaks {label}", f"{frequency_offset:.1f} Hz", f"{amplitude_offset:.1f} %"])
        
        ax.plot(signal1.freqs[peak1[0]], signal1.psd[peak1[0]], "x", color='black')
        ax.plot(signal2.freqs[peak2[0]], signal2.psd[peak2[0]], "x", color='black')
        ax.plot([signal1.freqs[peak1[0]], signal2.freqs[peak2[0]]], [signal1.psd[peak1[0]], signal2.psd[peak2[0]]], ":", color='gray')
        
        ax.annotate(label + "1", (signal1.freqs[peak1[0]], signal1.psd[peak1[0]]),
                    textcoords="offset points", xytext=(8, 5),
                    ha='left', fontsize=13, color='black')
        ax.annotate(label + "2", (signal2.freqs[peak2[0]], signal2.psd[peak2[0]]),
                    textcoords="offset points", xytext=(8, 5),
                    ha='left', fontsize=13, color='black')
        paired_peak_count += 1

    for peak in signal1.unpaired_peaks:
        ax.plot(signal1.freqs[peak], signal1.psd[peak], "x", color='black')
        ax.annotate(str(unpaired_peak_count + 1), (signal1.freqs[peak], signal1.psd[peak]),
                    textcoords="offset points", xytext=(8, 5),
                    ha='left', fontsize=13, color='red', weight='bold')
        unpaired_peak_count += 1

    for peak in signal2.unpaired_peaks:
        ax.plot(signal2.freqs[peak], signal2.psd[peak], "x", color='black')
        ax.annotate(str(unpaired_peak_count + 1), (signal2.freqs[peak], signal2.psd[peak]),
                    textcoords="offset points", xytext=(8, 5),
                    ha='left', fontsize=13, color='red', weight='bold')
        unpaired_peak_count += 1

    # Compute the similarity (using cross-correlation of the PSD signals)
    ax2 = ax.twinx() # To split the legends in two box
    ax2.yaxis.set_visible(False)
    similarity_factor = compute_curve_similarity_factor(signal1, signal2)
    ax2.plot([], [], ' ', label=f'Estimated similarity: {similarity_factor:.1f}%')
    ax2.plot([], [], ' ', label=f'Number of unpaired peaks: {unpaired_peak_count}')
    print(f"Belts estimated similarity: {similarity_factor:.1f}%")

    # Setting axis parameters, grid and graph title
    ax.set_xlabel('Frequency (Hz)')
    ax.set_xlim([0, max_freq])
    ax.set_ylabel('Power spectral density')
    psd_highest_max = max(signal1.psd.max(), signal2.psd.max())
    ax.set_ylim([0, psd_highest_max + psd_highest_max * 0.05])

    ax.xaxis.set_minor_locator(matplotlib.ticker.AutoMinorLocator())
    ax.yaxis.set_minor_locator(matplotlib.ticker.AutoMinorLocator())
    ax.ticklabel_format(axis='y', style='scientific', scilimits=(0,0))
    ax.grid(which='major', color='grey')
    ax.grid(which='minor', color='lightgrey')
    fontP = matplotlib.font_manager.FontProperties()
    fontP.set_size('small')
    ax.set_title('Belts Frequency Profiles (estimated similarity: {:.1f}%)'.format(similarity_factor), fontsize=10, color=KLIPPAIN_COLORS['dark_orange'], weight='bold')

    # Print the table of offsets ontop of the graph below the original legend (upper right)
    if len(offsets_table_data) > 0:
        columns = ["", "Frequency delta", "Amplitude delta", ]
        offset_table = ax.table(cellText=offsets_table_data, colLabels=columns, bbox=[0.66, 0.75, 0.33, 0.15], loc='upper right', cellLoc='center')
        offset_table.auto_set_font_size(False)
        offset_table.set_fontsize(8)
        offset_table.auto_set_column_width([0, 1, 2])
        offset_table.set_zorder(100)
        cells = [key for key in offset_table.get_celld().keys()]
        for cell in cells:
            offset_table[cell].set_facecolor('white')
            offset_table[cell].set_alpha(0.6)

    ax.legend(loc='upper left', prop=fontP)
    ax2.legend(loc='center right', prop=fontP)

    return similarity_factor, unpaired_peak_count


def plot_difference_spectrogram(ax, data1, data2, signal1, signal2, similarity_factor, max_freq):
    combined_sum, combined_divergent, bins, t = combined_spectrogram(data1, data2)

    # Compute the MHI value from the differential spectrogram sum of gradient, salted with
    # the similarity factor and the number or unpaired peaks from the belts frequency profile
    # Be careful, this value is highly opinionated and is pretty experimental!
    mhi, textual_mhi = compute_mhi(combined_sum, similarity_factor, len(signal1.unpaired_peaks) + len(signal2.unpaired_peaks))
    print(f"[experimental] Mechanical Health Indicator: {textual_mhi.lower()} ({mhi:.1f}%)")
    ax.set_title(f"Differential Spectrogram", fontsize=14, color=KLIPPAIN_COLORS['dark_orange'], weight='bold')
    ax.plot([], [], ' ', label=f'{textual_mhi} (experimental)')
    
    # Draw the differential spectrogram with a specific custom norm to get orange or purple values where there is signal or white near zeros
    colors = [KLIPPAIN_COLORS['dark_orange'], KLIPPAIN_COLORS['orange'], 'white', KLIPPAIN_COLORS['purple'], KLIPPAIN_COLORS['dark_purple']]  
    cm = matplotlib.colors.LinearSegmentedColormap.from_list('klippain_divergent', list(zip([0, 0.25, 0.5, 0.75, 1], colors)))
    norm = matplotlib.colors.TwoSlopeNorm(vmin=np.min(combined_divergent), vcenter=0, vmax=np.max(combined_divergent))
    ax.pcolormesh(t, bins, combined_divergent.T, cmap=cm, norm=norm, shading='gouraud')

    ax.set_xlabel('Frequency (hz)')
    ax.set_xlim([0., max_freq])
    ax.set_ylabel('Time (s)')
    ax.set_ylim([0, bins[-1]])

    fontP = matplotlib.font_manager.FontProperties()
    fontP.set_size('medium')
    ax.legend(loc='best', prop=fontP)

    # Plot vertical lines for unpaired peaks
    unpaired_peak_count = 0
    for _, peak in enumerate(signal1.unpaired_peaks):
        ax.axvline(signal1.freqs[peak], color=KLIPPAIN_COLORS['red_pink'], linestyle='dotted', linewidth=1.5)
        ax.annotate(f"Peak {unpaired_peak_count + 1}", (signal1.freqs[peak], t[-1]*0.05),
                    textcoords="data", color=KLIPPAIN_COLORS['red_pink'], rotation=90, fontsize=10,
                    verticalalignment='bottom', horizontalalignment='right')
        unpaired_peak_count +=1

    for _, peak in enumerate(signal2.unpaired_peaks):
        ax.axvline(signal2.freqs[peak], color=KLIPPAIN_COLORS['red_pink'], linestyle='dotted', linewidth=1.5)
        ax.annotate(f"Peak {unpaired_peak_count + 1}", (signal2.freqs[peak], t[-1]*0.05),
                    textcoords="data", color=KLIPPAIN_COLORS['red_pink'], rotation=90, fontsize=10,
                    verticalalignment='bottom', horizontalalignment='right')
        unpaired_peak_count +=1
    
    # Plot vertical lines and zones for paired peaks
    for idx, (peak1, peak2) in enumerate(signal1.paired_peaks):
        label = ALPHABET[idx]
        x_min = min(peak1[1], peak2[1])
        x_max = max(peak1[1], peak2[1])
        ax.axvline(x_min, color=KLIPPAIN_COLORS['dark_purple'], linestyle='dotted', linewidth=1.5)
        ax.axvline(x_max, color=KLIPPAIN_COLORS['dark_purple'], linestyle='dotted', linewidth=1.5)
        ax.fill_between([x_min, x_max], 0, np.max(combined_divergent), color=KLIPPAIN_COLORS['dark_purple'], alpha=0.3)
        ax.annotate(f"Peaks {label}", (x_min, t[-1]*0.05),
                textcoords="data", color=KLIPPAIN_COLORS['dark_purple'], rotation=90, fontsize=10,
                verticalalignment='bottom', horizontalalignment='right')

    return


######################################################################
# Custom tools 
######################################################################

# Simple helper to compute a sigmoid scalling (from 0 to 100%)
def sigmoid_scale(x, k=1):
    return 1 / (1 + np.exp(-k * x)) * 100

# Original Klipper function to get the PSD data of a raw accelerometer signal
def compute_signal_data(data, max_freq):
    calibration_data = calc_freq_response(data)
    freqs = calibration_data.freq_bins[calibration_data.freq_bins <= max_freq]
    psd = calibration_data.get_psd('all')[calibration_data.freq_bins <= max_freq]
    peaks, _ = detect_peaks(psd, freqs)
    return SignalData(freqs=freqs, psd=psd, peaks=peaks, paired_peaks=None, unpaired_peaks=None)
 

######################################################################
# Startup and main routines
######################################################################

def parse_log(logname):
    with open(logname) as f:
        for header in f:
            if not header.startswith('#'):
                break
            if not header.startswith('freq,psd_x,psd_y,psd_z,psd_xyz'):
                # Raw accelerometer data
                return np.loadtxt(logname, comments='#', delimiter=',')
    # Power spectral density data or shaper calibration data
    raise ValueError("File %s does not contain raw accelerometer data and therefore "
               "is not supported by this script. Please use the official Klipper "
               "graph_accelerometer.py script to process it instead." % (logname,))


def belts_calibration(lognames, klipperdir="~/klipper", max_freq=200., graph_spectogram=True, width=8.3, height=11.6):
    for filename in lognames[:2]:
        # Wait for the file handler to be released by Klipper
        while is_file_open(filename):
            time.sleep(2)

    # Parse data
    datas = [parse_log(fn) for fn in lognames]
    if len(datas) > 2:
        raise ValueError("Incorrect number of .csv files used (this function needs two files to compare them)")

    # Compute calibration data for the two datasets with automatic peaks detection
    signal1 = compute_signal_data(datas[0], max_freq)
    signal2 = compute_signal_data(datas[1], max_freq)

    # Pair the peaks across the two datasets
    paired_peaks, unpaired_peaks1, unpaired_peaks2 = pair_peaks(signal1.peaks, signal1.freqs, signal1.psd,
                                                               signal2.peaks, signal2.freqs, signal2.psd)
    signal1 = signal1._replace(paired_peaks=paired_peaks, unpaired_peaks=unpaired_peaks1)
    signal2 = signal2._replace(paired_peaks=paired_peaks, unpaired_peaks=unpaired_peaks2)


    if graph_spectogram:
        fig = matplotlib.pyplot.figure()
        gs = matplotlib.gridspec.GridSpec(2, 1, height_ratios=[4, 3])
        ax1 = fig.add_subplot(gs[0])
        ax2 = fig.add_subplot(gs[1])
    else:
        fig, ax1 = matplotlib.pyplot.subplots()

    # Add title
    try:
        filename = lognames[0].split('/')[-1]
        dt = datetime.strptime(f"{filename.split('_')[1]} {filename.split('_')[2]}", "%Y%m%d %H%M%S")
        title_line2 = dt.strftime('%x %X')
    except:
        print("Warning: CSV filenames look to be different than expected (%s , %s)" % (lognames[0], lognames[1]))
        title_line2 = lognames[0].split('/')[-1] + " / " +  lognames[1].split('/')[-1]
    fig.suptitle(title_line2)

    # Plot the graphs
    similarity_factor, _ = plot_compare_frequency(ax1, lognames, signal1, signal2, max_freq)
    if graph_spectogram:
        plot_difference_spectrogram(ax2, datas[0], datas[1], signal1, signal2, similarity_factor, max_freq)

    fig.set_size_inches(width, height)
    fig.tight_layout()
    fig.subplots_adjust(top=0.89)
    
    return fig


def main():
    # Parse command-line arguments
    usage = "%prog [options] <raw logs>"
    opts = optparse.OptionParser(usage)
    opts.add_option("-o", "--output", type="string", dest="output",
                    default=None, help="filename of output graph")
    opts.add_option("-f", "--max_freq", type="float", default=200.,
                    help="maximum frequency to graph")
    opts.add_option("-k", "--klipper_dir", type="string", dest="klipperdir",
                    default="~/klipper", help="main klipper directory")
    opts.add_option("-n", "--no_spectogram", action="store_false", dest="no_spectogram",
                    default=True, help="disable plotting of spectogram")
    opts.add_option("-w", "--width", type="float", dest="width",
                    default=8.3, help="width (inches) of the graph(s)")
    opts.add_option("-l", "--height", type="float", dest="height",
                    default=11.6, help="height (inches) of the graph(s)")
    
    options, args = opts.parse_args()
    if len(args) < 1:
        opts.error("Incorrect number of arguments")
    if options.output is None:
        opts.error("You must specify an output file.png to use the script (option -o)")

    fig = belts_calibration(args, options.klipperdir, options.max_freq, options.no_spectogram,
                            options.width, options.height)
    pathlib.Path(options.output).unlink(missing_ok=True) 
    fig.savefig(options.output)


if __name__ == '__main__':
    main()