"""
Example code by Jason demonstrating some pygama convenience functions.
"""
import numpy as np
import matplotlib.pyplot as plt
import pygama.analysis.histograms as pgh
import pygama.analysis.peak_fitting as pga
np.random.seed(0) # fix the seed s/t we can reproduce the plot
n = 10
data = np.random.normal(0, 1, n)
hist, bins, var = pgh.get_hist(data, range=(-5, 5), dx=1)
pgh.plot_hist(hist, bins, var, label="data")
pars, cov = pga.fit_hist(pga.gauss, hist, bins, var=var, guess=[0, 1, n])
pgh.print_fit_results(pars, cov, ['mu', 'sig', 'A'])
pgh.plot_func(pga.gauss, pars, label="chi2 fit")
nbnd = (-np.inf, np.inf)
pos = (0, np.inf)
pars, cov = pga.fit_hist(pga.gauss,
hist,
bins,
var=var,
guess=[0, 1, n],
bounds=[nbnd, pos, pos],
poissonLL=True)
pgh.print_fit_results(pars, cov, ['mu', 'sig', 'A'])
pgh.plot_func(pga.gauss, pars, label="poissonLL fit")
def peak(df, runDB, calDB, r, line, p=[1, 0], plotit=False):
cal = 0.04998 # calDB["cal_pass1"]["1"]["p1cal"]
meta_dir = os.path.expandvars(runDB["meta_dir"])
tier_dir = os.path.expandvars(runDB["tier2_dir"])
df['e_cal'] = p[0] * (cal * df['e_ftp']) + p[1]
#h = df.hist('e_cal',bins=2000)
#plt.yscale('log')
df = df.loc[(df.index > 1000) & (df.index < 500000)]
def gauss(x, mu, sigma, A=1):
"""
define a gaussian distribution, w/ args: mu, sigma, area (optional).
"""
return A * (1. / sigma / np.sqrt(2 * np.pi)) * np.exp(-(x - mu)**2 /
(2. * sigma**2))
line_min = 0.995 * line
line_max = 1.005 * line
nbin = 60
res = 6.3e-4 * line + 0.85 # empirical energy resolution curve from experience
hist, bins, var = pgh.get_hist(df['e_cal'],
range=(line_min, line_max),
dx=(line_max - line_min) / nbin)
if plotit:
pgh.plot_hist(hist, bins, var=hist, label="data", color='blue')
pars, cov = pga.fit_hist(gauss,
hist,
bins,
var=hist,
guess=[line, res, 50])
pgu.print_fit_results(pars, cov, gauss)
if plotit:
pgu.plot_func(gauss, pars, label="chi2 fit", color='red')
FWHM = '%.2f' % Decimal(
pars[1] * 2. * np.sqrt(2. * np.log(2))) # convert sigma to FWHM
FWHM_uncertainty = '%.2f' % Decimal(
np.sqrt(cov[1][1]) * 2. * np.sqrt(2. * np.log(2)))
peak = '%.2f' % Decimal(pars[0])
peak_uncertainty = '%.2f' % Decimal(np.sqrt(cov[0][0]))
residual = '%.2f' % abs(line - float(peak))
if plotit:
label_01 = 'Peak = ' + str(peak) + r' $\pm$ ' + str(peak_uncertainty)
label_02 = 'FWHM = ' + str(FWHM) + r' $\pm$ ' + str(FWHM_uncertainty)
labels = [
label_01,
label_02,
]
plt.xlim(line_min, line_max)
plt.xlabel('Energy (keV)', ha='right', x=1.0)
plt.ylabel('Counts', ha='right', y=1.0)
plt.tight_layout()
plt.hist(df['e_cal'], range=(line_min, line_max), bins=nbin)
plt.legend(labels, frameon=False, loc='upper right', fontsize='small')
plt.savefig(meta_dir + '/plots/lineFit_' + str(r) + '.png')
return peak, FWHM
def peak(line):
if (len(sys.argv) != 2):
print('Usage: fit_calibrated_peaks.py [run number]')
sys.exit()
# take calibration parameter for the 'calibration.py' output
with open("calDB.json") as f:
calDB = json.load(f)
cal = calDB["cal_pass1"]["1"]["p1cal"]
with open("runDB.json") as f:
runDB = json.load(f)
meta_dir = os.path.expandvars(runDB["meta_dir"])
tier_dir = os.path.expandvars(runDB["tier_dir"])
#df = pd.read_hdf("{}/Spectrum_{}_2.hdf5".format(meta_dir,sys.argv[1]), key="df")
df = pd.read_hdf("{}/t2_run{}.h5".format(tier_dir, sys.argv[1]))
df['e_cal'] = cal * df['e_ftp']
df = df.loc[(df.index > 1000) & (df.index < 500000)]
def gauss(x, mu, sigma, A=1):
"""
define a gaussian distribution, w/ args: mu, sigma, area (optional).
"""
return A * (1. / sigma / np.sqrt(2 * np.pi)) * np.exp(-(x - mu)**2 /
(2. * sigma**2))
line_min = 0.995 * line
line_max = 1.005 * line
nbin = 60
res = 6.3e-4 * line + 0.85 # empirical energy resolution curve from experience
hist, bins, var = pgh.get_hist(df['e_cal'],
range=(line_min, line_max),
dx=(line_max - line_min) / nbin)
pgh.plot_hist(hist, bins, var=hist, label="data", color='blue')
pars, cov = pga.fit_hist(gauss,
hist,
bins,
var=hist,
guess=[line, res, 50])
pgu.print_fit_results(pars, cov, gauss)
pgu.plot_func(gauss, pars, label="chi2 fit", color='red')
FWHM = '%.2f' % Decimal(pars[1] * 2)
FWHM_uncertainty = '%.2f' % Decimal(np.sqrt(cov[1][1]) * 2)
peak = '%.2f' % Decimal(pars[0])
peak_uncertainty = '%.2f' % Decimal(np.sqrt(cov[0][0]))
residual = '%.2f' % abs(line - float(peak))
label_01 = 'Peak = ' + str(peak) + r' $\pm$ ' + str(peak_uncertainty)
label_02 = 'FWHM = ' + str(FWHM) + r' $\pm$ ' + str(FWHM_uncertainty)
labels = [
label_01,
label_02,
]
plt.xlim(line_min, line_max)
plt.xlabel('Energy (keV)', ha='right', x=1.0)
plt.ylabel('Counts', ha='right', y=1.0)
plt.tight_layout()
plt.hist(df['e_cal'], range=(line_min, line_max), bins=nbin)
plt.legend(labels, frameon=False, loc='upper right', fontsize='small')
return peak, FWHM
print("File info: ", f.keys())
chn = 0
for ged in f.keys():
try:
dset = f[ged]['raw']
print("key: ", ged, "Data info: ", dset.keys())
except:
conf = f[ged]
print("Header info: ", conf.keys())
energies = dset['energy'][()]
#energies = f[ged]['raw/energy'][()]
#energies = dset['energy'][dset['channel'][()]==chn]
maxe = np.amax(energies)
h, b, v = pgh.get_hist(energies, bins=3500, range=(maxe / 4, maxe))
pgh.plot_hist(h, b, label=ged)
#bin_max = b[np.where(h == h.max())][0]
#print("chn %d Raw energy max %d, histogram max %d at %d " % (chn,maxe,h.max(),bin_max ))
#min_ene = int(bin_max*0.95)
#max_ene = int(bin_max*1.05)
#hist, bins, var = pgh.get_hist(energies, bins=500, range=(min_ene, max_ene))
#pgh.plot_hist(hist, bins, label="chn %d" % chn )
#chn = chn + 1
plt.xlabel("uncalibrated energy", ha='right', x=1)
plt.ylabel("counts", ha='right', y=1)
plt.yscale('log')
plt.legend()
plt.show()
#plt.savefig("./peak_chn%d.pdf" % chn )
#plt.cla()
def peak_352():
if (len(sys.argv) != 2):
print('Usage: fit_bkg_peaks.py [run number]')
sys.exit()
with open("runDB.json") as f:
runDB = json.load(f)
meta_dir = os.path.expandvars(runDB["meta_dir"])
#df = pd.read_hdf("{}/Spectrum_280-329.hdf5".format(meta_dir), key="df")
df = pd.read_hdf("{}/Spectrum_{}.hdf5".format(meta_dir, sys.argv[1]),
key="df")
def gauss(x, mu, sigma, A=1):
"""
define a gaussian distribution, w/ args: mu, sigma, area (optional).
"""
return A * (1. / sigma / np.sqrt(2 * np.pi)) * np.exp(-(x - mu)**2 /
(2. * sigma**2))
def radford_peak(x, mu, sigma, hstep, htail, tau, bg0, a=1):
"""
David Radford's HPGe peak shape function
"""
# make sure the fractional amplitude parameters stay reasonable...
if htail < 0 or htail > 1:
return np.zeros_like(x)
if hstep < 0 or hstep > 1:
return np.zeros_like(x)
bg_term = bg0 #+ x*bg1
if np.any(bg_term < 0):
return np.zeros_like(x)
# compute the step and the low energy tail
step = a * hstep * erfc((x - mu) / (sigma * np.sqrt(2)))
le_tail = a * htail
le_tail *= erfc((x - mu) / (sigma * np.sqrt(2)) + sigma /
(tau * np.sqrt(2)))
le_tail *= np.exp((x - mu) / tau)
le_tail /= (2 * tau * np.exp(-(sigma / (np.sqrt(2) * tau))**2))
# add up all the peak shape components
return (1 - htail) * gauss(x, mu, sigma, a) + bg_term + step + le_tail
hist, bins, var = pgh.get_hist(df['e_cal'], range=(345, 360), dx=0.5)
pgh.plot_hist(hist, bins, var=hist, label="data")
pars, cov = pga.fit_hist(radford_peak,
hist,
bins,
var=hist,
guess=[352, 1.05, 0.001, 0.02, 500, 1000, 40000])
pgu.print_fit_results(pars, cov, radford_peak)
pgu.plot_func(radford_peak, pars, label="chi2 fit", color='red')
#x_vals = np.arange(345,360,0.5)
#plt.plot(x_vals, radford_peak(x_vals, 353, 1.05, .001, 0.02, 500, 1000, 40000))
FWHM = '%.2f' % Decimal(pars[1] * 2)
FWHM_uncertainty = '%.2f' % Decimal(np.sqrt(cov[1][1]) * 2)
peak = '%.2f' % Decimal(pars[0])
peak_uncertainty = '%.2f' % Decimal(np.sqrt(cov[0][0]))
residual = '%.2f' % (351.93 - float(peak))
#chi_2_element_list = []
#for i in range(len(hist)):
#chi_2_element = abs((radford_peak(bins[i], *pars) - hist[i])**2/radford_peak(bins[i], *pars))
#chi_2_element_list.append(chi_2_element)
#chi_2 = sum(chi_2_element_list)
#reduced_chi_2 = '%.2f' % Decimal(chi_2/len(hist))
label_01 = '351.93 keV peak fit'
label_02 = 'FWHM = ' + str(FWHM) + r' $\pm$ ' + str(FWHM_uncertainty)
label_03 = 'Peak = ' + str(peak) + r' $\pm$ ' + str(peak_uncertainty)
label_04 = 'Residual = ' + str(residual) + r' $\pm$ ' + str(
peak_uncertainty)
colors = ['red', 'red', 'red', 'red']
lines = [Line2D([0], [0], color=c, lw=2) for c in colors]
labels = [label_01, label_02, label_03, label_04]
plt.xlim(345, 360)
plt.ylim(0, plt.ylim()[1])
plt.xlabel('Energy (keV)', ha='right', x=1.0)
plt.ylabel('Counts', ha='right', y=1.0)
plt.tight_layout()
#plt.semilogy()
plt.legend(lines,
labels,
frameon=False,
loc='upper right',
fontsize='small')
plt.show()