def runAndPlotOptFilter(event, pulse, pta=None, **kwargs):
cham = kwargs['cham']
ionpt = kwargs.get('ionpulsestarttime', 0)
if pulse.GetChannelName() not in kwargs['chanlist']: return
print pulse.GetChannelName()
optkamper = cham.GetOptimalKamper(pulse)
thewindow = optkamper.GetWindow()
#print thewindow.GetWindowSize()
#print thewindow.GetOutputPulseSize()
preproc = optkamper.GetPreProcessor()
#print preproc.GetInputPulseSize(), preproc.GetOutputPulseSize()
#print pulse.GetPulseLength()
if kwargs.get('runCalculation', True):
if pulse.GetChannelName().startswith('chal'):
optkamper.SetIonPulseStartTime(ionpt);
optkamper.MakeKamp(pulse)
#print thewindow.GetWindowSize()
#print thewindow.GetOutputPulseSize()
#print preproc.GetInputPulseSize(), preproc.GetOutputPulseSize()
#print pulse.GetPulseLength()
resultsMap = optkamper.GetResults()
for key, val in resultsMap:
print key, val.fValue
rawpulse = np.array(pulse.GetTrace())
thewindow = optkamper.GetWindow()
windowpulse = kutil.get_as_nparray(thewindow.GetWindow(), thewindow.GetWindowSize() )
windowoutput = kutil.get_out( thewindow )
optfilter = optkamper.GetOptimalFilter()
noisepower = kutil.get_as_nparray(optfilter.GetNoiseSpectrum(), optfilter.GetNoiseSpectrumSize())
templatepower = kutil.get_as_nparray(optfilter.GetTemplatePower(), optfilter.GetTemplatePowerSize())
hc2p = ROOT.KHalfComplexPower()
hc2p.SetInputPulse( optfilter.GetOptimalFilter(), optfilter.GetOptimalFilterSize())
hc2p.RunProcess()
optfilpower = kutil.get_out(hc2p)
hc2p.SetInputPulse(optfilter.GetInputPulse(), optfilter.GetInputPulseSize())
hc2p.RunProcess()
thispulsepower = kutil.get_out( hc2p)
hc2p.SetInputPulse( optfilter.GetOptFilterAndSignal(), optfilter.GetOptFilterAndSignalSize())
hc2p.RunProcess()
optfilter_andpulsepower = kutil.get_out(hc2p)
ampestimator = kutil.get_out( optfilter)
res = scipy.optimize.minimize(optkamper.Chi2Functor, [np.abs(windowpulse[ampestimator.argmax()]), ampestimator.argmax()], method='Nelder-Mead')
chi2_min_amp = res.x[0]
chi2_min_time = res.x[1]
print res
print 'chi2 start time after minimization', chi2_min_time
print 'chi2 amp after minimization', chi2_min_amp
print 'chi2 at mins', optfilter.GetChiSquared(chi2_min_amp, chi2_min_time)
print 'amplitude theoretical variance', 1./optfilter.GetAmpEstimatorDenominator()
print 'kdata: calcullate chi2 versus time at amplitude', chi2_min_amp
stepsize = 0.01
timearray = np.array([atime*stepsize for atime in range( int(pulse.GetPulseLength()/stepsize)) ])
chi2_time = np.array([optfilter.GetChiSquared(chi2_min_amp, time) for time in timearray])
print 'kdata: calcullate chi2 versus amplitude at time', chi2_min_time
stepsize = 0.01
amps = np.array([a*stepsize for a in range( 0, int(int(2*ampestimator.max())/stepsize)) ])
chi2_amp = np.array([ optfilter.GetChiSquared(a_i, chi2_min_time) for a_i in amps ])
print 'chi2 amp from scan', chi2_amp.argmin()*stepsize
print 'chi2 start time from scan', chi2_time.argmin()
theFig = plt.figure( kwargs.get('figure', 1))
theWidth = 1
plt.axis('off')
axes = plt.subplot(6,1,1)
plt.cla()
plt.plot(rawpulse, linewidth=theWidth)
plt.plot(windowoutput, linewidth=theWidth)
if kwargs.has_key('template'):
scalefactor = -1*ampestimator.min()
if ampestimator.max() > scalefactor:
scalefactor = ampestimator.max()
plt.plot( np.array(kwargs['template'])* scalefactor)
axes = plt.subplot(6,1,2)
plt.cla()
plt.loglog(noisepower, linewidth=theWidth)
plt.loglog(templatepower, linewidth=theWidth)
plt.loglog(optfilpower, linewidth=theWidth)
axes = plt.subplot(6,1,3)
plt.cla()
plt.loglog(thispulsepower, linewidth=theWidth)
plt.loglog(optfilter_andpulsepower, linewidth=theWidth)
axes = plt.subplot(6,1,4)
plt.cla()
plt.plot(ampestimator, linewidth=theWidth)
axes = plt.subplot(6,1,5)
plt.cla()
plt.plot( amps, chi2_amp, linewidth=theWidth)
axes = plt.subplot(6,1,6)
plt.cla()
plt.plot(timearray, chi2_time, linewidth=theWidth)
plt.show()
if kwargs.get('wait', True):
raw_input()
#now loop through the data again, applying the optimal filter that was just built.
#for i in range(f.GetEntries()):
# f.GetEntry(i)
for ii in range(e.GetNumBoloPulses()):
p = e.GetBoloPulse(ii) #returns a KRawBoloPulseRecord
if p.GetPulseLength() == 0: continue #skip empty pulses
if p.GetChannelName() == myChannel:
bas.SetInputPulse( p.GetTrace() )
y=bas.RunProcess()
#plt.subplot(3,3,ii+1)
x = kutil.get_out(bas)
coeffs = []
for i in range(lvl):
coeffs.append(pywt.wavedec(x, 'db2', level = i+1))
if max(abs(coeffs[i][0])) > 2*max(abs(coeffs[i][1])):
print 'for level', i+1, ':true'
else:
print 'for level', i+1, ':false'
plt.figure(2)
plt.subplot(lvl+1,2,1)
plt.plot(x)
tree.define_collection(name='pulsesy',
prefix='c1y_',
size='c1y_n',
mix=Pulse)
tree.define_collection(name='pulsesx',
prefix='c1x_',
size='c1x_n',
mix=Pulse)
# loop over "events" in tree
for event in tree:
print "%d" % event.date
bas.SetInputPulse( event.c1y_val )
bas.RunProcess()
pulsey = util.get_out(bas)
#plt.plot( np.array(event.c1x_val), util.get_out(bas) )
amp.Fill( np.argmin(pulsey) )
#raw_input()
#plt.cla()
f.close()
amp.Draw()
results.cd()
amp.Write()
raw_input()
def kamp(p, pta, **kwargs):
f = kwargs["kdatafile"]
channels = kwargs["channellist"]
cham = kwargs["chamonix"]
hc2p = kwargs["halfcomp2power"]
r2hc = kwargs["real2halfcomp"]
pulsePol = kwargs['pulsePol']
pulseLength, binSize, minSearch, maxSearch, _pulsetype = getPulseInfo(p.GetChannelName())
db = kwargs['templateDB']
if p.GetChannelName() not in channels:
return None
print 'Entry', f.GetCurrentEntryNumber(), p.GetChannelName()
pulse = p.GetTrace()
tempDft = cham.GetTemplateSpectrum(p.GetChannelName())
noisePower = cham.GetNoisePower(p.GetChannelName())
#print noisePower.size()
#print noisePower[noisePower.size()-1]
print 'number of noise events found', cham.GetNumNoiseEventsFound(p.GetChannelName())
optKamper = cham.GetOptimalKamper()
optFilter = optKamper.GetOptimalFilter()
optFilter.SetTemplateDFT(tempDft)
optFilter.SetNoiseSpectrum(noisePower)
optFilter.SetToRecalculate()
#optFilter.BuildFilter()
if p.GetIsHeatPulse():
optKamper.SetWindow(cham.GetHeatWindow())
optKamper.SetPreProcessor(cham.GetHeatPreProcessor())
win = cham.GetHeatWindow()
preproc= cham.GetHeatPreProcessor()
elif p.GetBoloBoxVersion() > 1.9:
optKamper.SetWindow(cham.GetIonWindow())
optKamper.SetPreProcessor(cham.GetBBv2IonPreProcessor())
win = cham.GetIonWindow()
preproc= cham.GetBBv2IonPreProcessor()
else:
optKamper.SetWindow(cham.GetIonWindow())
optKamper.SetPreProcessor(cham.GetBBv1IonPreProcessor())
win = cham.GetIonWindow()
preproc= cham.GetBBv1IonPreProcessor()
optKamper.SetPulseTemplateShiftFromPreTrigger( cham.GetTemplateShift(p.GetChannelName()) );
optKamper.SetAmplitudeEstimatorSearchRangeMax(maxSearch)
optKamper.SetAmplitudeEstimatorSearchRangeMin(minSearch)
plt.figure(1)
plt.subplot(7,1,1)
plt.cla()
plt.plot(np.array(pulse))
plt.title('raw')
plt.subplot(7,1,2)
plt.cla()
plt.plot( kutil.get_as_nparray(win.GetWindow(), win.GetWindowSize()) )
plt.title('window')
#winpulse = std.vector("double")()
#winpulse.reserve(win.GetOutputPulseSize())
#for i in range(winpulse.size()):
# winpulse[i] = win.GetOutputPulse()[i]
preproc.SetInputPulse(pulse)
preproc.RunProcess()
win.SetInputPulse(preproc)
win.RunProcess()
winpulse = kutil.get_out(win)
plt.subplot(7,1,3)
plt.cla()
plt.plot(np.array(winpulse))
plt.title('baseline removed and windowed')
# plt.subplot(4,1,2)
# plt.cla()
# plt.plot(np.array(tempDft))
# plt.title('tempDft')
hc2p.SetInputPulse(tempDft)
print 'template power', hc2p.RunProcess()
tempPower = kutil.get_out(hc2p)
plt.subplot(7,1,4)
plt.cla()
npp = np.array(noisePower)
plt.loglog(npp)
plt.loglog(tempPower)
#raw_input('hit enter to run filter....')
#wait, doesn't the optKamper do this?
#r2hc.SetInputPulse(win)
#r2hc.RunProcess()
#optFilter.SetInputPulse(r2hc)
#optFilter.BuildFilter()
optFilterResults = optKamper.MakeKamp(p)
kernel = kutil.get_as_nparray(optFilter.GetOptimalFilter(), optFilter.GetOptimalFilterSize())
hc2p.SetInputPulse(kernel, len(kernel))
hc2p.RunProcess()
kernelPower = kutil.get_out(hc2p)
plt.loglog(kernelPower)
plt.title('filter power')
hc2p.SetInputPulse(optFilter.GetInputPulse(), optFilter.GetInputPulseSize())
hc2p.RunProcess()
pulsePower = kutil.get_out(hc2p)
plt.subplot(7,1,5)
plt.cla()
plt.loglog(pulsePower)
plt.title('pulse power')
optOut = kutil.get_out(optFilter)
plt.subplot(7,1,6)
plt.cla()
plt.plot(optOut)
plt.title('amplitude estimator')
#calculate chi^2 as a function of t and see what it looks like.
chi2 = std.vector("double")()
for time in range(pulse.size()): #loop over time bins
chi2.push_back(optFilter.GetChiSquared(time))
plt.subplot(7,1,7)
plt.cla()
plt.plot(np.array(chi2))
plt.title('chi squared')
plt.show()
#raw_input()
plt.figure(2)
#plt.subplot(1,1,1)
plt.cla()
plt.plot(winpulse)
vr = db.view('analytical/bychandate',reduce=False, descending=True, startkey=[p.GetChannelName(), "2012-01-22 00:00:00.0"], limit=1, include_docs=True)
doc = vr.first()['doc']
vp = std.vector("double")()
vp.reserve(pulseLength)
exec(doc['formula']['python']) #defines 'template' function
#doc['formula']['par'][2] = doc['formula']['par'][2]/2.016
#doc['formula']['par'][3] = doc['formula']['par'][3]/2.016
#doc['formula']['par'][5] = doc['formula']['par'][5]/2.016
#doc['formula']['par'][0]=300 #put it close to zero, but away from the windowing function
for i in range( pulseLength ):
vp.push_back( template(i*binSize, doc['formula']['par']))
#plot the pulse templates for documentation
if p.GetChannelName().startswith('chal'): thePolarity = 1 #template heat pulses are already negative polarity
else: thePolarity = pulsePol.GetExpectedPolarity(p)
scaleFactor = thePolarity*max(abs(np.array(winpulse)))/max(abs(np.array(vp)))
optScaleFactor = thePolarity*optFilterResults['amp'].fValue/max(abs(np.array(vp)))
print 'simple scaling by', scaleFactor
print 'opt filter scaling by at amp', optScaleFactor
print vp.size(), vp[i], vp[vp.size()-1], vp[i]*scaleFactor
optPulse = std.vector("double")()
optPulse.resize(vp.size())
diffPulse = std.vector("double")()
diffPulse.resize(vp.size())
for i in range(vp.size()):
optPulse[i] = optScaleFactor * vp[i]
vp[i] = scaleFactor * vp[i]
diffPulse[i] = optPulse[i] - vp[i]
plt.plot(np.array(vp))
plt.plot(np.array(optPulse))
#plt.plot(np.array(diffPulse))
print 'optimal filter results'
for key, val in optFilterResults:
print key, val.fValue, val.fUnit
raw_input()