Technische Universität München Status report on SADC feature extraction
Mar 28, 2015
Technische Universität München
Status report on
SADC feature extraction
Technische Universität München
Outline
• Feature extraction fitting pulses– General idea– Parameter sets– Implementation
• AC noise suppression• Performance (comp. to std. alg.)
– Energy spectrum for e- calibration
– “0” mass spectrum
• Conclusion
Technische Universität München
Feature extraction using functional parameterization
• Describe pulses by analytic function (Silicon, J. Friedrich)
• Fix function parameters (calibration)• Fit/Calculation of time and amplitude
Technische Universität München
Typical pulse and its description
Technische Universität München
Result of fit (first parameter)
Using three parameter sets for ECAL02
Parameter sets depend on shaper not on cell type!!
Technische Universität München
Signals on two different channels
Analytic calculation of function variables due to large amplitude error not desirable
Technische Universität München
AC coupling
Technische Universität München
AC Coupling suppression (ACS)
• AC coupling signals have amplitude independent length
• Remove all signals with a half-width ≤ 5
• Problems for small signal introduce threshold to apply cut
Technische Universität München
40 GeV e- Beam – ECAL calibration(std: green, pf: red)
Technische Universität München
40 GeV e- Beam – ECAL calibration(std: green, pf: red)
Technische Universität München
40 GeV e- Beam – ECAL calibration(std: green, pf+acs: blue)
Technische Universität München
M(0) (PulseFit & AC suppresion) VS std. FE
• slightly improved energy resolution (~ 4%)
• slightly improved signal to bg ratio ~ (0.2%)
• reduced signal integral (~ 8%)
Technische Universität München
Deeper analysis:
Feature of shower reconstruction (“instable” against small changes )
M(0) PulseFit VS std. FE
• increased signal integral (~ 4%)
• decreased energy resolution (~ 50%)
• decreased signal to bg ratio (~ 20%)
Technische Universität München
Conclusion
• Alternative method for feature extraction available• Electron beam performance (without calibration
adaption) promising• Reconstruction problems for 0 observed
Improvement by adaption of shower reconstruction possible?
• Increased CPU usage (for physics runs 20% of total runtime)
• Fine tuning needed (dep. on shower reconstruction) For the moment std. reconstruction is the preferable
option!!!
Technische Universität München
Thank you.