Time Calibration of AMANDA Three Variations on a Theme of T 0 Kael D. Hanson Department of Physics & Astronomy University of Pennsylvania For the AMANDA Collaboration

March 27, 2002K. Hanson - Calor2002 – Pasadena, CA2 The AMANDA Collaboration 7 US and 9 European institutions, about 110 current members: 1.Bartol Research Institute, University of Delaware, Newark, USA 2.BUGH Wuppertal, Germany 3.Universite Libre de Bruxelles, Brussels, Belgium 4.DESY-Zeuthen, Zeuthen, Germany 5.Dept. of Technology, Kalmar University, Kalmar, Sweden 6.Lawrence Berkeley National Laboratory, Berkeley, USA 7.Dept. of Physics, UC Berkeley, USA 8.Institute of Physics, University of Mainz, Mainz, Germany 9.University of Mons-Hainaut, Mons, Belgium 10.University of California, Irvine, CA 11.Dept. of Physics and Astronomy, University of Pennsylvania, Philadelphia, USA 12.Physics Department, University of Wisconsin, River Falls, USA 13.Physics Department, University of Wisconsin, Madison, USA 14.Division of High Energy Physics, Uppsala University, Uppsala, Sweden 15.Fysikum, Stockholm University, Stockholm, Sweden 16.Vrije Universiteit Brussel, Brussel, Belgium

March 27, 2002K. Hanson - Calor2002 – Pasadena, CA3 The AMANDA Detector Currently operating AMANDA-II, 677 OMs, 20 Megaton geometric volume. Several artificial light sources deployed for calibrating AMANDA: –Nd:YAG surface laser Timing Geometry –N2 (UV) in situ lasers –UV LED ‘flashers.’

March 27, 2002K. Hanson - Calor2002 – Pasadena, CA4 Timing is Critical to AMANDA Event Reconstruction! Muon/cascade particle id and reconstruction depend crucially on relative timing in the OMs. Studies of reconstruction per- formance indicate ns resolution sufficient for muon track reco ( Biron, AIR ).

March 27, 2002K. Hanson - Calor2002 – Pasadena, CA5 Dealing with Discriminator Walk Electrical signals suffer large amount of dispersion in cables  pulse risetime corrections necessary. Timewalk effect is deter- ministic: pulse risetimes at surface are up to 100’s of ns but jitter is still at ns level. Pulse risetimes only signi- ficant for electrical channels; optical channels Cable TypeRisetime (typ.) Coaxial100 ns Twisted pair50 ns Optical fiber10 ns

March 27, 2002K. Hanson - Calor2002 – Pasadena, CA6 Variation 1: Laser T 0 T 0 measures the signal propagation time from OM to TDC –Cable prop. time –Front-end electronics Amplifiers (SWAMPs) Discriminators High power pulsed Nd:YAG laser at surface delivers 532 nm light via optical fibers to OM.

March 27, 2002K. Hanson - Calor2002 – Pasadena, CA7 Laser T 0 (continued) Acrylic “diffuser ball” at OM isotropizes laser light. –Each OM on strings 1-4 and strings equipped with diffuser ball. –Only even OMs on strings 5-10 have diffuser ball: neighboring OM used for those lacking diffuser. Fiber lengths determined separately using OTDR equipment.

March 27, 2002K. Hanson - Calor2002 – Pasadena, CA8 Laser Calibration Analysis YAG intensity controlled via ND filters and optical atten- uators. Sample T 0 in range of 1 – 5 photo-electrons. TDC leading edge plotted vs. 1/SQRT(ADC) –Y intercept is T 0 –Slope is timewalk coefficient Fiber lengths must be sub- tracted to obtain signal propagation time. Precision of laser cal esti- mated at ~ 5 ns.

March 27, 2002K. Hanson - Calor2002 – Pasadena, CA9 Stability of AMANDA T0’s Q: Is annual calibration sufficient? –Station closed for winter. –No HW changes unless catastrophic failure of equipment –Electronics in ice static –TDCs use crystal oscillator: very stable. A: YES!

March 27, 2002K. Hanson - Calor2002 – Pasadena, CA10 Laser T 0 Summary Laser T 0 remains the default AMANDA time calibration method. Very labor-intensive: full detector calibration done annually requires 1000 man-hours! Useful for debugging detector (channel mapping errors) after hardware work. Very easy to piggyback crosstalk mapping. Only method that currently obtains timewalk coefficients.

March 27, 2002K. Hanson - Calor2002 – Pasadena, CA11 Variation 2: Using CR Muons AMANDA-II receives ample supply (~100 Hz) of downgoing muons. If T 0 ’s known well enough to give track reco then possible to iteratively refine T 0 guesses. Premise: shift in T 0 will appear as offset in timing residual Timing residual = Measured hit time – Hit time expected from track parameters

March 27, 2002K. Hanson - Calor2002 – Pasadena, CA12 Outline of Algorithm 1.Reconstruct muon tracks using best knowledge of T 0 calibration constants. 2.Accumulate time residuals from tracks. 3.Determine time offset from residual distribution. 4.Apply offset as correction to T 0 constant. 5.Go to step #1. Repeat until T 0 converges to a fixed point.

March 27, 2002K. Hanson - Calor2002 – Pasadena, CA13 Offset Determination Timing residuals have complex structure: some suggestions for getting the ‘zero’ point: 1.Take maximum of distribution, 2.Fit gaussian around max, 3.Cross-correlation with template We chose 3 since it was overall most robust. Fast implementation using FFTs.

March 27, 2002K. Hanson - Calor2002 – Pasadena, CA14 Convergence Convergence is monitored by plotting width of distri- bution of offsets for each iteration. Applying only a fraction of offset at each iteration seems to stabilize method against oscillations: Terminal value of this width gives rough estimate of precision of calibration. Still not clear how close initial guess must be in order to ensure convergence.

March 27, 2002K. Hanson - Calor2002 – Pasadena, CA15 Testing Muon-T 0 Take standard AMANDA T 0 constants and shift by known amount (black line in figure). Run 25 iterations of muon-T 0 procedure. Corrections shown as red points.

March 27, 2002K. Hanson - Calor2002 – Pasadena, CA16 Muon-T0 Summary Systematic drift over string hampers adoption by AMANDA as primary calibration – however it has been used in conjunction with laser T 0 : –Incorrect laser T 0 ’s (fiber lengths incorrectly measured), –Fibers leaking at OM optical penetrator, –Can quickly check validity of T 0 ’s for any given run period throughout the year or even previous years’ data! –2001 calibration done using muon-T 0 to determine which offsets have changed: only run laser cal on those channels (~50 as opposed to 700). No special runs necessary (eliminates 1000 man-hour task in favor of 25 man-hour task). Does not (yet) calibrate timewalk coefficients. Eventually hope to use improved muon calibration.

March 27, 2002K. Hanson - Calor2002 – Pasadena, CA17 Variation 3: IceCube Clock Synchronization Baseline technology for IceCube is DOM: –Waveforms digitized in situ, stamped with local DOM time, and sent to surface as digital packet. –Each DOM (~5000 total) has independent clock oscillator. –Surface clocks are synchronized to GPS clock using high precision rubidium clock RAPCal ( R eciprocal A ctive P ulsing Cal ibration) method: –Surface electronics sends pulse on communication line at time T 1, –DOM digitizes pulse /w/ local timestamp T 2, sends mirror pulse at time T 3, –Surface electronics digitizes pulse /w/ timestamp T 4.

March 27, 2002K. Hanson - Calor2002 – Pasadena, CA18 RAPCal Waveform Analysis Pulses sent at time clock is latched, Received pulses arrive asynchronously, must fit WF to get higher precision than 33 MHz clock. Linear fit to leading edge extrapolated to baseline. T arrival

March 27, 2002K. Hanson - Calor2002 – Pasadena, CA19 RAPCal (continued) RAPCal calibrates: –Cable propagation delay (T 0 !) –Ratio of clock frequencies f SURF /f DOM RAPCal/DOM technology being tested now in AMANDA-II (18 th string DOMs) String 18 RAPCal done approximately 0.1 Hz, achieves time resolution of approx. 5 ns RMS.

March 27, 2002K. Hanson - Calor2002 – Pasadena, CA20 RAPCal Summary Hardware built into DOM and DOMHub (front- end surface electronics). Software runs as application in DOMHub. RAPCal must run in realtime: IceCube trigger depends on globally time-ordered hits. Simple linear fits currently implemented seem adequate to achieve desired time resolution of ~ ns. RAPCal in IceCube eliminates need for explicit T 0 calibration.