Attenuation measurement with all 4 frozen-in SPATS strings
Justin VandenbrouckeFreija Descamps
IceCube Collaboration Meeting, Utrecht, NetherlandsSeptember 15, 2008
Outline
• Motivation and background• Data set• Method• Amplitude vs. distance• Best fit and confidence regions• Attenuation length lower limit• Systematics
Justin Vandenbroucke Utrecht, Netherlands September 15, 2008
Motivation
• Two types of attenuation analyses– Inter-string: frozen-in sensors & transmitters– Pinger: frozen-in sensors & retrievable pinger in water
• Last year’s 3-string inter-string analyses inconclusive• Last year’s pinger analyses inconclusive• Improve with 4-string inter-string?• Now enough strings for single-transmitter analysis, to reduce systematics
Justin Vandenbroucke Utrecht, Netherlands September 15, 2008
• Is attenuation length at least a few hundred meters?
Background
Significantly improved inter-string data set taken
• 4 strings (D with improved transmitters and sensors)• Each sensor records 200 seconds @ 200 kHz• 40 Hz transmitter repetition: 8,000 pulses per T-S
combination! cf. 726 pulses previously• Optimized parameters
– Sampling frequency– RAM disk size– Transmitter repetition rate– Steering amplitude
Justin Vandenbroucke Utrecht, Netherlands September 15, 2008
Geometry: D transmitter to ABC sensors, all at 320 m depth
Justin Vandenbroucke Utrecht, Netherlands September 15, 2008
Data subset for this analysis
• Single depth (320 m) transmitters & sensors, to reduce systematics
• One transmitter only, to reduce systematics• All 3 ABC string sensors recording• All 3 channels per module recording• Select only the first 10 s: for longer duration we need
more precision in clock drift determination
Justin Vandenbroucke Utrecht, Netherlands September 15, 2008
Pulse averaging algorithm
Justin Vandenbroucke Utrecht, Netherlands September 15, 2008
(1) Stretch times according to clock drift(2) Wrap and re-sort by time(3) Re-bin at chosen sampling frequency
- all samples wrapped and sorted- average pulse after binning at 200 kHz
Azimuthal variation of transmitter?
Justin Vandenbroucke Utrecht, Netherlands September 15, 2008
transmittersensor
sensor
sensor
19°
29°
We use a small range of azimuths:
Analysis method
• Frozen-in transmitter removes many systematics plaguing the pinger analyses
• Single level to minimize systematics• String D transmitter (good quality)• Use amplitudes directly (no ratios assuming
negligible angular variation)• Determine clock drift by scanning over assumed drift
values, maximize Vpp, check Vmax and Vmin peak at same drift value
• Apply full confidence level treatment
Justin Vandenbroucke Utrecht, Netherlands September 15, 2008
Peak to peak amplitude vs. distance (linear scale)
Justin Vandenbroucke Utrecht, Netherlands September 15, 2008
Ln(amplitude * distance) vs. distance
Justin Vandenbroucke Utrecht, Netherlands September 15, 2008
“statistical” errors: std. dev. of avg. pulse Vpp
• pulse to pulse signal variation• pulse to pulse noise variation• residual clock drift
Attenuation coefficient confidence region
Justin Vandenbroucke Utrecht, Netherlands September 15, 2008
Best fit and sigma of each parameter from analytical methodEllipses from numerical method• best fit- best fit +/- 1 sigma- delta-chi-square = 1 (tangents contain 68.3% of either parameter alone)- delta-chi-square = 2.3 (contains 68.3% of parameter space jointly)
Attenuation length confidence region
Justin Vandenbroucke Utrecht, Netherlands September 15, 2008
Best fit: 1055 m
Attenuation coefficient PDF (Gaussian)
Justin Vandenbroucke Utrecht, Netherlands September 15, 2008
Attenuation length: probability distribution function
Justin Vandenbroucke Utrecht, Netherlands September 15, 2008
Attenuation length cumulative PDF and lower limit
Justin Vandenbroucke Utrecht, Netherlands September 15, 2008
Attenuation length > 363 m @ 68.3 % CL• statistical errors only• without constraining lambda positive
Adding 100% systematic error to the statistical error
Justin Vandenbroucke Utrecht, Netherlands September 15, 2008
Lower limit with 100% systematic error and lambda constrained positive
Justin Vandenbroucke Utrecht, Netherlands September 15, 2008
68.3% CL: attenuation length > 269 m90% CL: attenuation length > 168 m
Systematic effects
Justin Vandenbroucke Utrecht, Netherlands September 15, 2008
Effect Present? Comment
S chan. to chan. variation Accounted
Dominant systematic error: ~100% ?
IceCube cable shadowing Yes Effectively changes S sens. or T. azim.
Hole ice quality Yes Would appear as chan. to chan. S variation
Clock drift Small Removed, but sufficiently?
T azimuth response Small All S in roughly same direction
Background noise No Automatically in statistical uncertainty
T zenith response No Single depth
S zenith response No Single depth
S azimuth response No Each sensor channel used once
T module to module variation
No Single transmitter
Reflections interference No Frozen hole column
Transmission coef. with angle No Frozen hole column: no transmission coef.
Shear waves No No variation in fraction going to S waves
Saturation No None of these runs saturated
Variation in waveform shape No No pinger motion
T = transmitter S = sensor
Conclusions
• We now have high quality optimized 4-string inter-string data set• First analysis complete• Confidence interval and lower limit in addition to best fit• “Direct” method, complementary to “ratio” method• Claim: this analysis is least affected by systematics, of all (inter-string or
pinger) attenuation analyses to date• Is it good enough? No. We need improved pinger runs (see Delia’s talk) and
more inter-string analysis• To do
– Verify amplitude determination (clock drift correction algorithm)– Verify systematic error estimate– New data for few combinations with sufficient online drift determination– Repeat with transmitters at other depths on D (also ABC?)– Frequency domain analysis?
Justin Vandenbroucke Utrecht, Netherlands September 15, 2008
Cross check: Two independent amplitude determinations
Justin Vandenbroucke Utrecht, Netherlands September 15, 2008
Justin Freija
Ratio analysis: in progress
Justin Vandenbroucke Utrecht, Netherlands September 15, 2008
Pulse averaging improves signal to noise
Justin Vandenbroucke Utrecht, Netherlands September 15, 2008
- single raw pulse- average pulse
Pulse averaging: Effect of chosen re-binning sampling frequency on average pulse
Justin Vandenbroucke Utrecht, Netherlands September 15, 2008
- 200 kHz- 6 MHz
• after stretching times, wrapping, and re-sorting• BT6 to AS5-0; 3000 pulses