Study on ER6 using cWB

Online information

Analysis

• At Caltech cluster: L1H1 network
  home page
  good data quality periods
  2815 60s-segments analyzed
• At ATLAS cluster: L1H1V1 network
  home page
  good data quality periods
  60s-segments analyzed

Presentation

• LVC Stanford meeting August 2014
• Allsky teleconf 13/11/14
• Allsky teleconf 25/11/14
• Recap of burst call 14/01/15
• LVC Pasadena meeting March 2015

Online ER6 characteristic

Worked well

• zero lag performances ( run statistics)
  

  Mean processing time: ~ 50 s

  Mean (processing + launch delay) time:~ 110s

  Trigger delay of sending to gracedb:~ 150s

  	Difference between the GPS end of the analysis and the center segment GPS time

To be checked

• Background analysis
  1000 lags for 1200s-segments
  Jobs takes more than 20 hours to process
  Possible causes:
  

  Lower thresholds
  L1 glitchiness

  Necessary more dedicated studies offline

Offline studies

• Online background: standard cWB page
  Online analysis was made with lower threshold than default to allows more triggers for zero lag, (bpp and netRHO), this surely increase the running time
• Offline background using standard threshold: standard cWB page
  1000 lags on 1200s segments -> running time decreases
  

Running time performances

Performed several tests using different lags number. For this study we use the search R, and not the I, as for the online analysis

1200 s segments	1000 lags cWB page	500 lags cWB page	250 lags cWB page	100 lags cWB page
Running time
Memory usage
	Suppose to have 100% duty cycle: 6.5 hours for 1000lags 20m jobs 20 machines to maintain 1000 lags analysis	Suppose to have 100% duty cycle: 3.2 hours for 500 lags 20m jobs 2*10=20 machines to maintain 1000 lags analysis	Suppose to have 100% duty cycle: 1.5 hours for 250 lags 20m jobs 5*4=20 machines to maintain 1000 lags analysis
600s segments
600s segments	500 lags	250lags
Running time
Memory usage
	Suppose to have 100% duty cycle: 1.5 hours for 500 lags 10m jobs 9*4=36 machines to maintain 1000 lags analysis	Suppose to have 100% duty cycle: 0.9 hours for 250 lags 10m jobs 6*8=48 machines to maintain 1000 lags analysis

1200 segments - 32-1024 Hz - Resampling to 2048 Hz
1200s segments*	*1000 lags	250 lags
Running time
Memory usage

Background distributions

cWB page - TXT event list

Amplitude distribution

rate vs rho	rho distribution

Frequency distribution

Frequency vs rho	Frequency distribution	Bandwidth distribution	Bandwidth vs Frequency

Single detector Analysis

Because only L1 has real data, we perform a only L1 analysis.

cWB page - TXT event list

Loudest event (rho>1000)

ID	rho[1]	cc[0]	subnet	lag	slag	SNRnet	f	low	high	dT	size	res	run	GPS L1	GPS L1	SNR L1	SNR L1
1	39059.56	1.00	1.00	0	0	55677.6	73	32	864	3.406	1445	128	14	1102350765.49	1102350765.49	38729.8	38729.8
2	25760.21	1.00	1.00	0	0	36055.5	44	32	480	4.273	1089	128	7	1102121320.26	1102121320.26	25690.5	25690.5
3	4622.61	1.00	1.00	0	0	6557.4	49	32	2016	8.078	1943	128	27	1102369074.55	1102369074.55	4582.6	4582.6
4	4057.61	1.00	1.00	0	0	5744.6	54	32	1568	5.766	1894	128	36	1102380779.58	1102380779.58	4000.0	4000.0
5	3981.23	1.00	1.00	0	0	5656.9	42	32	224	3.289	497	128	120	1102801310.48	1102801310.48	4000.0	4000.0
6	3896.08	1.00	1.00	0	0	5477.2	58	32	672	5.852	1113	128	80	1102658740.65	1102658740.65	3873.0	3873.0
7	3527.64	1.00	1.00	0	0	5000.0	57	32	1568	5.859	1919	128	36	1102380793.10	1102380793.10	3464.1	3464.1
8	2442.11	1.00	1.00	0	0	3464.1	41	32	672	3.703	533	128	119	1102800994.61	1102800994.61	2449.5	2449.5
9	2356.46	1.00	1.00	0	0	3316.6	298	36	1492	2.500	1376	16	66	1102567394.61	1102567394.61	2366.4	2366.4
10	1854.38	1.00	1.00	0	0	2626.8	45	32	224	2.461	346	128	120	1102801273.65	1102801273.65	1843.9	1843.9
11	1621.13	1.00	1.00	0	0	2302.2	42	32	224	2.547	356	128	122	1102803761.68	1102803761.68	1612.5	1612.5
12	1524.67	1.00	1.00	0	0	2167.9	310	34	2046	2.375	2187	8	69	1102570997.13	1102570997.13	1516.6	1516.6
13	1297.16	1.00	1.00	0	0	1843.9	46	32	160	2.258	302	128	120	1102801292.85	1102801292.85	1303.8	1303.8
14	1138.51	1.00	1.00	0	0	1612.5	42	33	143	3.250	283	4	127	1102810603.26	1102810603.26	1140.2	1140.2

Other events

rate vs rho	rho distribution
Frequency vs rho	Frequency distribution	Bandwidth distribution	Bandwidth vs Frequency

LIB EVENTS

List of LIB events are on gracedb: Omicron+LIB

These events were no detected by cWB, we perform an analysis lowering the subnet, netRHO and netCC threshold, one using I search (as in online analysis) and one using R search, using regulators or not

GRACEDB	TIME	Frequency	Q	I search Freq - rho - cc	I search no regulators Freq - rho - cc	R search Freq - rho - cc	R search no regulators Freq - rho - cc
G124604	1102808224.2070	214.38	3.56	-	-	-	-
G124593	1102805651.3189	35.73	2.74	-	-	-	-
G124577	1102764101.0508	41.74	99.78	-	-	-	-
G124534	1102743920.3152	1195.03	4.32	1216 - 3.27 - 0.61 - ced	1216 - 3.19 - 0.59 - ced	1216 - 3.27 - 0.61 - ced	1216 - 3.24 - 0.60 - ced
G124526	1102670780.1133	624.74	23.13	-	-	-	-
G124520	1102661892.8486	794.49	32.94	-	-	--	-
G124496	1102650062.4419	1146.23	3.9	-	-	-	-
G124480	1102641451.6870	1172.99	14.64	-	-	-	-
G124465	1102565381.8145	722.66	14.75	707.14 - 2.78 - 0.54 - ced	707.14 - 2.77 - 0.54 - ced	707.06 - 2.78 - 0.54 - ced	707.14 - 2.77 - 0.54 - ced
G124456	1102477352.6910	1023.64	5.37	-	-	-	-
G124450	1102473882.8709	1156.32	62.49	-	-	-	-
G124429	1102414644.2021	380.57	11.31	380.28 - 4.18 - 0.65 - ced	380.28 - 3.77 - 0.65 - ced	380.92 - 4.21 - 0.67 - ced	380.13 - 3.98 - 0.67 - ced
G124427	1102414154.8320	746.07	49.72	-	-	-	-

PE plugin

The plugin has been applied only on 12 events for a bug in the code.

The plugin calculates central time and frequency using as weight the time series amplitude on whitened/strain reconstructed waveform.

cWB default time and frequency are calculated using as weight the time-frequency amplitudes of all the WDT decomposition levels used in the analysis

We compare here the value between "default" and plugin.

TIME

We report the histogram of the difference between the two estimation.

All events	Zoom at difference less than 10 ms
In each figure there is the difference between the plugin value and the default value. The right figure is a zoom on the x axis reporting value which difference is less than 10 ms. Top: plugin value calculated on strain waveform. Bottom: Plugin value calculated on whitened waveform. Left: L1 detector. Right: H1 detector.

FREQUENCY

Absolute difference	Percentage difference
In each figure there is the difference between the plugin value and the default value. Top: plugin value calculated on strain waveform. Bottom: Plugin value calculated on whitened waveform. Left: L1 detector. Right: H1 detector.