Characterization of PEM Channels Chirpiness of the Test Mass motion due to Noise Anthony Rizzi Staff Scientist, LIGO LIGO-G D Goals Push installation of PEM Verify PEM chain: installation and design From instrument through data collection and analysis Characterize PEM and other noise channels * PEM Channels Tiltmeters1,1,1Weather Station1,1,1Microphones8,1,1Temp. Sensors4,4,4AccelerometersC,Y,X= 8,1,1Residual Gas Analyzers1,1,1Dust Monitors10,2,2Seismometers1,1,1RF Receivers1,0,0Muon Detector1,0,0 Anthony: Read = useful chirp search Anthony: Read = useful chirp search * IFO Noise Curves How to Interpret PEM Channels Displacement spectral density. What do we care about? What interference does Non-gravitational wave source impinging on detector cause? Include data analysis method. Benchmark: #false GW/time interval: E.G., 1 false GW/year. Binary Inspiral: Chirp 1.4M Solar -1.4M Solar : * Optimal Filtering: Each Channel Introduces Chirps Anthony: Normalization: =1 Anthony: Normalization: =1 Mass Template m2 m Match between a and b: * Tools Modified GRASP Works with new Frames Works with DMT Modification of existing code Analysis tools added Two Beowulf systems: One 12-node Linux cluster One 6 node Solaris cluster (1, 4-processor node) Gaussian Noise * Gaussian White Noise Prob. Vs. Template No. Non-local Gaussian White Noise Prob. Vs. Template No.(local) Gaussian White Noise Prob. Vs. SNR 3438 files; probs=0; 4.67>SNR>3.3 Only one template per file included Gaussian White Noise Injecting Chirps: 1.4M-1.4Mper15 data segs. Mag~32k Prob. Vs. Template Number * Gaussian White Noise Injecting Chirps: Alternate between 1.2M,1.6Mper 15 segs.Mag: ~32k Prob. Vs. Template Number * Gaussian White Noise Injecting Chirps: 1.4M-1.4Mper15 data segs. Mag~8k Prob. Vs. Template Number 214 injected Gaussian White Noise Injecting Chirps: 1.4Mper15 mag~5.6k; SNR~3.5 Prob. Vs. SNR 3263 files Prob>.1 SNR>5 * Gaussian White Noise Injecting Chirps: Alternate between 1. 4M,1.4M every 15 segs. Magnitude: ~5.6k Prob. Vs. Template Number Fraction detected=35/217~1/6 Ground Motion will cause test mass motion that can mimic gravity waves. Transfer ground motion to test mass using stack transfer function and pendulum transfer function from Ed Daws document. Get Stack Model, use Mathematica to generate table. Looking for Chirps M Stack(f) Transfer function of Stack and Pendulum Gaussian Noise Thru Seismic Transfer Function (Stack, pendulum, accelerometer) Prob. Vs SNR for SNR>5 PROB> files all probs=0; 3.99>SNR>2.05 Experimental Setup Accel. big ADC Seismom. Anti-Alias Modified GRASP Accel small Pictures of ADC and Processor * Picture of Accelerometer and Seismometer Transfer Function of Accelerometer Background Seismic Spectrum (Accelerometer) Anthony: 3/10/00 seismic background measured with accelerometer with pumps off Anthony: 3/10/00 seismic background measured with accelerometer with pumps off Approximate LIGO Test Mass Motion 20Hz 100Hz (all calcs. are order of magnitude) Calculation of Ground motion Transfer function (G accel =1000V/g) : m/count (equivalent to strain/count) 20Hz 100Hz Note: there is gain of 2 in external Amp and all calcs. are order of magnitude. Head Room for Seismic Noise +/- 2v OR +/- 32K (2 15 ) counts Implies can only look at factor of ~ two in frequency at a time ~ (f 1 /f 0 ) 12 * 24 hours Seismic Data Set I SNR Vs Prob (large Wilcoxan accelerometer from South End on Weekend) files: 51 with 0 5) Only one template per file included ~38 events: High peaks suppressed 24 hours Seismic Data Set I SNR Vs Prob (same as conditions as above) Only one template per file included 24hrs Seismic Data Set I SNR Vs Prob Same conditions as previous with all templates shown SNR>5 prob >.1 24hrs Seismic Data Set II SNR Vs Prob (all points) Same conditions as previous, on 24hrs Seismic Data Set II Prob Vs Template No Same conditions as previous, on SNR>5,prob>.1 SNR>0, prob>0 279 events with SNR>5 PROB>.1 24hrs Seismic Data Set II SNR Vs Prob Small accelerometer (G=10) All probs=0 4.2>SNR> files Only one template per file included 24hrs Seismic Data Set II Time Series: Large Accelerometer Channel signals At sec Passed Gaussian Test Gaussianity Small Accelerometer Data Seg. Length # of Segs # of Non-Gaussian 0% 22% 23% 24% * Data taken by Chethan P. using our Gaussian Monitor Summary with Benchmarks Modified GRASP and tools available on DMT Gaussian Noise (SNR>5, prob>.1) (on 24hrs Matlab generated data) Benchmark: White: >0 false events/day Seismic Colored: >0 false events/day Signals at SNR~3.5 yield ~ 90% false dismissal rate Seismic Noise (SNR>5, prob>.1) (2 sets 24hrs data) E.G Hz width (note any multiple of ~ two in passband) : Absolute Benchmark > ~ 0 false events/day Relative Benchmark > ~ false events/day (~1-11/hour) Future Direction More, Seasonal and longer Seismic Data stretch Analyze Data for other Noise Sources (microphonetone generation, RF) Make experiment algorithmic so all nonGW signals can be characterized by the chirpiness benchmark. Incorporate non-PEM noises. Setup so chirp checking is continuous to allow long data stretch real time checking. Generalize to other types of GWs. More Involvement to get all channels covered: UL Group interest