DMT Monitor Verification with Simulated Data John Zweizig

DMT Monitor Verification with Simulated Data John Zweizig LIGO/Caltech LIGO-G 9900 XX-00 -M

Verification: Current Practices l Test trigger generation using IFO data » Noise is correct » Difficult to verify calculations l Manual ID of “non-gaussian” events » » Good feel for data, but Tedious – error prone Limited samples – not thorough Difficult to quantify efficiencies, resolutions Important but insufficient verification LIGO-G 9900 XX-00 -M

Verification with Simulated Data l Test with real IFO or generated noise » Intermediate result distributions know for generated noise » Can test effect of specific features (e. g. lines) l Inject known trigger sources » Measure efficiency vs. F, Amplitude, width, etc. » Measure efficiency for different waveforms (sine gaussian , gaussian noise burst, damped sine, etc. ) » Measure resolution of inferred parameters (t, F, etc. ). LIGO-G 9900 XX-00 -M

DMT Simulation Package l DMTGen Features: » Combines discrete signals with continuous background noise (generated or from frames) » Write output data to frames (looks like raw data) » Stand-alone program (no coding, compilation or rescripting) » Simple control syntax » Event parameters and signals recorded in output frames » Arbitrary filtration/delay » Fast: LIGO-G 9900 XX-00 -M

DMTGen Control Syntax l Parameter definitions » Set run parameter values l Filter Statements » Define filters, delays, etc. l Source definitions » Define continuous or discrete data sources » Specify timing, enable saving to frames l Channel definitions » Channels are a sum of sources, written to output frame LIGO-G 9900 XX-00 -M # These parameters define the times for the generated data. # Parameter Start. GPS 730000000 Parameter End. GPS 730000480 # # Define a source of background white gaussian noise # Source GS White. Noise(A=2. 0) # # Define a source of periodic noise bursts. These will be produced # with varying amplitudes (d. N/d. A ~ A^{-2}) and a width (sigma) of # 10 ms. The time of the burst will be random with an average rate # of 0. 2 Hz. Note the -simevent flags causes DMTGen to write a # description of each generated even to the output frame # Source GB Gauss. Burst(A=power(-2, 2), Sigma=0. 010) -rate 0. 2 -simevent # # Write a channel called "L 1: LSC-AS_Q" consisting of the sum of the # Background noise and data sources. # Channel L 1: LSC-AS_Q GS GB #

Discrete Signal Sources l Analytic functions » Sin. Gauss(A, F, Q, Phi, Width) » Damped. Sine(A, F, Q, Phi, Width) » Gauss. Burst(A, Sigma, Width) l l Single injections, constant or random time separation. Fixed, random function parameters, optionally recorded in frame. Event data recorded in Fr. Sim. Data structures (optional) Arbitrary filtration or delay applied individually. LIGO-G 9900 XX-00 -M

Waveform Parameter Generation l Parameter distributions » Constant or string » flat(min, max): d. N/dx ~ k » step(x 0, xmax, ): – x = x 0, x 0+2 , …, xmax » xstep(x 0, xmax, ): – x = x 0, x 0 2, …, xmax » gauss( , <x>): – d. N/dx ~ exp(–(x-<x>)2/2 2) » power(b, min, max): – d. N/dx ~ xb » exp(b, min, max): d. N/dx ~ e-bx LIGO-G 9900 XX-00 -M

Background Noise Generation l Continuous waveform sources » White. Noise(A) » Sine(A, F, Phi) » Frame. Data(Channel, Files) LIGO-G 9900 XX-00 -M

Match. Trig - Trigger Checking l Assign each generated event to the nearest trigger. » Triggers are read from Monitor xml output file or Data Base. » Event parameters are recorded by DMTGen in frames. l Plot trigger efficiency versus: » Amplitude, frequency, other generation parameters » Time since previous event l Plot reconstructed parameter resolution » Time, amplitude, frequency versus generated parameters LIGO-G 9900 XX-00 -M

Match. Trig – PSLmon results LIGO-G 9900 XX-00 -M

Writing Verifiable Trigger Code l l Start with finest ingredients (verified components) Look at every cut quantity » Histograms or spectra of intermediate results » Trip counters after each cut l l l Measure trigger efficiency versus generated signal parameters Measure t, F, A resolution Verify significance calculations: » Trigger rate for gaussian noise should be SR/erf( thresh) LIGO-G 9900 XX-00 -M

Summary l l We neeed to improve on current techniques of DMT trigger generator verification. DMTgen provides easily understood data for use in verifying DMT code. Match. Trig compares trigger results to generated event parameters – gives efficiencies and resolution. Verification of histograms of intermediate results is facilitated by using known input distributions. LIGO-G 9900 XX-00 -M