Draft of plan for study of DSO filter efficiency using simulated bursts
AJW (1/15/02)

Study DSO's response to (software) injected signals...

   - obtain tclsh scripts for running all 3 DSO's.
     - make sure we know what filterparams need changing

   - Obtain running time on an LDAS facility (preferably site).

     Obtain a test DB for storing triggers;
     make sure we can distinguish triggers generated using
     simulated data from triggers using real data.

     Run the DSOs exactly like they were run during E7.

   - OPTION: real E7 data (complete with glitches)
             or simulated Gaussian noise with E7 noise power spectra;
             in both cases, we have 3 IFOs to look at.

     Do we want to evaluate efficiency in presence of 
     Gaussian noise or outlier glitches?

     Want to do both! Start with the former.
     For a pub, must have REAL efficiencies using E7 data;
     this requires high statistics,
     so we must work towards having high statistics simulation runs.

   - Have access to minutes or hours of frame data
     containing all the channels needed by the DSO,
     locked streches, with relevant calibration,
     and some measure of how glitchy.

   - use stand-alone tools to acquire data, inject signals, 
     and deposit modified frame file on LDAS (via ftp)
   
   - OPTION: different waveforms.
     I think it's best to start with relatively narrow-band 
     bursts (damped sinusoids).
     
   - Damped sinusoids have 3 parameters:
     h_peak, f_central, and bandwidth (1/tau).

   - Absolute normalization of waveform
     (eg, h_peak = 1e-19, etc).

   - Filter waveform through E7 transfer function,
     including whitening, etc.
     Inject on top of data or simulated noise
     with E7 noise power spectrum.
     (Using E7 as a specific example; can do it for any TF).

   - Then we want to inject many bursts,
     varying h_peak, f_central, and bandwidth.
     
   - Check DB to see if burst was found 
     (eventually, within pre-defined windows in parameter space).

   - Represent efficiency vs h_peak, f_central, and bandwidth 
     with suitable plots.
   - At the same time, evaluate efficiency vs filterparam thresholds.

   - Repeat, with different waveforms...

   - Repeat, with data sources

   - Repeat, optimizing DSO parameters, thresholds, etc
     As we change the DSO parameters significantly,
     we must be prepared to run the DSO(s)
     through the entire E7 dataset again.

   - We may also want to study correlations between IFOs
     using these simulation tools.