Challenge

Data Challenge

Challenge activity for the Open Data Workshop course.

Data files may be downloaded from DCC G2300818 using these links:

• challenge1.gwf
• challenge2.gwf
• challenge3.gwf
• challenge3_2048hz.gwf <-- Downsampled version of Challenge 3 file

Workshop participants may submit solutions via the online course as individuals or in teams of up to 3 people (check for the deadline date on the website).

Challenges are ordered by difficulty. Entries will be rewarded a number of points that scales with the difficulty of the challenge.

Warning: you can submit solutions only once so we recommend answering all the questions in a Challenge before submitting.

Good luck to all!

Challenge 1 (1 point) -- Novice

Identify a loud binary black hole signal in white, Gaussian noise:

• Use the data file challenge1.gwf. The channel name is H1:CHALLENGE1.
• The data are white, Gaussian noise containing a simulated BBH signal.

Instructions:

• Load the data into memory. What are the sampling rate and duration of the data?
• Plot the data in the time-domain.
• Plot a spectrogram (or Q-transform) of the data, and try to identify the signal.

Question for Challenge 1:

• A. What is the time of the merger (in seconds from the start of the segment)?

Challenge 2 (2 points) -- Rookie

Signal in colored, Gaussian noise:

• Use the data file challenge2.gwf, with channel name H1:CHALLENGE2.
• The data contain a BBH signal with m1 = m2 = 30 solar masses, spin = 0.

Instructions:

• Plot a Q-transform of the data.
• Generate a time-domain template waveform using approximate SEOBNRv4_opt. with the same parameters as above. Plot this waveform.
• Calculate a PSD of the data, and plot this on a log-log scale. Use axes ranging from 20 Hz up to the Nyquist frequency.
• Use the template waveform and PSD to calculate the SNR time series. Plot the SNR time-series.

Questions for Challenge 2:

• A. From the Q-transform, what is the approximative time of the merger?
• B. What is the matched filter SNR of the signal?

Challenge 3 (4 points) -- Intermediate

Search for a loud event in realistic data:

• Use the data file challenge3.gwf with channel H1:CHALLENGE3.
• These are real LIGO data from O2, though we've adjusted the time labels and added some simulated signals.
• The data contain a loud simulated signal with m1 = m2 = 10 solar masses.

Instructions:

• Use the template waveform SEOBNRv4_opt and PSD to calculate the SNR time series.

Questions for challenge 3:

• A. What is the merger time of the loud signal?
• B. What is the matched-filter SNR of the loud signal?

Note: it's not needed to use bilby for this challenge.

Challenge 4 (8 points) -- Advanced

Realistic search and parameter estimation:

• We will use the same realistic data set challenge3.gwf with channels H1:CHALLENGE3 and L1:CHALLENGE3.
• These are real LIGO data from O2, though we've adjusted the time labels and added some simulated signals.
• All simulated signals have been added to both the H1 and L1 data.
• All simulated signals have spin = 0 and m1 = m2, with m1 somewhere in the range 10-50 solar masses.
• Watch out! These are real data, and so glitches may be present.

Questions for Challenge 4:

• A. Identify as many signals as you can. Any correct detection is +1 point but any false alarms will count -1 point against your score. For each signal you find, list:

  • The merger time
  • The SNR
  • Your estimate of the component masses

• B. Identify as many glitches as you can. Make a spectrogram of each one. List the times of any glitches you find.

• C. For the earliest event you found, use bilby to compute the posterior distribution for the mass. List the 90% confidence interval for the component mass.

  Hint:

  • Fix the spin and mass ratio to make this run faster.
  • You can also fix other angular parameters to zero, while leaving at least the chirp mass, sky position, time and distance free. This will still give an approximately correct posterior for the chirp mass.

Useful notes

Processing data from a local file with Bilby

sampling_rate = 2048  # needs to be high enough for the signals found in steps above
duration = 8  # needs to be long enough for the signals found in steps above
start_time = 100  # needs to be set so that the segment defined by [start_time,start_time+duration] contains the signal

interferometers = bilby.gw.detector.InterferometerList([])
for ifo_name in ['H1', 'L1']:
    ifo = bilby.gw.detector.get_empty_interferometer(ifo_name)
    ifo.set_strain_data_from_frame_file('challenge3.gwf', sampling_rate, duration, start_time=start_time, channel=f'{ifo_name}:CHALLENGE3')
    interferometers.append(ifo)

To load data in google co-lab. Run this code, and then 'restart runtime', and run it again

! pip install -q lalsuite
! pip install -q gwpy
! pip install -q pycbc
# -- Click "restart runtime" in the runtime menu

# -- download data
! wget https://dcc.ligo.org/public/0187/G2300818/001/challenge3.gwf

# -- for gwpy
from gwpy.timeseries import TimeSeries
gwpy_strain = TimeSeries.read('challenge3.gwf', channel="H1:CHALLENGE3")

# -- for pycbc
from pycbc import frame
pycbc_strain = frame.read_frame('challenge3.gwf', 'H1:CHALLENGE3')