cWB2G prc paper is on
trac repository. A very old version that should be probably rewritten.
Reading the old version, these are some comments:
- Check the earth coordinates of KAGRA and LIGO/India
Last paper should be Phys. Rew. D 90, 024053 (2014)
- Check if the sensitivity curves have changed
The same paper gives the same curves we are already using (as it seems comparing the figures)
- Analysis probably should be re-done with an appropriate cWB svn version and configuration parameters: establish which
- Decide which waveform to use and distribution
Past articles/Presentation/Poster
State of the art
Reed paper presents results for 2015HL and 2016HLV,
link at DCC. This paper should presents results from 3-fold network to 5-fold network, adding analysis using Invariant Topology detectors.
Last results on source reconstruction are in this
wiki page. In particular:
- The comparison of un-modeled search (row 22) with respect to a previous version link shows little difference, some better, other worse, but not so important. Improvements with respect to 1G still remains.
- Some not understood aspects remains, in particular on P search (row 21). This search performs a circular constraint adding a transformation on the pixel composing the cluster. In particular this transformation make a projection of the pixels likelihood according to the mean values of all composing the cluster. For an example see the link under the title "G & MASK". It is not still clear why the circular polarization waveform has worse results while linear polarization ones are better.
Results with last pipeline version
I made some simulations of SG for various Q for LHV network. Results report median value of erA[0] over each single sky pixel (Healpix=4) and construct Lars Histogram.
THIS RESULTS ARE MADE WITH FIXED NETWORK SNR ALONG THE SKY
Results show a dependency on frequency according to the sensitivity curve as in the figure:
According to the results, the following comes out:
- For lower frequency (up to 500 Hz) the curves seems following the behaviour of 1/f^2, like in Wen&Chen formula
- For upper frequency, the results are probably related to algorithm approximation, this is quite confirmed by test with upTDF=8 (instead of 4) for SGQ9
- For higher Q, performances worse, because it enters the group error, other than phase error. It would be nice to identify the correlation between the two effects
NOTES FOR INVARIANT DETECTOR PAPER
Starting proposal
- Networks: Starting point could be to analyze networks with L-shape detectors: from 3-fold to 5-fold, following the future starting-on on the detector (L1H1V1 - L1H1V1K1 - L1H1V1K1I1)
- cWB2G:
standard parameters
no regulators (delta=gamma=0)
P search
Single resolution (?)
- Waveforms:
Inject with a distribution uniform in volume
use the same as PhysRevD.83.102001: (SG235Q8d9, SG235Q3, SGC235Q9, WNB_250_100_0d1), for making general consideration of the behaviour of different network
CBC-like signals, but which parameters?
- Invariant detectors
Which IT? There should be no so much differences between IT90, IT45, ET, we can make a test comparison performances between these three and then choose one of them
Network: 3-fold with 1 IT, 3-fold with 2 IT, 4-fold with 1-IT, 4-fold with 2-IT
Paper section
How is the paper now and what should be written in each section:
- Introduction
standard introduction about EM follow up, and past literature
- Advanced network
list of present and future Adv, description of waveforms, results for 3-4-5 fold
- Beyond advanced detectors
description of 3G det, study of triangulation and antenna patterns (introducing IT section) (Join with following section?)
- Invariant detector topology
definition of IT, performances
- Optimal IT network
comparison of cartesian, big circle, tetrahedron networks with existing ones. (Maintain???)
- CBC
Results reporting a test with CBC
- Conclusions
MDC:
- Waveforms with duration and frequency distributed on interesting values:
- SineGaussian: f=[70,2000]Hz logarithmically distributed
Q=[3,100] logarithmically distributed
Linear and Elliptical (cos(iota) uniformly distributed) -> Two sets
- White Noise Burst: Uniformly distributed in central frequency [100,2000]Hz, duration [1,100]ms and bandwidth [10,1000]Hz
- BBH: could be the same population as Reed, Mtot=[15,25] uniformly distributed
should be use constraint on the chirp mass?
- I would not use Gaussian.
- Distance distribution: uniform in volume (For SG and WNB difficult to establish a range).