Interferometry

Related doc. Reco. workshop

Related code py_interferometers class

This is done by below command:

source ../setup.sh
python3 script_executor.py -k snr -s <station ID> -r <run number>                    # calculates SNR
python3 arr_time_table_maker.py <station ID>                                         # calculates arrivaltime table by ray tracing
python3 script_executor.py -k reco_ele_lite -s <station ID> -r <run number>          # performs vertex reconstruction based on AraCorrelator method

For sim:

source ../setup.sh
python3 sim_script_executor.py -k rms -s <station ID> -d <sim output path>            # both signal and noise. calculates rms
python3 rms_merge.py <station ID>                                                     # nosie only
python3 snr_maker.py <station ID> <rms results path>                                  # both signal and noise. calculates SNR by rms from noise event
python3 sim_script_executor.py -k reco_ele -s <station ID> -d <sim output path>       # both signal and noise. finds c_max of all elevation angle

User can find arrival time table from /data/ana/ARA/ARA02(or 3)/arr_time_table/arr_time_table_A2(or 3)_all.h5

Calculates the summed-cross correlation at a grid of pixels (theta, phi) for a source at fixed radius.

  • If signal is coming certain source, same type of antennas will have a similar WFs

  • By ray tracing algorithm, we can estimate arrival time differences of all air/ice positions

  • Cross-correlating WFs by removing arrival time differences, we can get correlation coefficient of all air/ice positions

  • Most highest coefficient score (Cmax) would be the reconstructed source position

5 radius (41, 170, 300, 450, 600 m), Direct / Reflect rays, D / R / D+R map, and default ice model were used for interferometry.

Detail Equations

Each cross-correlation ‘bin’ is normalized by product of sqrt of power.

Hilbert envelope is applied on each cross-correlation

Cross-correlation of all pairs are weighted by product of SNR.

Most highest coefficient score (Cmax) was chosen as reconstructed source position

../../_images/int_eq.png

Fig. 89 interferometry equations

Calpulser example

../../_images/int_hist.png

Fig. 90 calpulser example

Template Analysis

Related code ara_matched_filter class

User must generate the template by AraSim. Setup file for template: A2 and A3

This is done by below command:

source ../setup.sh
python3 arr_time_table_maker.py <station ID>                                         # calculates arrivaltime table by ray tracing
python3 sim_script_executor.py -k temp -s <station ID> -d <sim output path>           # both signal and noise. generates template that will be used for matched filter to data and sim
python3 sim_script_executor.py -k mf -s <station ID> -d <sim output path>             # both signal and noise
python3 script_executor.py -k mf -s <station ID> -r <run number>                     # perform matched filter

Searching for Neutrino signal by directly comparing with Neutrino template

  • Aiming to find low SNR signal

Simulated Neutrino templates

  • Preparing the expected Neutrino WFs that we can get from ARA

  • On/off-cone, shower, zenith gain pattern, and electronic response

Matched filtering technique

  • Cross-correlating data and template set while suppressing noise by noise model

Event-wise MF value (Mmax)

  • Summing up matched filter results (M(t)) after removing arrival time delay

Cons. of this analysis is, it is model dependent. Currently, I’m using the AraSim model. Single model dependency will be dealt with as an systematic uncertainty

Template Parameter

Four parameters for constructing template sets (total 24 WFs per channel)

  • Signal chain gain: 16

  • Zenith dependent antenna gain pattern: 4 (30, 50, 70, 90 deg)

  • On/Off-cone angle: 3 (0, 2, 4 deg)

  • Electromagnetic (EM) and hadronic (HAD) shower models: 2

Neutrino energy and the distance from the vertex to the detector, can be simplified using a scale factor.

  • Shape differences of Askaryan cut-off per energy/off-cone are not considered

Receiving angle, -60, -40, -20, and 0, and On/Off-cone angle, 0, 2, and 4, are used for template set

../../_images/mf_hist.png

Fig. 91 template parameter test results

Detail Equations

Cross-correlation between data and template in f-domain with Hilbert envelope

Suppress noise (weighting) in the f-domain by power spectral density (PSD) (averaged software triggered data)

  • Averaged software triggered data (Rayl. x sqrt(pi / 2))

Normalizes template and PSD amplitude (alpha_n)

Sums up correlation results (Msum) after removing arrival time delay (tau_n) from AraSim ray tracing

Most highest coefficient score (Mmax) was chosen for the event-wise result

Cons. of template analysis is, it is model dependent. Currently, I’m using the AraSim E-field model

../../_images/mf_eq.png

Fig. 92 matched filter equations

Event-wise Matched Filter

192 (8 chs * 24 templates) MF results are reduced to ‘single’ value (But total 2 by V/HPols)

Off-cone dim. is reduced in individual channel level

Channel dim. is reduced by summing up after removing arrival time delay

  • total 84 summed results by 2 shower, 7 zenith, and 6 azimuth

  • Summing is done by ~80 ns of rolling maximum

Among the 84 summed results, I choose 1 result that has maximum Mmax value

../../_images/mf_chain.png

Fig. 93 event-wise matched filter process

Simulation example

../../_images/mf_hist2.png

Fig. 94 simulation results

../../_images/mf_hist3.png

Fig. 95 simulation results with vertex reconstruction compare to interferometry