Robotic algorithm of HEND

1
Robotic algorithm of HEND detection of GRBs

• V. Grinkov, A. Sanin, S. Charyshnikov, A.
  Kozyrev, M. Litvak,
• I. Mitrofanov, V. Tretyakov
• Institute for Space Research
• K. Hurley
• Space Sciences Laboratory, University of
  California, Berkeley
• T. Cline
• NASA/Goddard Space Flight Center

2
Gamma-Ray Burst HEND Subsystem
Profile data are transmitted continuously by 60
s profiles Time resolution Outer 0.25 s
Inner 1 s Energy range Outer 30 1000
keV Inner 60 2000 keV
Spectra of both detectors consist of 16 channels
and have time resolution of 20 s
3
This is the position where the GRB will be
registered
GRB phase
Afterglow phase
t
0
1 second
1 month
1 minute
1 hour
1 day
4
Three main steps of the algorithm are as follows
I. Background Model, to know the background
variations
II. Search for Events, to find possible
candidates to be GRBs
III. Correlation Analysis, to eliminate
non-burst events
5
I. Background model

1. Smoothing
2. Weight determination
3. Weighed smoothing
4. Return to Step 2

The initial profile is smoothed with 70-bin
window to derive a smoothed one
6
I. Background model

1. Smoothing
2. Weight determination
3. Weighed smoothing
4. Return to Step 2

In future smoothing each bin will have its
weight which is determined by the weight
function (upper left) according to the
difference between smoothed and initial profiles
7
I. Background model

1. Smoothing
2. Weight determination
3. Weighed smoothing
4. Return to Step 2

Derived with the same window, new smoothed
profile substitutes the old one
8
I. Background model

1. Smoothing
2. Weight determination
3. Weighed smoothing
4. Return to Step 2

Steps 2-3-4 are repeated 20 times until the
smoothed profile represents a good background
estimate (green line at the picture)
9
II. Search for events
1. Search for triggered bins and combine them
into events 2. Discard short events 3. Stretch
boundaries
All bins exceeding 6 std. dev. above the
background are treated as triggered. Neighboring
triggered bins are combined into events.
10
II. Search for events
1. Search for triggered bins and combine them
into events 2. Discard short events 3. Stretch
boundaries
Events shorter than 3 bins are discarded
11
II. Search for events
1. Search for triggered bins and combine them
into events 2. Discard short events 3. Stretch
boundaries
35 bins are added to each side of the
event. Some events will merge.
12
III. Correlation analysis
The GRB gives signals both in the outer and
inner SC (left),
while the anomalous signal is reflected only in
the outer (down).
The blue lines represent flux in the outer, and
the green lines in the inner, detector
13
Statistics of the registered events
Time period Known GRBs GRBs found manually GRBs found by the algorithm Number of false events
Feb 02 - Oct 02 (quite) 92 41 6 0
Nov 02 - Jan 03 (noisy) 34 7 1 14
Mar 03 - Apr 03 (noisy) 30 9 1 10
14
The whole set of GRBs found by the algorithm
15
Conclusions

1. An algorithm to search for bursts in HEND data
   was built.
2. The algorithm triggered 8 real and 24 false
   events during both quite and noisy data.
3. As an improvement of the algorithm spectral data
   procession must be added.