Home

PSS User Guide - To Parent Directory

1. 156 COMPAN 157 The Poo Flee diu n p DOREM ERR IIR a dle ea vara RE 158 Tore d log ile S o fd ER ul 161 Es mple of use OF daa ebd doen ce 162 EV EHES stmt ote M 163 166 The PSS databases data analysis organization eene 173 Genetal structure oL data bas s cu Gate Limbo 173 Th h reconstructed data base aie iiini 175 ed esta eed ra ded T 176 Thenotmalized spectra cata scm ot dete de a EH aa 176 Thepeak map ooi ice toco a e edle Ced 176 Tbe caritidate databases neta ee ebat fond 176 The comcidence databases oput da n Rte ba CA 176 Darabase nas Falls ob fos i alas defen CP cd HE o Ub ann 176 SEVEN LOCS T 176 Jal SUS ai es te Gna td dn dra sr pedo Diae nS 176 oat aep qe em an quad a eve 177 File Systemi 177 Orgatization of WOE RR ER AD GNU a ona 178 Basic analysis Pet OOS qoe quitus 178 DATA Y
2. MO Qe a SEE 133 ut UI apad aca pag A NS 135 Basic e ed os 135 ETT POPOLE Ss i 136 aA SR 138 Hough transform dete dieta ede erai d de omit 138 SELVICE TOUTINES 139 Matlab setvice foutities codon 139 140 M CD 140 General parameter 141 141 143 SOULE SUL CLUES 144 antenna 145 Structute menit 146 StEUCEUEG s 147 structures iei e i rene e e e 148 ente aetatem edt edu 149 ua AERA 150 c 151 hmap o e EA NEU d LR 152 153 SS 154 candidate SEAS ELE oi e tao it et ith t itat t eet arte i tereti 155 evetit SAUCO mu ient
3. msn ut maga femme cu dbi punt tulum neca bli ra HR 90 Type s auarus 91 a etd bu c ausi Eat ta Gti 92 Operating Or Bless ca Mop EG is tad aet Satins C CU E Cu p NR ERU 92 on Candidate o qe iie ee ten qa edi obitu pt 93 Opetating on candidate males sace qup are eda ge o arb Ro da e SERRE 93 Other ERR TR RUNE RM pce da 94 Analyzine the BASE ohio o qo de 95 Searching for coincidences in the 96 Coincidence oie ERN 97 Coherent follow up 98 n a gy N 98 Extract tbe band inan n 98 Create uniformly sampled gd corrected for source Doppler and spin down 100 Elininate bad pektodss ine ata ice Gerh tp cet de ebat io R eoi tiat 101 Eliminate bl CVC ES actu sie US ERRARE RN EU TREE 103 Create the Wiener filtered cla 104 aaa 104 Check the value wil AUS EE UOT 104 Some miscellanea Sna
4. 4 044 decre quorum RR Only when all the periods have been selected we should press the End selection menu option and the new gd is created together with a file with the selected periods e g 102 Operating on gb gd or array Producing gou gd or array tini tfin iini ifin 1748462 882278 1749764 369128 1738078 1739379 2270617 689045 2292445 264693 2260233 2282060 1060921 112908 1062486 537651 1050536 1052102 2448896 803872 2451522 136453 2438512 2441137 3653292 076651 3656007 448343 3642907 3645622 3667115 787082 3671312 270605 3656731 3660927 3721670 072888 3748823 789805 3711285 3738439 4080296 473596 4082828 398386 4069911 4072443 The file can be edited to cancel ot add petiods The new gd is cleaned and we should operate it for the following step There are two utilities that operate on the generated file gout re apply null gin file where gin is the original and file is the null period file produced by the preceding program gout is the nullified file gdcell extract null gin file where ginis the original gd and file 15 the null period file produced by the preceding program is a cell array containing the gds with the nullified periods A plot with the original data and the nullyfied data in read is produced Eliminate
5. This is the beginning of a list of 500 sources There are 3 frequency parameters 3 amplitude and polarization parameters the amplitude is normally in units of 10 2 is the parameter describing the contents of the linearly polarized power so O means circularly polarized radiation and is the angle of the linear polarization with the local celestial meridian 2 position parameters Note that O the frequency is at epoch 1 1 2000 0 00 UT while the frequency of candidates is referred to the beginning time of the data the position is in equatorial coordinates while the candidates position is in ecliptical coordinates the sin down first order parameter is Hz day as for candidates are the characteristics of the 500 sources 3 Defined as the difference between semimajor and semiminor axis of the normalized polarization ellipse 111 Frequency 1200 1000 800 400 200 50 100 150 200 250 300 350 400 450 50 sources Sources position declination 0 50 100 150 200 250 300 350 right ascension 112 Amplitude 100 150 200 250 300 350 40 450 500 sources 50 113 x10 First order spin down parameter sources files are created by an m file driver called crea fakesource file m There is also a function to read these files fcand sour pss readsourcefile infile infile output file interactive if not present fcand can
6. VIR hrec 20041203 004502 sds 03 Dec 2004 00 45 02 000000 h err 0 000000 1598 000000 s HOLE at 03 Dec 2004 01 48 12 000000 VIR hrec 20041203 021450 sds 03 Dec 2004 02 14 50 000000 h err 0 000000 VIR hrec 20041203 054310 sds 03 Dec 2004 05 43 10 000000 h 4kHz err 0 000000 50 000000 s HOLE at 03 Dec 2004 06 15 39 000000 VIR hrec 20041203 061629 sds 03 Dec 2004 06 16 29 000000 h 4kHz err 0 000000 13172 000000 s HOLE at 03 Dec 2004 07 31 19 000000 VIR hrec 20041203 111051 sds 03 Dec 2004 10 51 000000 h 4kHz err 0 000000 57 000000 s HOLE at 03 Dec 2004 11 25 32 000000 VIR hrec 20041203 112629 sds 03 Dec 2004 26 29 000000 h 4kHz err 0 000000 23392 000000 s gt HOLE at 03 Dec 2004 11 39 52 000000 VIR hrec 20041203 180944 sds 03 Dec 2004 18 09 44 000000 h err 0 000000 106 000000 s HOLE at 03 Dec 2004 18 22 40 000000 VIR hrec 20041203 182426 sds 03 Dec 2004 18 24 26 000000 h 4kHz err 0 000000 51 000000 s HOLE at 03 Dec 2004 21 02 00 000000 VIR hrec 20041203 210251 sds 03 Dec 2004 21 02 51 000000 h 4kHz err 0 000000 42 000000 s HOLE at 03 Dec 2004 21 03 06 000000 Summary 133 files start 3 571748 days 129 holes 03 Dec 2004 00 45 02 000 of total duration 141025 000000 s 000 end percentage duration duration duration duration duration duration duration duration duration
7. fft length frin initial frequency nfr number of frequency bins eS spectral resolution in normalized units t0 initial time mjd Defined as the difference between the semimajor and the semiminor axis of the normalized polarization ellipse 116 np _ number of periodograms thresh threshold typically 2 win window type 0 gt 1 gt pss o doptab Doppler table see Doppler effect page 123 nois level table containing the Doppler data depends on antenna and year a matrix n 4 or n 5 with first column containing the times normally every 10 min second col x in c units in equatorial cartesian frame third col y fourth col 2 pmstr np array with noise levels in standard deviations if the dimension is not exact the value 1 is given for all the periodograms simulation level no sid modulation direct frequency computation 1 sid modulation direct frequency computation res 1 no sid modulation fft frequency computation sid modulation fft frequency computation o tfspec if 1 time frequency spectrum otherwise only peak map default 0 Output peak map structure output peak map structure elements np number of periodograms frin initial frequency nfr number of frequency bins dt sampling time s Ifft fft length dfr frequency bin width reS spectral resolution in normalized units tO time mjd thresh
8. periods ou qan aa SUM gh S qma GU 31 EVO ES das edita periti ris 33 Filtenneina ds framework sesini apte DAR t SUIS EAM DAS 33 De iSc 33 ee LEURS NS 34 Ey nt periodicities quiste eo queda Que AUN 35 SEDB EP 37 D 37 PEIOCedufe is atate nb edit 38 Software PAP Sn AI a RE P EI enne node 39 39 PISS 39 Software Matlab environment 40 Time frequency data quality 41 rie 43 Peak map creation 43 Other peakmap 45 Rough clednt rm b er EP REP d qp 50 Brequenicy domain cleamttie procedures ioni E TW 51 Peakimapanal s Si 54 Er ener es Pestis din ry 57 Effectiveness of the procedure 59 Hough transform c
9. 06 Dec 2004 14 28 21 000000 3790 000000 5 chs 12500 000000 chs 1949 000000 chs 4490 000000 s chs 881 000000 chs 803 000000 chs 776 000000 chs 9454 000000 chs 15 000000 chs Tobs 0 456985 28 Data selection MatLab environment Basic sds operations If the sampled data are in the sds format it is easy to perform a variety of tasks Here we will speak of higher level tasks lower levels are discussed in the programming guides Among the others of particular interest for the PSS 8 sfc_ sfc_open file that outputs the sfc structure of the file chss ndatatot fsamp t0 t0gps sds_getchinfo file that shows the UTC time and outputs channels in a cell array the total number of data the sampling frequency and the initial time both in MJD modified julian date and gps g sds2gd selt file chn t creates a gd from file channel number chn and t initial time duration if the parameters are not present asks interactively 595 spmean frbands file chn fftlen nmax creates an 595 file named psmean sds containing the spectral means for many different bands frbands is an Nx2 matrix containing the bands if it is not present it can be input as a text file like for example 0 20 20 48 48 52 52 70 70 98 98 102 102 200 200 500 500 1000 1000 2000 file and chn the file and the channel number fftlen is the length of the FFT and nmax the maximum number of output data
10. 189 PUD int 189 fa Oe baia i Ei fa EA 196 dong adeo San La tO 197 ania aah 198 Gereral 198 199 eA AS 199 199 Matlab procedires eS espe 200 200 Preparation proceddfess 200 Check GS de PORE cllc deu d ere 200 Analysis EU OUR 201 C procedutes resume odio tu rasta di tard t a ED CE rex 202 Reference re eoe reip 203 Some Basic formulas dion dna 203 Proposed 205 papers Tto gal Ss dieta h etis 206 Introduction The PSS software 1s intended to process data of gravitational antennas to search for periodic soutces It is based on two programming environments MatLab and C The first is basically oriented to interactive work the second to batch or production work in particular on the Beowulf farms There are also programs developed in Matlab then compi
11. 85 Implementation of the Supervisor The steps outlined in the last paragraph can be translated in a number of services which form the SV program and which we have already introduced in the previous section a data prod Production of input data files starting from the SFDB b job sub Job submission workload management and retrieval of output files c data out Management of output files d farm mon Monitoring of PSS jobs and slave nodes e mast mon Monitoring of master node The mapping between the SV steps and services is the following SV activity Service node status monitoring farm mon master status monitoring mast_mon input data production data_prod job submission job_sub job status monitoring farm_mon output file copy data_out In the following figure a schematic view of the SV services and their relation with the cluster environment is given In the implementation of the services we sometimes rely on the PBS batch system This has been conceived for job distribution and workload management and gives to the user a set of commands also for job node queues monitoring In particular we use the Batch Interface Library IFL which provides a set of user callable functions with approximately a one to one correlation with client commands 86 Candidate database and coincidences The database FDF The candidates are stored in text files with many check informations The
12. 1 3 1 1340 1 3 2 1 1 3 3 10 mask crea cleaningmask peakfr nofr 10 0 1 99 1886 0 0 0111 0 The read_virgolines function reads lines file in Virgo format crea cleaningmask produces also figures to check the mask creation 51 0 200 401 600 800 1000 1200 1400 1600 1800 2000 Final peakfr b mask DUNT AI LU AMA ML 200 400 500 800 1000 1200 1400 1600 1800 2000 Hz 52 Final peakfr b mask r 496 497 498 49 500 501 502 Final peakfr b mask 7 990 1000 1010 1020 1030 1040 53 Peak map analysis In a old vbl file the first channel is the parameters as the detector velocity in AU day the second the frequency bins number the third the equalized amplitudes other channels can contain short spectra and other In the new vbl file the first channel is the parameters as the detector velocity normalized to the second the short spectrum the third the index for the peak vector the fourth the peak vector i e the frequency bins number the fifth the equalized amplitudes To take the data of the peak map the following function can be used x y z tim1 show peaks frband thr time file where frband min max frequency 0 all thr min max threshold mjd 0 all time min max time mjd 0 all file input file 2 peaks data time frequency amplitude It shows also the
13. 4 13 21 4 43E 21 2 77 20 1 73E 19 4 43 17 4 43 17 2 77 16 1 73 15 205 206 Papers and tutorials
14. cand candidate 7 n matrix typ type of show freq freq0 sky sd cr mh pars parameter structure depending on type out output result depending on type Other functions k psc kfiles lam bet which of the 242 files to take vector version lam bet ecliptic coordinates degrees candi psc sortcand cand sorts candidate arrays of different types Analyzing database 95 Searching for coincidences database There are some Matlab procedures to search for coincidences between two candidate sets C 1 cand2 psc coin dircand1 dircand2 freq searstr resfile seatch for candidates in a complete type 2 data base producing 2 candidate matrices dircandl dircand2 freq searstr resfile candidate first directory with last candidate second directory with last min max frequencies 0 all frl fr2 frn if n gt 2 multiple of 10 Hz start search structure result file type 2 psc db is in files of 10 Hz each group in 242 sky patches 1 cand2 psc coin single dircand1 dircand2 searstr resfil analogous to the preceding one but operating on a single file type 2 db psc coin 2fil file1 file2 searstr resdir similar to the preceding ones but operating on two single files C 1 cand2 psc coin list dircand listcand searstr resfile seatch coincidences between a list of candidates for example a list of known or fake
15. conj fft data norm where norm dt nfft if go2time 1 fclose TIME if write filefreq 1 fclose FREQUENCY fclose PEAKMAP FINAL fopen filepeak w final file for the peakmap PEAKMAP fopen appo dat r appo file for the peakmap Now write in peakmap the correct number of FFTs in the file nfft_tot fft_read numbers of read ffts fprintf FINAL Sd n nfft_tot se in ascii fwrite void amp nfft tot sizeof int 1 FINAL per Cristiano binario while c 2getc PEAKMAP EOF putc c FINAL 191 fclose fclose FINAL fclose AREST return errorcode fft2spectrum sqrt double header _ param gt tsamplu 2 0 header_param gt nsamples to remove the normalization noise spectral density which depends on the sampling time and on the number data on which the FFt is done Both these parameters will change printf fft2spectrum f n fft2spectrum if k 0 beg mjd REFERENCE MJD printf First file mjdtime gps time s and ns 15 10f sd n header param mjdtime header param gps sec header param gt 9 nsec total band gd gt dx gd gt n 2 the data are Real and Imag so there is a factor 2 printf Initial frequency of the data Bandwidth positive only total length sf f ld n gd gt ini total_band gd gt n if h
16. 69 8 007 9870777 lt 16 70 9 9866 01699600 7007 99999999107 V0 99 829888886996tv00 cccCCCCLVEO 70 9 764460288067 9 7007 8LLLLLLLLZO vvSTS 70 275881198 705556700 80200777 lt 6 V0 92LSI9196682cELTS00 68888888 T0 vv GTG VO 9L90LOLLSTE99 6 S00 SvvVVVV6900 vVVGTIGS v0 918cEvVOESCVZCV6OSSOO 00000000000 vVSIS 21 spuooes 34611 zsod sod xsod ae nbue3oeW 124 Notes and observation 1 this code 1s based on the PSS astro library whose documentation is given in the PSS astro UG doc and PSS astro PG doc 2 it uses the JPL ephemeris file DE405 and files from IERS for fine corrections of the time and nutation These fine corrections use of UT1 and not UTC application of corrections to the nutation angles DPsi and DEpsilon are provided by IERS only up to date with the file eopc04 62 now In case these files are not present for the time lag ot part of it provided by the user then the code does not use these fine corrections and a message appears the screen and also in the file which is created the deinstein corrections is at a level less than one part on one million 4 Asshown with details in the documentation the Einstein effect 1s only a small correction to the final Doppler effect but given the fact that it can be evaluated using only information on
17. This line lists index A unique number given to each element of the table Index numbers are sorted numerically The index number is printed next to every function name so it is easier to look up where the function in the table time This is the percentage of the time that was spent in this function and its children Note that due to different viewpoints functions excluded by options etc these numbers will NOT add up to 100 self This is the total amount of time spent in this function children This is the total amount of time propagated into this function by its children called This is the number of times the function was called If the function called itself recursively the number only includes non recursive calls and is followed by a and the number of recursive calls name The name of the current function The index number is printed after it the function is a member of a cycle the cycle number is printed between th function s name and the index number For the function s parents the fields have the following meanings self This is the amount of time that was propagated directly from the function into this parent children This is the amount of time that was propagated from the function s children into this parent called This is the number of times this parent called the function the total number of times the function 74 was called Recursive calls to the function ar
18. continuing Task Program file Mb day in out single jobs candidate create db cand 4 min 750 collection reshape db2 psc type 2 db using 242 x 50 1050 Hz psc reshape db2 T hour n or reshape db3 coincidences psc coin cand cand 13 Programming environments Matlab For the MatLab environment the Snag toolbox is used It contains more than 800 m functions February 2004 and has PSS as one of the projects regarding the gravitational waves in snag projects gwpss It is almost completely independent from other toolboxes There are two useful interactive gui programs in Snag snag provides GUI access to Snag functionalities It can be used stand alone or in conjunction with the normal Matlab prompt use of Snag At the Matlab command prompt type snag A window appears gd s window al gt type 1 dim 1000 0 dx 1 000000 gt vector 21 2 a2 type 1 dim 1000 ini dx 1 000000 gt a1 a1 3 c1 gt type 1 dim 32768 0 dx 0 000100 gt chirp c2 type 1 dim 32768 ini 0 dx 0 000100 gt gauss 5 c3 type 1 dim 32768 0 d 0 000100 gt 1 ccl gt type 1 dim 32768 ini 0 dx 0 000100 gt worm ofc1 gl type 1 dim 1000 0 dx 1 000000 gt sin g2 type 1 dim 1000 ini 0 dx 1 000000 gt gauss 1 99 processing It has text window where ar
19. gives the sidereal response in energy the parameters O antenna an antenna structure source a source structure the number of points in sidereal day 128 Other analysis tools Sensitivity evaluation h_eff sens dist med effect_bg chn x y from a short spectra file normally in sfdb run folders of pss computes the effective h background noise the pss sensitivity the horizon and the mean for each spectrum in 3 bands Here ate the plots for C7 effective h background noise T a PPP T quier rrr mother rho 2 4 9 4 2 2 amm Me cogs 21111 SMO REE 2 42 44 444 4 4 084848468 0 200 400 600 800 1000 1200 1400 1600 1800 2000 129 pss sensitivity 4 4 2 0 4 144 nor dfEeRRI M RI M eA M IM Swe oo de age gies sive cic eft wc es b See 21 0 200 400 600 800 1000 1200 1400 1600 1800 2000 1
20. in pss hough drive c comment the call to houghBuild and uncomment the call to houghRad and houghBuildAdap compile the code with make MakefileAdaptive In the future we could decide to use only float type for Hough maps so that no data types difference will exist between standard and adaptive computation However tests have shown that a strong loss of performance happens on machines with a second level cache L2 less than if float data type is used This means at least 1MB L2 cache is in any case needed for the adaptive Hough transform to be fast enough The program assumes that the detector velocity vector read from the input file 15 in ecliptical coordinates however the possibility it is expressed in equatorial coordinates is also foreseen but requires to comment few lines and uncomment few other lines in the function readPeaksinpss hough drive c and obviously re compile the program The program refers frequencies to the beginning of the observation period However it is foreseen the possibility to refer frequencies to an epoch in the past e g to Jan 1 2000 at 00 00 00 this requires to uncomment a few lines in the function houghBuild in pss hough c or function houghBuildAdapinpss hough adaptive c and re compiling the program In the present implementation for each search frequency the LUT 15 re computed and for each spin down value a check is done for every time if the LUT has to be re computed The check consis
21. n gd time dx puts ATT Time data do not contain the normalization from fft 2 to Spectrum because this depends on puts the sampling time which is known and the number of data on which YOU will then to the FFT which is unknown puts zz fft data conj fft data norm where norm dt nfft fclose OUTH1 if go2time 1 fclose TIME if write filefreq 1 fclose FREQUENCY fclose FINAL fopen filepeak w final file for the peakmap PEAKMAP fopen appo dat r appo file for the peakmap Now write in peakmap the correct number of FFTs in the file nfft tot fft read last fft which has been read fprintf FINAL dWMn nfft tot se in ascii fwrite void amp nfft tot sizeof int 1 FINAL per Cristiano in binario while c 2getc PEAKMAP EOF putc c FINAL fclose PEAKMAP fclose FINAL fclose AREST return errorcode 195 08 It is translated to vbl format p082vbl with the pss batch diary with the script cd y pss virgo pm VSR1 v2 conv 08 conc list txt is the format p08 format Header nfft int32 sfr double sampling frequency 1 int32 original length of fft divided by 2 inifr double oo Block header mjd double npeak int32 velx double AU day vely double AU day velz double AU day oo Short spec
22. put a big number and all the concatenated files will be analyzed m sds2mp file t creates an mp multi plot structure from an sds file the command can be issued without parameter and asks interactively For example it can be applied to the spectral mean sds file created by sds spmean The mp structure can be showed by plot m 3 m is the mp and 3 means log y obtaining on 4 data of the 29 10 channel 1 min max freq 0 000000 20 000000 channel 2 min max freq 20 000000 48 000000 channel 3 min max freq 48 000000 52 000000 channel 4 min max freq 52 000000 70 000000 channel 5 min max freq 70 000000 98 000000 channel 6 min max freq 98 000000 102 000000 channel 7 min max freq 102 000000 200 000000 channel 8 min max freq 200 000000 500 000000 channel 9 min max freq 500 000000 1000 000000 channel 10 min max freq 1000 00000 0 2000 000000 10 10 20 30 40 50 else among other choices 20 20 60 70 80 90 10 1 10 1 channel 1 min max freq 0 000000 20 000000 channel 2 min max freq 20 000000 48 0000 10 nyt NE 19 ul I E 10 i 10 40 0 20 40 60 80 100 10 20 40 60 80 100 E l l channel min
23. strcpy caption c6 strcpy outdestbase gsiftp virgo se romal infn it storage virgo gwave4t cristiano hough cand strcpy outdestbase2 gsiftp gridse pi infn it flatfiles SE00 virgo OUTPUT strcpy outdestbase3 gsiftp se01 lcg cr cnaf infn it storage gpfs virgo3 scratch PSS hough cand virgo strcpy outdestdir local strcpy detector virgo strcpy fileinput home Palomba InputData strcpy lfnpath local 70 strcpy 1 strcpy command file nocommandfile The parameter values are set through the following options d detector name gt f lt input file gt x lt input file logical file name with path gt e lt duration s gt s lt spin down age yr gt t lt threshold for candidate selection units of sigma gt p lt starting search frequency Hz gt b lt name of the candidate database gt r maximum allowed frequency shift n of bins for re computing LUT gt c lt caption gt o lt output destination gt local retrieve locally W lt command file gt Usage example pss_hough f file doppler band 0 x grid virgo vsr1 file doppler band 0 c vsrl o local e 1048 576 p 49 735069 s 4000 000000 t 3 800000 A command line help is obtained with pss hough help Performance issue Performance is an important issue for the PSS hough library The computation of the Hough transfor
24. FrBand Hz 19 799390 23 800610 Fraction done Sky grid position done n lambda beta deg 0 001418 1 000000 28 428098 Average Median and std of the hough map 187 358149 197 528903 43 597255 Number of candidates per Sky patch 49 000000 FIRST STEP candidates frequency Hz lambda deg beta deg spin down Hz s Candidate Amplitude CR 89 59 XN 5 O 20 49836424 0 0000 28 4281 7 62699775 12 395 162 4 766 21 80302732 0 0000 28 4281 6 10159820 11 384 591 4 524 22 73776853 0 0000 28 4281 0 00000000e 00 377 203 4 355 22 10618894 0 0000 28 4281 3 81349887e 11 375 209 4 309 RESULTS around the found candidates for one Sky patch Refinement factor in Spin down In the Sky Grid 5 2 around the original position lambda and beta deg 09 000000 28 428098 Average Median and std of the hough map 187 358149 197 528903 43 597255 Spin down step Hz s Ncandidates per patch 1 525400e 12 49 000000 frequency Hz lambda deg beta deg frequency Hz spin down Hz s Cand Amplitude 001 DDb 89 59 5 59 5 69 20 49844969 0 0000 29 5139 1 22031964 11 447 301 5 962 0 786 1 388 21 80296629 0 0000 29 5139 5 33889842 11 205 011 0 405 0 786 1 388 22 73778074 359 6078 28 1208 6 10159820e 12 381 483 4 453 0 786 1 388 22 10605467 0 3922 28 1208 3 05079910e 11 302 401 2 639 0 786 1 388 20 49771726 359 6078 28 1208 3 96603883 11 208 130 0 476 0 786 1 388 21 68985594 359 6078 28 1208 5 03381851 11 230 063 0 980 0 7
25. OUTH1 188 if errorcode 1 printf Error in reading real data into file d return errorcode j errorcode fread void amp ipw sizeof float 1 OUTH1 if errorcode 1 printf Error in reading imag data into SFT file Wn jii return errorcode freg float header param firstfregindex kk 1 0 header gt sden rpw rpw ipw ipw norm kk Note header_param gt normd is the normalization from FFT to spectra header_param gt normw is the normalization for the window header_param gt tbase is the duration in seconds of 1 FFT Peak maps p05 This is a non standard format for the peak maps It is translated in standard format with piapeak2vbl These files contain 3 arrays the bins the equalized values and the mean value of the spectrum The vbl files only the bins and values Program to produce the peak map from the SFDB files Author Pia X x X KKK KKK KK KK ke ke ke ke e e e A KK Ttast version 2 9 February 2005 xxx e ce e e ce e x include lt stdio h gt include lt math h gt include lt string h gt include lt stdlib h gt PSS libraries x 189 include pss PSS lib pss math h include pss PSS lib pss serv h include pss PSS lib pss snag h include pss PSS lib pss sfc h k Antenna libraries x includ
26. VIR_hrec_20070728_175136_300 SFDBO9 O pss virgo sfdb VSR1 v2 deca_20070727 VIR_hrec_20070729_105745_400 SFDBO9 O pss virgo sfdb VSR1 v2 deca_20070727 VIR_hrec_20070730_045326_500 SFDBO9 O pss virgo sfdb VSR1 v2 deca_20070727 VIR_hrec_20070730_193526_600 SFDBO9 O pss virgo sfdb VSR1 v2 deca_20070727 VIR_hrec_20070731_085006_700 SFDBO9 O pss virgo sfdb VSR1 v2 deca_20070727 VIR_hrec_20070801_042216_800 SFDBO9 O pss virgo sfdb VSR1 v2 deca_20070727 VIR_hrec_20070803_133416_900 SFDBO9 O pss virgo sfdb VSR1 v2 deca_20070727 VIR_hrec_20070804_125026_1000 SFDB09 O pss virgo sfdb VSR1 v2 deca_20070727 VIR_hrec_20070805_005026_1100 SFDB09 O pss virgo sfdb VSR1 v2 deca_20070727 VIR_hrec_20070805_163616_1200 SFDB09 O pss virgo sfdb VSR1 v2 deca_20070806 VIR_hrec_20070806_013606_1 SFDBO9 O pss virgo sfdb VSR1 v2 deca_20070806 VIR_hrec_20070806_133606_100 SFDBO9 O pss virgo sfdb VSR1 v2 deca_20070806 VIR_hrec_20070807_014426_200 SFDBO9 O pss virgo sfdb VSR1 v2 deca_20070806 VIR_hrec_20070807_194426_300 SFDBO9 O pss virgo sfdb VSR1 v2 deca_20070806 VIR_hrec_20070808_130726_400 SFDBO9 It creates an sbl file with an user header read by sfdb09_us read_sfdb_userheader sbl_ containing the information of the sfdb09 header Note in sbl_ dt there is the normal time between 2 ffts The program is rapid 2 months of data are processed in about a quarter of an hour There is also the possibility to do different simulations in some cases much slower F
27. at 2005 10 14 5 22 07 3 virgo sfdb gt sd p sds cb ETOS c4 no5 0Hz 1 1 frame pm C metadata cand 15 sim 25 sd 2 metadata 25 software windows r3 source Snag linux CJ DopplerD ata 1 analysis The h reconstructed database Third level is frame 87 sds Fourth level can be runs ot years The naming is derived as easily as possible by the original antenna name 175 sfdb database The normalized spectra database The peak map database The candidate database See Candidate database and coincidences page 87 The coincidence database See Candidate database and coincidences page 87 Database Metadata Server docs It should contain information on the available servers how to teach them what is supposed to contain Analysis docs This should contain the analysis batch log and where the results are stored 176 Antenna docs It should contain at least Antenna basic information e g the position Channel names e Basic information on the runs File System utilities 177 Organization of the workflow Basic analysis periods short runs up to about fifteen days the analysis is done for each full run This applies to the sds sfdb and peakmap creation The subsequent analysis Hough map candidates coincidences could be independent of this basic divis
28. basic periodic simulation function is sim sds sim str fdb str doptab that creates one or more sds data files It is based on two structures Sim Str simulation structure with elements 7 type stationary 1 non stationary a nss non stationarity structure if type 1 ant antenna structure only for for signal simulation sour source structure only for for signal simulation 7 Ifft fft length for simulation t0 initial time mjd dt sampling time s str a file database structure with elements folder database folder head filename header p es hrec tail filename tail p es crab ndat total number of data fndat number of data o doptab the Doppler table see Doppler effect page 123 fileout zrealsim sds sds in chn sim str fdb str doptab that adds simulated periodic source signal to a real data file The parameters are o Sds the input sds file to be used o chn the channel number o Sim Str a simulation structure with elements 108 ant antenna structure only for for signal simulation sout source structure only for for signal simulation Ichunk chunk length for simulation fdb str a file database structure with elements folder database folder o doptab the Doppler table see Doppler effect page 123 These functions to simulate the periodic source signals use d source data ps chunk source antenna data do
29. beg mjd header param gt mjdtime header param gt tsamplu header param nsamples day2sec new start time to evaluate the beginning phase if iband extract 1l printf initial freq band dx Sf n initial freq gd gt if iband extract 1 gd band ini double initial freq if iband extract 1 i band band extract gd gd band band Creation of the peak map if ipeak extract 1 peakmap from band header param if ipeak extract 1 peakmap from ar gd header param peakmap gd gd short header param end of the creation of the peak map Write if required the frequency domain file if write filefreq 1 if iband extract 1 kkf 0 for iif 0 iif gd gt 11 2 fprintf FREQUENCY f n header param mjdtime kkf gd band ini tkkf gd band dx gd band y iif gd band gt 116 94 band y iif 1 gd band y iif 1 if iband 1 kkf 0 for iif 0 iif gd n iif t 2 fprintf FREQUENCY f n header param gt mjdtime gd gt init kkf gd gt dx gd gt y iif gd gt y iif t gd gt y iif 1 gd gt 11 1 if go2time 1 amp amp k 0 strcpy capt gd h reconstructed if iband extract 1 if go2time 1 amp amp k 0 gd time crea gd gd band n 0 gd band dx capt gd if iband 1 if go2time
30. big events This is accomplished by the program gout gsigmf ada clip c gin den lenstat cr typ whete gin input gd or array den density of log bins bins per unit interval typically 10 lenstat length of stationarity number of samples cr critical ratio for clipping typ type of clipping 0 gt clipped value default 103 1 gt 2 gt interpolate NOT YET IMPLEMENTED 3 eliminate data YET IMPLEMENTED The output is the clipped gd and the estimated varying time standard deviation This functions operates on complex gds Create the Wiener filtered data The data produced are now cleaned but their variance changes with time and so the sensitivity of the antenna in that frequency band In order to optimize the detection we should use a Wiener filter that gives a gain proportional to the quadratic SNR This however changes the spectral shape of the signal so the wotk of the spectral filter is affected So as a first step we use a filter that depends only on the noise supposing constant the signal This is accomplished by gw nois ps_wiener g thr source g gd with time series data as obtained by pss band recos thr cr threshold typically 4 5 source source if present if source 1 gt whitening filtered gd 18 standard deviation of the noise It is also possible to create whitened data On the created gw gd we perform athe power spect
31. by the function psc reshape 1 The problem of the 1 database in case of big databases 15 the long coincidence time performed by the psc coin 1 function Type 2 database 90 Better results can be achieved with the 2 database In this case a single sorted file is created for each 10 Hz group named e g pss 1390 cand or pss 0100 cand Then from each of these 24 10 2 242 files The number 1 240 are 15 15 degrees patches in the sky the 241 is the 15 degrees radius south pole zone and the last the 15 degrees radius north pole These new files are created by psc reshape db2 So the data for each 10 Hz band are both in a single file possibly but not necessarily frequency sorted and in the 242 sky patch files allowing best results for each type search Regarding the sky patch files note that now there is an asymmetry between the two coincidence groups and the second coincidence group should be a little larger in area in order to perform coincidence with neighbors also at the edge of the patches Type 3 database Another database structure is the type 3 Also in this case a single sorted file is created for each 10 Hz group named e g pss cand 1390 cand or pss cand 0100 cand Then from this a file for each spin down value is created e g pss cand 0230 sd0002 cand or pss cand 0240 sd 022 cand for negative spin down These new files ate created by psc reshape db3 So
32. creates a channel structure for events nch is the number of channels evch crea_ev n chstr tobs creates an event channel structure simulating n events in the time span tobs and with the channel structure chstr evch crea_evch chstr evstr creates an event channel structure from a channel structute and an event structure fil dir fid Jsave ech ch direv fil mode capt save a channel structure in an ascii file that can be edited ch is the channel structure direv and fil are the default folder and file mode 15 0 for standard 1 for the full evch capt is the caption fil dir save ev ev direv fil mode fid capt save an event structure in ascii file that can be edited ev is the event structure direv and fil are the default 33 folder and file mode is 0 for standard 1 for the full evch fid is the file identifier or 0 capt is the caption save evch evch saves an evch structure in Matlab format load evch interactively loads an evch structure in Matlab format eo sort_ev ei time sorts an event structure syy out ev_sel in ch t a l selects events on the basis of the channel time occurrence amplitude and length out are the input and output evch structure ch ifitis an array it is the probability selection of different channels if 0 the channel disappear otherwise is not used O ta l ifitisan array of length 2 defines the interval of acceptance if the first element 15 grea
33. date and viceversa example mjd s2mjd 25 Apr 2004 18 44 11 produces mjd 53120 7806828704 and str mjd2s mjd produces 25 Apr 2004 18 44 11 000004 str mjd2s mjd converts a modified julian date to string time example v mjd2v mjd and t v2mjd v converts a modified julian date to vectorial time month day hour minute second and viceversa tgps mjd2gps mjd and mjd gps2mjd tgps converts mjd to gps time and viceversa tai mjd2tai mjd and mjd tai2mjd tai converts mjd to tai and viceversa tdt tai2tdt tai conversion from TAI to Terrestrial Dynamical Time tsid sid tim mjd long sidereal time in hours o mjd modified julian date days o long longitude positive if west of Grenwich degrees 120 Astronomical coordinates ao do astro coord cin cout ai di astronomical coordinate conversion from cin to cout Angles and tsid in radiants Local tsid and latitude is needed for convetsions to and from the horizon cootdinates O cinand cout can be horizontal azimuth altitude equatorial celestial equatorial right ascension declination ecliptical ecliptical longitude latitude galactic galactic longitude latitude epsilon 23 4392911 deg is the aberration to right asc and declination if they are referred to the standard equinox of year 2000 epsilon 23 4457889 deg if they are referred to the standard equinox of year 1950 121 Source and Antenna str
34. double 141 148 FFT length fftlen inta 149 156 initial frequency bin of the file inifr int8 157 164 delta lambda float 165 168 delta beta dbet float 169 172 delta sd Hz day dsd1 float 173 176 delta CR dcr float 177 180 delta mean Hough map mh dmh float 181 184 delta h dh float 185 188 number of spin down nsd1 int4 189 192 89 epoch 1 epoch 2000 0 epoch initim epoch int4 193 196 free now 15 int4 197 256 Any candidate frequency at epoch of initial time or epoch uint4 1 4 2000 depending on epoch in units of microHz lambda index uint2 5 6 beta index from 90 uint2 7 8 591 2 9 10 CR index uint2 11 12 mh index uint2 13 14 h index uint2 15 16 Note that the single candidate have 7 parameters and 8 16 bit words this is the why that we use in Snag raw candidate 8 N uint2 vectors used only for rapid disc access and fine candidate 7 N double matrices with the seven parameters for each candidate lambda and beta index start from 0 value If the data base contain 10 9 candidates each file should contain about 500 000 candidates So the mean value for the dimension of the file is 500000 16 256 8 MB and the total dimension of the data base should be about 16 GB To perfotm the coincidence analysis the files are supposed to be sorted This can be accomplished
35. max freq 48 000000 52 000000 channel 4 min max freq 52 000000 70 0000 10 n 10 10 10 10 0 20 40 60 80 100 10 20 40 60 80 100 1 1 channel 5 min max freq 70 000000 98 000000 channel 6 min max freq 98 000000 102 000 10 10 1 NUN RN M 102 10 40 0 20 40 60 80 100 40 20 40 60 80 100 1 1 channel 7 min max freq 102 000000 200 00000 channel 8 min max freq 200 000000 500 00C 0 10 Yit LE N it 10 4 it 102 102 i 10 0 20 40 60 80 100 10 20 40 60 80 100 d 1 1 channel 9 min max freq 500 000000 1000 0000 channel 10 min max freq 1000 000000 2000 10 N 107 m n 10 10 i 0 20 40 60 80 100 0 20 40 60 80 100 The abscissa is in houts from the 0 hours of the first day crea ps sdsfile chn lfft red creates an sbl file containing power spectra of data similar to that produced for the sfdb from channel number chn a big FFT length and a length of the power spectra Ifft red 30 Choice of periods MatLab environment The choice of the periods on which the SFDB should be created and then are to be analyzed can be done by the use of the Virgo data quality information and of the basic instruments like those shown in the preceding section In Snag there ate some useful interactive functions that helps in choosing prtiods 5 1 absc typ y x file the easyest one w
36. plot of the points 54 50 14 501213 5014 50 08 4 43 50 06 Hz 50 04 50 02 bin EY 4958 49 96 A tim vel vbl peakmap frband thr time file is similar to show peaks but produces a gd2 containing the time frequency data frband min max frequency 0 all thr min max threshold mjd 0 all time min max time mjd 0 all file input file A output a gd2 sparse with velocities tim times vel velocities vbl_ file header It can be used for the Hough map snag procedures crfr crtim peakfr peaktim npeak splr peaklr mub sigb sphis spfr spti m zana peakmap frband thr time res file analyzes a vbl file peakmap frband min max frequency 0 all thr min max threshold 0 all time min max time mjd 0 all 55 5 resolution reduction in frequency in low res maps 0 low res file input file crfr crtim mean cr vs freq and time spfr sptim mean spectrum vs freq and time peakfr oeaktim number of peaks vs freq and time splr peaklr t f peak maxima and number low res sphis peak CR histogram mub sigm mean std of peaks in a spectrum npeak total number of peaks 56 Time frequency cleaning The idea 15 to histogram the peak map in a lower resolution 2D grid for example 12 h starting from 6 for time and 0 01 Hz for frequency then define the maximum allowable value of this histogram for the normal data e g 80 and finally cut away the peaks in n
37. pudo In the following picture the relation between the main steps in the data analysis procedure and the Supervisor services is outlined In particular monitoring services i e farm and mast mon not tied to a specific point of the analysis chain but act on the whole space delimited by the box 83 84 Outline of the supervisor We call master the node where the active SV is running it is the machine from where most of the management is done we call slave or Computing Node CN a machine where calculations are done the master will be also a CN We call Storage Element SE the machine where data both input and output ate stored We assume the cluster is on a private network with the masters both primary and secondaty as the only interface to the LAN Secondary masters CN which can replace the master if this fails Moreover some of the CN can be also part of a grid farm We distinguish two types of jobs PSS_jobs which are directly managed by the SV and N Jobs which run on the CN In general we can think a PSS job as made of several N job plus some information Now we list the main steps in the SV activity 1 m supervisor start SV starts on the master master monitoring start SV starts the master monitoring service on secondary masters secondary masters are ordered according to the rank active master has rank 0 the second one has rank 1 and so on node statu
38. relevant information contain e ecl long A ecl lat frequency v spin down d amplitude CR global Uncertainty on Uncertainty Starting from these files practical database is created The parameters archived in the files are for example FrBand Hz 19 800000 23 800000 VETO Average Median and std of the hough map 187 752249 197 579648 46 121870 VETO Threshold thresC n sigma 372 239727 VETO Sel of peaks to be vetoed with 0 s d CR Average CR std 16 897884 10 074178 VETO frequency spin down Candidate Amplitude CR 19 999951 7 626998 12 2175 503186 43 097796 20 556470 7 626998e 12 1822 882285 35 452380 20 999951 7 626998 12 874 638571 14 892855 21 337476 7 626998e 12 421 780901 5 074136 21 999951 7 626998e 12 834 369456 14 019753 22 999951 7 626998 12 1104 227264 19 870726 6 first file of the decade storage gpfs virgo4 virgo4 RomePSS virgo pm VSR2 v3 256Hz hp20 CLEAN peak 20090707 256Hz 1 cl vbl original sampling time s 3 906250e 03 lenght of the FFTs 2097152 Peakmaps frequency step Hz 1 220703e 04 Beginning mjd time of the analysis 55019 680150 Used mjd epoch 55111 702702 Subband analyzed Candidate Amplitude 1 7807138 Coarse spin down min max step Hz s 6 864298e 11 7 626998e 12 7 626998 12 Frequency over resolution factor 10 89 89 59 59 59 5 58 8 59 59 59 5 59 87 Total FFTs number 3616
39. that the value 0 is an escape character used if more than 27 2 zeroes should be represented in such case the datum is put in a side array of uint32 In practice there are 5 different cases sparse non binary the 0 runs and the non zero elements sparse binary only the 0 runs of the sequence sparse derived binary only the 0 runs of the derived sequence non spatse non binary normal vector a float per element non spatse binary one bit per element 22 PSS SFC files The PSS Periodic Source Search project uses many different types of data to be stored Namely h reconstructed sampled data raw and purged Short FFT data bases e Peak maps Hough maps PScandidates Events Each of these has a peculiar type of SFC h reconstructed sampled data raw and purged This type of data are normally stored with simple SDS Short FFT data bases The data are stoted in a SBL file In the user field there are other information like length number of samples of the time series e I Interlacing size number of interlaced samples D sampling time of the time series S window used on the time series The blocks contain one half of the FFT of purged sampled data one short power spectrum one one minute mean vector a set of parameters as O O O O e I number of added zeros for errors holes or over resolution time stamp of the first time datum mjd e t
40. the PSS software see the sub section pulse regarding pulse detection gw sim with data simulation radpat for the radiation pattern and response of antennas sidereal response of an antenna sky coverage The pss folder It contains about 200 functions Feb 2007 divided 1n cohe for coherent analysis hough for Hough transform hough exp containing the Hough Explorer functions other service routines mainly for peculiar file access procedures batch procedures PSC periodic source candidate functions for analysis and coincidence sfdb short FFT database and peak map 16 sim simulation theory statistical and physical theoretical analysis 17 environment contains a library and some module The library contains pss routines to operate with the snag objects GD DM DS RING pss_math basic mathematical routines pss serv service routine among the others vector utilities string utilities bit utilities interactive utilities simple file management pss gw physical parameters management pss astro astronomical routines pss frame routines for frame format access pss 187 routines for 187 format access pss sfc routines for the sfc file formats management pss snf routines for snf format management partially obsolete The othet modules ate pss bench for comp
41. the data for each 10 Hz band are both in a single file possibly but not necessarily frequency sorted and in the n different spin down files allowing best results for each type of seatch For coincidence purpose all the database files are supposed to be sorted in frequency 91 Basic functions The candidates ate stored in vectors of dimension 8 N as they are read from files or matrices of dimensions 7 N in physical units remember that the frequency uses 2 variables Normally the matrix is wrapped in a structure with other information like the epoch Operating on files psc wheader fidopss cand head writes a psc file header candstr psc rheader fid reads the header of candidate file producing a structure A nread eofstat psc_readcand fid N reads the candidates of candidate file fid file identifier N desited number of candidates A candidate vector nread number of obtained candidates eofstat 1 end of file cand psc_type n1 n2 notype file produce a candidate matrix and types candidates from file n1 n2 min max sequence number notype 70 type file candidate file head check psc file checks the consistency of a pss candidate file psc reshape 1 dirin pss candidate files post processing type 1 obsolete psc reshape db2 op dirin dirout infr1 infr2 type 2 pss candidate files post processing op operation 0 sort input files 92 1 gt to 242 files first coincid
42. the detector velocity divided by c The eighth column contains the Einstein effect gravitational redshift so f 1 p v deinstein The Doppler table is produced by the function doppler table fileesubsamp fileout with 125 file created by crea table subsamp subsampling one every o fileout output file can be not present The following function puts the Doppler table the x y and z component of the velocity of the detector in single precision in an dds file or sds in 4 byte precision doptab2dds doptab sdsfile capt or doptab2sds with doptab the Doppler table sdsfile the sds file capt the caption of the sds file These sds files that are tiny can be read to produce the matrix doptab the procedure to do it is doptab doptab_from_dds subsamp tim file or doptab_from_sds with subsamp the subsampling factor tm min max time mjd the 44 file like virgo doptab dds of 60 Normally the Doppler tables are computed and stored for long periods e g 2000 2020 but for practical purposes typically for spline interpolation it is more practical and fast to use shorter tables This is accomplished by extracting a sub table from a big table by subdoptab reduce doptab doptab tmin tmax were doptab is the original table tmin and tmax the time interval of the sub table The following function is a simple use of dopta
43. 0000 0 000000 psi 0 000000 0 000000 alpha 0 000000 360 000000 delta 90 000000 90 000000 2 400 sources fr 50 000 1050 000 0 000 0 001437 542 0 000 0 000 amp 0 000100 0 500000 eps 0 000000 0 000000 psi 0 000000 0 000000 alpha 0 000000 360 000000 delta 90 000000 90 000000 5 1 542 amp eps psi alpha delta 1 50 2010 1 3501e 003 0 0000 000 1991e 003 0 0000 0 0 6 78 6 76 2 50 6941 0 0000 000 0 0000 000 4009 004 0 0000 0 0 200 89 62 12 3 51 1881 3 8150 004 0 0000 000 4438 002 0 0000 0 0 194 19 14 11 4 51 3515 0 0000 000 0 0000 000 2793 004 0 0000 0 0 128 52 38 01 5 52 9395 2 3869 004 0 0000 000 4 6262 001 0 0000 0 0 10 29 32 71 6 53 0087 0 0000 000 0 0000 000 3098 003 0 0000 0 0 212 45 34 62 7 53 6771 1 3127 003 0 0000 000 8 2358 002 0 0000 0 0 273 64 51 68 8 56 0266 7 3225 004 0 0000 000 2 1763 004 0 0000 0 0 241 99 28 50 9 56 5089 3 4526 004 0 0000 000 3 9523 001 0 0000 0 0 316 17 37 88 10 57 2503 1 4212 003 0 0000 000 9581 004 0 0000 0 0 281 76 61 42 11 60 3573 1 1635 003 0 0000 000 1067 002 0 0000 0 0 31 75 3 80 2 Defined as the difference between semimajor and the semiminor axis of the normalized polarization ellipse 110 12 63 6015 4 9812 004 0 0000 000 3 2904 001 0 0000 13 64 0545 1 3731 003 0 0000 000 3 7991 004 0 0000 127 37 3 55 235473 16 15
44. 1 amp amp k 0 gd time crea gd gd n 0 gd dx capt if go2time 1 puts from gd or gd to gd time and call gdfreq2time If subsampled data have to be divided by isubsampling if 1 printf band n gd time n ld 514 n gd_band gt n gd_time gt isubsampling gd n gd gt for ii 0 ii gd gt time y ii gd band y ii isubsampling gd time dx gd band dx the function gdfreq2time then 193 evaluates the proper dx if iband 1 printf gd n gd time n 5194 ld n gd gt n gd_time gt n for 11 0 1 lt gt 1 gt 11 gt gd time dx gd dx the function gdfreq2time then evaluates the proper dx iii gdfreq2time 9 time header param ivign if k 0 len gd_time gt n len4 len 4 len2 len 2 inter 0 if header param gt typ 2 inter 1 data overlapped by the half Remove window if needed if header param gt wink gt 0 wingd2gd gd_time header param gt wink printf FFT number gd_time gt dx header param gt typ sf n k gd_time gt dx header param gt typ if k 0 write some information on the file at the beginning fprintf TIME s 59 s 59 s n Beginning mjd days Gps Gps ns Sampling time s fprintf TIME s 58 58 s s s n S Beginning fr
45. 1544 01388889 51544 02083333 51544 02777778 51544 03472222 0 000100557372253 0 000100537101901 0 000100514848156 0 000100490649671 0 000100464548820 0 000100436591616 0 000016371087845 0 000016427927612 0 000016483926814 0 000016538999573 0 000016593061781 0 000016646031264 0 000006927597163 0 000006932417453 0 000006937237591 0 000006942057575 0 000006946877404 0 000006951697077 0 000000000335206537 0 000000000335208710 0 000000000335210877 0 000000000335213040 0 000000000335215197 0 000000000335217350 123 1 December 2009 since New format dat virgo table20 the following 1 5 example of the tables inning beg 018 225 00000000070 00 91 004 996 09 08 0894204 60000000007 20t t99LCGG9GT9v608118 000000000 00 9 lt 8 4 0762066 lt 78 1 1004660000000007 00 974969072620897028 8926128 000000000 0 9211999 88 OZILICSEE000000000 COt9LEIVOVEEETVGISS8CB L96v1c8E 000000000 00 90 7074 0969789268 60821204 6000000000 cOt9LVIOOECL901S89 8 9v901cS5 000000000 LLV80cCSE 000000000 00 924071 8 lt 614 16 8 7069025 6000000000 020 9211940982826 lt 6 8 0000000000 0000 OV 90 2 96 0691 1891619 1 666 90 960069920712698470 679 00 9 0149028 289727 277
46. 2 50 37 o apply frequency domain anti aliasing filter and sub sample if 20 kHz data high events identification and removal create a stream with low and high frequency attenuation identify the events on this stream starting time and length with the adaptive procedure After this this stream is no more used smooth removal of the events in the original stream purged stream estimate the power of the purged stream every minute or so creating the PTS power time series for each of the 4 sub bands of the SFDB not for the first band apply anti aliasing and subsample create the windowed interlaced FFTs both simple and double resolution the simple resolution are archived for the following steps the double is used for the peak map create a low resolution spectrum estimate VSPS Very Short Power Spectrum e g length 16384 38 Software C environment Routines pss sfdb 39 Software MatLab environment There is also a software to create a SFDB using Matlab This be used for checking and particular purposes crea sfdb sdsfile chn Ifft red can be used also interactively without arguments sdsfile is the first sds file to be processed chn is the channel number is the length of the non reduced ffts red is the reduction factor for requested band normally 1 4 16 64 A file sfdb sbl is created with the fft collection There are al
47. 4 4 28 8 2474e 005 s fft 2048 3 93 1 6667e 004 s fft 4096 3 38 3 1250e 004 s fft 8192 3 12 6 2500e 004 s fft 16384 5 99 2 5833e 003 s fft 32768 399 5 1667e 003 s fft 65536 11 91 2 3500 002 s fft 131072 11 21 4 7000 002 s fft 262144 13 18 1 1700 001 s fft 524288 10 41 1 9500e 001 s fft 1048576 12 48 4 9250e 001 s fft 2097152 10 09 8 3600e 001 s fft 4194304 10 04 1 7425 000 s fft 8388608 9 86 3 5780 000 s fft PSS bench FFT fourl report on Tue Feb 17 17 38 49 2004 Computing System Xeon 2 66 GHz Operative System MS Windows XP Clock Frequency 2658 RAM 2048 Mbytes 136 Comment length 1024 2048 4096 8192 16384 32768 65536 131072 262144 524288 1048576 2097152 4194304 8388608 16777216 Disk Cache Number of CPUs Computing power Nodes System identif SCSI 3 256 kB 2 2658 Mflops per CPU 1 Sf Lower values are better results efficiency loss 840 729 5355 169 5 Su 43 505 54 59 20 64 69 JU 79 60 26 80 14 10 99 59 60 34 58 61 78 76 66 28 eel Got os TREE ME rep cm WOON WA WW C CS C2 CO CX C2 gt s ff s ff s ff s ff s ff s ff s ff s ff s ff s
48. 600 4 75 06 1 19 01 2 99E 05 7 9 08 5 01 E 04 Jupiter 3 72 08 7 39 08 1 25 01 2 10 07 8 32 08 5 97 05 Saturn 9 28 08 4 06 08 2 75 00 1 86E 08 gt 1 83 08 2 01 06 Sky resolution dv _ 1 21 pep 1 y whete 203 2 Ig 2 0 is number of frequency bins covered by the Doppler effect and is angle between the radiation direction and the rotation axis of the orbital epicycle o First order spin down grid 50 Fo p Av 1 Tsp 2 2 with from 0 to E D coin defining Vo o the spin down decay time o Number of points in the parameter space 3 j 107 tot At j At e At the sampling time e g 0 00025 0 001 0 004 and 0 016 s for the four bands T the observation time e g 4 months 10 s the duration of the FFTs e g 41000 4000 8000 and 16000 s the minimum spin down time e g 10 years where T nin With 10 years and chosen according to the standard bound only the first order spin down parameter should be used and we have N 10r Tar qm 10 years At 4000 5 4 months At tot min min 204 Proposed parameters Here is a summaty of the proposed choices for the parameters of the search for periodic sources on the V
49. 67 123 34 0 3 10 fid logfile PROVA logfile comment fid commento 1 logfile par fid ASD 1234 567 0 logfile input fid FILE IN CHANNEL 1 logfile output fid FILE OUT scratch file logfile ev fid PYW 4 data prec logfile comment fid commento 2 logfile stop fid logfile close fid 160 And this is the created file PROVA 20051230 171452 0 PSS PROVA job log file started at Mon Jan 30 14 01 57 2006 commento 1 ASD 1234 57 INPUT FILE IN CHANNEL 1 OUTPUT FILE OUT scratch file gt gt 123 434 123 533 123 670000000 1 233400e 002 commento 2 stop at Mon Jan 30 14 01 57 2006 To read logfiles Some functions has been developed in MatLab to read log files Here are some out read_log filout filin produces an output structure out that contains 5 objects that can be sub structures or cell arrays each item of which refers to log line 5 objects are cell array for the comments par a structure for the parameters a structure for the events Stat a structure for the stats cell array for the errors It shows also the comments and the parameters on the screen and can put them on a file 161 out read_log_noev filout filin similar to the preceding one but skipping the events it can be much fast
50. 7 90 66081 07 966949679 CO0t9vEt6EOTVE6GTISSSZGITT Y 90 90c62002S80L1v0c8096 9 00 906804904 88192972777 90 9 9289 8941 2010094679 9 9 V 90 9097698 2680418714679 00 90601 4606 767928 72777 90 99789921148181969 679 0 96772 lt 6276411926 2777 90 96 lt 8 667872 617 lt 1 679 00 981 88201 266 6442002777 90 98cLL6V8 ELTL6TCL9C6 9 c0 t99896 c618ELSV7TCZY 90 6768 lt 1810616679 2 98 1 9 90 94984 lt 870 lt 1174180 2679 00 9 4667940876972602777 60 6686876891048898897 0 96 604696189466794878 60 90020982292869607897 0 6 674861870087 60 60 1 06046 46606197 0 9 06 6189 7 229771878 0 9099 LLYOVYVOEOETVL9 0 9 lt 0 861020 407780878 40 90201 94024 4697606997 0 96 97 660494891720878 60 6 01 114007072 07997 0 779690206425676696 78 S0 9GcGLV86VG8LOECL8S9 970 9 0 9 9 8 0 9898TvILOCE6T9GTEEGS9 0t962LVLSVLEOOEGSEET8L 8 0 990L 988985865086LLY9 0 98E602vL801L6v0E8LL S0 910v8S88199L2L881cVv9 0 9881896ccGS8092 L2CLL 8 S0 99cELEL8CSTILC96V9 9 T 0 9807208806 98207299 78 0092 lt 776 28 6 0670 9 v0 900vEZ9VLE8E8LTYLCOO SvvYVYVYVOO0 VVGIG 70 9708694067497017600 000000042490777 lt 4 V0 92001 LE00 LCLSVE0O 0 70 9 72069529
51. 80 Fon ee UN MM M M PS O _ E a MM MENU NNUS ME T olet 06 TUN MC TIR E EE epbpeed TRE TRETEN BEE i PAES REE ede Reed mE onum 2 3 4 0 200 400 600 800 1000 1200 1400 1600 1800 2000 130 Mean of bands each spectrum 0 100 200 300 400 500 600 131 Sidereal analysis Here we present some procedure to detect circadian and sidereal periodicity in the noise data Data preparation We start from the sfdb data also containing only the very short spectra in the form of mean of periodograms Then create the list of the SFDBOO file with dir b s list txt and edit it in the standard way In Matlab execute iok piavespet2sbl list filesbl Then from the created sbl file create the subband gd2s by a script like e g calc cent rawExpl for i 1 7 str sprintf Expl raw07 02d i nl num2str i 1 25 n2 num2str i 25 eval str sbl2gd2 sel phys expl07 sbl 1 0 60000 1 1 87 eval save str str eval clear str end This set of files are used for all analyses 132 Tests and benchmarks The PSS bench program The interactive program The program starts from the command console At the begin
52. 86 1 388 22 73787840 0 0000 29 5139 5 33889842 11 239 638 1 199 0 786 1 388 22 77130125 0 3922 28 1208 1 83047946 11 360 222 3 965 0 786 1 388 88 The old database Old Hough Periodic Source Candidates are stored in particular huge data bases named In this case only one spin down parameter is considered PSC DB is a collection of files and folders contained in a folder with name PSC DBxxxxxx This folder contains areadme txt file data base creation log file psc log e adoc file psc doc with the documentation e apsc dat file that is a script with peculiarities of that DB like the starting time the sampling time the length of the ffts the number of spin down parameters the antenna coordinates ue 20 folders named 0000 0100 0200 1900 each of these folders contain 10 folders named 00 10 20 90 and each of these last contain 10 files one for each hertz of starting frequency The name of the files refer to the name of the PSC_DB and to the covered frequency range for example pss_cand_1394 cand or 0101 This type of database is called type 1 461 Each file has the following structure Header rotocol old 1 normal 2 coin 274 grou 3 4 219 group di pror ini T4 caption 128 5 132 initial time mjd initim double 133 140 sampling time 51
53. AR short power spectrum FFTs real and imaginary part Q O Open the file OUTH1 fopen filename r Read the header_param 186 fread void fread void fread void fread void fread void fread void fread void fread void fread void fread void fread void fread void fread void fread void fread void fread void fread void fread void fread void amp header gt sizeof double 1 OUTH1 fread void amp header param detector sizeof int 1 OUTH1 amp header param gps sec sizeof int 1 OUTH1 amp header param gps nsec sizeof int 1 OUTH1 amp header_param gt tbase sizeof double 1 OUTH1 amp header param firstfregindex sizeof in1 OUTH1 amp header param nsamples sizeof int 1 OUTH1 amp header gt sizeof int 1 OUTH1 amp header gt sizeof int 1 OUTH1 amp header param n flag sizeof int 1 OUTH1 amp header gt sizeof float 1 OUTH1 amp header param mjdtime sizeof double 1 OUTH1 amp header param nfft sizeof int 1 OUTH1 amp header param wink sizeof int 1 OUTH1 amp header param normd sizeof int 1 OUTH1 amp header param normw sizeof
54. For a given detector and given time interval expressed in mjd it gives an output file with the following information 1 minute step detector position vector p in light s in rectangolar Equatorial J2000 coordinates referred to SSB velocity vector v normalized to c in rectangolar Equatorial J2000 coordinates referred to SSB frequency vatiation deinstein due to the relativistic Einstein effect a given detector and source emitting at s the observed frequency obs is evaluated using the following formula f_obs 5 1 p v f s deinstein where p is the unitary position vector of the source in rectangular equatorial 2000 cootdinates The code asks for the name of the detector Presently only the detectors Virgo Explorer and Nautilus are considered but we plan to insert soon all the others it is only a question of inserting them in the detector structure Detector gt virgo explorer nautilus Initial mjd final mjd from 1 Jan 1991 48257 e g 48300 48500 The output is a file with the name table dat This is a typical output Old format up to December 2009 An example of the tables is the following tableVirgo_2000 2010 dat beginning virgo lat 43 631389 long 10 504444 azim 10 00000 deg velx C vely C velz C in rect Equatorial coord deinstein Observed frequency at the detector source_freq 1 Source_pos x vel C deinstein 51544 00000000 51544 00694444 5
55. Initialize 0 00 11 10 0 00 1 0 00 0 38 lutBuild 0 00 11 10 0 00 1 0 00 0 00 readInfoPeakmap 0 00 11 10 0 00 1 0 00 0 00 readParameters the percentage of the total running time of the time program used by this function cumulative a running sum of the number of seconds accounted seconds for by this function and those listed above it self the number of seconds accounted for by this seconds function alone This is the major sort for this listing calls the number of times this function was invoked if this function is profiled else blank self the average number of milliseconds spent in this ms call function per call if this function is profiled else blank total the average number of milliseconds spent in this ms call function and its descendents per call if this function is profiled else blank name the name of the function This is the minor sort for this listing The index shows the location of the function in the gprof listing If the index is in parenthesis it shows where it would appear in the gprof listing if it were to be printed Call graph explanation follows 72 granularity seconds 9 index time self 1 100 0 0 00 0 04 0 00 0 19 0 0 0 00 children TT 10 10 49 38 00 00 00 each sample hit covers 4 byte s called 1 1 1 1 1 1 Lu 1 1 for 0 09 of 11 10 name spontaneous 1 houghBuild 2 lutBuild 6 writeMaps 7 readParameter
56. PSS Software for the Periodic Source Search User Guide Version 23 05 2013 Updated version http grwavsf roma1 infn it pss docs PSS_UG pdf Contents NM Uu SD m EPI NT SER Fus Pu D DR En dU qi 8 10 10 Old pipelines ibd purto at rire tug ray 12 Upto test UR oe ele Oan ap db ARM 12 Programming environments 14 a baka 14 project 16 Thepss a 16 A ct AE E EE aca oes tots 18 SEC gro ipi 19 o 19 Compressed data tit ve ton a d ce ca ee dte Cte 20 20 Sparse vector FORMA nihii 21 PSS Saal neh 23 Data ea roc a mesh a e i ral e o Rl s 26 LOI dba tiat dibus ditata 26 EU LM D EE 29 HOS a uo cespite od detur a aper un sia 29
57. SIS SUC Siz 178 178 Tes DIE ooo a UN EEA ELDER 178 e DEP 178 peakmap creato finnarna Een b embed ene e pub det 178 Main deci eI ete pe 179 Bri Epid e Em 179 fu Ru 179 Pinal dite 179 ott basiert eb suu Treen rrr 179 179 to dtd biet en 180 pulsar een mee ita a T Rm 180 Eon otto ok qmd ot t tedio t PER 180 wigan veda C 181 Doppler reigue teh las 181 ICM ira M M E 181 Theory Astronomical TIMES 182 Theory Contributions to the Doppler effecta emit d bote e rape oda 183 ed un Ca dtu d eti DIE Ct Dad 185 Windows Frame acorde axe ta teni dV RU T re SEU NEU UHR UA clar 185 Non standard file formats ax EUR HI 186 Spi 186
58. TH1 amp header gt 8 sizeof int 1 OUTH1 amp header gt sizeof int 1 OUTH1 COLO VE LOC LE ELE fread void Derive from the header parameters to read next data howmany header param nsamples number of samples stored howmany2 howmany 2 dimension of the total FFT 187 dx header_param gt deltanu frequency resolution lenps int howmany2 header_param gt red Read the averages of spectra with time ps float malloc size_t neader_param gt red sizeof float for ii O ii lt header_param gt red ii errorcode fread void amp rpw sizeof float 1 OUTH1 if errorcode 1 printf Error in reading power ps data into SFT file d n ii return errorcode ps ii rpw free ps Read the AR subsampled short power spectrum ps_short float malloc size t lenps 2 sizeof float for ii O iiclenps 2 ii errorcode fread void amp rpw sizeof float 1 OUTH1 if errorcode 1 printf Error in reading short FFTs into SFT file d return errorcode ps short ii rpw free ps short Note the vectors ps and ps short need to be defined only if one needs to use them Read the FFTs data 0 gt 2 gt 2 for iiz0 ii 2 howmany ii 2 errorcode fread void amp rpw sizeof float 1
59. Theory Contributions to the Doppler effect The Doppler shift that is the frequency doppler observed at the detector for a given source whose intrinsic frequency supposed to be constant is source gt frequency is mainly the result of 1 Earth revolution around the Sun 2 Rotation and 1 relativistic delay Einstein effect 2 light deflection in the Sun s field Shapiro effect To get an idea of the relative weight of these effects we have the code and we have written the separate contributions We have used detector Parkes PSR 437 supposed to be at rest with ra and dec given at Epoch 2000 Their coordinates are both defined in the header file daspostare h Let us consider the quantity x source gt frequency nu_doppler source gt frequency At the MJD 49353 0833 January 1th 1994 we get 1 Revolution 2 916691723 10 5 2 Rotation 4 3508614 10 7 3 Deinstein 3 3540 10 10 4jDshapito 4 65 10 13 Thus the total value of the considered variable z is 2 87314952 10 5 including Einstein and Shapiro We have done not formal comparison of these results with C Cutler Potsdam The results of the comparison have been 1 Cutler Revolution 2 91669173 10 5 2 Cutler Rotation 4 3508340 107 7 3 Cutler Deinstein 3 3439 10 10 4 Cutler Shapiro 4 73 107 13 Cutler total value is 2 87215000 10 5 We have indicated in red those numbers tha
60. V1 h 4096Hz 20090 t tinit tfin or 0 gt all 5 frmin frmax dt output sampling time gdout name of the output gd in absence a default name Extracts a narrow band of signal from a collection of sds files also for more decades function out deld rough clean in miny maxy nw ROUGH CLEAN rough cleaning of array or gd oe in input out deld rough clean in miny maxy Dt array or gd miny maxy permitted interval nw window in samples def 4 105 Cleans a gd an array for burst disturbances function gout rescale gd gin newscalex newscaley RESCALE GD rescales the abscissa and ordinate of a gd newscalex a b newscaley a b if present Rescaling of x and y units for gds Asimilar function is present also for 0425 function dop del gd dopplerfromsfdb tim vv pp source cont GD_DOPPLERFROMSFDB computes the percentual doppler shift and the delay from the data in the sfdb files Extract vv pp tim for example 4 vv pp tim pss band recos vela pulsar 3 551 1024 5 vv pp tim velocity position and time 5 source Source structure 5 cont 0 gt begins from the time 0 of the first day default function p092vb1 piafile filout p092vbl piafile filout 09 nfft 110632 sfr
61. age maxage the AR is re evaluated EVF FAC is the factor for which the threshold is multiplied to write less EVF in the log file stop at Wed Feb 1 12 39 22 2006 162 Parameters GEN BEG 53580 583183 GEN NSAM 2097150 000000 GEN DELTANU 0 000954 GEN FRINIT 0 000000 EVT CR 6 000000 EVT TAU 600 000000 EVT DEADT 1 000000 EVT EDGE 0 150000 EVF THR 2 500000 EVF TAU 0 020000 EVF MAXAGE 0 020000 EVF FAC 2 000000 Time events With gt gt ta d zread logtev we obtain the 8109 time events in the log Here 15 the plot event amplitude h days 163 jua a days 164 s days 5 uoneinp days 165 duration s 11 588 11 57 11 572 11 674 11 676 11 678 11 58 11 682 11 684 Frequency events Then with gt gt t f a cr d read_logfev we obtain the 379960 frequency events in the log Here is the plot 166 2000 1800 1600 1400 1200 F 1000 frequency 800 H 9 600 400 200 frequency E 167 frequency 107 08 107 07 107 06 107 05 107 04 a ae o 107 02 107 01 107 sus 106 99 106 98 3 2 3 4 3 6 3 8 4 42 days 4 4 4 5 48 5 2 168 frequency 1030 1020 1010 1000 990 970 960 1006 E Pese 7 we 5 1004 E 1002 freque
62. arameters 141 142 It contains const_ structure constant class type what pi 0 3 1415926535897932384626433832795 e 0 2 7182818284590452353602874713527 0 light velocity 299792458 0 gravitational constant 6 67259 11 deg2rad 0 degree to radiants conversion Eorbv 0 Earth orbital velocity Erotv 0 Barth rotational velocity at the equator SY s 0 sideteal year seconds SD s 0 sidereal day seconds 143 the angles are in degrees source structure parameter class type what name 1 a 1 rigth ascension ecliptical longitude deg d 1 declination ecliptical latitude deg va 1 variation of a 1 variation of d pepoch 1 position epoch eps 1 contents of linear polarization psi 1 polarization angle psi lt 0 clockwise rotation also if eps 0 and psi 90 100 1 epoch typically 1 1 2000 10 1 initial original frequency 1 initial first derivative of the frequency frequency vatiation day 1 second derivative of the frequency variation per day fepoch 1 frequency parameters epoch h 1 h amplitude snr 1 signal to noise ratio coord 1 0 gt equatotial 1 gt ecliptical n 1 number of soutces chphase 4 phase of the last chunk used by pss chunk 5 Defined as the differen
63. ation needed to understand how the program works More details can be found in the programming guide create db PG doc The candidate database consists of a hierarchy of directoties coveting the frequency range from 0 to 2kHz The standard configuration consists of 20 directories each covering 100 Hz each containing 10 directories each covering 10 2 Each 10Hz directory contains one standard or more files containing candidates for that frequency range Program schematic description The program gives two possibilities 1 create a new database and fill it with candidate files 2 fillan existing database each candidate it checks on the basis of its frequency if it goes into the currently open candidate file ot in a currently close already existing file or in a new file to be created The new candidate file is created if needed and some parameters are computed and written in the candidate file frequency index respect to OHz in units of step steps in lambda beta spin down Other parameters read from the first file produced by the Hough code The number of candidate files written and the total number of candidates are shown at the output Input parameters Options create db help f lt starting frequency Hz gt g lt FFT duration s gt d lt frequency band Hz gt l list of files to concatenate gt q lt numbert of files to concatenate gt 4 lt input files path g
64. b doppler doptab source t computes the percentage Doppler shift for the Earth motion It works with a single time or with a single soutce doptab table containing the Doppler data depends on antenna and year a matrix n 4 or n 5 with 1st column containing the times normally every 10 min 2nd col x in c units in equatotial cartesian frame 3rd col y 4th col 2 5th col vx in c units in equatorial cartesian frame 6th col vy 7th col vz 8th col de einstein source pss structures or a double array 1 2 with first col the sources alpha in deg 126 second col the source delta in deg o t time array TAI in mjd days This function uses the tables created by pss astro interpolating with spline Hete is the results for 2003 1 0001 1 0001 1 0001 Y UA 1 0001 1 0001 1 0001 There is also another function that computes Doppler effect dop gw doppleri v source that uses the velocity vector of the detector and the soutce Another example is v_x v_y v_z v_detector doptab t that creates the three arrays containing the three components of the detector velocity at the times of the array t 127 Sidereal response Because of the radiation pattern of the antenna the response to a gravitational source vaties depending on the sidereal hour A number of functions dealing with these problem are developed sr sid resp antenna source n
65. ble setting the variable WRITE to 1 in pss hough h and re compiling the program to produce some ASCII files containing the last computed Hough map derivative extended Hough map covering in lambda from PI to 3PI i e before folding and Hough map Also the cumulative Hough map of the whole job can be written it is meaningful only if the sky grid does not change during the job As typical input files cover a small frequency range a fraction of Hertz see in Appendix for more details the analysis of a large frequency band implies the submission of a large number of jobs each analyzing an input file The management of this submission job status checking retrieval of the output etc is done by the Supervisor program PSS_SV which is described in detail in another guide Function prototypes long readParameters int argc char argv char char char double float float double char int char char char void setParam char double double float double long long int double double int long long int float double int void gridInitialize double void trigInitialize void lutBuild float int void DrawLeftCircle float float float float int int float void houghBuild FILE int double double int float float char double int void houghBuildAdap FILE int double double int float float char double int float void readPeaks double do
66. ce between the semimajor and the semiminor axis of the normalized polarization ellipse 144 antenna structure parameter class type what name long lat azim azimuth alt altitude ovet sea level incl inclination type ii bar _ structute itf_ structure n number of antennas 145 data structure parameter class type what dt 1 sampling time s sf 2 sampling frequency tobs 1 total observation time days 10 1 initial time mjd iniwin 1 double array statting time of windows finwin 1 double array ending time of windows nwin 1 number of windows t 4 time mjd 146 structure This is sub structure used in or in conjunction with sfdb_ tfmap_ and tfpmap_ structures It describes the working band It can be also used alone parameter class type what len 1 fft length number of samples tlen 2 fft time length N 1 number of ffts onev 1 take one fft evety df 2 frequency bin res 1 resolution interl 1 interlacing wind 1 window type frin 1 initial frequency tin 1 initial time unit 1 unity 147 band structure This is a sub structure used in sfdb_ tfmap_ and tfpmap_ structures It describes the working band It can be also used alone parameter class t
67. cedures These are templates of batch Matlab procedutes to be edited 200 do base run sp creates a standard spectrum from sds data two run sensitivity computes the sensitivity of coincidences of two runs Some coefficients computed with pss run basic analysis xls save pia hist equalized spectra peak histogram search 10Hz peaks with ef search 10Hz peak with epoch folding convert peakfiles converts peak files with piapeak2vbl candidate coinc 1 single file candidate analysis crea lintfmap driver of crea linpm linear peak maps clean conv peakfile cleans p05 peak files using vbl files aa analysis anti aliasing analysis big coin does big coincidence jobs merge coin merge coincidence matrices Preparation procedures Check procedures 201 Analysis procedures 202 C procedures r sum o Limiton where is the radiation frequency and and R are the parameters of the rotational Tyrer Reference Some basic formulas _ 1 2 epicycle Here is table of some epicycles of interest Period Radius 20 Ry D May Ime velocity oppler 2kHz 5 Rotation equator 86160 6 38 06 4 65 02 0 033928 3 10E 06 1 49E 03 Rotation 43 N 86160 4 66 06 3 40 02 0 024781 2 27E 06 1 74 03 Earth orbit 31557600 1 49 11 2 97 04 0 005906 1 98E 04 3 56E 03 Moon 2505
68. cepta To each point new straight line in the parameter space with equation q y m x corresponds That is we can draw a line in the parameter space for each pair x y and all these will intersect in a point m q identifying the parameters of the original line If noise is present several clusters of points will appear in the parameter space In our case the original data are the points in the time frequency peak map and the 62 transformed data are points in the source parameter space 4 1 position frequency and frequency derivatives Assuming that there is no spin down the relation among each point in the time frequency plane and the points in the parameter space is given by the Doppler effect equation ven fk fo where f is frequency of a peak v is detector velocity vector and nis versor identifying the direction of the source From this equation we find that the locus of points in the sky to which a source emitting a signal at frequency at f could belong if a peak at frequency f is found is a circle with radius given by A foz C and centered in the direction of the detector velocity vectot Due to the frequency discretization in fact we have an annulus delimited by two circles Peaks belonging to the same spectrum produce concentric annuli while moving from one spectrum to the following the center of the annuli moves on t
69. ch mode 1 sets batch mode 1 ask item choice of the directory home federica hrec v2 input directory 3 ask channel choice dL 20kHz channel 4 ask item DataType file name block hrec the block 5 ask output directory home federica sds output directory 2 ask item File choice hrec 710517600 3600 gwf the file 6 creates the sds file 26 2 ask item File choice hrec 710521200 3600 9wf 6 creates the sds file 2 ask item File choice hrec 710524800 3600 gwf 6 creates the sds file 12 exit To create easily the batch file create a list of the files to be converted and then edit it In Windows the command is dir b gt list txt and can be issued with the command file to list bat it creates a file list txt to be edited to create the batch command file To start the program a batch command can be created containing something like D SF_Prog C NET Frameutil Release frameutil lt batchwork txt gt out txt e Simple batch mode Alternatively the program can be run with the argument of a text file that contains the configuration and the file names in this way gt frameutil batch txt The text file any name is possible has the following format folder in example D Data pss virgo sd frame folder out example D Data pss virgo sd sds channel example h 4kHz antenna acronym example VIR data type example hrec frame filel example HrecV 806888400 01 Aug 2005 01n40 600F gw
70. ciotti Moreovet the most distant allowed frequency from the search one is given by Y en fo z 9 C 2 because the frequencies refer to the center of the bins then moving of half a bin from S extreme WE go to the end of the Doppler band In this situation we have where the second order term oc of y has been neglected Then the maximum variation of circle radius can be written as ALAS Ah C yee y Jo c Now let us express the change in the search frequency as an integer number of frequency bins Afo impose that corresponding maximum variation of the circle radius is less than fraction of the sky pixel 3 1 s a with 1 From here we find a condition on the maximum shift on the search frequency in otder to have a small variation on the circle radius 68 pa dvd c fo where how much small depends on the chosen value of 6 For a given frequency band the most stringent limit is obtained using the minimum frequency of the band and the maximum value of the v c ratio which is B 1 0222 10 A reasonably conservative value for 6 could be 0 1 With this choices and using standard parameters for the frequency and sky resolution we find that 1 lt 28 would be enough on the wall frequency band 0 2kHz Larger values are allowed at smaller frequencies debugging and test purpose it also possi
71. didate matrix coordinates are converted to ecliptical sour complete matrix 114 Time frequency map simulation The functions to simulate time frequency spectra are obtained by the same functions that simulate the peak maps with the input keyword tfspec set to 1 115 Peak map simulation There ate two simulation functions one easier that can be used for the first incoherent step when the data chunks have the length of no more than some hours and another for the subsequent analysis when the data chunks can be of more than one day There some function to create files to store peak map data pss pm O pm file that stores the peak map structure pf in the file file in a very easy format Low resolution simulation pm pss sim pm lr ant sour pmstr doptab nois level tfspec where o ant isan antenna structure with elements lat latitude in degrees long longitude in degrees azim azimuth type antenna type 1 gt bar 2 gt interferometer sour source structure array elments used right ascension degrees d declination degrees 5 contents of linear polarization psi angle of linear polarization 10 un Dopplered frequency at start time coefficient of first power spin down Hz day ddf0 coefficient of second power spin down Hz day 2 Snr signal to noise ratio linear supposing white noise pmstr peak map property structure sampling time s
72. ding if the last bit is used for the sign Linear coding can be for X 8 or 16 signed or unsigned integer coded Linear and logarithmic coding can be adaptive So totally we have 16 different LogX formats 7 linear 4 logarithmic 3 look up table 1 float and 1 constant float 11 of which can be adaptive Sparse vector formats Sparse vector is a vector where most of the elements are 0 We call density the percentage of non zero elements Sparse matrixes are formed by sparse vectors Sometimes binary mattices the non zero elements are all ones and sometimes they are also aggregated In this last case the binary derivative 0 if no variation 1 if a variation is present is often a sparse vector with lower density value We represent sparse vectors with the run of 0 coding It consists in giving just the number of subsequent zeros followed by the value of the non zero element In the case of binary vectors the value of the non zero element is not reported Examples M2000 003200000 0230000000030 0002 21 codedas 0 1 2 5 32 6 23 8 3 0 3 2 binary case 1000100000010000000011000 coded 36803 aggregate binary case 111000000001111000000 111110000000000111 binary derived as 100100000001000100000100001000000000 100 coded as 02735492 In practice the number of subsequent zeroes is expressed by an unsigned integer variable with b 4 8 16 or 32 bits one is added to the coded values in such a way
73. double sampling frequency Lfft2 int32 original length of fft divided by 2 lnifr double oe Block header mjd double npeak 532 1 double 5 vely double velz double double 5 double posz double 2 Short spectrum Einstein 2 spini double spdf double splen int32 sp float splen values Data npeak times bin int32 ratio float xamed float mean of H 106 function conv p09 conc folder filelist CONV P09 CONC o AP AP o o o o AP AP o oe folder folder containing the decades t file in folder filelist file lis to obtain the list do 20070607Npm VSR1 20070607Npm VSR1 20070607Npm VSR1 20070607Npm VSR1 20070607Npm VSR1 20070617Npm VSR1 20070617Npm VSR1 20070617Npm VSR1 without blank lines dae gt vise _20070607 1 20070607 2 20070607 3 20070607 4 20070607 5 720070617 1 20070617 2 20070617 3 09 09 209 09 po 09 09 09 converts from p09 to vbl and concatenates and edit to arrive at the list is automatically sorted by the function 107 Theory and simulation Snag pss gw project PSS detection theory Sampled data simulation The
74. dule containing the main function and all the functions which ate common at most modulo the data type of the hough map and hough map 63 derivative to standard and adaptive hough transform parameters computation construction of the LUT reading of the peakmap input file selection of candidates pss hough h headet file containing type definitions global variable declarations function prototypes constants declaration of structures pss hough c module containing functions related to the computation of the standard Hough transform pss hough adaptive c module containing functions related to the computation of the adaptive Hough transform Makefile MakefileAdaptive makefiles for compiling and linking the library respectively for the standard and adaptive Hough transform The library can be divided from a logical point of view into four parts Building of the Look Up Table LUT For each search frequency new LUT is computed As a matter of fact the same could be used for more frequencies but for the moment 1 as long as observation times ate short and then computing times are not too much long we take a conservative approach Given a search frequency and a spin down value we use the same LUT for all the times for which the frequency shift respect to the original one is lower than a given number of frequency bin 4 in our current default choice This value is somewhat arbitrary and again cons
75. e included in the number after the name This is the name of the parent The parent s index number is printed after it If the parent is a member of a cycle the cycle number is printed between the name and the index number If the parents of the function cannot be determined the word lt spontaneous gt is printed in the field and all the other fields are blank For the function s children the fields have the following meanings self This is the amount of time that was propagated directly from the child into the function children This is the amount of time that was propagated from the child s children to the function called This is the number of times the function called this child the total number of times the child was called Recursive calls by the child are not listed in the number after the name This is the name of the child The child s index number is printed after it If the child is a member of a cycle the cycle number is printed between the name and the index number If there are any cycles circles in the call graph there is an entry for the cycle as a whole This entry shows who called the cycle as parents and the members of the cycle as children recursive calls entry shows the number of function calls that were internal to the cycle and the calls entry for each member shows for that member how many times it was called from other membe
76. e physical amplitude instead of the statistical amplitude The format of the file can be sbl or vbl depending if a normal or compressed form is chosen peak map file contains all the side information of the SFDB parent file Peak map creation MatLab environment The peak map can be created by peakmap sblfil thresh typ where sblfil is the SFDB file thresh is the threshold and typ is the type 0 normal with amplitudes 1 compressed with amplitudes 2 normal only binary 3 compressed only binary Here there is an image of the peak data with zooms 43 44 PR So Ra 2 5 x10 6 6 64 6 2 5 8 5 6 54 10 H 10 16 Y ze j 1 6 15 6 2 6 25 63 6 35 Other peak map creation procedures Various techniques have been studied to construct effective peak maps The main problem 15 that of the presence of big peaks that obscure the area around The basic procedute is the following ind fr snr1 amp peaks snr sp find fpeaks in tau thr inv based on the gd findev function idea that is adaptive mean computation with o in input gd or array o tau AR 1 tau in samples o thr threshold o inv 1 gt starting from the end 2 gt min of both index of the peak of the snr array o fr peak frequencies o snri peak snr o amp
77. e pss ante h k k x SFDB libraries x include pss sfdb h int main void ANE int nmax number of FFT to be read int errorcode char capt gd 12 GD gd GD with SFDB data GD gd short GD with the short SFDB data GD gd time GD with h reconstructed GD gd band GD with SFDB data over a chosen sub bandwidth int go2time iii iband extract ipeak extract char filetime MAXLINE 1 char filefreq MAXLINE 1 char filepeak MAXLINE 1 FILE TIME FILE FREQUENCY FILE FINAL INE C int ivign 0 1 if data are windowed before going back to the time domain int istart i band float total band initial freq band int len len4 len2 inter int simulation float freq sin amp sin double beg mjd int isubsampling int write filefreq kkf iif double fft2spectrum double phase phase of the sinusoidal signal added in the simulation int iw int fft read int nfft tot HEADER PARAM header param Structure with parameters which have been written in the sfdb file header param HEADER PARAM malloc sizeof HEADER PARAM strcpy capt gd SFDB data gd crea gd 0 0 0 capt gd gd short crea gd 0 0 0 capt gd printf Number of FFTs to be read 190 scanf Sd amp nmax printf file with the peakmap output binary n
78. e These data could be stored in an SDS file with many channels but for the necessity of easy random access needed for such data bases a peculiar format will be used 25 Data preparation Format change C environment In order to use some Snag features in a more proficient way the frame format data must be must be converted in the sds format This can be accomplished by the interactive program FrameUtil exe It can be run in two ways interactive or batch e Interactive can be used also in batch mode in some systems d SF_Prog C NET Frameutil release FrameuUtil exe iD Data pss virgo sd frame t 8068884060 01 2005 01 40 60 qu 1 1D DataN pss virgo sd f rame HrecU 8 66888 466 61 Aug 2665 61 h40 6BBF quf i Working on D Data pss virgo sd f rame HrecU 866888 466 61 Aug 2665 01h40 6 BBF quf gt channel h_4kHz2No5 Output file prova sds gt GPS time is Sun Jan 66 66 66 66 1988 Frame_Util menu Folder choice File choice Channel choice DataType file name block Output directory sds file creation Time check Batch mode lt 0 1 gt SFC Util Dump file standard on stdumpfile txt gt Dump file Con dumpfile txt gt Exit T 2 s 4 6 rut 8 7 Which item 2 Very useful are two dump file facilities that give resume of all the frames The program can be used also in batch creating a batch file as this 8 ask bat
79. e listed gds that have been created and some buttons For a more extensive description of this see the Snag2 UG 14 B data browser that is a Snag application part of the gw project to access and simulate gravitational antennas data It is started by typing data browser or just db if the alias 1s activated This opens a window Snag Data Browser Bronie window Data Browser Status access Iype by file data type frames file dat6064 channel 2 5kHz hp white filtering type filters processing type with a text window and some buttons The text window shows the status of the DataBrowser as is due to the default and user s settings The parts that are developed in this environment are labeled MatLab environment 15 Inside Snag gravitational wave project has been developed important of this project is the DataBrowser showed in the preceding sub section Other patts of this project are 9 9 9 9 astro on astronomical computations coordinate conversion doppler computations time with a set of functions dealing with the time Among the others O conversions between mjd modified julian date gps and tai times sidereal time conversions between vectorial and string time formats sources about gravitational sources pulses chirps and periodic signals pss specifically for
80. e of long runs with the following format described lines the decade or run the original sampling time the length of the original FFT the frequency step the spin down epoch the spin down min max and step the number of FFTs Then for each frequency decade the frequency band with a gt Fr Band at beginning the number of original peaks of the peak map the mean value of the Hough map the total number of found h peaks one line for each h peak with frequency spin down Hough map peak amplitude On the basis of this file we can modify the peakmap files in order to flag the disturbance peaks This is achieved changing the INDEX vector that gives the frequency to its negative values at the flagged peaks 82 Supervisor Basics We call Supervisor SV the ensemble of services and programs we are developing for farm job management It is built on top of a batch system which controls job scheduling and load balancing PBS see later for a more detailed description The batch system can be customized according to the user s needs the SV in some cases uses the PBS API functions Let us see how the SV enters in the general scheme for the hierarchical procedure developed for the periodic sources search Schematically these are the main steps in the data analysis procedure 1 build SFDB construction of the SFDB from the h reconstructed data peak selection selection of
81. eader param gt detector 0 printf Minus mode plus mode calibration freq sf sf n header param gt freqm header param gt freqp header param gt frcal printf Do you want to ADD sinusoidal signals l yes n scanf d amp simulation if simulation 1 printf Frequency Amplitude times EINSTEIN scanf Sf Sf amp freq sin amp amp sin printf Do you want to extract a bandwidth 2 l yes n scanf Sd amp iband extract if iband extract 1 if header param detector 0 printf Minus mode plus mode calibration freq sf sf n header param gt freqm header param gt freqp header param gt frcal printf Initial frequency bandwidth positive band only n scanf Sf f amp initial freq amp band gd gd gd n double initial freq gd gt dx capt_gd if header param detector 1 amp amp iband extract 1 only for itf can be generalized to bars printf Peakmap only on the sub bandwidth 1 0 on the whole bandwidth n scanf Sd amp ipeak extract if simulation 1 add sinusoids param freq sin amp mjd amp phase if header param gt typ 2 beg mjd header param gt mjdtime header param gt tsamplu 2 header param nsamples day2sec new start time to 192 evaluate the beginning phase if header param gt typ 2
82. ean conv peaktile ybi vbi creation using 2 08 08 2 hours 750 new 08 clean piapeak P new vbl mes create input file p08 input 3 min 450 creation 750 3000 Sky Hough 4 16 pss hough p08 cand hrs independent of time f fdot Hough pss cand map 10 continuing Task Program file Mb day in out single jobs candidate create db cand 4 min 750 collection reshape db2 psc type 2 db using 242 x 50 1050 Hz psc reshape db2 T hour n or reshape db3 coincidences psc coin cand cand All batch procedures are submitted by pss batch diary 11 Old pipelines Up to August 2008 Task Program filein file time day Mb day out mo Extract tcsh frame frame 10 min 1500 frame to hrec 595 frame util frame files 10 min 1400 conversion files sds reshape sds sds 15 min 1400 sfdb crea_sfdb sds sfdb 30 min 3000 peakmap crea peakmap sfdb p05 15 min 450 piapeak2vbl or 05 to vbl by 05 30 300 convert peakfiles cleaning mask clean conv yi bl creation using 05 05 2 hours 750 05 clean piapeak new vbl oa create input file 05 input 3 min 450 750 3000 Hough 4 16 pss 05 independent map hours of time 12
83. ed spectrum The details are in the following sub sections Extract the band in an 0 file This step is done by iok sbl sfdbO9 band2sbl freq cont folder pialist filesbl simpar whete freq cont folder pialist filesbl frequency band min max rough n whole band n is the number of blocks per file output file content 1 only data 2 data and velocity position double 3 data velocity position double and short spectrum band folder containing the input pia files pia file list obtained with dir b s gt list txt file sbl to create if simX sbl simulation simO sbl only sinusoid classical sim1 sbl only sinusoid frequency domain sim2 sbl data plus sinusoid sim3 sbl white noise 98 sim4 sbl only source naive sim5 sbl only source fine simpar simulation parameters e g frequency amplitude iok 0 gt error header of the output file An example is the following iok sbl sfdbO9 band2sbl 22 25 22 5 2 O pss virgo sfdb VSR1 v2 list2 txt v_B sbl In list2 txt there is the list of the sfdb09 files to work on E g the list is created by dir gt list2 txt then edit and check O pss virgo sfdb VSR1 v2 deca_20070727 VIR_hrec_20070727_025306_1 SFDBO9 O pss virgo sfdb VSR1 v2 deca_20070727 VIR_hrec_20070727_171336_100 SFDBO9 O pss virgo sfdb VSR1 v2 deca_20070727 VIR_hrec_20070728_051336_200 SFDBO9 O pss virgo sfdb VSR1 v2 deca_20070727
84. ence group contains sort 2 gt to 242 files second coincidence group contains sort dirin input directoty root dirout output directory root infrlinfr2 min max value of frequency groups def 1 200 psc reshape single op prefilout filinidirout pss candidate files post processing type 2 single input file op operation 0 sort input files 1 gt to 242 files first coincidence group contains sort 2 gt to 242 files second coincidence group contains sort prefilout first piece of output file names filin input file dirout output directory root Operating on candidate vector fr psc_readfr candvect reads frequencies from vector candidate in the files the data are read as UInt16 candvect contains single candidates information as groups of 8 uint16 the first 2 uint contain the frequency information mcand psc vec2mat vcand head nn converts a candidate vector 8 n to a candidate matrix 7 n vcand candidate vector head candidate header structure nn how many candidates at most Operating on candidate matrix candout psc_sel candin fr lam bet sd cr candidate selection 23 candin input candidate matrix 7 n fr frmin frmax 0 no selection lam lammin lammax 0 no selection bet betmin betmax 0 no selection sd sdmin sdmax 0 no selection ct crmin crmax 0 no selection out psc show cand typ pars shows PSS candidates the data can be previously selected with psc
85. eq Half band Samples in one stretch Subsampling 2 if data were overlapped fprintf TIME s 15 10 5 param gt mjdtime header param gps sec header param gps nsec gd time dx if iband extract 1 fprintf TIME s 1d 1d dWMn gd gt ini total_ band gd gt n gd gt n gd gt n header gt if iband extract 1 fprintf TIME s sd n S gd_band gt ini gd_band gt n gd_band gt dx 2 0 gd_band gt n gd gt n gd_band gt n header param gt typ for 11 0 11 lt len inter len4 ii fprintf TIME n iitistart iitistart gd_time gt dx gd_time gt y ii for ii 0 ii lt len inter len4 11 fprintf TIME 15 10f header param gt mjdtimetii gd time dx day2sec gd time gt 11 fft2spectrum if k 0 for ii inter len4 ii lt len inter len4 fprintf Xn di4distart iijistart gd time dx gd time y ii for ii inter len4 ii lt len inter len4 ii fprintf TIME 515 10 sf header param gt mjdtimet tii gd_time gt dx day2sec gd_time gt 11 fft2spectrum istart len inter len2 in the first FFT written 3 4 len data In the others written len 2 data if go2time 1 printf gd gt n gd time n gd_time gt dx 514 3194 f n gd 194 n gd time
86. equency which takes into account the spin down Each read peaks is translated into one or more peaks of the LUT in this way we take into account the fact that the LUT circles are computed for the largest possible velocity and then are 64 more numerous than the true ones i e it may be possible that one peak of peak map corresponds to two circles of the LUT Calculation of the Hough Map HM The HM is a histogram in the sky coordinates For each selected peak two semi circles are read from the LUT one corresponding to the current peak and one corresponding to the peak immediately before in the LUT This is done because each peak in a discretized space produces an annulus of pixels delimited by a pair of circles The center of each semicircle is properly shifted depending on the time index and the corresponding right half is computed using simmetry then for each peak we have four semicircles Semicircles with radius larger than 90 degrees are computed taking the corresponding opposite circles with lambda rotated by 180 degrees and beta inversed To the pixels of the semicircle of each pair left right assigned values 1 and 1 respectively for the external one and the reverse for the inner or real values properly determined if the adaptive Hough map 15 computed This procedure is applied for each peak of the peak map and for each time the reference frequency is propetly shifted i
87. er ta out read_logtev filout filin this is specific for the SFDB creation jobs It reads the time event found tfa out read_logfev filout filin this is specific for the SFDB creation jobs It reads the frequency event found Example of use of logfile data For the construction of the C6 SFDB we have a logfile of 21 MB The output of read log noev is File DASF DatAnpss datan Reports crea sfdb 20060131 173851 log started at Tue Jan 31 17 38 51 2006 even NEW a new FFT has started PAR1 Beginning time of the new FFT PAR2 FFT number in the run even EVT time domain events PAR1 Beginning time in mjd PAR2 Duration s Max amplitude EINSTEIN even EVF frequency domain events with high threshold PAR1 Beginning frequency of EVF PAR2 Duration Hz Ratio in amplitude max average PAR4 Power EINSTEIN 2 or EINSTEIN average if duration 0 when age gt maxage stat TOT total number of frequency domain events par GEN general parameters of the run GEN is the beginning time mjd GEN NSAM the number of samples in 1 2 FFT GEN DELTANU the frequency resolution GEN FRINIT the beginning frequency of the FFT EVT CRis the threshold EVT TAU the memory time of the AR estimation DEADT the dead time s EVT EDGE seconds purged around the event EVF_THR is the threshold in amplitude EVF TAU the memory frequency of the AR estimation EVF MAXAGE Hz the max age of the process If
88. ervative it can be shown that using the LUT for 27 different frequency bins would be enough and even more at low frequencies Given a frequency the circles corresponding to all the possible peaks in the Doppler band of the chosen frequency are computed The ecliptic coordinate system is used As a consequence the circles have centers in a narrow belt around the ecliptic This belt is discretized so that ordinates of the circle centers are taken a discrete set of 11 different values while the abscissa is assumed to be zero Also the sky is discretized in a number of pixels then two circles with the same center distinguishable only if their radii differ by at least one pixel each circle center and for each allowed radius there 15 a loop the ordinate values and correspondingly the values of the abscissa as real values only for the left semi circle are computed This is done for performance reasons since it is then very simple to determine the corresponding right semi circle using simmetry arguments The computed abscissae are stored into an array which 15 the look up table Another 15 created containing the index along the vertical direction of the pixels corresponding to the minimum and the maximum ordinate of each circle and cumulative difference between them Reading of the peak map each time read the portion of the peak map input file corresponding to the Doppler band around the actual fr
89. ery time the search frequency crosses a ten Spin down depends on the assumed spin down decay time it is computed once 65 Critical ratio 0 001 fixed Amplitude adimensional 1028 Mean number count 1 Fixed The output file can be copied locally i e on the directory where the executable has been run or to a remote destination a Storage Element in the grid terminology through the grid command lcg cp The number of frequencies to be explored is set by the user and cannot overcome the frequency range covered by the input file while the number of first order spin down values is computed on the basis of the minimum spin down decay time which is assigned by the user according to the relation 2 Nus 27 07 where fj 15 the maximum searched frequency default 2KHz N ppr is the number of FFT i e total observation time divided by the duration of a single FFT multiplied by 2 for Na taking into account the overlapping T nis the minimum decay time at fj default 4000 years Of is the frequency bin width inverse of the FFT duration An over resolution factor of 2 has been also included The above equation is obtained starting from the definition we use for the minimum decay time So du The maximum frequency variation in the observation time is Af am t a Then the number of spin down values of the first otder is considering a factor of 2 of over resolut
90. ese log files have general standard format Here are the main features thenameisjobtype yyyymmdd hhmmss log BP they text files ASCII BP they are structured in o explanatory lines starting with header with starting date 22 parameters a parameter for each line starting with and then the name of the parameter and its value special parameters namely arrays or matrices written in blocks in user mode These blocks are marked by a first line with SPECIAL gt label and last line with ENDSPECIAL gt label body with events starting with gt XXX gt with XXX a 3 character code of the event type partial statistics and r sum starting with gt gt gt XXX gt with XXX a 3 character code of the stat type errors starting with ERROR followed by a string footer with date final statistics and r sum Example of log file 158 PSS program name job log started at date and time possible explanation 11 any number everywher PARI PAR2 e R NPUT OUTPUT PARn XXXXX filein comments fileout comments 2 2 2 gt 1 gt 1 gt 1 gt gt gt FDS 2 2 1 1 gt gt gt 05 gt gt gt gt VCX gt ERROR SSSSSS file strin
91. f frame file2 frame file3 If we have a set of sds files we can concatenate them i e put in each of them the correct values filspre and filspost so the data can be seen as a continuous stream and one can access at any of them pointing to any file of the chain also not containing the given datum This concatenation is performed for example by the function 595 concatenate folder filelist where folder is the folder containing the files and filelist a file containing the file list similar to the above list txt in the cotrect otder Be sure that the files in the list are in the correct order 27 mote complex operation on a set of 545 files is performed by sds reshape listinimaxndat folderin folderout that constructs a new set of sds files with different maximum length and concatenates them In this way a mote efficient data set is built correct order maxndat is maximum of data for channel folderin and folderout are the folders containing the input and output data listin is a file containing the file list similar to the above list txt in the When all the files of a run are produced and concatenated possibly with 545 reshape they should be checked by check 545 conc outtile that analyzes the set of files producing a report in outfile or on the screen if outfile is absent Here is an example of one of these reports c5 data check
92. ff s ff s ff s ff s ff s ff ct ct ct ct ct ct ct ct ct ct ct ct ct 2 reo re vr vr vr ve vr vr vo vr vv HY 137 SFDB Hough transform 138 Service routines Matlab service routines antenna source data pss par ipss par 8 source tab 139 pss lib pss rog 140 General parameter structure the parameters used for this software are organized in nested structures Here all these structures ate described Any parent structure can lack some children Main pss structure This is the container for the other high level structures The parameters are classified in 4 classes O fixed 1 primary or input 2 derived 3 analysis results 4 other Structures Type Use const Mathematics Physics and Astronomy constants may contain G2 source periodic source parameters atrays may contain antenna y detector physical parameters atrays data data parameters fft_ fft parameters band _ band description may contain sfdb y short fft data base parameters may contain tfmap time frequency map atrays may contain time fr k tfpmap e frequency peak map may contat hmap y hough map parameters atrays cohe _ coherent follow up parametets ss supetvisor parameters may contain candidate y periodic source candidates atrays may contain event event parameters atrays computing_ computing p
93. g Starting with XXXXJ XXXXJ XXXXJ XXXXJ XXXXJ stop at date and time string final results There is a collection of function to create log files It is in pss serv a module of pss lib and contains FILE logfile open char prog int logfile comment FILE fid char comment int logfile error FILE fid char errcomment int logfile par FILE fid char name double value int prec int logfile ipar FILE fid char name long value 6 int logfile special FILE fid char label 6 int logfile endspecial FILE fid char label 159 int logfile input FILE fid char inpfile char comment int logfile output FI E fid char outfile char comment int logfile ev FILE fid char type int nvalues double values int prec int logfile stat FI values int prec int logfile stop FI E fid char type int nvalues double E fid int logfile close FI fid Note that for the functions logfile par logfile ev and logfile stat we need not only the values of the parameters for the output but also the precision namely prec format 0 900 1 31 2 96 6f 3 91 10 6e 100 01 Here is an example of use void main FILE fid double data 4 int prec 4 data data data data prec prec prec prec 123 434 123 533 123
94. g utilities for KNOWN SOULE bbb 105 Theory 108 108 assiepasta qut edis quai Neq RM GT QI RR RE EUER E 108 Sampled abd simulator 108 ELA 110 145 Peak map SU UIA MOM odes bea bo lev teh Preseason bte y 116 Tow tesolution snnulatioti etie 116 High resolution o cR 118 Candidate 119 Time and astronomical functions 120 dum dtu cundcm dimmi 120 Astronomical coofdifiates n aste ttti 121 Source and Antenna str ctufes sai nio iti et i Eee Pos 122 Doppler effect pes binas 123 Sidereal UA 128 Other analysis tools csi YARIS KA YVES 129 Sensitivity evaluation dose ut genu m teri 129 Siciereal analysis du epa Bes t e dta e a 132 DAD preparation eun duse Quos Qua ADU ODORE LUE 132 Tests and benchmatks 5 t er EE E 133 592 DUE SHORT Ally ud durae AS 133
95. he celestial sphere around the ecliptic Moreover also the circles radius changes in time because of the vatiation of the modulus of the detector velocity vector For a given value of the source reference frequency f and spin down calculating the annuli corresponding to all the peaks in the time frequency map and properly summing them in order to take into account the source spin down we compute the Hough map For each value of the source reference frequency we have several Hough maps depending on the number of possible different values of the spin down parameters The number of spin down parameters that must be taken into account depends on the minimum source decay time 7 Which we search for In order to limit the overall 0 needed computing power to a reasonable value Tai 10 yr could be a reasonable choice for the analysis of the whole frequency band over months This would imply that only the first spin down parameter should be considered For shorter observation period like C6 or C7 Commissioning Runs we can adopt a smaller value for the minimum decay time The total number of Hough maps that are built is given the product of the number of frequencies to be searched for and of the number of values of the first order spin down parameter V On each map a number of candidates is selected putting a suitable threshold Implementation Pss hough is divided in the following files pss hough drive c mo
96. here typ 0 1 2 3 indicates the type of plot simple logx logy logxy y is a gd or an array x is simply if y is a gd the abscissa array file if present is the file to put the output i e the starting and ending points of the chosen periods xx is an n 2 array with the bounds of the chosen periods This program is very simple to use you can directly choose the start and stop abscissa of as many periods you want when you chose the stop point the chosen period is colored in red and you are prompted if you choose another period Figure No 1 Vindowv Heln APS Q Do you want to choose a period le Edit Nei Inser 4 39 4 4 1 1 1 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000 The problem with this function is that it is not possible to zoom the plot for more precise choice 31 sel absc permits the use of the zoom and so an high precision choice It uses global variables 1119399 9399 19399 as the beginning times ending times and number of periods The data of the periods are put in a file named f119399 txt The input variables typ y and x have the same use of the function sel absc Figure No 1 gd Plot BE Stop period End selection Data Selection Menu procedure select Start period and choose the starting point select Stop period and choose the stop point iterate the first two step for a
97. ime stamp of the first time datum gps time e fraction of the FFT time that was padded with zeros e D velocity and position of the detector at time of the middle datum vx eq vy eq vz eq px py eq pz eq coordinates in Equatorial reference frame fraction of c Middle datum means the Lo 12 1 datum 23 position of the detector at time the middle datum x eq 2 coordinates in Equatorial reference frame meters of SSB Peak maps The data are stored in a VBL file The header is similar to that of the SFDB but a peak vector takes the place of the FFT The format of the peak vector is a sparse binary vector so the real length of each block is not constant There are two versions of the vbl peak maps derived from the non standard format p05 and p08 upto August 2008 There ate 3 channels the velocity of the vector Cartesian in the Ecliptic reference frame fraction of c the frequency bins of the peaks integer the equalized peak values e starting from August 2008 There ate 5 channels the velocity of the vector Cartesian in the Ecliptic reference frame fraction of c the short spectrum the index of the peaks array for direct access to PEAKBIN the frequency bins of the peaks integer can be negative for vetoed peaks the equalized peak values The channels have the following parameters CH name lenx dx t
98. int 1 OUTH1 amp header param frinit sizeof double 1 OUTH 1 amp header_param gt tsamplu sizeof double 1 OUTH1 amp header param deltanu sizeof double 1 OUTH1 amp header gt eq sizeof double 1 OUTH1 amp header gt eq sizeof double 1 OUTH1 amp header gt eq sizeof double 1 OUTH1 amp header gt 5 1 sizeof double 1 OUTH1 amp header param spare2 sizeof double 1 OUTH 1 amp header gt eq sizeof double 1 OUTH1 amp header gt eq sizeof double 1 OUTH1 amp header gt 2 eq sizeof double 1 OUTH1 amp header param n zeroes sizeof int 1 OUTH1 amp header param sat howmany sizeof double 1 OUTH1 amp header gt 5 1 sizeof double 1 OUTH1 amp header param spare2 sizeof double 1 OUTH1 amp header param spare3 sizeof double 1 OUTH1 amp header gt 4 sizeof float 1 OUTH1 amp header gt 5 sizeof float 1 OUTH1 amp header gt sizeof float 1 OUTH1 1 fread void fread void fread void fread void fread void fread void fread void fread void fread void fread void fread void fread void fread void fread void fread void fread void fread void amp header gt 7 sizeof int 1 OU
99. ion NE min 2 fme N 2 mas J max pu Teer 2 N rrr m T min 2 2 The spin down step is simply of Na The sky grid is computed starting from the following relation for the mean number of bins in the Doppler band of a given frequency 5v 5v Nay z Lrrr 8c 4c where is the ratio of detector to light velocity s 10 j L is the FFT length The c resolution is the inverse of this the point is that covering the whole Doppler band means covering the whole sky taking as frequency the upper ten to the current frequency and also 20 over resolution factor 410 4 66 log file is produced containing several information on job and environment where it has run parameters used machine characteristics OS cpu type etc timing etc If the adaptive Hough transform must be computed the user has to specify a detector default Virgo this is needed to calculate the detector response which is a function of the detector geometry and its position and orientation on the Earth At the moment only the Virgo and Nautilus detectors ate supported To switch from the standard to the adaptive Hough transform computation some modification to the soutce code must be done by hand in the header file pss hough h change the data type of hough t and houghD t from short int to float in the main function
100. ion For longer runs the analysis is divided in segments of 10 days decades Each decades contains all the data contained in the original frame files starting from the 0 hour UTC of the first day to the 24 hour of the tenth day So we act as the real run 15 divided in sub runs of about ten days The data sds sft peakmaps should be put in separated sub folders with the name convention deca yyyymmdd As an example this are the first decades for the 5 1 Decade Beginning End 1 18 May 2007 21 02 46 28 May 2007 03 59 46 2 28 May 2007 03 59 46 07 Jun 2007 01 55 06 3 07 Jun 2007 02 48 36 16 Jun 2007 01 46 26 Initial analysis steps hrec extraction sds creation and reshaping sfdb creation peakmap creation 178 Main analysis Data quality Hough map candidates Final analysis Coincidence Coherent analysis 179 Other tasks organization Known pulsar investigation Known location investigation 180 Doppler effect computation pss astro PSS astro is a C code for the computation of various astrometric quantities coordinate transformations evaluation of Doppler effect It uses a JPL C code to read and manage the ephemerides file DE405 It also uses NOVAS Naval Observatory Vector Astrometry Subroutines a C code for the computation of a wide variety of common astrometric quantities The JPL Planetary and Lunar Ephemerides can be dow
101. irgo data Two types of search are proposed here as e g one normal limited to the decay time of 10000 years with computing cost of about one teraflops with high sensitivity and the other low sensitivity limited to the decay time of 100 years with computing cost of about one half of the previous one In this second search the sensitivity is about a factor 2 lowet in h Normal Low sensitivity Hz 2 1 00 22 S 10368000 months 4 10368000 months 4 Tmin S 3 16 11 years 10000 3 16 09 years 100 At s 0 00025 0 001 0 004 0 016 0 00025 0 001 0 004 0 016 Vmax Hz 2000 500 125 31 25 2000 500 125 31 25 1000 4000 8000 16000 100 400 800 1600 Litt 4 00 06 4 00 06 2 00 06 1 00 06 4 00E 05 4 00E 05 2 00E 05 1 00E 05 1 04 04 2 59 03 1 30 03 6 48 02 1 04 05 2 59 04 1 30 04 6 48 03 2 64 14 2 64 14 1 65 13 1 03 12 1 14 13 1 14 13 3 57 11 1 11 10 2 28 17 5 71 16 1 78 15 5 57 13 9 86 16 2 46 16 3 85 14 6 02 12 Gflops 8 81 02 2 20 02 6 88 00 2 15 01 3 80 02 9 51 01 1 49 00 2 32E 02 Gflops tot 1 11 03 4 77 02 4 43 25 3 13 25 2 64 25 2 22 25 7 88 25 5 57 25 4 69E 25 3 94 25 64 64 step factor
102. l Times good introduction to the definitions of the Astronomical times is given here http www gb nrao edu rfisher Ephemerides times html TAI is the International Atomic Time UTC is the Cootdinated Univetsal Time UTC TAI number of leap seconds TDT is the terrestrial dynamic time It is not yet used for planetary motions calculations but it is only used to calculate TDB which takes into account relativistic effects TAI 32 184 UTC number of leap seconds 32 184 it is tied to the atomic time by a constant offset of 32 184 seconds 15 the Barycentric Dynamic Time approximately TT 0 001658 sin 0 000014 sin 26 seconds where g 357 53 0 9856003 JD 2451545 0 degrees and JD is the Julian Date UT1 is the Universal Time and it is a measure of the actual rotation of the Earth Hence it is not uniform Itis the rotation of the Earth with respect to the mean position of the Sun UTC is incremented by integer seconds leap seconds to stay within 0 7 seconds of UTI Then he difference between UT1 and is greater than this Planetary motions are computed using TDB UT1 should be used to evaluate Greenwich Mean Sidereal Time but UTC can be used as a good approximation of UTI In the code we can use either UTC or UTI by setting a flag Sidereal time is the measure of the Earth s rotation with respect to distant celestial objects 182
103. lassical sinccssissocsssesonsateatvosassndvessnassesdesosivasentadeckenossovsenensnadasendos 62 ru 02 ImplemehntatloBs nF Fiesta Pesos vt 63 Function Prototypes 69 Us r assigned 70 Performance ISSue ntu 71 R sultsof d 71 76 77 77 CLEATS SC 77 80 teacher OR tarts tah 80 Hough transform dot 81 THEO 81 Jt como dung dos ae ep la 81 82 SUPEI SOD Gite DOMINI B RU DATEN RUFEN ERN RENS SUA DUE 83 BASICS E AUN ono 83 Outline ofthe SUDEEVISOE Lab sa Sates 85 Implementation of the Supervisor scs dee tepore eed ieri 86 Candidate database and coincidences 87 The database Y s edad Ue THU AP a 87 The old database Old oni deat ende d ros 89 Type
104. lch channel selection array 0 excluded channel dt period n number of bins of the phase diagram pd mode 0 simple events 1 density normalization mode 1 1 mode 2 bin width s 2 amplitude mode 1 2 2 0 1 2 normal abs square long longitude for local periodicities local solar and sidereal natm number of harmonics 36 SFDB Theory The maximum time length of an FFT such that a Doppler shifted sinusoidal signal remains in a single bin is C 1119 Az SEU where and are the period and the radius of the rotation epicycle and is the T T max maximum frequency of interest of the FFT Because we want to explore a large frequency band from 10 Hz up to 2000 Hz the choice of a single FFT time duration is not good because as we saw 1 T max Vg so we propose to use 4 different SFDB bands Short FFT data base Band 1 Band 2 Band 3 Band 4 Max frequency per band 2000 0000 500 0000 125 0000 31 2500 Observed bands 1500 0000 375 0000 93 7500 23 4375 Max duration for an 2445 6679 4891 3359 9782 6717 19565 3435 Max len for an FFT max freq 9 7827 06 Length of an FFT max freq 8 3886E 06 Length of the FFTs 8388608 4194304 2097152 1048576 FFT duration 2097 15 4194 30 8388 61 16777 22 Number of FFTs 9 5367 03 4 7684 03 2 3842E 03 1 1921E 03 SFDB storage GB 160 00 40 00 10 00 2 50 Storage for sampled data GB 80 00 Total disk storage GB 29
105. led by the Matlab compiler and run on the Grid environment The input gravitational antenna data on which the PSS software operates can be in various formats as the frame Ligo Virgo format the R87 ROG format or the sds format that is one of the Snag SFC formats The data produced at the various stages of the processing are stored in one of the Snag SFC formats The candidate database has a particular format There ate some procedure to prepare data for processing There is a basic check for timing and basic quality control A report is created Then the Short FFT Database SFDB is created This is done in different way depending on the antenna type For interferometric antennas it is done for 4 bands obtaining 4 SFDBs For bar antennas it is done for a single band The SFDB contains also a collection of very short periodograms It has many uses in particular it is used for the time frequency data quality From the SFDB the peak map is obtained it is the starting point for the Hough transform The Hough transform the incoherent step of a hierarchical procedure that is the main part of our procedure is normally run on a Beowulf computer farm A Supervisor program creates and manages the set of tasks The Hough transform produces a huge number from hundreds of millions to billions of candidate sources each defined by a starting frequency a position in the sky and a value of the spin down parameter These are stored i
106. ll periods select End selection If a start is followed by another start it is cancelled If a stop is followed by another stop it is cancelled If one begins with a stop the first start is the very beginning If one ends the selection with a start the last stop is the end of data You can zoom at will between steps 32 Search for events MatLab environment Filtering in a ds framework The ev ew structures The event management is done by the use of the ev ew Snag structure see the programming guide Snag2_PG pdf An event is defined by a set of parameters like the starting time of the event in days normally mjd the time of the maximum in days normally mjd the channel the amplitude the critical ratio the length in seconds The difference between the ev and ew structures 15 that the first describes a single event so set of events is array of structures the second describes a set of events The ew structure is normally more efficient but the ev structure 15 more rich it can contain also the shape of the event The two function ew2ev and transform one type in the other losing the shape if present set of events 15 associated to a channel structure that describes the channels that produced the events constituting a new event channel structure evch There is a number of auxiliary functions to manage events chstrzcrea chstr nch
107. m is the heaviest part of the data analysis procedure and it is then important to have a software as much efficient as possible The present implementation of the Hough transform computation is the fastest we have developed up to now Possibly faster versions will be released in the future In the next subsections results of the timing and profiling of the library are given and discussed Results of gptof Machine grwavcp romal infn it Processor Intel Pentium 4 L2 cache 1MB RAM 1GB The code has been compiled with options pg 03 ffast math funroll loops fexpensive optimizations 71 Program parameters Search frequency 330Hz Number of spectra 4911 corresponding to 6 months of data RAD 0 non adaptive Hough map The following profile has been obtained with the unix linux tool gprof Flat profile Each sample counts as 0 01 seconds cumulative self self total time seconds seconds calls s call 11 name 48 02 5 33 5 33 103601 0 00 0 00 DrawCircleNext 45 95 10 43 5 10 106354 0 00 0 00 DrawCircleNextOpp 3 42 10 81 0 38 6270 0 00 0 00 DrawLeftCircle 1 71 11 00 0 19 1 0 19 0 19 writeMaps 0 36 11 04 0 04 1 0 04 10 53 houghBuild 0 27 11 07 0 03 1 0 03 0 03 candidates 0 09 11 08 0 01 799 0 00 0 00 DrawCircleInit 0 09 11 09 0 01 1 0 01 0 01 hmd2hm 0 09 151 220 0 01 1 0 01 0 01 readPeaks 0 00 11 10 0 00 169 0 00 0 00 DrawCircleFin 0 00 11 10 0 00 1 0 00 0 00 gridInitialize 0 00 11 10 0 00 1 0 00 0 00 hough
108. n a database and when there are independent data analysis for different periods for different antennas a coincidence search is performed on them The survived candidates are then followed up to verify their compliance with the hypothesis of being a periodic gravitational source to refine their parameters and to compute other like polarization An impottant part of the package is the simulation modules This guides ends with a report of vatious tests done of some parts of this package More information can be found on the programming guides and other documents Snag2 PG PSS PG PSS astro PG PSS astro UG PSS Hough PG Supervisor PG Typical pipelines e Times are only indicative e Often the transfer time between computers is higher than processing time Basic e After each step check and pre analysis time should be considered and this could be much longer In the table grid operations are in red Matlab operations are in blue Task Program filein file time day Mb day out hrec Extract tcsh frame frame 10 min 1500 extraction frame to hrec sda sds frame_util frame 10 1400 files conversion files sds reshape sds reshape sds sds 15 min 1400 sfdb crea_sfdb sds Sfdb 30 min 3000 peakmap crea peakmap sfdb p08 15 min 450 p082vbl 08 to vbl by p08 30 300 convert peakfiles cleaning mask cl
109. n different platforms is to compute the equivalent number of clock cycles needed to increase by 1 the number count in a given pixel of a Hough map In the following table some results are reported for different values of the search frequency f The other parameters are the same as those given at the beginning of the previous section Pentium 4 Pentium 4 Xeon Opteron 64 bit CPU GHz 2 8 2 8 2 4 2 2 1 2 1 1 512 1 compiler gcc 3 2 6 1 gcc 3 2 gcc 3 2 f 20Hz 40 34 41 29 f 80Hz 43 34 64 32 f 320Hz 69 57 79 70 76 Appendix In this Appendix we describe two programs that are to be meant as part of the PSS hough library one crea input file produces the input files used by the Hough program starting from the full peak map file the other create db collect candidates produced by the Hough jobs in a database that is used in the next steps of the analysis crea input file This is a program developed to produce input files for the Hough program It starts from the time frequency peak map and produces a set of files each being a portion of the original peak map covering a small frequency range This is needed for two reasons 1 ifinput file for the hough program is put in the grid sandbox and then is copied to the working node at submission it cannot be larger than a few MB at most to avoid network problems 2 even if input to the Hough program is already at the site
110. n order to take into account the source spin down At the end we have a Hough map derivative HMD which is then integrated to produce the final Hough map For each assumed value of the source spin down and for each soutce reference frequency a HM 15 obtained An HM 15 simply an array of int or float for the adaptive containing the number count of each pixel in the sky Selection of candidates In each HM pixels where the number count is above a given threshold are then selected as candidates Candidate parameters are written in candidate files From the user point of view the library takes at input a binary file containing a portion of the time frequency peak map and produces a binary file containing selected candidates The user can choose a number of parameters which will be described in detail in the next section Each candidate is described by a set of parameters frequency position in the sky in ecliptical coordinates first order spin down value critical ratio depending on the number count of its sky cell in the Hough map which it has been selected from and on the mean and standard deviation of the number count mean value of the number count of the corresponding Hough map and adimensional amplitude of the corresponding gravitational wave All these quantities are expressed as integer numbers of unitary steps which value is chosen as described in the following table Frequency 10 Hz fixed Position recomputed ev
111. ncy 998 169 600 F 550 500 ta frequency 400 F 350 frequency days 170 9 frequency Duration 10 20 10 10 10 10 Amplitude 171 172 55 databases and data analysis organization General structure of PSS databases The general organization of the PSS databases i e the folder tree is the following First level antennas and general metadata general server and analysis e Second level data categories and antenna metadata Data categories Sd sampled data h reconstructed or equivalent sfdb sfdb data equalized spectra pm peak maps cand candidate metadata antenna metadata analysis analysis reports e Third level optional different data types e Other level optional internal organization of sub databases The naming of files is such that they are alphabetically ordered for the basic key normally time A possible implementation is the following AntennaName_DataType_StartTime_Characteristics FormatExt where e AntennaName for example Vir Nau Exp e DataType for example raw m hrec 173 sfdb StartTime in the format YYMMDD hhmmss Characteristics depends on the category e FormatExt depends on the data format frame r87 sds The pss directory structure 174 K pss as
112. ning the computer information appears d SF_Prog C NET PSS PSS_bench Release PSS_bench exe Jof x Computing System Keon 2 66 GHz Operative System MS Windows XP Clock Frequency 2658 RAM 2948 Mbytes Disk 5051 3 2 256 Number of CPUs CANT Computing power 2658 Mflops Nodes 21 System identif 2 sf If it is not correct please upgrade local_header h lt CR gt The computer information resides in a header named local_header h and should be changed anytime that the benchmark is installed on a new computer It should be re compiled and linked for different configuration important is the no optimization Then the main menu appears It shows the different classes of benchmarks that can be run 133 Computing power 1680 CPU Nodes System identif sf If it is not correct please upgrade 1_ lt CR gt C benchmarks Basic benchmarks FFT benchmarks Hough benchmarks Coherent procedure benchmarks Basic report FFT report Vector report Hough report Disc report 1 2 3 E 6 8 9 2 Exit Which item 2 Then the menu of the particular class of benchmark appears For example for the basic benchmark 5 Which item 1 Basic benchmarks chosen Basic benchmarks Integer Float Float new Double Sine function Vectors Crazy Vectors Disk access Existing file acce
113. nloaded from the site ftp navigator jpl nasa gov pub ephem export by choosing the appropriate way depending on the operating system unix is the directory for unix users and ascii for non unix users The instructions written is the description file linked at http ssd jpl nasa gov eph info html ot directly at navigator pl nasa gov pub ephem export usrguide Each ascii file contains 20 years of data while each unix binary file contains 50 years of data Anyway we have experienced some problems with the binary UNIX files while ASCII files work properly even on an UNIX machine Then the procedure we have used to download the ephemerides DE405 for the years 1980 2020 is download from the directory ASCII the files ascp1980 405 2000 405 e download from the directory FORTRAN the code asc2eph f which must be used to convert the ASCII files into binary files and to merge them to form a single ephemetis file Then inasc2eph f set the NRECL parameter to 4 compile and link 977 asc2eph f 477 o asc2eph out binmerge o e run the code with the following syntax on Unix cat header 405 ascp1980 405 ascp2000 405 asc2eph out on windows using command console copy header 405 ascp1980 405 ascp2000 405 infile 405 andthenrun asc2eph out infile 405 181 This procedure produces binary file jpleph We have renamed it to Jpleph 405 Theory Astronomica
114. or sfdb09 files created before 15 Apr 2010 use sfdb09 band2sbl rad2 99 Create uniformly sampled gd corrected for source Doppler and spin down This step is done by g vv band recos source filein nnO gdname whete source frequency alpha delta source Hz deg easy source 0 gt no correction source structure fine source filein input file sbl file created at the preceding step nnO imposed reduced fft length a sub multiple of original fft length gdname name ofthe saved gd or nothing or 0 g output gd detector velocity in the SSB in c units pp detector position in the SSB in light s The data produced are complex In the case of fine source well known source it uses the parameters of the source Snag structure Vela case a 1 288358812083750 002 d 44 823645805527782 a 49 680000000000000 d 29 899999999999999 pepoch 54157 f0 22 382394913120621 410 3 125054539948422e 011 9910 2 524864601014520e 021 fepoch 54157 t00 51544 eps 1 psi 0 h 1 000000000000000e 023 snr 1 fi 0 coord 0 chphase 0 100 for every the structure in practice frequency position epochs is updated by the function sourupd new_posfr sour t The position of the Detector the SSB is computed at each time sample by the values of the position and velocity at the central point of the data chunk that produced SFT this information i
115. ot allowable spots The cut away procedure simply change to negative the sign of the frequency of the peaks In practice the procedure starting from the dirty vbl is the following Functions Script create the peakmap file list with e g dir vbl s b gt list _VSR2_pm txt and edit to eliminate blank lines etc edit and execute basic m this creates the PMC files peakmap maps This should be done 5 8 sub bands each time create the PMH files list edit and execute PMH fixing the threshold thresh this creates the veto array and the vetocell cell array apply vbl this creates the new vbls that have anyway the full information of the cut peaks concatenate new vbl files if needed with conc check the effect with basic tfclean vbl read peakmap integralF fromhist plot peaks p2rgb p2rgb colormap color points ana peakmap basic and ana peakmap basic 9 peakmap basic Dfr 5 VSR VSR4 EDIT for 1 17 26 5 EDIT band i 1 i Dfr strtit sprintf s band 4 1 4 1f VSR band savestr sprintf PMH s 03d 03d VSR band 57 tim vel vbl read peakmap list VSR l pm txt band 0 0 eval x y v savestr plot peaks A 1 0 1 strtit eval save savestr savestr end ana PMH Analyzes and creates time frequency veto structu
116. peak amplitudes o peaks sparse array o snr the total snr array Applying this procedure to a plot like 45 10 1034 TNNT eo _ TEM SS ee 1038 ER RR JN mE 1038 TES EU AER ncs isse eer m M NM 1442111 0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000 We have with inv 0 46 And with inv 2 47 x 107 i fr s peaks snr sp find fpeaks based on the gd findev nl function idea that is adaptive mean computation non linearly corrected with o in input gd or array o tau AR 1 tau in samples o maxdin maximum input vatiation to evaluate statistics o maxage max age to not update the statistics same size of tau o thr threshold o inv 1 gt starting from the end 2 gt min of both 48 index of the peak of the snr array o fr peak frequencies o snri peak o amp peak amplitudes o peaks spatse atray o snr the total snr array Applied to the same spectrum obtains 10 10 10 10 10 10 10 55 3 65 3 7 375 3 8 3 85 3 95 4 x 10 49 Rough cleaning procedure These procedure is setup to eliminate spu
117. peaks in the periodograms obtained from the SFDB incoherent step the peaks are the input of the HT candidate selection candidates in the output of the HT are selected coherent step initial data are corrected longer FFTs are calculated and a new peak map is produced Steps 2 5 ate repeated until the length of FFTs is equal to the total observation time The SV enters in peak selection after peak selection a service running on the master node data prod will produce input files for the HT each input file will consist of lists of peaks corresponding to a given frequency band plus the components of the detector velocity vector incoherent step coherent step in these phases a service job sub is dedicated to job submission and workload management on the farm nodes Note concerning the coherent part of the analysis that at least the first coherent step will be performed on several processors because the computing power needed though small with respect to what we need for the first incoherent step is not negligible candidate selection after candidate selection a service running on the master node data out copies output files two kinds of files one containing candidates and the other containing information on the jobs on the storage steps 2 5 a service running on the master is dedicated to the monitoring of farm nodes and of PSS jobs a setvice running on secondary masters is dedicated to the monitoring of the master node
118. ptab were source antenna and data are the pss structures and doptab is the correct Doppler table see Doppler effect page 123 The data are simulated in chunks during which the frequency and amplitude is standard but the chunks can also last a single sample it is reasonable that last at least one or more seconds The produced chunks can be added to real data or data simulated with sim sds or other 109 Fake source simulation Fake sources can be simulated and added to real data This is important to define the sensitivity of a set of antenna data Fake sources ate created with the MatLab function fcand pss creasourcefile sourstr outfile where sourstr structure array for source creation each structure contains number of sources fr min max frequency at epoch 2000 1 1 0 0 581 min max first spin down parameter Hz day sd2 min max second spin down parameter Hz day amp min max amplitude eps min max contents of linear polarization psi min max linear polarization angle alpha max alpha equatorial coordinates delta min max delta equatorial coordinates outfile output file fcand candidate matrix coordinates are converted to ecliptical all angles are in degrees Sources are frequency sorted An example is the following Fake Source List 1 100 sources fr 50 000 1050 0001 91 0 000 0 000 542 0 0000 0 000 amp 0 000100 1 000000 eps 0 00
119. r assigned parameter aun asas si nea RR LG 70 lesson 71 fo A E 71 76 77 I RE I anaes TT E ESI AESAAT 77 Introduction This is the User Guide for the PSS_hough library PSS hough is a C library which performs the first full sky incoherent step the hierarchical procedure developed by the Virgo Rome group for the search of periodic gravitational signals in the data of the Virgo interferometer The input consists of a peak map file the output of a candidate file performance reason the program does not use as input the whole peak map produced in the preceding step of the analysis but small portions of the peak map which are produced at the beginning with the program crea input file which is described in the Appendix Once the candidates have been produced they are organized in a candidate database through the program create db which is also described in the Appendix Theory The Hough transform is a robust parameter estimator for patterns in digital images It can be used to estimate the parameters of a curve which best fits a given set of points The basic idea 15 that of mapping the data into the parameter space and identifying the For instance suppose our data are distributed as straight line In this case the parameters to be determined are the slope m and the inter
120. res e conc vbl files Effectiveness of the procedure on peakmaps Using check PMH Resolution 12 hours 0 01 Hz Threshold 80 On VSR2 blue all red clean 241748723 all peaks 206402619 cleaned perc 0 853790 X 40 VSR2 data peaks vs time 0 H H 0 50 100 150 200 days x 10 VSR2 data peaks vs frequency 8 T T T T 7 59 VSR4 blue all red clean 122104240 all peaks 102300423 cleaned perc 0 837812 X 10 VSR4 data peaks vs time 0 20 40 60 80 100 days x10 VSR4 data peaks vs frequency 0 0 20 40 60 80 100 120 140 Important it doesn t cancel gravitational signals Case of the Pulsar 3 60 before Peakmap VSR2 band 105 0 110 0 Peakmap map histogram VSR2 band 105 0 110 0 200 days after Peakmap VSR2 band 105 0 110 0 Peakmap map histogram VSR2 band 105 0 110 0 100 120 140 160 180 days ut 110 100 90 80 70 60 50 40 30 20 10 100 61 Hough transform classical C environment MET T 62 Eu E M 62 cosa ao dd d VERAS ______________ _ 63 Function prototypes RES CEU VUE AH EPIRI maaan nue 69 Use
121. rious lines It is based only on the peak frequency histogram d data pss virgo pm c6 gt gt spfr sptim peakfr peaktim npeak splr peaklr mub sigpb ana peakmap 0 0 0 0 peakmap c6 1 vbl gt gt mask crea cleaningmask old peakfr 5 1 6181 gt gt clean piapeak mask peakmap c6 1 dat gt gt piapeak2vbl 4000 4194304 4194304 peakmap c6 1 clean dat peakmap c6 1 clean vbl 50 Frequency domain cleaning procedure These procedure is based on the frequency histogram of peaks and the a knowledge on lines basic function is mask crea_cleaningmask peakfr nofr nomfr yesfr thrblob thratt with peakfr peakfr peak frequency histogram gd nofr xcluded frequencies mat nfr 3 with centr freq band att band 0 gt no 0 gt only attention band nomfr xcluded frequencies and harmonics mat nfr 3 with centr freq band att band 0 gt no yesfr not excluded frequencies mat nfr 2 with centr freq band 0 gt thrblob threshold on unexpected blobs def 1 2 thratt threshold on attention band def 1 6 An example of use is by the procedure crea_mask crea_mask spfr sptim peakfr peaktim npeak splr peaklr mub sigb ana_peakmap 0 0 0 0 peakmap C7 vbl nofr read_virgolines 10 2000 21 12 01 1 1 950 nofr nlt 1 2 1 nofr nl 1 3 6 nofr nl 2 1 1002 nofr nl 2 2 1 nofr nl 2 3 4 nofr
122. rs of the cycle 75 Index by function name 12 DrawCircleFin 8 candidates 6 lutBuild 9 DrawCircleInit 13 gridInitialize 15 readInfoPeakmap 3 DrawCircleNext 10 hmd2hm 16 readParameters 4 DrawCircleNextOpp 2 houghBuild 11 readPeaks DrawLeftCircle 14 houghInitialize 7 writeMaps Comments It is clear from the profiling that most of the time is spent in the functions DrawCircleNext and DrawCircleNextOpp This is due to two facts 1 these functions are called a number of times equal to the number of circles which must be drawn The time spent in each call is about 50 microseconds 2 In these functions a large array the Hough map derivative is accessed in a random way and this results in a large number of cache misses If we could access it in a more regular way a gain in performances of at least a factor of 2 could be obtained We will try to reduce this bottleneck in the next releases of the library Compiling the library with the compiler performances improve by about 20 Similar results are obtained taking a search frequency in the highest frequency band 500Hz 2kHz For lower bands lt 31 25Hz 31 25Hz 125Hz the weight of the functions DrawCircleNext and DrawCircleNextOpp gradually decreases because the dimension of the Hough map derivatives becomes smaller and consequently decreases the number of cache misses A useful way to compare the code performances o
123. rum with the highest resolution Spectral filter Check the value with distribution 104 Some miscellanea Snag utilities for known source lfftout dtout bandrecos check dtout0 filein For the choice of the lfftout BANDRECOS CHECK checks for the best parameters for pss bandrecos dtouto output preferred sampling time filein reduced band sbl file pss_bandrecos parameter Shows many other parameters function g sds_bandextr 1 dt gdout 5 505 BANDEXTR extracts a band from an sds stream creates a gd and a new unique sds o AP file the first oe decade O NpssNvirgoNsdNsdsNVS 707 131945 sds 2 O NpssNvirgoNsdNsdsNVS 717 010625 sds 9 O NpssNvirgoNsdNsdsNVS 728 002625 sds 2 O NpssNvirgoNsdNsdsNVS 807 002625 sds good for very narrow bands LIMITATION the sds are supposed to contain a single channel g sds bandextr file t band dt lfftout sds file or list or 0 interactive choice the list is used in case of more decades It is created by dir b s gt list txt and then edited taking only the first file of the each R2Nsds 20090707NVIR V1 h 4096Hz 20090 R2Nsds h 20090717NVIR V1 4096Hz 20090 R2 sds_h 20090728NVIR V1 4096Hz 20090 R2Nsds 20090807NVIR
124. s 16 readInfoPeakmap 15 gridInitialize 13 main 103601 103601 106354 106354 1 1 799 799 Ty T 169 169 1 1 103601 103601 103601 106354 106354 106354 6270 6270 62710 main 1 houghBuild 2 DrawCircleNext 3 DrawCircleNextOpp candidates 8 DrawCircleInit readPeaks 11 hmd2hm 10 DrawCircleFin houghInitialize 4 9 12 14 houghBuild 2 DrawCircleNext 3 houghBuild 2 DrawCircleNextOpp 4 lutBuild 6 DrawLeftCircle 5 main 1 lutBuild 6 DrawLeftCircle 5 main writeMaps 1 7 799 169 169 169 houghBuild candidates 8 houghBuild DrawCircleInit 9 houghBuild 2 hmd2hm 10 houghBuild 2 readPeaks 11 houghBuild 2 DrawCircleFin 12 73 0 00 0 00 1 1 1 13 0 00 0 00 1 gridInitialize 13 0 00 0 00 1 1 houghBuild 2 14 0 0 0 00 0 00 1 houghInitialize 14 0 00 0 00 Lil main 1 15 0 0 0 00 0 00 1 readInfoPeakmap 15 0 00 0 00 1 1 main 1 16 0 0 0 00 0 00 1 readParameters 16 This table describes the call tree of the program and was sorted by the total amount of time spent in each function and its children Each entry in this table consists of several lines The line with the index number at the left hand margin lists the current function The lines above it list the functions that called this function and the lines below it list the functions this one called
125. s inside SFDBO9 and in the sbl files If the is not provided the programs computes the shortest possible value Eliminate bad periods This step is accomplished interactively by the help of the function gd absc typ g gdoutname The input data can be a gd or array The output is a gd or array typ 0 linear 1 logx 2 logy 3 logx y 5 ordinate or gd or array gdoutname name of the output gd The program is not standard the out is not produced directly by it but by a set of three other programs that communicate with it by global variables start sel 9399 m stop sel 9399 m and end null 9399 m The help appeats at start as 101 Data Selection Menu procedure select Start period and choose the starting point select Stop period and choose the stop point iterate the first two step for all periods select End selection If a start is followed by another start itis cancelled If a stop is followed by another stop itis cancelled If one begins with a stop the first start is the very beginning If one ends the selection with a start the last stop is the end of data You can zoom at will between steps When a petiod is selected to be nullified it appears in red Figure 6 gd Plot File Edit View Insert Tools Desktop Window Help Data Selection mem 4 4 4 4 2 4
126. s monitoring SV periodically checks the status of all CN and produce report master status monitoring on each secondary master a service periodically checks the status of all the mastets of greater rank and if no active master exists the node where it is running becomes the new primary master and start farm management input data production The active SV produces input data on the master job submission The active SV starts the PSS job creates a PSS job folder on the SE and manages the workload of the CNs job status monitoring The status of a PS S job is periodically monitored querying the node s where it is running saved on a file and distributed among all the secondary masters This is needed if the active master crushes and must be replaced by another one The status of a PS S job contains the information listed in ref A output file copy The final results of each N job are stored in the PSS job folder on the SE a copy is kept also on the node where the job run ref A the status of a PSS job contain the following information the name of the submitting machine the master the submission time on which CNs one ot more and which of these if any have the sleeping SV the workload distribution i e the N jobs name submission time input data all the important events end time of a IN ob and the exit status crashes problems these information are collected in a file and also in a C structure
127. scanf s filepeak PEAKMAP fopen appo dat w appo file for the peakmap AREST fopen arest dat w printf Do you want to create the file with data in time domain 1 Nn scanf d amp go2time if go2time 1 printf file time domain output ascii scanf Ss filetime TIME fopen filetime w printf Time domain file opened s n filetime printf Do you want to create an ascii file with ALL data in freq domain one 1 2 spectrum after the other 1 scanf d amp write filefreq if write filefreq 1 printf file frequency domain output ascii n scanf s filefreq FREQUENCY fopen filefreq w printf Frequency domain file opened s WMn filefreqg istart 0 printf FFT number to be written on the two test files 0 first one 2 n scanf d amp iw for k 0 k lt nmax k printf FFT number k 1 n k t1 errorcode readfilesfdb gd gd short header param k iw amp fft read if errorcode 1 printf Error in reading data into SFT file n printf Lette fft numero n k if go2time 1 printf gd gt n time n gd time dx 514 sf n gd gt n gd_time gt n gd_time gt dx puts ATT Time data do not contain the normalization from fft 2 to spectrum because this depends on puts the sampling time which is known and the number of data on which YOU will then to the FFT which is unknown puts zz fft data
128. so some useful routines iok piaspet2sbl folder piafilelist filesbl extracts the very short spectra from the Pia SFDB files folder folder containing the input pia files piafilelist pia file list obtained with dir b gt list txt filesbl file sbl to create 40 Time frequency data quality MatLab environment The time frequency data quality analysis is done using the set of power spectra created together with the SFDB b the set of power spectra created by the function crea ps c the high resolution periodograms obtained directly from the SFDB FFTs In the cases and b the time frequency map be imported in a gd2 with the function g2 sbl2gd2 sel file chn x y where file is the sbl file containing the power spectra for example an sfdb file chn the channel number in the sbl file x a 2 value array containing the min and the max block y a 2 value array containing the min and the max index of the spectrum frequencies From this gd2 an array can be extracted and on it the map2hist and imap hist be applied h xout map2hist m n par linlog creates a set of histograms one for each frequency bin of the various spectral amplitudes of that bin at all the times m is the time frequency spectral map n is the number of bins for the histograms linlog 0 1 determines if the histogram is done on the value spectral values of the spectra or on their logarithms par is the set of parameters to do
129. sources and a type 2 candidate db 96 Coincidence analysis coincidence file can be analyzed by the program psc anacoin It produces a psc structure whose elements are the following psc psc structure level level 1 candidate 2 coincidence candidate n n groups coincidence candidates width normalized coincidence width Sear search structure diffr frequency difference diflam lambda difference difbet beta difference sdl difference N number of candidates initim n initial time fftlen n fft length dfr n frequency quanta dlam n lambda quanta dbet n beta quanta dsdl1 n 541 quanta dcr n cr quanta dmh n mh quanta dh n h quanta nfr frequencies numbers nlam n lambdas numbers nbet n betas numbers nsdl1 n sdis numbers fr n frequencies lam n lambdas bet n betas sdl n sdis cr n crs critical ratios mh n mhs mean Hough Hu n hs h amplitudes 97 Coherent follow up Known source The basic pipeline for known source starting from the decade SFDBs is the following Extract the band in an sbl file From this create uniformly sampled corrected for the source Doppler and spin down Eliminate bad periods Eliminate big events Create the Wiener filtered data Do the maximum length power spectrum Apply the matched filter and the coherence Check the value at the expected frequency vs the distribution of the filter
130. ss Create junk file Main menu Which item The benchmarks started from the main menu can be also reports 134 reports Basic report Poo bench report on Tue Feb 17 12 07 17 2004 Computing System Xeon 2 66 GHz Operative System MS Windows XP Clock Frequency 2658 RAM 2048 Mbytes Disk SCSI 3 2 Cache 256 kB Number of CPUs 2 Computing power 2658 Mflops CPU Nodes Bs E System identif sf Comment Lower values are better results Basic tests Integer roughl rough2 sum product 11 615460 16 187220 4 598340 27 005280 Float roughl rough2 sum product 13 290000 6 219720 1 674540 1 249260 Double roughl rough2 sum product 19 934999 7 894260 0 850560 0 398700 Sines 203 337006 Vectors length 100000 gt rough sum 7 043700 19 084440 Crazy length 100000 gt no crazy crazy 4 173060 40 667400 FFT tests Fourl 54 664391 length 1048576 Hough tests LUT 12 260115 nalpha ndelta 720 360 135 FFT report PSS bench FFT fftw report on Tue Feb 17 15 07 42 2004 Computing System Xeon 2 66 GHz Operative System MS Windows XP Clock Frequency 2658 RAM 2048 Mbytes Disk 5651 3 2 256 2 Number of CPUS S 2 Computing power 2658 Mflops per CPU Nodes SW System identif sf Comment Lower values are better results length efficiency loss 102
131. t c lt file caption gt b lt DB base name gt with e file frequency width gt default 10Hz w DB name caption e g PSC DB 7 lt gt t Maximum spin down age at 2kHz yr gt default 4000 p Number of spectra default 556 Usage example 78 lOreate db f 49 735069 g 1048 576 d 1100 1 list cand sim olean txt q 2100 i PSC DB c7 sim clean c 27 sim clean BPSC DB GcT7 10 w new t 4000 p 556 The program is compiled with cc create db c o create db Function prototypes lm void readParameters int char float float char int char char char void createDBStructure float float char void populateDB char int char char char void extractDir int char char int computeDecina int int computeStep int long long int float float float int 79 MatLab environment pss explorer pss hough 80 Hough transform f f dot Theory The Hough transform Implementation As said in the previous section 81 Peak map cleaning The first step of the f f dot Hough transform is the analysis without Doppler shift that identifies the disturbance lines This is done for short runs and decades in case of long runs The found h peaks are archived in ASCII files one for each analysis e g one per decade in cas
132. t are different from Cutler s Hence the difference between our result and Cutler result is for the considered variable z 4 8 10 12 We recall that this comparison has been done not formally on August 2000 In particular we t know if Cutler now 15 183 still getting these numbers We have done this and other comparisons using TEMPO 184 Programming tips Windows FrameLib FrameLib 6718 To construct the FrameLib lib we have to comment the line include lt unistd h gt in FrlO c The functions dup open close write read Iseek are not defined They are used if it is not defined FRIOCFILE 185 Non standard formats SFDB FORMAT of the SFDB files Format of the SFDB files files are binary in blocks Each block has header 2 Averages ofthe short power spectra with time in units EINSTEIN Hz 3 short power spectra obtianed with subsampling factor which is written in the header They contain the amplitude which has to be squared to get the very short power spectrum Units EINSTEIN Sqrt Hz 4 FFTs real and imag part in units EINSTEIN Sqrt Hz The header contains all the relevant information which regards the detector and information on how the Data Base has been done Parameters written in the header are put in the structure HEADER PARAM which is defined in pss sfdb h To read the files one has to read in order HEADER Averages of the short power spectra with time
133. t name without extension default pss batch diary Saves the previous report in old and updates the rep fil Check and possibly edit the rep file It records the operation on a file Example of a record START convert peakfilesl at 21 Aug 2006 18 14 37 by sf convert peakfilesl 2 1 4000 4194304 4194304 peak100sfakeC7 p05 peak100sfakeC7 vbl 2 1 4000 4194304 4194304 peak100fakeC7 dat peak100fakeC7 vbl STOP at 21 Aug 2006 21 33 56 There is another procedure that is useful for documenting interactive and also batch work with Matlab that uses the Matlab Diary It saves all the inputs and outputs of the Matlab session diary out snag diary onoff filnam or simply diary onoff lor 0 to turn on or off if different or absent toggles filnam filname default diary xxxxx in the snagout directory 198 There is also an automatic Snag data analysis report creator function pub dir snag pub mfile typ outputdir or options mfile the m file to publish typ format doc html latex ppt xml outputdir or options in case of options the structure with elements format outputDir should end with default tmp showCode true or false imageFormat dir output directory It uses the Matlab publish function Data preparation Candidates Coincidences Issues On particular problems 199 Matlab procedures r sum Batch pro
134. te datum The zero in the case of mixed sign data is coded as 111 11 while 000 007 is the code for the number m 1000 00 is m 0111 11 is the maximum value and 111 110 the minimum 20 mean percentage error in the case of a gaussian white sequence is in case of Log16 better then 104 Also a linear coding is possible x mt b E Also in this case the coded datum 0 always codes the uncoded value 0 also if such a value doesn t exist In case of linear coding if the data are mixed sign really or imposed and X is 8 or 16 is a signed integer otherwise it is an unsigned integer normally in the first case m is 0 In case of data dimension X less than 8 4 2 or 1 the sub byte coding the logarithmic format is substituted by a look up table format In such case a look up table of 2 1 fixed thresholds t 0 lt k lt 2X 2 in ascending order must be supplied Data lt to are coded as 0 data between ti 1and coded as k and data greater than the last threshold are coded as 11 1 In the case of linear sub byte coding the coded data are unsigned Here is a summary of the LogX format 45 8 4 2 1 0 of bits Coding float 2 linear 2 linear linear linear linear constant 2 logarithmic 2 logarithmic look up look up look up Logarithmic coding can be done using X or X 1 bits for the exponent depen
135. ter than the second they defines the interval of rejection chstr stat ev evch statistics for events dd ev_dens evch selch dt n event densities o evch event channel structure o selch selection array fot the channels 0 exclude 1 include O dt time interval o number of time intervals ev plot evch type plots events type is simple amplitude colored length colored both stem3 amplitude stem3 length stem3 both CN OF de C5 po rs co Coincidences To study coincidences between events a set of functions is provided dcp ini3 len3 dens coin evch selch dt n type coinfun creates delay coincidence plot and finds coincident events ini3 len3 are the initial times and lengths dens is the event density if used In input o evchis an event channel structure 34 selch is a selection array with dimension of the number of channels that defines which channels are to be put in coincidence every channels can be 0 excluded 1 putin the first group 2 putin second group 3 putin both O dt is time resolution s 15 the number of delays fot each side O type an array indicating the coincidence type type 1 1 only event maxima 2 whole length coincidences type 2 1 normal 2 density normalized type 3 density time scale s for density normalization O coinfun if exists external coincidence function is enabled The coincidence function is gt 0 if the even
136. the detector that is it does not depend on the source position we have decided to store also this information The Einstein effect gives a contribution which is toughly speaking 5 orders of magnitude lower compared to the revolution and 3 orders of magnitude lower compared to the rotation On the contrary the Shapiro effect which is even smaller depends also on the source and thus cannot be evaluated at this stage There is a function in the PSS astro code to add the Shapiro effect if a very high precision is needed The Shapiro effect gives a contribution which is toughly speaking 3 orders of magnitude lower compared to the Einstein effect 9 The time is TT Terrestrial Time the form of MJD Interpolation from a table at 10 minutes one yeat about 10 MB in matlab format 3 gives rise to an error on each component less than 10 13 so obtaining a maximum relative etror of the order of 10 9 The problem for the exact computation of the Doppler effect 15 that it cannot be done in advance because the future Earth rotation 15 not completely known so there is an error on this component that is about 0 01 of the total motion that is for the position an error of about 300 m 1 us per each leap second The Doppler table used by the software is a 8 x n matrix with the first column containing the times in TT days then 3 columns containing the components of the position in light seconds and then 3 components with
137. the histograms ifm is mx2 array it contains the min and the max of the histograms o ifm 0 the bounds of the histograms are computed automatically by taking the minimum and the maximum of all the data 1 the bounds of the histograms are computed automatically by taking the minimum and the maximum of every bin hist h x y is used to plot the histogram map h is the histogram map x is the frequency value y the spectral values if the scale is unique Here are the two maps for the two cases of par 0 and 1 41 70 80 FL DU ENTE ta dim T WINE omen ve ele a 1000 2000 3000 4000 5000 6000 7000 8000 90 100 1000 2000 3000 4000 5000 6000 7000 8000 42 From SFDB we can obtain the peak map i e a time frequency map containing the relative maxima of the periodograms built taking the square modulus of the short FFTs To obtain the peak map the procedure is the following reada short FFT of the data from the database using this construct an enhanced resolution periodogram equalize the periodogram using for example the ps autoequalize procedure o find the maxima of the equalized spectrum above a given threshold for example 2 The stored data can be just 1 and 0 binary sparse matrix or also the values of the maxima of the not equalized spectra this in order to evaluate th
138. threshold typically 2 Win window type 0 1 gt pss t periodograms time V 3 petiodograms detector velocity PM map sparse matrix or t f spectrum matrix 117 High resolution simulation 118 Candidate simulation MatLab environment basic candidate simulation function is 8 crea pssfakecand that creates a candidate database The folder structure should exist It can be copied from the template that exists in the metadata sub directory The input data are o dircand the candidate root directory containing the folder structure containing up to 2000 files in 200 folders grouped in 20 parent folders o N the number of candidates to create o band wich band 1 2 3 or 4 o pss cand head pss candidate file header structure and simulation parameters This function no return values can be launched by the m file batch fakecreation to be edited that creates two pss candidate databases The creation of a pss database of 1 08 candidates takes about 7 minutes on a 3GHz two cpu computer 119 Time and astronomical functions MatLab environment Time In Snag there is a number of time conversion functions The basic time used in Snag is MJD Modified Julian Date Other time used are TAI and GPS time the form of time normally UTC time is also string or vector t s2mjd str and str mjd2s mjd str mjds2s mjd for multiple operations converts string time to mjd modified julian
139. tructure of all the blocks is the same but the length can be different gbl general block data format it is not a format but practically a sequence of superblocks each following one of the preceding formats it is a repository of data not necessary well structured for an effective analysis but good for storage exchange etc set of files can be internally collected i e ordered serially or in parallel using the internal file pointers for example subsequent data files or to put together different sampling time channels externally collected i e logically linked by a collection script file as it happens for internal collecting embedded in a single file with a toc at the beginning or at the end This is the case of the gbl files file can be wrapped by adding one or more external headers for example describing the computer which wrote the file 19 The SFC data formats are presented in Snag2 Programming Guide Snag2 PG pdf Compressed data formats These formats are not compulsory they are not used for the first development phase LogX format This is a format that can describe a real number float with little more than 16 8 4 2 or 1 bits X indicates this number of bits It uses normally a logatithmic coding but can use also linear coding and in particular cases the normal floating 32 bit format In the case that all the data to be coded are equal only one data is archived plus the stat
140. trum in Einstein 2 5 spini double spdf double splen int32 sp float splen values oo Data npeak times bin int32 ratio float xamed float mean of H 196 09 It is translated to vbl format by p092vbl or 092 3 or with the pss batch diary with the script cd y pss virgo pm VSR1 v2 conv 09 conc list txt conv 09 conc 3 list txt is the format p09 format Header nfft int32 sfr double sampling frequency 11162 int32 original length of fft divided by 2 inifr double Block header mjd double npeak int32 velx double v c vely double v c velz double v c posx double x c posy double y c posz double z c Short spectrum Einstein 2 5 spini double spdf double splen int32 sp float splen values oo Data npeak times bin int32 ratio float xamed float mean of H 197 Standard Reports the standard reports for the pss data analysis There are also automatic reports General In Matlab the submission of batch procedures should be carried out by the use of the function pss batch diary pss batch diary mscript user scriptin outdir repname mscript m file script to be submitted user user name scriptin gt 0 gt print script default outdir output directory default pssreport repname repor
141. ts are compatible the inputs are len1 len2 amp1 amp2 that are the lengths and amplitudes for the two coincident event It produces the plot of the delay coincidences and its histogram dcp in3 len3 vec coin in1 len1 in2 len2 dt n coinfun a1 a2 is one of the coincidence engines used inside coin It considers the length of the events dcp in3 vec coin nolen in1 in2 dt n coinfun a1 a2 is one of the coincidence engines used inside coin It considers the length of the events as dt corr evch dt mode computes and visualize the correlation matrix between all the channels mode 1 15 for symmetric operation mode 0 is for time atrow coincidence causality It produces also the map of the matrix search evchi dt identifies cluster of events and labels the events with the cluster index Event periodicities An impottant point in the event analysis is the study of periodicities This is performed by the following functions sp ev_spec evch selev minfr maxfr res that performs the event spectrum in input we need 35 evch event channel structure selch channelt selection array 0 excluded channel minfr minimum frequency maxft maximum frequency res resolution minimum 1 typical 6 period evch selch dt n mode long narm event periodicity study phase diagram in input we need evch event channel structure se
142. ts in measuring the frequency shift at a given time and a given spin down value and comparing it with the maximum allowed shift as a number of bins This value can be set by the reasonable value can be obtained in the following way and probably this computation should be done directly in the code The problem 15 that the 15 computed initially for a given frequency When the program analyzes a different frequency it should refer to computed for that Howevet as we will show the error done using the wrong LUT i e computed for a different frequency is small as long as the difference between the frequencies is small This means that we can use the same LUT for a small range of different frequencies To estimate this number let us compute the variation on the radius of circles as a function of the frequency and at the end let us impose it is less than some reasonable value Let us start from the Doppler formula 52 74 coso 67 where f is search frequency f is the frequency of the k th peak If we move to the search frequency fy Af keeping f fo fixed the circle radius will change as Je sin g Ag 2 8 fo Hence ped Af fo a The maximum radius variation takes place for a fixed f when sing is minimum f is maximum i e for circle of small radius that is for thick annuli anelli
143. uble double char double int double float float int int int float void DrawCircleNext float int int int int float float houghD t 92 69 void DrawCirclePoles float int int int int float float houghD t void DrawCircleInit float int int float houghD t void DrawCircleNextAdap double float float int int int int float float houghD t void DrawCircleInitAdap double float float int int float houghD t void DrawCircleNextOppAnomalousAdap double float float int int int int float float float houghD t void hmd2hm houghD t hough t hough t void writeMaps houghD t hough t hough t int candidates FILE float double short int hough t void houghRad char float fl 43 oat radpat interf oat radpat interf eclip oat radpat bar oat radpat bar eclip double localSiderealTime double void print cpu time void get sys info void get uname void hmcum hough t void hmcum2 hough t void writehmcum User assigned parameters These are the parameters that the user can set at command line tau f0max 4000 thr 3 8 FOstart 49 735069 format 1 nfreq 100 rec fact 4
144. uctures All the functions that need antenna or source information use the following structures source structure interactive function SOUrzi source o a right ascension degrees at pepoch o d declination degrees at pepoch velocity marcs y o vd velocity dec marcs y O eps contents of linear polarization psi angle of linear polarization respect to the source meridian o 10 frequency at fepoch o frequency first derivative epoch 0 frequency vatiation Hz s o frequency second derivative epoch 0 40 variation Hz s 2 snf signal to noise ratio O pepoch position epoch o fepoch frequency epoch antenna structure interactive function Sourzi antenna type 1 interferometer 2 lat latitude degrees long longitude degrees heig heigth m azim azimuth degrees incl inclination degrees 5 Defined as the difference between semimajor and semiminor axis of the normalized polarization ellipse 122 The computation of the motion of the Earth respect to Solar System Barycenter is computed by pss astro an adaptation and enhancement of a code from JPL and NOVAS software Among the other possible uses of pss astro there is the production of a table depending on the location of the antenna and the year that contains the velocity vector of the detector at certain times The program in C is crea table
145. uter benchmarks pss math basic mathematical routines pss_sfdb for short FFT data base and peak maps creation and management pss hough for hough transform pss_cohe for the coherent step of the hierarchical search pss_ss Hough tasks management and supervision The parts that are developed in this environment are labeled C environment 18 SFC formats Basics basic feature of the file formats collected is the ease of access to the data The ease of access means the softwate to access the data consists in a few lines of basic code the data can be accessed easily by any environment and language the byte level structure is immediately intelligible nounneeded information is present the number of pointers and structures is minimized structure fits the needs the access is fast and possibly direct the need for generality is tempered by the need for easiness The collection is composed by 595 simple data stream format for finite or infinite number of equispaced samples in one more channels all with the same sampling time simple block data format in a more general case a block can contain one or more data types any block have the same structure i e the sequence and the format of the channels is the same and the same length i e the number of data in a block for a certain channel is always the same vbl varying length block data format where the s
146. variable It best applies to sets of homogeneous numbets Let us divide the data in sets that are enough homogeneous as a continuous stretch of sampled data The conversion procedure computes the minimum and the maximum of the set and the minimum and the maximum of the absolute values of the set checks if the numbers are all positive or negative or if are all equal then computes the better way to describe them as a power of a certain base multiplied by a constant plus a sign So any non zero number of the set is represented by x m b ot if all the number of the set have the same sign x S m b where S is the sign one bit m is the minimum absolute value of the numbers in the set b is base computed from the minimum the maximum absolute value of the numbets of the set is the positive exponent 15 or 16 bits for Log16 7 or 8 bits for Log8 and so on The coded datum 0 always codes the uncoded value 0 also if such a value doesn t exist m b and a control variable that says if all the number are positive negative or mixed are stored in a header The data bits contain S and E or only E minimum and maximum values can be imposed externally as saturation values In case of mixed sign data in otder to have automatic computation of m and b an epsval a minimum non zero absolute value should be defined If this is put to 0 this value is substituted with the minimum non zero absolu
147. where the job will run it proves not efficient to open several times the full big peak map and read the small portion that a single job will analyze The frequency range covered by each input file is chosen on the basis of a compromise between the need to not have too many files to be processed and to not have jobs lasting too much The choice adopted in the analysis of and C7 during 2006 has been 0 1Hz for and 0 5Hz for C7 For C6 this means more than 10000 jobs to cover the frequency range 50 1050Hz and for C7 2100 jobs over the same frequency range Input parameters file with the list of peak map files gt f lt overall start search frequency g start search frequency for the current job u numbet of files to be produced q number of frequencies covered by each file gt o location of output files 5 lt type gt normal use input peak map default flat all peaks random uniform peak distribution Usage example lOrea input file 1 lista 017 olean txt f 49 735069 g 48 609733 u 2100 q 500 o home palomba InputData c7 clean prova create db TT This is program developed to perform 2 operations using candidate files produced the Hough jobs 1 creation of the candidates database structure 2 fill the database with candidates files candidate database structure is described in the PSS UG to which the reader can refer to for more details Here we simply give the basic inform
148. ype 1 DETV 3 double 2 SHORTSP length of short spectrum sh sp frequency step float 3 INDEX length of short spectrum sh sp frequency step int 4 PEAKBIN length of half FFT frequency step int 5 PEAKCR float starting from November 2010 There 6 channels the velocity of the vector Cartesian in the Ecliptic reference frame fraction of c the short spectrum the index of the peaks array for direct access to PEAKBIN the frequency bins of the peaks integer can be negative for vetoed peaks the equalized peak values the mean for the peaks 24 The channels have the following parameters CH name lenx dx type 1 DETV 3 double 2 SHORTSP length of short spectrum sh sp frequency step float 3 INDEX length of short spectrum sh sp frequency step int 4 PEAKBIN length of half FFT frequency step int 5 PEAKCR float 6 PEAKMEAN float Hough maps The data are stored in SBL or VBL files The parameter to be stored in each block containing a single Hough map are length of the record the parameters of the hough map amin da na dmin da nd the spin down parameters nspin spinl spin2 the number of used periodograms and the type interlaced windowed the initial times and length of each e the type and the parameters of the threshold PS candidates and Events
149. ype what 1 initial frequency of the full band Bf2 1 final frequency of the full band 10 1 supposed initial unshifted frequency 1 first derivative 1 fO second derivative 1 band f0 source f0 fft df natb 3 natural sub band working band knatb 1 widening factor of natural sub band bf1 3 sub band initial frequency bf2 3 sub band final frequency 148 5196 structure parameter class type what 149 tfmap structure parameter class type what 150 tfpmap structure parameter class type what 151 hmap structure parameter class type what 152 cohe structure parameter class type what 153 ss Structure parameter class type what 154 candidate structure parameter class type what 155 event structure parameter class type what 156 computing structure parameter class type what 157 The Log Files Any production job produces a log file Th

PSS User Guide - To Parent Directory

Contents

Download Pdf Manuals

Related Search

Related Contents