Документ взят из кэша поисковой машины. Адрес оригинального документа : http://xmm.vilspa.esa.es/calibration/documentation/epic_cal_meetings/200804/kirsch_08_04_09_tj_mgfk.pdf Дата изменения: Tue May 13 17:49:29 2014 Дата индексирования: Wed Apr 13 13:42:46 2016 Кодировка:

Time jumps -end of story

M.G.F. Kirsch1, E. Kendziorra2, M. Freyberg3
2

ESA XMM-MOC, ESOC Germany IAAT, University of Tuebingen, Germany 3 MPE Garching, Germany
1

Marcus G. F. Kirsch
XMM-Newton Page 1


menu · time jumps - recap situation · · · · refined frame times refined jump search SAS algorithm proposal for SAS setting list of still present time jumps

· lets randomise the times now

Marcus G. F. Kirsch
XMM-Newton Page 2


pn time jumps - definition

Photon a

Photon b

CCD frames

n*frame time

If the time difference between the two arrival times of photon a and photon b is not a multiple of the frame time TIME JUMP
Marcus G. F. Kirsch
XMM-Newton Page 3


SAS time jump corrections

· SAS algorithm tries to detect and corrected time jump · not found jumps can be spotted by EPIC IDT SW at ESAC (now implemented on generic epicmon account · list of non SAS found / corrected time jumps is provided
Marcus G. F. Kirsch
XMM-Newton Page 4


how to detect a non SAS corrected jump

· take all frame times in array: T · =(T(i+1)-T(i))/frametime · =difference of to next full frametime · plot () · if >tolerance time jump
0

Time jumps
Counting mode


Marcus G. F. Kirsch
XMM-Newton Page 5


archive processing
· processing of all public and non public available archive data at ESAC --> implemented frame times are not correct --> refine frame times using this analysis before time jump analysis can be followed up
FF LW SW

Marcus G. F. Kirsch
XMM-Newton Page 6


relative difference of measured frame time old new

· Oscillator stable, otherwise relative differences should show same shift · Different slopes for different modes --> sequencer clocks have not been counted fully correct --> errors < 10-5 frame times per frame (< 20 clocks, 1 clock = 40 ns) · Slight drift in frame times (see MJF Escorial proceedings) Marcus G. F. Kirsch · refined frame times on that basis not violating XMM-Newton quantumization of oscillator clocks Page 7


search for jumps · with that correction still residual drifts caused by temperature changes and aging of the quartz · can mimic in case of large event differences (counting mode) time jumps

Marcus G. F. Kirsch
XMM-Newton Page 8


refined (non SAS) search algorithm

· take individual drifts into account · aequivalent to search with variable tolerance depending of event difference

Marcus G. F. Kirsch
XMM-Newton Page 9


ton archive data until revolution 1250 we could lower the number of remaining time jumps in all XMM-Newton data 16 % to 2.2 % using SAS_JUMP_TOLERANCE = 22.0. Table xx is giving an overview of the improvements:

status
Table 1 non SAS

detected time jumps in %

Mode Old FT + old algo New FT + old algo New FT new algo FF 5.8 0.6 1.9 eFF 20.8 20.2 2.3 SW 37.2 12.5 6.6 LW 39.0 1.6 1.4 Timing 12.0 1.9 1.1 Burst 52.9 4.2 3.2 Sum 15.6 4.9 2.2
SAS_JUMP_TOLERANCE 22

· Proposal: ­ switch on new algo default in next SAS ­ randomise times after further checks by MJF ­ if no checks done by MJF with 3 month --> randomise
Marcus G. F. Kirsch
XMM-Newton Page 10


Marcus G. F. Kirsch
XMM-Newton Page 11