Question ID = 0092
Question ID = 0092
Assigned Member (担当者)N. IsobeName and Institute (質問者/機関)Dr. PAIZISPAIZIS Ada (INAF-IASF, Milano)Question (質問)(1) I have recently started the analysis of Suzaku data and have read the user manual (both the ABC guide and the Suzaku Technical Description). While everything seems to be very clearly written I could not find any reference to the bkg spectrum rebinning. I.e. the xisresp file is rebinning the source pha and rsp but for the bkg rebinning we are supposed to do it manually using rbnpha, is this correct? It could be useful to add this information, I understand that it is pretty intuitive but I had to go into the perl program (xisresp) to see how to rebin the background according to the mode chosen (fast,medium ecc). It could be useful to have a pragraph with the channel selection to be used in each case, currently this is not there. Or at least I could not find it..(2) I have another question as well: the xisresp tool is producing the .rsp file (rmf and arf merged) but as far as I can see this is not working for addascaspec that expects rmf and arf separately (can you confirm?). If you confirm, then basically in the guide the procedure says to use a tool that produces output that is not compatibile with what is written in the folowing. Since I could not get addascaspec to work with .rsp (but maybe this is my mistake) I changed xisresp and made it produce rmf, save it, rebin it and then produce arf and save it. Then I could use addascaspec. (3) I have another question about XIS data analysis, sorry. I have extracted XIS lightcurves in the standard way, using xselect. But when I try to subtract the bkg lcurve to the source one with lcmath I get strange results. to make a simple test I took the source+bkg file (in attach) and I tried to subtract it to itself expecting to obtain about 0cps but this is not the case, the resulting lcurve is around -2 -3 counts/sec. I tried the same simple test with a PIN curve and things work correctly so my question is: are you aware of any keyword or so that could be "disturbing" lcmath? Answer from(回答の文責)R. Fujimoto (Suzaku Help Desk)Answer (答え)(1) Yes, you are right. Please manually rebin the background spectrum using rbnpha. If you use "fast", for example, prepare a file which contains the following contents:0 699 2 700 2695 4 2696 4095 8 Here we assume the name of this file is chanfile.txt. Then, run rbnpha with the following arguments % rbnpha infile=< original background pha file > binfile=chanfile.txt outfile=< output filename > properr=yes If you just type "xisresp", you will see the help which tells you how the PHA file is rebinned. (2) Currently, addascaspec requires arf file, and the manual is not consistent yet (as of Mar 20, 2008). Update of addascaspec is scheduled soon, which can handle rsp without arf. Until then, please manually make rmf and arf. We are sorry for the inconvenience. Thank you for your useful comments. Note: Updated version of addascaspec is now available from 2008-03-21, at http://suzaku.gsfc.nasa.gov/docs/suzaku/analysis/watchout.html. (3) We checked your light curve and confirmed the phenomenon you pointed out. It seems that since lcmath applies deadtime correction only to background lightcurve, it gives negative rates. When we ran lcmath as > lcmath xis2_3x3_s.lc xis2_3x3_s.lc outfile.lc 2.01356755e+00 1 the result became nearly zero. Here, 2.01356755e+00 = 1/4.96630967E-01, 4.96630967E-01 is the value of keyword DEADC in the header of xis2_3x3_s.lc. We will report this problem (possibly a software bug) to the ftools team. For the time being, however, we recommend that you use lcmath with a scaling factor for input lightcurve as shown above. Answer from ftools helpFrom: Alex PadgettDate: Thu, 10 Apr 2008 I have tracked this problem down, and it is definitely a bug in lcmath. There is a release of HEASoft coming in the next few days, but unfortunately there is not time to test the solution before this. It will be fixed in the next release in mid-May (along with the documentation). There is a simple, yet un-elegant workaround for this problem: 1. Apply dead time corrections to each light curve (Rate and Error). 2. Set the DEADC keyword to 1.0 in each light curve. 3. Set the DEADAPP keyword to T in each light curve. This will produce a properly subtracted light curve. Status （詳細なステータス）2008-03-12 Accepted2008-03-14 Answer1 2008-03-20 Answer2 2008-03-28 Answer3 2008-03-28 Done 2008-04-16 Add answer from ftools help 
|