Here is my summary about my suggestions which values to choose, I first list the problems, then the strings. T0: ----- In general I suggest to AVERAGE (!) the measurements, which we trust. This means usually std,mtd and smd. In some cases we have to rescale r97 or ariels values. ALPHA METHOD: ------------- In case of alpha I think we should correct the std and mtd alphas (using the function alpha'= alpha - 25 * MIN( dt , 20.) (see fit in the macro) dt is my time difference of M-PE and 1-PE. (use ONLY for TWISTED PAIR OMs) After that we have to rescale std and mtd values. alpha'' = beta * alpha'(std,mtd) I suggest beta = 0.655 for string 5-10 and 0.764 for string 1-4 It may be even better to use only MD values, because there we know to make the slope fit with the right ADC. Pat (PC), please decide on that. INTERSTRING NORMALISATION: -------------------------- In the normalized values already on string 3 one observes a shift of a few ns between smd,mtd,std which is due to the normalisation on the first channel of string3, and different cableng in the Main daq. I suggest to normalize only on the MD value and shift the mtd and std values. If I look at the strings I get the following shift values: string: 1 2 3 4 5 6 7 8 9 10 mtd-smd: -12 -7 -6 -7 -15(*) -19 -15(*) -5(*) -5(*) -10(*) std-smd: -7 -6 -4 -3 -8(*) -15 -16 -2(*) 0(*) -8(*) std-mtd: +6 +2 +2 +2 +5(*) +4 +10(*) +5 +5 +5 (*) large errors >From this I conclude, that in addition to an error in the normalization (differences string 3) Different cable lenghts inside the main DAQ can NOT be neglected. Do we have a map or can we soemhow obtain how the channels were wired? At least a list, which swamp was where could help, since some channels of one string sit at competely different places/crates. I think using just mtd or std is NOT ok. Well, if we cannot find out these individual shifts, I suggest to SHIFT STRINGWISE: std=std+x, x=4ns for s=2,3,4,10 x=2ns s=8,9 x=8ns s=1,5 x=15ns s=6,7 mtd=mtd+y, y = x + 4ns for all strings. ARIELS DATA: -------------- In general Ariels method yields WRONG corrections for string 1-4 and improvements for string 5-10. My personal conclusion is: The method is powerful enough to detect a wrong tdc scale, as seems to be the case for the 97 calibration. So it more or less corects that. However, it turns out WRONG as soon as it corrects single OMs in the recursive mode. I now doubt, that it yield nothing else than rescaled 97 t0's. So my suggestion is,to test if a linear fit to the 97-98 T0 differences (see plots) and a rescaling of the old 97/T0 does not give similar or better corrections. A method like ariel is important to find systematic differences between MD and TD. We should develop this method better by using MC. SUMMARY ---------- Alphas: String 3: average mtd+std+smd OM 9 : mtd+std , no smd, OM 10: old r96 OM 17: dead String 1: average all, except OM 3,18 : smd only OM 19 : mtd,std only string 2:average all, except OM 14: only smd OM 8,12,19,20: dead string 4: OM2 :r96 OM1 : std and mtd only OM 16: std smd ( smd may be fishy) OM 21-26 mtd only, scale like string 5-10 !!!! string 5: even oms: std+smd+mtd, odd OMs mtd+smd except: 1,9,11,17,19,21,22,23,28,31: mtd only 13:std only 15: only smd 3,27,33,34: std+mtd 26: std+smd 35:std+smd+mtd 10: dead string 6: even oms: std+smd+mtd, odd OMs mtd (only) 16,28: std+mtd 19,23: mtd+mmd+r97 26:mtd only string 7: even oms: std+smd+mtd, odd OMs mtd+smd 1,2: smd+std 7,9,27,29,31,33,34,35:mtd 8,32:std+mtd 14,30:dead string 8: even oms: std+smd+mtd, odd OMs mtd (only!) 7,10,30:mtd+std 11,13,17,19,23,27,35:mtd+smd 31:std+mtd+smd 16,26,34:mtd 1,3,5,21,33:dead string 9: even oms: std+smd+mtd, odd OMs mtd+smd 6,25:mtd 22:std+mtd 33:r97 1,2,5,34:dead string 10: even oms: std+smd+mtd, odd OMs mtd+smd 4:std+mtd 9,10,33:mtd 24:smd 1:dead T-ZEROS: ---------- string3: average smd,std,mtd, except: OM 9 : mtd+std OM 10: r96, NO shift OM 17: dead string1: average all, except OM 3,18 : smd only OM 19 : mtd,std only OM 9: mtd+std OM9 shows a strange 50ns shift already in r96-o96 but more than 150ns between std,mtd and smd. smd is supported by ari. There is no obvous reason why smd should be wrong, except that 150 ns is too much for cabeling. This OM has to be tested independently. string2: average all, except OM 14: smd OM 8,12,19,20: dead ARI indicates shifts seen by smd in OM 9 and 14 compared to o96, 9 is supporet by std. all other ari seem wrong. string4: average all, except OM2 :r96 OM1 : std and mtd only OM 16: std (mtd seems 10ns wrong) OM 14,15: std+smd (mtd seems 10ns wrong) OM 21-26 mtd only - The SMD shift in OM 1 is not suported by ari...However it is not clear why smd is 30ns off - OM 21 to 26 are NOT OK. a) I need the correct length of the additional fibre!!!!!!!! b) No corrects for the light path to the distant channels 21,22,23,25,26 is applied yet. I assume the fibre ends at channel 24, so one has to add the geometrical distance. What are the spacings ????? string 5: even oms: std+smd+mtd, odd OMs mtd+smd except: 1,9,11,17,19,21,22,23,28,31: mtd only 13:std onlyes 15,35: only smd 3,27,33,34: std+mtd 26: std+smd 10: dead - OM 3 may had a 100ns wrong delay (corrected in latest macro) in case of an old t0_calc.kumac, add 100ns to std and mtd - OM 35 seem to be 100ns wrong in std AND mtd. SMD value is supported by r97 and ari, and the general slope. string 6: even oms: std+smd+mtd, odd OMs mtd (only) 16,28: std+mtd 26:mtd only 19: ari+90ns 23: mtd - mtd-ari is about 70ns -> 90=70+20 string 7:even oms: std+smd+mtd, odd OMs mtd+smd 1,2: smd+std 7,9,:mtd 8:std+mtd 14,30:dead STRANGE behaviour above OM 26 there are double structure in the LE distributions of mtd and std. fibres m16 and m18 seem to produce this effect (17 is too weak) The only usable may be ariel!. OMS to trust above 26 (???): 27: mtd (may be too high, but suportet by ari!) 28: ????? 29,30,31,33,34,35: ari+90 ns 32: std 36: std+smd maybe all channels are on one line, as indicated by: std:32,36 smd:36, ari:29,30,31,33,34,35,36 32,36 seem to reliable (fibres s16 and s18 should be OK._ At least 27 is very doubtfull, an alternative be to use: ari(n-1) + 0.5*(ari(n+1)-ari(n-1)) +90 ns test: 32: std+15=2586 ~= 2585 = ari(31) + 0.5*(ari(33)-ari(31)) !!! 27: 2326 (ari(29-26)), 2325 (ari(29-25)), 2324 (ari(30-25)), ..... -> 2325 ns 28: -> 2377 ns the difference between 2 neighbour Oms is about 52 ns (I like the second solution better). string 8: even oms: std+smd+mtd, odd OMs mtd (only!) 4: strong oszillations, std strange -> use: mtd+smd 7,20,30:mtd+std (20:smd bad) 10,16,26,34:mtd (also all odd, incl 19 because smd bad) 11,13,17,23,27,35:mtd+smd 31: std+mtd+smd 1,3,5,21,33:ari+102ns (=94+8) (DEAD ch) string 9: even oms: STD+MTD, odd OMs MTD 3,6,7,9,11:smd+mtd 4,8,10,12:smd+std+mtd 33:ari+70ns 1,2,5,34:dead, ari+120ns Different shift for OMs1-12 (0ns) and OM 13-36 (55ns) between smd and std/mtd ! smd is not suportet by ARI! but ari somehow agrees with mtd-> Do not use smd above 12! (or shift by 55ns ??????) note, that also the slopes of r97 and ari make steps ! string 10: even oms: std+smd+mtd, odd OMs mtd+smd 9,10,33: use mtd 24:smd 4,6:std+mtd AGAIN strange 50ns shift between std/mtd and smd. (This time after OM 8). mtd even and odd have been measured at different times but show the same effect. smd is supportet by ARI ! std is in contradiction to ari!. smd(6) seems to be off anyway. std/mtd show a line (support by r97 does not count,since also td) smd/ari show a step ! Incase of preference to SMD: subtract 45ns from mtd, for OMS: 2,3,4,5,6,7,8 subtract 55ns from std for even OMs 2,4,6,8 In case of preference to STD/MTD add 45ns to SMD for OMs 2,3,5,7,8. (after doing the inter string correction ) I have no idea which one to prefer. Maybe SMD is better here ! Conclusion ----------- following questions remain: - Was o96 done with a timing DAQ (TD) or with MAin DAQ? (I think o97 was done with a TD) - what is the spacing and delay for oms 81-86, where is the fibre? - what to do about the strange shifts especially: -string10 2-8 -string9 13-36 -string1 OM9 -string2 OM9 -string4 OM1 - string 7 above 27 is weired but may be ok. - still correct the transittimes for string 1-4 for the transit time in ice Ekki, can you do that, e.g.find out, which fibre needs correction? - alphas for OM 81-86 need a correctly rescaled alpha in the macro.