Oct 16-17/25 PionLT/KaonLT Analysis Meeting Notes ------------------------------------------------- (Notes by GH) Today: PionLT will be discussed first Please remember to post your slides at: https://redmine.jlab.org/projects/kltexp/wiki/Kaon_LT_Meetings Thursday: Present ----------------- Regina - Garth Huber, Alicia Postuma, Muhammad Junaid, Nathan Heinrich, Nermin Sadoun FIU - Pete Markowitz Virginia - Richard Trotta CUA - BC Manoj Ohio - Julie Roche JLab - Dave Gaskell Nathan ------ PionLT - wrapping up CoinLumi study - evaluating EDTM Total Livetime systematic uncertainty - ROOT refused to fit an exponential to EDTM_TLT vs CoinRate, so kept to a logarithmic fit - does a separate fit for each study and plots residuals of EDTM_TLT-fit - resulting histogram is slightly asymmetric - standard deviation 0.55% - *NB* Dave: concerned that the scatter in EDTM_TLT at constant CoinRate for various settings means this is not the right variable. Suggests a method that should not depend on choosing the right variable - plot NormYield vs CoinRate for each CoinLumi setting separately - fit CoinRate with only a constant, should be the same as the simple Error-Weighted_average of the NormYield data - then renormalize each NormYield with this constant, so that the weighted average for each setting is unity - make a histogram of the residual NormYield-1 and find the standard deviation - Garth: concerned that this will double-count the statistical error, i.e. this distribution will include statistical uncertainty of these points in addition to systematics. What we want is only the systematic, which we can apply to the physics data in addition to the statistical error - Dave suggests a modification: - look at the consistency of ChiSquare with 1 for each setting - if the ChiSquare is 1, this would indicate only statistical errors and we could argue that there is no additional systematic - any deviation ChiSquare>1 would indicate a systematic uncertainty - Additional detail on the CoinBlocking correction: Denominator makes use of the cut: CoinBlocking_Total CTime.CoinTime_RAW_ROC2 > -150 && CTime.CoinTime_RAW_ROC2 < 250 && ALL_COIN_events && H.hod.goodstarttime == 1 && P.hod.goodstarttime == 1 && H.dc.ntrack > 0 && P.dc.ntrack > 0 Numerator uses: CoinBlocking_Good CTime.CoinTime_RAW_ROC2 > 20 && CTime.CoinTime_RAW_ROC2 < 140 && ALL_COIN_events && H.hod.goodstarttime == 1 && P.hod.goodstarttime == 1 && H.dc.ntrack > 0 && P.dc.ntrack > 0 where ALL_COIN_events is: g.evtyp >= 4 && g.evtyp <= 7 - This selects only coin events and makes sure that there was good signal in both hodoscopes and drift chambers. - i.e. events with CoinTime=10^38 are removed - apparently the issue is that some prescaled singles come into the COIN stream by mistake. The DAQ trigger logic really needs to be reporgrammed to allow such double counting - modifications made to Report Files for CoinBlocking correction - beam current offset and its uncertainty applied correctly now - LH2 boiling correction uncertainty also in report file Junaid ------ PionLT Q2=3.85 LT-sep - tested Nathan's replay changes, seems OK - replaying Q2=3.85, W=2.02 setting for next LT-separation - still working on TT model for W=2.62 LT-sep - still see double-bump oscillation in Data/MC ratios Alicia ------ pi+ BSA PLB updates - Haiyan Gao had a staff member help with the resubmission - the paper is now submitted, but the references are corrupted - Alicia contacted the staff member and Haiyan on getting the references fixed, but no reply --- Friday: Present --------------- Regina - Garth Huber, Alicia Postuma, Vijay Kumar, Nacer Hamdi, Nermin Sadoun, Muhammad Junaid Virginia - Richard Trotta JMU - Gabriel Niculescu JLab - Dave Gaskell CUA - Sameer Jain, Tanja Horn, Chi Kin Tam, BC Manoj FIU - Pete Markowitz Ohio - Julie Roche Glasgow - Kathleen Ramage Richard ------- KaonLT Q2=4.4, W=2.74 LT-sep - functional forms had no Q2-dependence in them - now scaling L/T/LT/TT terms by (Q^2_event/Q^2_central)^n where n= -2.18(L) -1.89(T) -2.01(LT) -2.34(TT) Chi Kin ------- KaonLT high Q2 data analysis - Q2=4.4, W=2.74 - compared MM data after PID cuts with Richard's analysis - no acceptance, HGC hole or diamond cuts - the distributions look identical - then compare NormYield/mC with Richard - has some significant differences in some (t,phi) bins - investigating what is different, will check the diamond cuts - Q2=3.0, W=3.14 - undid shift in SIMC MM distribution - adjusted t-bin width - compare HMS distributions for Data and MC - weird SIMC distributions observed earlier was due to unphysical cross-section values at large -t - now applied a t-cut to restrict SIMC to the range of the data - agreement looks much better now - still running into bad parameter fits for TT term - changed from p4/(|t|^p5)*exp(-|p6*t|) to p4*exp(-p5*ln|t|-|p6*t|) - not working well, needs to look into some other form - Garth: taking the exponential of a logarithm seems weird - *NB* suggests to try simply checking with a calculator some functional forms and see what might be more promising, rather than random changes Sameer ------ KaonLT CoinBlocking correction - following Nathan's steps - revised CTime.CoinTime_RAW_ROC2 cut values: -20 to +80 Q2=2.1, W=2.95, high epsilon, center SHMS runs 4891-4912 - most runs have correction ~0.9, but there are some weird outliers with correction ~0.7, and a sharp shift in correction at rate ~2.5kHz - *NB* Richard: cut still looks too tight, suggests to increase to +110 and also loosen negative side - *NB* need to investigate the outliers and the sudden shift - also need to add the loose denominator cut that Nathan mentions above Vijay ----- PionLT Summer 2019 CoinBlocking correction - also following Nathan's steps - using a 135ns wide cut, the trigger oscillation traces have 140ns - the cut looks slightly generous but should be good - correction factor ~0.9, with high epsilon ~0.95 and low epsilon at 0.88 - for denominator, will apply a wide cut -600 to +900 Nacer ----- KaonLT Q2=0.5 LT-sep - last week, had Data/MC ratios near 1 but with double-bump oscillations at high epsilon - the fit model had LT=TT=0 - now trying TT=(p7*exp(-p8*|t-tmin|)/((Q^2+p9)^2)*sin(thetacm) and similar function for LT - ratios get much worse after 3 iterations, oscillations continue to grow - Dave: all analyses are encountering interference function problems, is there an inconsistency in how phi is calculated between hcana and SIMC? - Richard: all recon_hcana calculations applied to SIMC are taken from hcana xq = fX.Vect(); bq = fB.Vect(); xq *= rot_to_q; bq *= rot_to_q; //Calculate Angles of q relative to x(detected proton) and b(recoil neutron) th_pq = xq.Theta(); //"theta_pq" ph_pq = xq.Phi(); //"out-of-plane angle", "phi_pq" th_nq = bq.Theta(); // theta_nq ph_nq = bq.Phi(); //phi_nq thetapq = th_pq; phipq = ph_pq; Richard's version of recon_hcana at: https://github.com/trottar/simc_gfortran/blob/LTSep_Analysis_2024_kaon_xsects/recon_hcana/recon_hcana.C - Gabriel: it's important to verify that data and SIMC phi distributions are the same - Nacer looks up a comparison histogram, they indeed look identical - Dave: maybe there's an inconstency between how phi-vertex used in the model is calculated compared to phi-recon used in the iterations? - *NB* Garth: suggests to make a plot comparing phit from SIMC with phi from recon_hcana. Expect to see a difference in resolution because phit is the truth variable, but they should not differ in gross features - *NB* Dave: suggests to apply recon_hcana script to phit when calculating model cross sections in SIMC - *NB* Dave: looking at Richard's version of recon_hcana - this version has cuts in it, was not expecting to find any cuts there if this is for general use - *NB* Gabriel: this version also has offsets applied in it, what happens if the offset is applied repeatedly in each iteration, that would cause huge problems - it's better to apply the offsets in a transparent fashion rather than being in this code, to avoid them being applied more than once - Richard: I uploaded a clean version of my recon_hcana script to my master branch. There is only the hard coded kaon mass in this version (just search 'HARD CODED' in cpp script). https://github.com/trottar/simc_gfortran/tree/master/recon_hcana If anyone finds any other issues or if we'd like to add calculations for the vertex variables let me know. These variables are already defined (unchanged), but easy enough to recalculate. It is also easy to add new versions of variables (e.g., phipqi->phipqi_test) for comparisons. New variables just need to be initiated in the header, added as a branch, then calculated in the loop in the cpp file. - *NB* it's important that everyone double check their version of recon_hcana compared to this clean version! BC Manoj -------- - still learning how to replay data, no report yet --- Next Meeting ------------ New 2-meeting/week structure - Thurs: Oct 23 @ 16:00 Eastern/14:00 Regina - KaonLT will go first - Fri: Oct 24 @ 11:00 Eastern/9:00 Regina - we will continue where we left off