Algorithm works both on real data and MC.
- 1. Algorithm
- The program matches the RPNT created by ranger at the begining of OTR-PC segment(pc) and the end point of SVD segments(svd). First "raw cuts" are applied :
- Raw comparison of SVD and PC vertical parameters at PC Z :
|svd->y + (pc->z - svd->z)*svd->ty - pc->y | < 35.7- Raw comparison of SVD and PC horizontal parameters at z = 455
| svd->x+(455-svd->z)*svd->tx - pc->x-(455-pc->z)*pc->tx |<18.6- Momentum is estimated by requirement that both parts have equal TX in the middle of the magnet. The covariance of TX is not taken into account. The difference must be smaller then 0.01 mrad.
- Svd and PC is prolongated to the Z = 455 (svdA, pcA ) and there more accurate cut on X is applied :
| svdA->x - pcA->x | < deltaX
where deltaX could be choosen by MARPLE/SOFT kuip command- Svd is prolongated to Z of pc (svdB) and there the accurate cut on Y is applied :
| svdB->y - pc->y | < deltaY
where deltaY could be choosen by MARPLE/SOFT kuip commandThe magnet is supposed to be ON.
- 2. Selection of objects to be matched
- SVD : taken from RSEG ArteTable
- rseg->fit = 0
- (rseg->cmp & Rsegc::bitsign) == 0
- rseg->cmp == Rsegc::vxd || rseg->cmp == 65537(CATS)
- -30 < rseg->ze < 800
- the end point of RSEG is considered for match
- OTR : taken from RPNT ArteTable
- The loop over RSEG is performed : rseg->fit = 0
(rseg->cmp & Rsegc::bitsign) == 0
rseg->cmp & Rsegc::patt
- the RPNT connected to the given RSEG is taken if : rpnt->zf > 680 && rpnt->zf < 860
- from 2 RPNT connected to the given RSEG and passed selection the one with smaller rpnt->zf is accepted
- 3. Match Parameters
- There are 5 different possibilities is available They are steared by MARPLE/SOFT <N> kuip command.
The efficiency and ghost rate in the table refer to the numbers obtained on Monte Carlo sample for values not multiplied with FdataX, FdataY.
MARPLE/SOFT deltaX deltaY efficiency ghost rate 1 12.2*FdataX 18.7*FdataY 98.9% 68% 2 6.0*FdataX 8.6*FdataY 97.4% 34% 3 or 0(default) 2.3*FdataX 4.0*FdataY 91.8% 11% 4 1.2*FdataX 2.1*FdataY 80.0% 4.7% 5 0.6 *FdataX 1.1*FdataY 54.4% 2.2% The additional factors FdataX and FdataY to multiply were estimated from the correlation plots :
FdataX = dX*sigma(dX_data)/sigma(dX_MC) = 0.492/0.44
FdataY = sigma(dY_data)/sigma(dY_MC) = 1.122/0.794
The correction was obtained on
J/psi triggered data: run 15127 (reprocessed with rc0306r8)
and on Monte Carlo sample: single inelastic interaction, geometry 99.0812. If one runs the matching on the same MC data he should see the efficiency and ghost rate higher that it is declared in the table (For given MC, default set: Efficiency = 93.7%, ghost Rate = 11%). But it is expected that on data it should be of this order.- 4. Call Sequence
- -
- RECON/MATCH marple VDS PC
- from user.kumac, if standard reconstruction for real data is called
or- void matchsvdpc_();
- from usevnt_() if no reconstruction is called
- 5. Recomendations
The algorithm was tested and optimized on run 15127 with ARTE-03-07-r4. The programs used for subdetector parts were CATS and RANGER (soft ranger parameters). Even that CATS provides a covariance matrix in this release, the usage of $\chi^2$ cut is not tested and therefore not used. The algorithm can not cut on delta TX since it is used for momentum estimation. The TY in principle could be set, however in present setup it was proven to be inefficient after cut on delta Y. The efficiency is expected to scale with the width of the deltaX, deltaY correlations.
Another important note that the momentum saved in match segment is ok for match but is NOT GOOD for analysis. One should use outcome of refit procedure to get proper momentum estimation.
If one re-estimate efficiency and ghost level of package he/she should remember that the correlation on real data looks wide than on MC therefore the correction factor should be applied as described above.
- Acknowledgment
- W.Hulsbergen helped a lot in optimization of parameters and building the scheme. The performances of algorithm was heavily checked by P.Conde and results are available in the minutes of tracking group.
generated by doc++