| e949.pass2 |
The Photo Veto study is matured, established and controlled analysis in E787/E949. We are not worrying about the study, and good rejection and acceptance are already achieved. In this review process, we would rather concentrate on understanding the changes in PV from E787-98 to E949-02 and determining the improvements in E949, We also confirm if there are loopholes or not.
and please download the latest one: | slides | printer-friendly slides |
- page 6: BV/BVL plots: Earlier hits at around -20 nsec to 0 nsec.
o not observed in other subsystems, no such hits in the acceptance plots.
--> real photons in BV/BVL (ADC correct, TDC screw up)
o correlation to other hits ? hits in other subsystems ?
o Kentaro made new plots in page 7, after imposing the offline BV cuts
--> All the PV cuts at in-time should be imposed.
==> The range in ntuple will be extended to [-100nsec,...]
- page 9: strong correlation between the CT sector and BV Photon sector (90 %)
o proposed mechanism:
lower energy photon is overlapping to the CT, and
higher energy photon has photonuclear interaction and
leaves small visible-energy due to protons/alphas (not neutrons).
o We might be able to impose a tighter offline PV cuts
to the sectors in the opposite side of CT.
o Nobody studied the BV sector distributions in E787, and
we might have missed the mechanism in E787.
<-- impose the full PV cuts (except for the BV cut)
<-- tail in the timing plot in the later side ?
<-- overlapping photon cuts to the CT (RS de/dx cut) ?
<-- Impose Kpi2 momentum and range box cuts.
In the etot distribution, you should see the 2nd peak at 20MeV above
( See p86 of E787 TN 383 ).
<-- energy/timing distribution of these hits ?
compare the plots "in the band" and "off the band"
<-- Timing distributions should be plot in Log.
==> more studies on these BV hits
- page 10: two peaks in the module distribution (and CT stopping sector dist).
o This is observed in the pnn1 events, not in Km21.
o Kp21 distribution at bottom left. online PV cuts are imposed. --> statistics not enough
FITPI cut (ipiflg=0) is imposed.
o Distribution is flat for Kp22 after online PV cuts.
o cos3d 1-dim plot: Kentaro's new plots, see page 4.
top right: cos3d distribution is shifted.
o Online PV bits are not guilty... Is it due to Refined Range Mask ??
<-- no such effect in QOD plots, (dominated by km2)
<-- no such effect observed in Kinematic study.
==> We should do systematic check with kp22 data and pnn1or2, by imposing the cuts one by one.
- page 12: BV/BVL high occupation in end#2 (downstream PMTs) in single TDC hits
o time window for upstream-downstream matching: 15 nsec.
o earlier hits at 20nsec ?
o We might have offline rejection power to the single end hits in general,
which are missed by the meantimed online PV in E949.
o Was the downstream rate high in the real case ??
<-- No, the rate is high in the upstream.
==> check single rate in km21
- page 15: BVL, small excess in top left --> page 17: Peak "B"
o proposed mechanism:
high energy, and matching in both-end timing
--> two photons hit in the same BVL sector.
o This mechanism was suggested by T.Nakano/T.Sasaki in their E787-96 photon study.
o Veto these hits by setting up another BVL category (with high threshold)
or just adjust the BVL parameters to cover the region ??
o No such hits are observed in BV
- The shower does not reach BV ??
- These photons are caught by BVL, with its extended Z coverage ?
- Effects of muons in BVL ??
==> Make the plot after imposing all the offline PV cuts
(including the standard BL cut)
and see if we have rejection power by vetoing peak "B" events.
- Why so many BV/BVL hits with large visible energy in the scatter plots ?
Did the online BVL/BV bits work ? or were they not-well calibrated ??
<-- should be checked with Kp21, Kp22. See energy distributions of intime hits.
<-- CT timing at the trigger level (Tdot2) might be screwed up ...
See trs vs Tdot2_in_TDC to check the screwed up.
- page 15 bottom right: BVL, big energy to the later timing
o proposed mechanism: muon -> electron ecays in BVL ?
o Should we store hits in later timing in the ntuples ?
<-- not interesting. Concern is the deadtme.
o We are not going to use the electrons in BVL
in TD analysis.
- BVL and RS have better timimg resolution, thanks to RSTDCs.
- pages 22-25 Endcap plots ... no surprise to the high energy hits.
The narrow time window (with better timing resolution)
is the key of the endcap analysis.
- page 26 .. Range Stack PV
o strong correlation between the CT sector and Photon sector, as in BV
<--- same mechanism (photonuclear interaction).
o The online HEX cut is imposed to the opposite sectors
(but the threshold might be too high ??)
- Collar, MC: photon rejection power to pnn1 in Joe's study
- Do we need further time walk correction to all the subsystems ??
Steve:
1) It seems that our time walk correction (pulse-height correction to
timing) is not optimized for almost all systems. Why is this? Have they
all been calibrated?
BV: time-walk of +2ns at 0.2MeV and still has +0.5ns at 1MeV (Tamaki?)
BVL: time-walk of +1ns at 0.2MeV (not-so-bad)
EC: time-walk of +4ns at 0.5MeV and still +1.0ns at 1MeV (Takeshi?)
Do we have a CCD time-walk correction? 0.5 MeV pulses have many
noise pulses taht are not from CsI, but at 1 MeV they should be
reasonable, yet we still have a 1ns time-walk
RS: time-walk of +0.5ns at 1MeV (seems pretty good..., Dima?)
IC: time-walk of -1ns at 1MeV (overcorrection? Jim?)
- Jim said we need slewing correction to the TG-CCD timing.
So do we need the correction to Endcap-CCDs as well ?
Concern is the Endcap hits hits with largr visible energy
(time window is +-2nsec, while the slowing is 1nsec).
- Do the correction in KOFIA/CFM, or in the ntuples ??
--> The effect is small in PV.
A KOFIA code change might introduce some effects to other parts ..
==> Test the slewing corrections in the ntuples at first.
- Don't we have slewing correction to BV/BVL TDC hits ??
Ask Jim.
- DPV and AD are unused in the PV study, nor in the beam study.
==> Turn them on to the ntuples (Jingliang, Kentaro).
- RS-discriminator threshold level
15mV in E787 ==> 0.3MeV
25mV in E949 ==> 0.5 MeV
- TDC information is primarily used, but if there is no TDC hit
the TD hits are used (same as in the TD studies).
- Is the slewing correction etc used in TD studies ??
PV in E949 2002 (revised by Joe on June 17);
TOFF TWIN EMIN RLAST ALAST ALASTC ALASTL18
======================================================================
BV 1.25 6.00 0.20 19.39 0.966 0.955 0.954
BL 0.75 2.00 0.00 1.23 0.995 0.993 0.992
EC 0.25 2.25 2.00 2.60 0.983
RD -0.50 1.75 0.00 6.06 0.956
TG 0.00 1.50 9.60 1.12 0.997
IC 0.25 2.25 0.40 1.38 0.995
VC -1.25 2.50 1.60 1.07 0.998
CO 2.75 1.25 0.00 1.08 0.998
CM -0.50 1.50 2.80 1.01 1.000
EC1 0.50 2.25 0.80 1.25 0.991
BVS -9.50 9.50 4.80 1.06 1.000
RDS 1.50 0.50 15.20 1.00 1.000
(RDB 0.00 1.00 3.80 1 1 <--- will be dropped)
RD1 2.50 1.50 2.40 1.00 1.000
RD2 0.75 2.00 0.40 1 1
RD3 -0.50 0.00 2.80 1 1
BV1 2.50 3.00 2.40 1.00 1.000
BV2 -1.50 0.00 6.00 1 1
BV3 -4.00 11.00 7.60 1.04 1.000
EC2 -3.00 0.00 19.20 1 1
BL1 -1.00 4.00 12.60 1.02 0.999
-
PV in E787 1998 (page 15, Table 2 of the 1/3 note);
Detector TOFF TWIN EMIN RLAST ALAST
=======================================================
BV 3.75 6.25 0.20 5.61 0.981
BL < none >
EC 0.50 3.00 2.20 3.35 0.981
RD 2.40 4.25 0.8 2.30 0.967
TG -0.50 2.25 3.8 1.12 0.998
IC -1.00 2.50 0.40 1.21 0.997
VC -1.25 3.50 0.40 1.08 0.998
CO 2.00 3.50 0.40 1.13 0.997
CM -0.50 1.50 2.00 1.01 1.000
EC1 0.50 2.75 0.40 1.34 0.986
BVS -8.00 2.50 0.00 1.00 1.000
RDS -7.50 2.50 3.60 1.00 1.000
( RDB < none > )
RD1 0.50 2.50 0.40 1.01 1.000
RD2 -3.00 5.00 0.40 1 1
RD3 -3.50 0.00 1.60 1 1
BV1 8.00 5.50 2.00 1.00 1.000
BV2 3.00 1.00 0.00 1 1
BV3 -6.50 0.00 3.60 1.00 1.000
EC2 0.50 1.50 6.80 1 1
BL1 < none >
- 'BL1': both-end ADC, signle end TDC.
- The role of BVL is really not that large for the pass2 cut,
indicating that the online cut was perhaps too tight.
<-- The "rejection" sample for BVL is the events
which have survived the offline PV cuts of all the other PV subsystems.
The BVL hits and BV hits should have correlations; for example,
the photons that start the EM shower in BL should leave
large visible energy in BV as well and
have already been vetoed by the offline BV cut.
- > First, there have been major changes to basic reconstruction routines
> and/or hit selection algorithms that feed into the photon veto, like the
> items mentioned above in relation to RD information, and also the replace-
> ment of swath by swathccd. Second, we are operating in a much more noisy
> environment in e949, which the optimization code tends to handle by
> raising energy thresholds. Finally, except for RD, none of these subsytems
> has any kind of a major contribution to the total rejection, which by
> definition means that the parameters are not well defined, i.e. they could
> change by a large amount and not make a noticable impact on the total
> rejection and acceptance.
o Why the threshold for TG is so high (E787 3.8MeV -> E949 9.6MeV) ?
I suppose the "gamma" fibers were identified by swathCCD.
<-- small statistics and small rejection power in these subsystem
o Is the TG PV cut related to 2 beam background study ??
<-- the beam cuts are applied as the setup cut.
o Compare the RD PV cuts between E787 and E949.
E949: RD -0.50 1.75 0.00 6.06 0.956
E787: RD 2.40 4.25 0.8 2.30 0.967
<-- The cut paramters are highly correlated to other subsystems
<-- In the parameter-space searching, there might be some "local minimum" points...
the solution might not be a unique best place
<-- If the time windows and energy thresholds are reasonable, that should be okay.
o The PV cuts and the inside-the-box functions are tuned up with 1/3 sample, and
tested with 2/3 sample. (No further tuning using 2/3)
- In the "blank-out" method, the six BV/BVL
sectors nearest in phi to where the charged track exited the
rangestack are "blanked-out" to avoid vetoing on track hits. A
correction is then applied to the measured BV/BVL acceptances.
- In the "Layer 18" method,
the acceptance was measured with the Km21 muons
that stopped in Layer 17 or 18 (and have a funny double-humped
cos3d distribution, moreso than for e787).
"Blank-out" Method "Layer 18" Method
========================================================
ON_HEX 235541/251930=0.935 20523/21990=0.933
ON_BV - 20220/20523=0.985
ON_EC 232587/235541=0.987 19974/20220=0.988
ON_BL - 19528/19974=0.978
-------------------------------------------------------
0.923 0.888
-------------------------------------------------------
RSHEX/2 221602/232587=0.953 17620/19528=0.902
TGPVCUT 219993/221602=0.993 17432/17620=0.989
-------------------------------------------------------
0.946 0.893
-------------------------------------------------------
PVCUT02 192694/219993=0.876 15253/17432=0.875
0.863*
* Blank-out correction included
- RSHEX/2: two pass1 cuts (energy in the stopping hextant)
5% difference between two methods.
<-- Layer 18 method might be wrong.
<-- Should it be in the "photon veto" category ??
- Was Level 1.2 imposed in advance ?? <--- No. That should be made.
| JM's plots |
- Correration (?) observed for the first time: PV rejectioin vs dip-angle (cos3d).
o No asymmetry observed after BV cuts ... related to BV Z funny distributions ??
o Correlated to RS de/dx cut (overlapping photon) ?
o target Z correlation ?
(stop in the downstram in tgt, longer pathlength in RS?) ?
o TGT de/dx cut ??
o PV Acceptance vs cos3d is very flat.
- "crack (or dead material)" in a certain dip angle ??
...who knows...
==> Kentaro's study on the upstream/downstram single-end hits might give the answer
to this correlation ??
- The best achieved PV rejection is x20 higher than the E787 published paper on pi0 -> nu nubar;
this is a good news for pi0 -> nu nubar at E949.
- Is the improvement due to the high-energy photon rejection in BVL/BV
or low-energy photon in the beam region ?
a. BV/BVL earlier hits
==> The range in ntuple will be extended to [-100nsec,...]
b. strong correlation between the CT sector and BV/RS Photon sector
==> more studies on the BV hits oppsite of the CT.
c. two peaks in the BV module distribution
==> We should do systematic check with kp22 data and pnn1or2, by imposing the cuts one by one.
d. Geometrical mystery: BV/BVL high occupation in end#2 (downstream PMTs) in single TDC hits
==> check single rate in km21
e. BVL Peak "B" events
==> Make the plot after imposing all the offline PV cuts
(including the standard BL cut)
and see if we have rejection power by vetoing peak "B" events.
f. so many BV/BVL hits with large visible energy; did the online BVL/BV bits work ?
==> check with Kp21, Kp22 (see energy distributions of intime hits)
==> check trs vs Tdot2_in_TDC (screwups in the CT trigger ?)
g. further time walk correction to all the subsystems ?
==> Test the slewing corrections in the ntuples at first.
h. DPV and AD
==> Turn them on to the ntuples.
i. Acceptance measurement on RSHEX/2
==> Level1.2 should be imposed in advance.
j. Correlation in PV rejectioin vs dip-angle ??
==> The correlation is not serious.
Studies on "Geometrical mystery" might give the answer ..