L0trigger and new generator (lhcbtrig)
From: dyk@mail.cern.ch
Date: 2/4/00
Time: 8:13:31 PM
Remote Name: 137.138.142.33
Remote User:
Comments
Since the TP days we have changed quite a few items concerning the
L0trigger rates. In this mail I would like to quickly review the TP
situation, and then give the situation from SICB v220 onward.
 TP situation.
. We generated only hardscattering processes, assuming a total cross
section of 80 mb.
. We used the 3x3 code for electron and hadron triggers.
. The bandwidth of 1 MHz of L0YES was subdivided as follows:
Average number of interactions/crossing=crosssection*luminosity/rate,
where luminosity=2*10**32, and the rate of filled bunch crossings
per second is 30 MHz. This gives 0.533 interactions/crossing.
#interaction Probablity/interaction rate(MHz)
0 0.587 17.6 (+10 due to empty bunches)
1 0.313 9.39
>1 0.100 3.00
Hence we need a L0mbiasretention of: 1/(9.39+3)=8.1% to get a
1 MHz L0accept rate.
(However, the L0mbiasretention fraction was usually quoted as
a fraction of events which were accepted by the pileup veto.
The pileup veto accepted 95% of single and 20% of multiple
interaction crossings, hence: 1/(0.95*9.39+0.2*3)=10.5%.
The h/mu/e then took: 6/2/1% of this 10.5% leaving the
remaining 1.5% for singlephoton, mumu, LHChalomuons etc.
I suspect that the pileup veto loss of 5% (or maybe more for
Bchannels) has not been taking into account for the final
trigger acceptance in the TP.)
The pt/Et thresholds used for the TP were:
parameter (rmucut = 1.0, ! L0 muon Pt cut
> relcut = 2.34, ! L0 electron Pt cut
> rgacut = 4.0, ! L0 gamma Pt cut
> rhacut = 2.4, ! L0 hadron Pt cut
 Situation since SICB v220:
. We generate hardscattering+diffractive(single and double)+elastic
processes. The total cross section is 102.4 mb. This is evtyp=61.
The cross section of 102.4 mb is coherent with respect to the one
adopted at the TP time: the main difference is that we model the
inelastic events taking into account all the effects and the 80 mb
are now shared in the following (more realistic) way:
 sigma(hard scattering) 55 mb
 sigma(diffractive) 25 mb
and in addition we have 22.4 mb of elastic events.
(Evtyp=51 is the hardscattering only part)
. We moved to the 2x2 algorithm for the calorimeters. (Bank T2X2).
Note that the "energyscale" is slightly changed due to this move.
(We did NOT yet move to the new pileup code, which has two passes to
to recognise two pileup events with very different multiplicity)
. We started producing tapes with pileup events according to the
expected luminosity: DST1=no pileup.
DST2=1.56*10**32 and 2*10**32 (see note 1)
DST3=5*10**32
DST2 tapes are produced assuming a falling luminosity during a fill
with a lifetime of 10 hours and a fill time of 7 hours, requiering
the AVERAGE luminosity to be 2.10**32 (or 1.56*10**32 see note 1).
DST3 tapes try to simulate "worst conditions", hence a flat luminosity
of 5.10**32. (Which will hopefully last for minutes rather than hours
during the real experiment).
Note 1: For the data produced with v220 the totalcrosssection*L was
assumed to be 80 mb * 2.10**32, but since the correct
totalcrosssection is 102.4, (80mb+24.4mb elastic) this
effectively means that L=1.56*10**32 for these tapes.
.DST2 tapes for v222:
(For v220 data I do not give the pt/Et threshold, since
they are not so relevant at L=1.56*10**32, the first DST2 tapes of
v222 have this low L too, hopefully to be fixed soon).
The new number of interactions/crossing=0.683 (But note that on our
tapes a fair fraction of the events is "empty", elastic, backward
singlediffractive, even hardscatter with very low multiplicity)
#interaction Probablity/interaction rate(MHz)
0 0.505 15.16 (+10 due to empty bunches)
1 0.345 10.35
>1 0.150 4.49
Hence we need a L0mbiasretention of: 1/(10.35+4.49)=6.5% to get a
1 MHz L0accept rate.
I have retuned the pt/Et thresholds to get roughly the same division
of bandwidth as for the TP. Eventually Clair's allocation of
trigger rates (TP, section 12.3.5) has to be revived again to do
this properly, now I just required h:mu:e:g:mumu=6:2:1:0.1:0.1, and
10% of the bandwidth is left for LHChalomuons and randomtriggers
and etc.
I get the following rates for all the triggers: (rate=number of
events which pass both the threshold cuts, and are not
vetoed by the pileup or "activity" veto (see note 3 below),
devided by the total number of events on tape, each event enters
only once):
mu = 0.010355 (mu+(mu+e)+.... )/6.5%=23%
e = 0.00419132 (e+(mu+e)+(h+e)+(mu+h+e))/6.5%=11%
mu+e = 0.000493097
h = 0.0357495 (h+(h+mu)+... )/6.5%=64%
h+mu = 0.00345168
h+e = 0.00221893
h+mu+e = 0.000493097
gamma = 0.000739645 /6.5%= 1%
mumu = 0.000739645 /6.5%= 1%

Sum = 0.0584 %, hence leaving .7% for LHChalomuons etc.
The pt/Et thresholds used for the above are:
parameter (rmucut = 1.17, ! L0 muon Pt cut
> relcut = 2.67, ! L0 electron Pt cut
> rgacut = 4.43, ! L0 gamma Pt cut
> rhacut = 2.65, ! L0 hadron Pt cut
> rmmcut = 3.45, ! L0 sum pt mumu cut: Note 2
> etocut = 50., ! L0 min etot cut : Note 3
Hence: the cuts are very similar to the TP values as expected,
maybe slightly higher (i.e. less efficient for B's?).
Note 2: I introduced a mumu cut, which requires two muons to be
reconstructed, and the sum of their pts>3.45 GeV. This
trigger ignores the pileup veto decision. Mainly for
rare Bdecays. Forced to be small=1% of L0.
Note 3: I require some minimal "activity" in the spectrometer,
for now etot(calorimeter)>50 GeV. Note that about 20 MHz
of crossings will not have "hardscattering" pp interactions,
but for these events we do not want to have the muon trigger
fire when it finds a LHChalomuon. The 10 MHz of empty
bunches could be vetoed by the trigger supervisor.
The source of this "activity" signature should be discussed,
Etot, number of hits in pileup veto, both, ??.
This 50 GeV cut lost me one event in 10k B>pipi events,
while 39% of the events on tape have Etot<50 GeV.
The pileupveto finds more than 1 vertex in 12% of the events,
or .12/(1.39)=20% of the "active" events, which is roughly
the same as in the TP days: (.05*9.39+.8*3)/(9.39+3)=23%.
.DST3 tapes for v222 onward, i.e. 5*10**32 flat rate.
Number of interactions/crossing=1.707
#interaction Probablity/interaction rate(MHz)
0 0.181 5.44 (+10 due to empty bunches)
1 0.310 9.3
>1 0.509 15.28
Hence we need a L0mbiasretention of: 1/(9.3+15.28)=4.1% to get a
1 MHz L0accept rate.
No data available yet. Hence I have reweighed the DST2 tapes to get
a mixture which looks like 5*10**32 (but low statistics).
The corresponding pt/Et cuts are:
The pt/Et thresholds used for the above are:
parameter (rmucut =2.8 , ! L0 muon Pt cut
> relcut =3.2 , ! L0 electron Pt cut
> rhacut =3.4 , ! L0 hadron Pt cut
> etocut = 50., ! L0 min etot cut
(Photon and mumu did not have enough statistics to give a "tuned" cut)
At 5*10**32, 26% of the events have Etot<50 GeV, while 18% are
vetoed by the pileupveto, i.e. pileup=.18/(1.26)=24% of "active"
events.
Especially the increase in the muon threshold is somewhat surprising,
maybe due to statistics(1000 events left after reweighing all the DST2
data), but clearly the muons have a much longer/flatter
tail in their pt distribution. Maybe very large pT muons should be
deappreciated? We will know the moment we get DST3 data.
Once the luminosity is in the PASS bank per event, I'll release a
AXTRIGGER replacement which will return the level0 decision
according to the generated conditions automatically.
Regards, Hans Dijkstra.
