TRACKER choice: Comparison of tracking capability and systematics 1. Not all information is available – only one group gave results (bravo!) and they are.

Report
TRACKER choice:
Comparison of tracking capability and systematics
1.
Not all information is available – only one group gave results (bravo!) and
they are not complete – I use yesterdays results.
2.
For this reason the ‘critical comparison’ will be critical of them – which is
unfair.
3.
Conclusion will be that we may have one device that is good enough, but
we are not *sure*. We need thorough documentation and discussion of
results before final endorsement of any device.
Alain Blondel
Quantities to be measured in a cooling experiment
cooling effect at nominal input
emittance ~10%
equilibrium emittance = 0.25 cm
curves for 23 MV, 3 full absorbers, particles on crest
Alain Blondel
Emittance measurement
Each spectrometer measures 6 parameters per particle
x y t
x’ = dx/dz = Px/Pz
y’ = dy/dz = Py/Pz t’ = dt/dz =E/Pz
Determines, for an ensemble (sample) of N particles, the moments:
Averages <x> <y> etc…
Second moments: variance(x) sx2 = < x2 - <x>2 > etc…
covariance(x) sxy = < x.y - <x><y> >
Covariance matrix
 s 2x

 ...

M =  ...
 ...

 ...
 ...

s xy
s xt
s xx'
s xy '
s 2y
...
...
...
...
...
...
...
s t2
...
...
...
...
s 2x'
...
...
...
...
s 2y '
...
s xt' 

s yt ' 

s tt ' 
s x't ' 

s y 't ' 
s t2' 
6D

 de t(M xytx 'y 't ' )
Evaluate emittance with:
 4 D  de t(M xyx 'y ' )   2
Getting at e.g. sx’t’
is essentially impossible
with multiparticle bunch
measurements
Compare in with out
Alain Blondel
requirements on spectrometer system:
1.
must be sure particles considered are muons throughout
1.a reject incoming e, p, p
=> TOF 2 stations 10 m flight with 70 ps resolution
1.b reject outgoing e => Cerenkov + Calorimeter
2. measure 6 particle parameters
i.e. x,y,t, px/pz , py/pz , E/pz
=>
3. measure widths and correlations …
resolution in all parameters must be better than 10% of width
at equilibrium emittance (correction less than 1%)
s2meas = s2true+ s2res = s2true [ 1+ (sres/ strue)2 ] (n.b. these are r.m.s.!)
4. robust against noise from RF cavities
Alain Blondel
b  30 cm in the detector solenoid
Beam size at eq. emittance is sigmax= sqrt ( 30. 0.25) = 2.7 cm
Transverse angle is sigma x’ = sqrt(0.25/30) = 0.09 radians i.e. Pt = 18 MeV
resolution in all parameters must be better than 10% of width
at equilibrium emittance (correction less than 1%)
requires resolution better than 1.8 MeV/c in Pt.
THIS MUST BE THE RESOLUTION ON THE PREDICTION OF THE PARTICLE
ENTERING/EXITING THE COOLING CHANNEL (not what it was in the tracking device)
This situation is quite usual in HE experiment (e.g. track extrapolation to muon chambers)
Alain Blondel
TRANSVERSE MOMENTUM RESOLUTION spt
= 110 keV
RESULTS
From proposal:
PL= 200 MeV/c
Pz resolution degrades at low pt :
resolution in E/Pz is much better behaved
measurement rms is 4% of beam rms
Alain Blondel
Limit to precision: multiple scattering in the last plane of detector:
e.g. 1.9 mm of plastic (sci-fi) or 0.2 mm of mylar (TPG end window – to be verified)
Sci fi: qMS= 15MeV/c / bp L/X0 ~ 0.09 L/X0 =0.006
Or Pt = 1.25 MeV/c (rms will be somewhat worse)
There is no way this can be 110 KeV/c…. There must have been a mistake
TPG same calculation gives 0.41 MeV/c .. *Not* 0.035….
1.4 Safety factor
What is the loss performance in the spectrometers which will compromise the
measurement? How large is the safety factor?
recalled the requirement in the LOI (1/10 of the beam size at equilibrium emittance)
beam size in transverse momentum is about 10 MeV at equilibrium emittance
we have 110 keV Pt resolution with sci-fi 270keV multiplexed.
35 keV for TPG
Safety factor is large, but systematic errors (alignment!) have not been taken
into account.
This was wrong.
Alain Blondel
No Multiple Scattering
Sci-Fi (M. Ellis)
• Point resolution
determined exclusively
by d/12
•Ganging times 7
makes resolution 7
times worse
•This is the
configuration that had
been (accidentally)
presented at previous
meetings.
Alain Blondel
No Ganging
 Multiple scattering is
now on in all future
plots
 Point resolution is
dominated by effect
of m/s with no ganging
 Momentum resolution
is still good enough for
a cooling measurement
Tails come from MS… and are real!
•Sigma (Pt) of 1.13 matches the b.o.e.
calculation that gives 1.25 (but please
use rms -- and careful with tails?)
Alain Blondel
7 Fibre Ganging
 Position and
momentum
resolution have
degraded slightly,
but no longer by
factor of 7
 Use of “projector”
in Kalman package
will allow
improvement in
resolution per plane
?
Alain Blondel
iii. Tracking performance:
The detector response and point resolution must be based on
experimental demonstration.
[…]
A viable scenario for alignment and calibration must be described.
Alain Blondel
Nominal Configuration
 Dead Channels (0.25%) based on D0 experience
 Official background source equivalent to estimate of
rate based on Lab G tests.
 All relevant physics processes simulated
 7 times Ganging
 Full simulation, digitisation, pattern recognition and
track fitting/extrapolation
Main worry (and deep reason for TPG preference) is that traking with only
5 points makes it delicate to estimate precisely the efficiency (which will
NOT be 100%) and to be certain that it is not a biasing factor.
My wish: I would like a prototype that shows that we can really achieve
the ‘nominal conditions’: 100% hit efficiency(?), 0.25% dead channels in
the MICE tracker, and that a tracking algorithm is able to recover from
the resulting inefficiency.
Alain Blondel
Pattern Recognition Efficiency
What would be this result if the efficiency and dead
channel were those achieved in the prototype ?
 Number of space
points in fitted track
(or zero if no track
fitted)
 Most of the time 5
space points are made
and fitted in track
 With no dead channels
or background:
99.9 ± 0.1 %
 Overall tracking
efficiency in nominal
situation:
99.0 ± 0.5 %
Alain Blondel
Why do we bother with a TPG?
Results
Strip
number
up&down stream, 1 muon, 1 graph = 1 layer
Sampling number
Alain Blondel
Why do we bother with a TPG?
5 points per track: a
little noise and inefficiency and you are lost.
Alain Blondel
Typical cosmic-ray event
By M.Ellis
This is very nice, but we need an analysis of Cosmic ray results => eficiency etc…
Alain Blondel
TPG
Alain Blondel
At first sight TPG looks like a killer:
-----
better momentum reconstruction (factor 3)
less X0
many points and unbeatable pattern reconstruction capability.
it is also less expensive.
-- only design default: long integration time ~50 microsecond during which
background accumulates, this is compensated by lower Xo and higher
number of time slices.
Alain Blondel
However, the TPG is a more delicate object as we could see very
clearly:
1. Construction of GEMs and hexaboard is new technology. How many
prototypes will be needed to reach acceptable performance?
2. Aligment and control of distortions is a notoriously difficult problem.
Existing ideas are based on deposition of permanent radioactive sources
on the field membrane, followed by measurements with B field
on/off/reversed. (It is not possible to align the chamber with B-off
tracks as will be done in Sci-fi.) This also can be done in the foreseen
test, but has not been done as of today.
These ideas have not been worked out and have not been demonstrated
experimentally. Demonstration of alignment procedure is essential.
Alain Blondel
Conclusions
The Sci-fi group has carried out many of the requirements !
There remain several serious questions related to
Efficiency, light yield, dead-channels and their effect on
resolution & efficiency
The Pt resolution is not far from the upper limit that we fixed
ourselves and we must make sure that no unwanted effect will
push it over the edge.
=> Document work and have cross-checks performed!
The TPG is clearly very promising but has encountered difficulties
in the construction and the situation of MICE in Italy did not help.
Alain Blondel

similar documents