236
IEEE TRANSACI'IONS ON NUCLEAR SCIENCE,
VOL.
37,
NO.
2,
APRIL
1990
RESULTS ON HEAVY QUARK SELECTION
THROUGH AN IMPACT PARAMETER TRIGGER
M.
Adamovitch,
Y.
Alexandrov,
S.
Gerasimov,
S.
Kharlamov, L. Malinina, M. Zavertiaev
Lebedev Institute of Physics, Moscow, USSR
F. Antinori, M. Dameri, R. Hurst,
B.
Osculati, L. Rossi, G. Tomasini
University
di
Genova and
I",
Genova, Italy
J.L. Bailly, A. Buys, F. Grard, P. Legros
Department of Physics, Mons, Belgium
A. Forino, R. Gessaroli, P. Mazzanti, A. Quareni, F. Viaggi
INFN
and University of Bologna, Bologna, Italy
C. Meroni,
N.
Redaelli, D.Torretta
INFN
and University of Milano, Milano, Italy
D. Barberis, W. Beusch,
F.
Bourgeois, A. Corre, M. Davenport, J.P. Dufey, B.R. French, A. Jacholkowski,
K.
Knudson, J.C. Lassalle, F. Muller, R. Pegaitaz
CERN, Geneva, Switzerland
Abstract:
this paper describes the implementation of a
trigger aimed at selecting those nuclear interactions which
contain a charm particle. The trigger uses a fast emulator of
the PDP-11 which calculates the Impact Parameter of the
decaying charged tracks originating from the charm decay. A
enrichment factor of
15
is thereby obtained.
1
-
INTRODUCTION
The new trigger technique presented here aims at selecting,
among numerous interactions produced by a hadronic beam on
a fixed target, those which show evidence for a secondary decay
close
to
the production vertex. This is achieved by searching,
among the secondary tracks, at least one that exhibits an
impact parameter (i.e. the distance between the primary vertex
and the direction of
the
track) above a minimum value.
This "Impact Parameter Trigger"
(IPT)
has been developed
and used by
the
WA82 experiment at CERN, running on the
OMEGA spectrometer. The physics motivation was the
collection
of
a large sample of heavy quark state decays
produced in hadronic interactions. Using a
a-
beam of
2
-
IMPACT PARAMETER TRIGGER: PRINCIPLE
AND SET-UP
The decay products of a relativistic particle of lifetime
2,
issued from a primary vertex, have a mean value of the impact
parameter given
by <d>
=
c
*
2-
For D or
B
mesons,
<d>
=
300
pm.
With this order of magnitude, a selection on impact
parameter values requires the use of an accurate detector, such
as a telescope of silicon microsmp planes. Accuracy of the
same order is also required on
the
primary vertex position.
In the design of the trigger system the following sources of
Any trajectory deflection or secondary interaction
between
the
beam
telescope and the downstream detector, including
its first plane.
contamination have
to
be considered:
-
-
Decays
of
strange particles.
-
Ghost tracks due to spurious combinations of
measurements in the downstream detector.
340
GeV/c, this
has
1989)
a
Of
55
.
lo6
events
to
now
(October
by
the
In*
The
Off-
The implementation
of
the IpT for the WAS2 experiment
aims
at
simplifying
the
trigger algorithm
as
much
as
in order to keep the decision time within values compatible
with the other dead time sources (interaction trigger and data
read-out). It is also designed to reduce the effect of some of the
contamination sources.
line analysis of one part (8.106 triggers) has allowed
us
to
KaTc*
Do
+
Ka7
Do
+
Kaax.
Some
750
decays
D'
A report on this mgger and its first results can be found
in
ref. [ll. Physics results
are
presented in refs [2] and
[3].
We present here the details on the implementation of the
impact parameter trigger
and
On
the software Part of the
decision logic.
this context, the beam and seconm tracks
are
measured
in a plane parallel to the magnetic field
of
the spectrometer
where they appear as straight tracks. The vertical coordinate
axis
Z
is parallel to the field,
the
X
axis is parallel to the
0018-9499/90/O4OO-0236$01.00
8
1990
IEEE
237
beam. The detector planes are perpendicular to the
X
axis and
measure the
Z
coordinate.
So
we shall only use the
2
component of the impact parameter.
Searching for a track segment
with
an impact parameter
value can then start. We require that this value be higher than
100
pm
and lower than 1000
pm
(in the 1989 run, the low
value was 50pm). To reduce the number of spurious
combinations, this segment must be composed of 3 points. In
addition, the contamination due to multiple scattering of low
energy tracks can
be
reduced by requiring a confirmation of the
The beam is measured with two
20
Pm Pitch microstrip
planes separated by
500
mm
from each other, the last one
being at 25
mm
in front of the target.
segment by one of the OMEGA proportional chambers which
is located at 2
m
downstream of the target. The horizontal
acceptance of this proportional chamber corresponds
to
a
It
is important to remove as many measurements as
possible that are due to the tracks coming from the primary
vertex: when searching for a track with impact parameter,
momentum
Of
p
'
3*5
GeV/c.
the number of combinations is reduced and
so
is the time
spent in checking for their validity. The probability of
3
-
THE TRIGGER PROCESSOR MICE
spurious combinations is
also
much smaller.
The impact parameter trigger algorithm is implemented in
.~
a program executed by a fast processor.
For
this purpose we
use MICE (standing for MICroprogrammable Engine), a
processor developed at CERN in 1980 [4]. It was
part
of the
OMEGA data acquisition system, until 1988, and was used for
front-end buffering and on-line filtering.
This suppression is simplified by using a thin metallic
target (1.3 mm) together with a special geometric arrangement
of the downstream microstrip detector. Figure
1
shows this set
up:
it is made of 3 planes of microstrips perpendicular to the
X
axis, having a pitch proportional to their distance from the
target.
For
all the tracks coming from the primary vertex, the
hit number, measured from the
Z
position of the beam, is the
same in the 3 planes.
MICE
has a cycle time of 105 nsec per micro-instruction.
It is based
on
the Motorola
~10800
MECL LSI
high
performance bit slice processor family. The standard
4
operations involved in the execution of an instruction (fetch,
decode, execute, increment PC) are pipe-lined and executed
simultaneously for consecutive instructions. The micro code
emulates the
PDP-
1 1 MACRO assembler language; this
feature allows the user to develop applications on a host, like
a VAX, that provides the appropriate tools. It is possible to
use the high level languages for which a suitable compiler
exists. MICE is fast: for example, the association of a
measurement to a plane and its normalisation takes some
2.6
psec, including the loop control.
MICE has a main memory of
56
Kbytes and a micro-code
storage
area
of 1024 words of 120 bits. The limited size of the
memory is not inconvenient as long as the machine
is
used
as
a de-randomizer
or
a mgger element. The
full
data acquisition
pSZ=SOpm
a15rm
pSk20prn
a6rm
Fig.
1
-
Layout
of
the
three
microstrip
detectors
Conversely, these hits will have different values if they
belong
to
a track emitted from a secondary decay, in the region
between the primary vertex and the first plane.
In fact, the overlap of corresponding strips, as seen from
the vertex, cannot
be
achieved exactly, firstly for geometrical
reasons, that are related
to
the finite size of the strips and of
the target, and secondly because of the inaccuracy in the
Z
position of the planes.The precise position of the planes is
determined using beam tracks. Consequently, we have to allow
for
a tolerance of at most one strip when removing
corresponding hits.
occupies 10 Kbytes only and the size of an event is rarely
exceeding
8
Kbytes.
Two interfaces are provided for data acquisition and
communication with the host: one for the REMUS and RMH
data acquisition systems ("read mostly" systems in
use
at
CERN since 1975, see
[5])
and the other for the connection
to
the host via CAMAC. The latter provides a
full
set of
functions in order to control MICE and execute block
transfers; it is implemented
as
a CAMAC module. Both of
these interfaces can execute
DMA
block transfers in cycle steal
mode. A set of coaxial connectors can be used to receive
interrupts
or
to send signals to the trigger logic.
All
the
interrupts are on the same level of priority.
238
addition, the cycle steal mode of these transfers leaves plenty
of
CPU time to execute tasks such as filtering of events or
software triggering. limits.
-
Search, among the remaining hits, for an aligned mplet of
points that give an impact parameter value within the
During the
1989
run, MICE was only used as a 2nd level
trigger and did not participate to the data acquisition (fig. 2).
A description of the new OMEGA data acquisition system
can
be found in
[5].
I
Main
RMH
int. trigger read-out
Ir
fl
ss
PP
bb
uu
MICE
trigger
data
rr
read-out trigger
I
reset
Down loading,
Start/stop.
calibration
CAMAC
U
URHst.
.earn
REMUS
FASTBUS
URH
UME
32
-
Extrapolate this segment to the proportional chamber and
search
far a hit within a tolerance
(4
x
2
mm
pitch).
The execution stops whenever a condition rejecting the
event
is encountered. In such case, MICE generates a signal
that
resets
the frontend system. These conditions
are:
-
-
A
time-out during data read-out.
An
empty microstrip plane after removal of hits pointing
to the interaction vertex.
-
An
insufficient removal of hits.
-
No triplet found with an impact parameter within the
specified limits.
-
No confirmation of the triplet in the proportional
chamber.
If the event is accepted, MICE generates a signal that
starts
the read-out of the whole detector. A block of data containing
some information on the IPT is added.
4.2
-
The search for a triplet of points with a good impact
value
This section illustrates one aspect of the optimisation
process of algorithms. One tries to avoid the combinatorial
method which generates
too
many hypotheses that have to
be
checked for an impact parameter value.
In what follows, a measurement, or "hit", is the number of
Fig.
2
-
Block
diagram
of
the data
acquisition
hardware
a micro-strip counted from the
z
position of the interaction
vertex.
4
-
THE IMPACT PARAMETER PROGRAM
4.1
-
The algorithm
The electronic interaction trigger, that includes a
multiplicity selection, ensures that the data in the microstrip
planes used for the
IPT
fulfil the following conditions: exactly
1 hit in each of the beam planes; at least 3 hits, and at most
16
in each of the
3
downstream planes and in the proportional
chamber.
On receipt of an interrupt generated by the interaction
trigger, the
RMH
system is used to read the data from the
IF'T
microstrip planes. The various steps of the trigger algorithm
are
then executed, namely:
-
Decode the
beam
microstrip information and calculate the
beam parameters.
Decode the
data
for the planes downstream and normalise
the hit values to the
Z
position of
the
interaction vertex.
-
Remove corresponding hits, scanning the planes by pairs.
-
Determine whether this removal was sufficient, the exact
criteria being: NP1 .NP2*NP3/(N1 -Nz-N~)
>
5
,where
NPi is the number of hits in
the
plane
i
before the
removal of hits and Ni this number but after removal.
-
With the geometrical arrangement of the planes as
described before, a segment defined by a hit in each of 2
planes, say I1 in plane 1 and I3 in plane 3, has an impact
parameter d which is proportional to the difference I3
-
11:
d
=
d13 I13
-
111.
The factor d13
is
a constant for a given set-up and is the
elementary impact increment, corresponding to a difference of
one. For planes
1
and 3
it
is given by:
d13
=
P3
'
Xl/(x3
-
xl),
with p3 being the pitch of plane 3,
Xi
and X3 the position of
the
planes. It is assumed that
:
XI/pI
=
X3/p3.
Considering
the
measurement I3 in plane 3 and the high
cut on the impact parameter value, IPmax, one can easily
determine
the
lowest and highest limits (hit numbers) where to
search for a point in plane
1
(fig. 3). These values, 11-low
and 11-high, are given by:
11-low
=
I3
-
IPmax
/d13
11-high
=
13
+
IPmax
/d13
The limits for the low cut on
the
impact value, Ihin, are
obtained by substituting IPmin to IPmax in these relations.
239
2
-
Code optimisation, such as use of the registers whenever it
is possible, selection of the branching instructions taking
into account their time asymmetry, and use of the special
instructions provided by the emulator.
4.4
-
IPT
program performance
The real time for executing the trigger algorithm is
350
pec,
averaging
on
all types of events (accepted and rejected). The
lowest and highest values are approximately 100psec and
800 psec. Adding the time to read the trigger data, the total
average time is 470
psec.
In the course of 3 runs (from 1987 until 1989) the total
number of events selected by this program amounts to
55
.
106.1f one adds another 15.106 events produced durin
tests and setting up runs, this total corresponds
to
about 10
events submitted to the IPT selection. The program proved to
be
very reliable as it never crashed nor entered into a loop.
8
I
2
3
Plane
target
Pitch
Pl
P2
P3
Fig.
3
-
Schematic
of
the boundaries
in
the first plane
An example of the rejects due to the various selection
stages is given in Table 1. The numbers are typical of a 1
spill sample.
To find a triplet, the algorithm is then:
Repeat until a valid triplet
is
found
or
until all data is used
Take next hit I3
in
plane
3;
compute the limits
in
plane 1;
search
for
a hit I1
in
plane 1 within these limits;
if
found
then
compute a prediction in plane
2
for the 3rd
point (It
is
1/4.11
+
3/4-13
in
this
set-up);
search
for
a hit
in
plane
2
within a tolerance;
end if;
End
repeat
Note that the search
for
hits in planes
1
and
2
can be
limited to those that are greater than the value of 11-low
calculated at the previous iteration.
Also note that the impact parameter value does not need
to
be calculated since the search is limited to a range of hits
which gives, a priori, a valid value
4.3
-
Characteristics
of
the
program and optim’sation
The IFT program is written in PDP MACRO-Assembler.
It includes the handling of all the interrupts and signal
generations,
as
well
as
the display of information at the end of
burst. The trigger algorithm is coded
in
some 750 instructions
and occupies 3.7 Kbytes, i.e.
7
%
of the
MICE
memory. It
relies entirely on integer arithmetic
The computing time per event could
be
reduced to a
Optimisation of the algorithms, in order to achieve a
dependency proportional to the number of hits in the
planes, rather than a combinatorial dependency. Sect, 4.2
shows, as an example, the algorithm to
search
for a triplet
having a suitable impact value.
-
Optimisation of the program structure to avoid
unnecessary data movements and detect the conditions of
the rejects as early as possible.
minimum by carefully optimising the code at several levels:
-
The total rejection factor of the IPT is approximately 8
when a minimum impact value of
50
pm
is
required and 14
for a minimum of 100
pm.
TABLE
1
-
Details of the rejects with a range of impact
parameter value from
50
to
1000
pm; the values in parentheses
correspond to a low cut at 100
pn.
after removal
tional chamb.
To determine the enrichement factor of the IPT, we have to
compare the number
of Ds
that can be extracted from a
set
of
events having passed the IPT with the
number
that is obtained
from a sample of events selected only by the interaction
trigger. As mentioned earlier, we found 750
Ds
in the 8.106
IPT events analysed
so
far (fig. 4a and ref [3]). A sample of
1.5.106 interaction triggers has been processed by the same
off-line chain of programs: 9
D
events were found (fig. 4b).
Taking into account the uncertainties on these numbers, one
arrives at the value:
Enrichment factor
(100
pm)
=
15.6
k
6
240
'
Beam
Interaction trigger
Interaction trigger
with full dead time
350
-
300
250
200
150
100
50
-
-
-
-
-
'
Multiplicity
Impact parameter trigger
Full
read-out
2.0
2.1
0
1.7
1.8
1.9
00.D.
+
invariant
mass
(GeVI
Fig.
4
-
(a)
Ds
extracted from a set
of
IF'T
events
3.0
2.5
2.0
1.5
1
.o
0.5
0
1.7
1.8
1.9
2.0
2.1
DO.D
+
invariant
mass
IGeVl
Fig.
4
-
(b)
Ds
identified
in
the raw data
5
-
PERFORMANCE OF THE OVERALL TRIGGER
Table 2 presents typical rates at the different stages of the
full trigger, as well as the dead times due
to
these selections.
The equivalent full dead time at the 1st interaction trigger level
is 240 psec/event, thus giving a total dead time ratio of
70
%.
TABLE
2
-
Trigger rates and dead times
Trigger stage
I
time/even t
(in
2.6
sec)
~
2 106
25 000
7000
1
830
217
217
Equiv:
240
58
470
2 200
6
-
CONCLUSION
The values of the 2 factors of enrichment (15) and rejection
(14), show that the IPT in its present implementation, while
giving a good enrichment factor, performs also a correct
rejection since it does not eliminate many good events that
would have
been
retained by the more sophisticated selection
of the off-line programs. It could of course
be
improved to
reach a higher rejection factor.
One can envisage to improve the selectivity by searching
for 2 tracks, with an impact parameter, which are crossing
in
the region between the target and the detector.
Another important improvement would
be
to reduce the
IPT
time which is, when combined with the trigger data read-
out, the main source of dead time. In future runs the reduction
will
be
achieved by applying the technique of the contiguity
processor
[6].
7
-
REFERENCES
J.F. Baland et al., Results
on
the Impact Parameter
Trigger for the selection of short lived particles,
M. Adamovich et al., Preliminary results on
$
production properties in the interactions of
340GeV/c
x-
on a target of Si and
W.,
24th
Rencontres de Moriond, March 1988.
M. Adamovich et al., Charm hadroproduction with an
impact parameter trigger, International Symposium
on Heavy Quark Physics, Cornel1 Univ., Ithaca,
June 1989.
J.
Anthonioz-Blanc et al., MICE, a fast user
programmable emulator of the PDP- 1 1, CERN-
DD/80- 14.
F.
Antinori et al., Rejuvenation of a
data
acquisition
system for fixed target experiments in a large
multiuser spectrometer at CERN. These
Proceedings.
G. Darbo et al., The DELPHI contiguity trigger
processor, a hardware and software overview, these
Proceedings.
Nucl. Phys. B1 (1988) 303-310.