The trigger processor Triggering is performed in several levels of complex-ity, where the earliest decisions are the most primitive VII... The dead time due to background events is negli
Trang 1North-Holland, Amsterdam
Section VII Triggering systems
TRIGGER USING TRACK SEARCH AND KINENIATICAL ANALYSIS
FOR RARE DECAY CHANNELS AT HIGH RATES
The CP-LEAR Collaboration
Dominik Andreas TRÖSTER'), Ronny ADLER'), John Richard FRY 8) , Theodoros GERALIS 7l,
Claude JACOBS 3), Emanuel MACHADO 4), Panagiotis PAVLOPOULOS 3), Charles RHEME 6), Daniel SACKER'), Guido TARRACH'), Panagiotis TSILIMIGRAS 3) and Edmond John WATSON 3)
Basel University, Switzerland, 21 Boston University, USA, 3)CERN, Geneva, Switzerland 4)LIP Coimbra, Portugal, sl Delft Univer-sity, The Netherlands, 6)Fribourg University, Switzerland, 7) Ioannina University, Greece, 81Liverpool University, England, 9)Marseille CPPM, France, 10) Saclay CEN DPhPE, France, 11JManne Siegbahn Institute, Sweden, 12)PSI, Switzerland 13) Thessaloniki Univer-sity, Greece, 14)Zürich ETH, Switzerland
We report on the definition and construction of the trigger system for the CP violation experiment with tagged K° (PS195) at LEAR, the low energy antiproton machine at CERN The beam is assumed to have a continuous intensity of 2 x 106antiprotons per second The requirements for fast and efficient rejection are stringent
1 Introduction
With improving accelerator technology, both the
en-ergy and intensity of available beams increase. LEAR,
the low energy antiproton ring at CERN, is optimized
to deliver a pure do antiproton beam of low energy
Each spill lasts about an hour; the time lapse between
spills is several minutes This quality of beam enables
high statistics experiments, e.g to search for small
viola-tions of known symmetries This provides an alternative
source of knowledge complementing high energy
experi-ments
The PS195 experiment, also termed CP-LEAR [1],
uses a classical 4m detector like the ones used in
collid-ing beam experiments It should be capable of handlcollid-ing
the low energy antiproton beam of LEAR with
intensi-ties of up to 2x 106 s-'. The experiment aims to
measure CP violating parameters by tagging the
strangeness of neutral kaons from the
antiproton-pro-ton annihilation pp - (K+m-K°) or (K-m+K°) with a
branching ratio of 4x 10 -3 per antiproton The event
size is in the kilobyte range, hence unwanted reactions
must be rejected early and efficiently Otherwise, the
data rate to be written to mass storage media would
exceed present technological capabilities
Several triggers for physics, monitoring, and
calibra-tion purposes need to be implemented The trigger must
be sufficiently modular to offer the flexibility and
potential for later refinement and enhancement It is
required that the trigger be easily testable at any time
between spills in order to reduce the mean time to
repair
0168-9002/89/$03 50 C Elsevier Science Publishers B.V
(North-Holland Physics Publishing Division)
2 The detector The PS195 detector is mounted in a solenoid of 3 6
m length and 1 m radius producing a magnetic field of 0.43 T parallel to the antiproton beam (the z axis) to measure charged particle momenta The detector is sim-ilar to a colliding beam experiment Moving radially outwards from the center with a pressurized hydrogen gas target, there are tracking devices (TD) (r = 9-62 cm), particle identification devices (PID) (r = 62-75 cm), and the electromagnetic calorimeter (= 75-100 cm) The whole is surrounded by the aluminium coil A drawing is found in the contribution by Rickenbach [2] The TD consist of two proportional chambers (PC), six drift chambers (DC), and a double layer of streamer tubes for fast information on the z coordinate The chambers are sectorized into 64 parts, each one group-ing several wires, whilst the tubes are sectorized into 32 planks The PCs have a pitch of 1 wire/mm The DCs have a single layer of twin wires for on-line left/right ambiguity suppression There is one doublet per cm; the pitch using the TDCs is 0.5 mm
The PID [2] consist of a Cherenkov detector sand-wiched between scintillators They are all read by PM tubes and thus the PIDs are the fastest devices The electromagnetic calorimeter is required for pho-ton detection It is not used for the trigger purposes
3 The trigger processor Triggering is performed in several levels of complex-ity, where the earliest decisions are the most primitive
VII TRIGGERING SYSTEMS
Trang 20
2
-2
-3
3.1 Trigger concept
I ,
Late Intermediate ~ decisions Early
i Log(hme(nsl)
Fig 1 The proportion of background events per good event as
a function of time and decision level The dead time due to
background events is negligible already after the intermediate
decisions This plot is based on simulations using the actual
algorithms of the trigger processor
ones simply rejecting obviously unusable events Every
level is optimized to improve the signal/background
ratio while keeping dead time low
The decision are taken in three levels : early,
inter-mediate, and late Early decisions operate on detectors
with very fast response: the beam counter and the PIDs
Intermediate decisions operate on the TDs, but just
performing coincidences for fast pattern recognition
Late decisions use track parameters and kinematical
constraints with full resolution
There are decisions which are done in parallel for
other annihilation channels or calibration purposes and
for minimum biased events If any one of these parallel
triggers agrees with the event, it is accepted
As time goes by, and trigger levels are passed, the
signal to background ratio improves from 0.01 to 300
according to Monte Carlo simulations and as shown in
fig 1 The overall trigger efficiency for good events is
about 0.15, taking account of losses due to dead time,
detector inefficiency and solid angle
The trigger decision time is greatly reduced by the
use of tables for wire number translation, trigonometric
functions etc., and it uses constants for comparisons
and computation of kinematical quantities These can
be programmed remotely, updated, and verified at any
time This gives the trigger the required flexibility in
order to search for other decay channels
The trigger processor is a "pipelined dataflow
ma-chine" This is a novel type of processor, a block
diagram is shown in fig 2 The term "pipelined" refers
to the presence of registers between the functional units
These registers decouple them, store intermediate
re-sults, and allow independent dialogue with the periph-ery Due to their programmability, the registers are an important tool for test purposes The term "dataflow" indicates that functional units are put at fixed places, but addresses and data (tagged for validity and particle type) seek their way through Several tracks can be followed through the detector at the same time
In each decision level, several tests can be performed
in parallel and serially The trigger sequencer element (TS) takes account of the outcome of these tests and decides whether the current event must be evaluated any further or whether it may be rejected The TS is an operating system built in hardware It is a sequencer which operates at 100 MHz
3.2 Early decisions
Early decisions are taken within 60 ns since the charged kaons in the relevant momentum range
Fig 2 The trigger processor block diagram The top left column represents the front-end electronics containing dis-criminators, TDCs and ADCs, and with which the next col-umn, consisting of track follower elements, is performing a dialogue Processing goes from top to bottom The next col-umn, consisting of delay elements, collects the data and groups them for each track Its output is routed to several elements (SELECT, TRACK P, CALIBRA, MOM, COMPENS) for parametrization and stored in appropriate registers These data are then used to calculate missing masses (MIM) and time of
flight (TOFU)
Trang 3(300-750 MeV/c) can be identified by using the PIDs.
In the early level, the existence of a kaon candidate is
checked by requesting, for any segment, coincident hits
in the scintillators (to make sure that the track of the
kaon candidate crosses the Cherenkov detector between
them) and absence of a hit in this particular Cherenkov
Simultaneously, the multiplicity of inner scintillator hits
is evaluated This is followed by the multiplicity of kaon
candidates If there are fewer than two hits in the inner
scintillators or no kaon candidate, the event is rejected
If the beam counter signals another incoming
antipro-ton, the event is rejected, in order to avoid overlapping
signals in the TD Otherwise, a "general strobe" is
generated At this level, only events with kaons or slow
pions (with a momentum of less than 300 MeV/c) can
provide a "general strobe"
3.3 Intermediate decisions
Intermediate decisions are taken within 250 ns In
the intermediate level, there are several activities Early
information of the TDs is used to detect the presence of
tracks and to count "primary tracks" in general, and
"kaon tracks" in particular (Primary tracks come from
charged particles emerging of the annihilation, whilst
secondary tracks stem from the decay of neutral primary
particles Kaon tracks are primary tracks with the
sig-nature of a kaon candidate in the PIDs The other
tracks are considered to be pions.) Simultaneously, the
transverse momentum of the kaon candidate is
esti-mated
Primary tracks have information inside a given range
around the position of the corresponding PID in at least
one of the PCs, one of DC(1) or DC(2), and one of
DC(5) or DC(6) If this is not the case, the track could
not be followed reliably through the TDs in the later
processor stage The logic is done with coincidences on
the crude hit map data, with the detector split into 64
segments If there are fewer than two primary tracks or
no kaon track, the event is rejected at this stage
The transverse momentum of the kaon is crudely
estimated by requiring coincidences of sets of wires in
DC(1) and DC(6) with the kaon candidate signature in
the PIDs; in detail: some 15 wires of DC(6) are in front
of each PID module, at least one of them must have
fired Assuming circular tracks originating from the
center of the target, one of three to nine wires in DC(1)
depending on the momentum, must have fired in front
of this wire in DC(6), too If such a coincidence if
found, it is assumed that a charged kaon is present
Otherwise, the event is rejected A threshold of 300
MeV/c on the "kaon candidate" transverse momentum
will ensure that all the events with slow pions faking
kaons are rejected, whereas events with charged kaons
(P<0.84) will survive
This trigger stage is implemented with ASICs
(appli-cation specific integrated circuits) One ASIC contains all circuitry for the coincidences related to one PID module, requiring some 650 gate functions It is made in TTL compatible CMOS and returns the result within 30
ns A solution using standard components would not be faster nor cheaper nor safer, but much more space-con-suming
3.4 Late decisions
Late decisions are taken within 5 Ws In the late level, the tracks are followed in two passes The first, crude pass uses the segmentation in 64 of the TDs in order to associate hits into tracks The fast pattern recognition algorithm is based on the low track multiplicity of the
pp annihilation at rest (maximum 6) and has an ef-ficiency of 98% per track
The electronics of the TDs performs real-time digiti-zation (within 700 ns) The precise information (1 mm and 0.5 mm bin equivalent, respectively) is available through a dedicated controller, of which each TD (and PID) has one The raw track information is then used to acquire the precise digitizations using a simple dialogue mechanism - the second pass This dialogue is done simultaneously in successive TDs for all tracks (for example, track 1 in DC(3), and track 2 in DC(4)) Using these data, the primary tracks are parametrized (momentum, charge, starting angle) through different algorithms via LUTs (look-up tables) and linear ap-proximations
The precise kaon momentum is checked against a minimum value If its momentum is too low, it is considered to be a fake kaon The event is rejected If not enough tracks can be parametrized, the event is rejected, too If there are two or four tracks, the charges must be balanced There must be at least one track of each charge sign, but not more than two of the same sign
The kaon track is then combined with the pion tracks of opposite charge to get the missing mass of the primary tracks If within the resolution it does not match with the neutral kaon mass, the event is rejected The primary tracks are double-checked (in TOFU) for their identification: the amplitudes in the scintilla-tors are used to correct the time of flight From the particle identification and momenta, the expected time
of flight difference in the scintillators is evaluated and compared with the measured value If these quantities
do not match within their resolution, the event is re-jected
At this level, the event is considered to be a candi-date "good event" and is written to tape
3.5 Trigger monitoring
Due to the remotely controllable pipeline registers, the trigger processor is easy to test Characteristic data
VII TRIGGERING SYSTEMS
Trang 4can be inserted at any place to debug the equipment
between two programmable elements
An autonomous readout processor, called spy, is
provided on top of the whole trigger processor The spy
has up to 64 parallel data input channels Each channel
can acquire trigger processor data independently and at
full rate, i.e 30 MHz Any cable between two trigger
elements can be fed to the spy for reading out the data
on this cable This is very useful in the startup phase
and it is vital in the production phase, since the
ex-istence of intermediate results on tape speed up off-line
evaluation considerably The pattern recognition and
track finding are time consuming tasks when
pro-grammed on a standard computer The existence of a
first guess for the tracks will offer more time for fits
which are not foreseen on the trigger processor
4 Elements
The electronic modules built for the CP-LEAR
trig-ger processor are called elements The most dedicated
ones are described below
4.1 Triggersequencer
The trigger sequencer element is the most dedicated
one and will be briefly described first It contains the
philosophy of the trigger behaviour which can probably
be applied to many other experiments
It contains an algorithmic state machine which runs
at 100 MHz It has inputs for launching the trigger
processor (beam, test); for manipulating the fate of the
event under evaluation (accept, good forever, cancel
keep mode, panic, skip) ; for synchronization (busy,
kinematics done, drift time past, follow done, pretrigger
done) It controls any activity of the trigger, from the
acceptance of an antiproton to the end of the readout,
according to the state diagram in fig 3 To avoid
confusion, if more than one event happens before the
end of the early state, the sequencer resets via the
multi(hit) state
4.2 Simple logic
The terminator/converter and the
differentiator/de-glitcher take the role of "glue" to other logic (NIM,
TTL) and standardize signal shapes
There are several types of coincidences (32X2
in-puts, 16X3 inputs, 16X(2 inputs, 2 veto) and similar
OR gates There is a maskable 8 input AND/OR/
NAND/NOR with logic fan All these elements have
about 8 ns delay The multiplicity unit for adding 36
1-bit inputs gives the precise count and signals even/
odd, n=2, n=4, and n = 2 or n = 4 within 20 ns
4.3 Programmable logic
Fig 3 The trigger sequencer state diagram It shows an infinite loop which is characteristic of any operating system The bubbles show states in which the experiment can be while running, the arrows show possible transitions When switched
on, the system starts in panic state in order to set all equipment into a known - idle - state While idle, any kind of event is
accepted, but not more than one at a time If the event survives the different trigger levels, it is finally read out
There is an adder with four inputs and an arithmetic element with a 16-bit ALU (arithmetic logic unit) (15 ns) and an independent 16 X 16-bit multiplier (30 ns) The adder operates on 12-bit numbers and has outputs
Some elements are memory or register programma-ble, others are programmable via their inputs The first kind includes pipeline registers, timer/scalers, ap-proximately equal windows (an element with the dual functionality of approximately equal comparator and window comparator), look-up tables, and flexible mem-ories They are controlled via LORCOM, the system test network which is based in CAMAC The other kind
is represented by the data selector, the delay, the reg-ister control, the timing generator, the track follower, and the trigger sequencer The spy is somewhere in between
The data selector is an extendable multiplexer which selects data according to their validity The delay ele-ment (see fig 2) is used for synchronization of data acquired at different times The register control puts track parameters into the right slots of the missing
Trang 54.4 The CP-VME form factor
43 cm
16 cm
Fig 4 The CP-VME crate dimensions The crate is 9U high
and is shown with the optional 6U high VME partition Each
station may dissipate 4.5 A at -5 V, 3 A at -2 V, 1.5 A at +5
V on average Of the three buses, the topmost is dedicated to
power distribution
mass/invariant mass calculator The timing generator
provides the system clock, 10-30 MHz, by respecting
the speed of the slowest active element
In order to match our specific requirements, the
elements used in the trigger have a particular shape
They are fastbus high, but only VME deep
Ap-propriately, this format is termed CP-VME
Most logic requires between 64 and 110 contacts on
the front panel On the other hand, by making use of
the high circuit density offered by 100K ECL logic,
VME depth is good for most, if not all, applications
Following the VME standard allows us to mix VME
cards and CP-VME cards in one crate, as shown in
fig 4 The crate offers the two low buses as standard VME buses The upper bus is for power distribution
5 Summary The design of a trigger processor for real-time physics evaluation has been sketched The task is feasible, but involves the design of lots of new hardware (and soft-ware) According to Monte Carlo studies, the efficiency and timing requirements are met It has been shown that a new size of printed circuit boards is advanta-geous
Acknowledgements
We would like to thank the electronics workshops of Basel, the ELD groups of CERN, Coimbra, and Fri-bourg for their highly valuable suggestions, and their efficiency We are much in debt to the whole CP-LEAR group for the time spent in fruitful discussions The summer students S Andouche, B Edholm, P Lahary,
K Peters, S Vlachos each did a good job
References [11 L Adiels et al., Proposal for the experiment PS195, CERN/PSCC/85-6/P82, PSCC/85-30/P82/Add 1, PSCC/85-43/P82/Add 2, PSCC/86-34/M263, PSCC/87-14/M272
[2] R Rickenbach et al., these Proceedings (Int Conf on Advanced Technology and Particle Physics, Como, Italy, 1988) Nucl Instr and Meth A279 (1989) 305
VII TRIGGERING SYSTEMS
(power onlv)
CPVME
IEI
(o tionall i