5.
Front-end Electronics, trigger and DAQ
5.1 Trigger
Trigger is the fast real-time event selection and control system for BESIII. It
selects physics interested events from enormous background and suppresses
background to the level that the DAQ system can sustain. As the BEPCII will operate
with a beam structure of almost continuous bunches 8ns apart that will achieve a peak
luminosity of 1033 cm-2 s-1 , which will produce much more good events and
backgrounds, a new trigger system must be built to match the new beam structure and
a higher event rate.
5.1.1 Backgrounds and rates
1)
Backgrounds
A good understanding of the expected trigger rate for luminosity of 1033 cm-2 s-1 is
desirable to determine the architecture for the trigger and front-end electronics. A
good way to estimate background rate is simulation of lost beam particles and
synchrotron radiation which needs a lot of time and good simulation tools. Here we
estimate with our experience with BESII. The main backgrounds in BESIII will be
the machine backgrounds due to electromagnetic processes, Bhabha and radiative
Bhabha scattering from collisions of the electron and positron beams, as well as
Coulomb scattering and Bremsstrahlung in the residual vacuum chamber gas and
finally Touscheck scattering in the single bunches.
The beam current of
BEPCII will be 40 times as
high as that of BEPC (I=2.2
A, 1.1×1013 electrons in the
beam) with beam lifetime to
be only half of that of BEPC
( τ BEPCII =3.5~3.8hr).
Therefore the electron lost
rate, correspondingly the
beam background also, will
be 80 times of that of BEPC.
Electron and positron lost
rate is estimated to be ( dn /
dt ) = 8.7 × 108/s. Suppose
the lost electrons and
positrons are distributed Figure 5.1-1. background rate from one wire in
uniformly around the BEPCII first layer of MDC vs beam current
ring, the number of lost e+, ehitting the detector is ( dn / dt ) BESII ≈ 1.3×107/s, i. e., a background event rate of
13MHz. This is a conservative estimation, because the lost of electron and positron
would be mainly around the quadruple, not uniformly along the beam pipe. Figure
5.1-1 is a measurement of background in MDC-II versus beam current in BEPC.
With the experience of BESII trigger, we estimate the beam background event rate
that may pass the trigger as Nbeam = 1500Hz. The full size of BESIII will be roughly
the same as BESII so the cosmic ray event rate hitting on BESIII will be 170 m -2 s-1
× 3 m × 3 m = 1500 Hz. In BESII, 95% of cosmic ray background is suppressed by
TOF with a time window of 40ns. While in BEPCII the bunch spacing is too small to
use a time window. By applying vertex detector as in BESII, this rate can be
suppressed to 40 Hz.
2) Physics events
The design of BEPCII will be optimized at J/ψ energy with highest luminosity L=1
×1033cm-2s-1. The event rate will be 1500Hz and 420 Hz at the J/ψ and ψ´(2S)
resonance. The Bhabha event in the detector coverage ( |cosθ| < 0.95 ) will be NBB =
L * σBB
= 550 Hz
which is another source of background. As Bhabha
event will be used for detector calibration and luminosity measurement, it can not be
fully ejected. We use a prescaler to reduce bhabha trigger event to an acceptable rate.
3)
Total event rate
Assuming we adopt the same trigger criteria as in BESII, BESIII would suppress
event rate from 1.3 × 107 Hz to maximum event rate about 3000Hz (accidental
coincidence influence not included which will be studied).
5.1.2 The principle of BESIII trigger
As we know the purpose of a trigger system is to select physics interested events
from enormous backgrounds and to suppress backgrounds to a level that the DAQ
system can sustain. The BESIII DAQ is quite safely designed for a maximum
throughput of 3000 good events per second. The trigger system must therefore
reduce the rate of various backgrounds and bhabha events discussed above down to
1500 Hz while keeps high efficiency for J/ψ and ψ´(2S) decays.
1) Requirement of BESIII trigger(pipelining)
The aim for a trigger system design is to keep the total dead time as small as a few
percentage, which corresponds to an acceptable loss of luminosity as it is in BESII.
Formerly this is achieved by designing a trigger system with levels. The lowest level
is very fast and reduces much of the backgrounds, but does not introduces dead time.
High level is slow but because the rate is low so the dead time introduced is low. But
this scheme can not be used in BESIII because we face different constrains.
Unlike in BEPC where the bunch spacing is 800ns which leaves trigger system
enough time to process various subdetector signals and make decision before next
collision, in BEPCII the bunch spacing (not finally decided yet but likely to be) is
only 8ns, it is not possible to generate a trigger in such short time.
This situation is further complicated by the fact that the arrival time of the Time of
Flight (TOF) signal has an intrinsic spread of 30ns due to the different decay
products of the J/ψ and ψ ' resonance with different momenta and due to different
hitting position at the scintillater which makes it not possible to identify a single
bunch.
The TOF signals will be ready in 30ns after the collision. While in the drift
chamber, the drift time is about 600ns, only after that time the wire signals can be
used for trigger. Again in the
electromagnetic calorimeter of BGO,
because of its slow rising time and
trailing time, only after 1.5μs
calorimeter signal can be used for
trigger.
Pipeline must be used in trigger to
overcome these constrains. From these
constrains we know that it is not
possible to determine in hardware from
which bunch an event originated. This
fact gives us the freedom to set trigger
sampling period several times longer
than the bunch spacing, easing
considerably the design of the pipeline
Figure 2. BESIII event flow chart
components. The period will be chosen
as 50ns which is six times the bunch
spacing, so the signals from different sub-detectors will be binned into 50ns wide
time slices and be processed in each pipeline step.
2) BESIII trigger
Considering all the above mentioned constrains and requirement, the BESIII trigger
system will be in two level scheme, level 1 for hardware trigger and level 2 for
software event filter as shown in Figure 5.1-2. The signals from different subdetectors are splitted into two, one is digitized and stored in the pipeline in front-end
electronics (FEE), the other is used in the level 1 hardware trigger to be processed to
make a trigger decision. There is a latency between the level 1(L1) trigger signal and
the originating event time. We set the trigger latency 2.4μs (an integer multiple of
800ns is a proper choice to adapt to different BEPCII operation modes). When there
is an L1 signal, the DAQ moves the data in the FE pipeline to buffer, and packs them
into an event, and send them to farm machine where the level 2 filter filters the
backgrounds events further.
5.1.3 implementation
1) System Overview
A schematic view of the BESIII trigger system is shown in Fig. 5.1-3. Electronic
signals from sub-detectors are received and processed by the appropriate circuits in
separate VME crates to yield basic trigger primitives such as the hit count in TOF,
the hit count and topology in VC, track count in the drift chamber, as well as the
cluster count and topology in the electromagnetic calorimeter. The information from
these sub-systems is correlated by global trigger logic (GTL) which generates an L1
strobe every time a valid trigger condition is satisfied. The L1 signals are
conditionally passed by the data flow control circuitry to the fan-out modules for
distribution to the data acquisition system(not shown). A L0 signal is also produced
for FE electronics in case of demand as the FE E have not been finalized.
2) Vertex
The scheme of Vertex Chamber has not finally determined but is very likely fiber
type. Experience from BESII showed that VC is very effective in suppressing the
Figure 3. Schematic diagram of BESIII trigger
cosmic ray background, so the count of hit will be used, and possibility of tracking
and timing will be also studied.
3) MDC Tracking
There will be axial and stereo layers of sense wires in MDC-IV. The tracking
component of the MDC-IV trigger will use only axial wires in first phase. The tracker
examines the complete set of 1704 wires, 16 axial layers (4 super layers each with 4
layers), for all possible valid patterns caused by tracks having transverse momentum
greater than 150 MeV/c. The tracker consists of 2 parts: segment-finding and trackfinding. The former will locate in the FEE board to reduce connection cables. The
result is sent using LVDS signals to the track-finding logic. The maximum drift time
is about 600ns which must be taken into account in the design of the segment-finding
and track-finding logic. Field programmable gate arrays (FPGAs) will be used to
perform the pattern lookup at the pipeline clock.
4) Electromagnetic calorimeter
The BESIII electromagnetic calorimeter (ECAL) will comprise 19,200 BGO
crystals in the detector barrel part, arranged 96 along the beam(θ) by 200 in
azimuth(φ). For the triggering, the signals from 16 crystals (4θ × 4φ) are summed as
basic trigger element, therefore there are 24 × 50 trigger cells. The summed signal in
a trigger cell (cell sum) is discriminated by both a high level comparator (roughly
twice minimum ionizing energy) and a low comparator (about half minimum
ionizing), resulting in one high bit and one low bit. All of these 1200 high and low
trigger bits are read out and made available to further software filter. For the
hardware trigger , 60 of the basic trigger bits will be OR-ed together (12 in z and 5 in
φ) to arrive at a 20-segment topology. The calorimeter energy from a shower is
usually shared among several crystals. As a result, the efficiency for setting a
threshold bit in a particular region of the calorimeter depends both on energy and
position: a shower shared by crystals spanning a boundary can be below threshold in
both regions. Monte Carlo simulation will be studied to reduce the complications
associated with boundaries in the calorimeter by creating overlapping “tiles” by
forming analog sums of signals from larger groups of crystals.
5) Time of Flight
The Time of Flight counter of BESIII consists of 192 scintillators in two layers
with 96 in each layer. Photo multiplier signals are discriminated first and then the
number of hits is calculated which is used to produce the spare L0 trigger signal.
These hits are matched with MDC tracks, possibly also VC tracks, and ECAL
clusters. The matched number of tracks are used in the final decision.
6) Trigger Timing and Control
The clock frequency of storage ring of BEPCII is 500MHz. The clock of FEE
pipeline and trigger pipeline will be derived from this clock. The pipeline clock will
be 20MHz and all the processing of the sub-detectors signals in pipelined trigger will
be under control of this clock.
7) Global Trigger Logic
The primitive signals from sub-detector are fed to the Global Trigger Logic(GTL).
They are matched there to define a valid event trigger L1. The L1 signal is
synchronized to the trigger pipeline clock to have a fixed latency of 2.4μs with
collision.