The D0 upgrade will be a key element of the attack on physics in the new high luminosity Main Injector environment. The upgrade plan [1], first proposed in October 1990, has been endorsed by the PAC and Director's Review. It builds on the strengths of D0, full coverage in calorimetry and muon detection, while enhancing the tracking and triggering capabilities. Construction is now underway on a number of the detector systems. It is thus an appropriate juncture at which to review the D0 upgrade detector systems and the physics performance we will achieve with the upgrade.
In the next section we summarize the design of the major upgrade detector systems sufficient to give a broad overview, without excess detail. An overall view of the D0 detector is shown in Fig. 1 with the primary detector systems indicated. A major element of the upgrade is the replacement of the inner tracking systems, required because of the expected radiation damage to those detectors by Run II and also to improve the physics capabilities of the D0 detector. The upgraded tracking system consists of an inner silicon vertex detector, surrounded by four superlayers of scintillating fiber tracker. These detectors are located inside a 2 Tesla superconducting solenoid. A scintillator based central preshower detector with wavelength shifter readout is located between the outer radius of the solenoid and the inner radius of the central calorimeter cryostat to provide electron identification and to compensate for energy losses in the solenoid. A detailed view of the D0 upgrade tracking system is shown in Fig 2. In the forward region, the design of the preshower detector is presently under study. We had considered small angle tracking detectors, but eliminated them in favor of extended fiber barrels and small angle silicon disks due to cost considerations. The higher event rates in Run II have led us to add new muon trigger detectors covering the full pseudorapidity range. Electronic upgrades are driven by the need to handle a smaller bunch spacing and provide pipelining of the various front end signals from the tracking, calorimeter, and muon systems. The front-end electronics for all these systems will be replaced. The addition of new trigger elements and front-end electronics requires a new trigger control system.
The momenta of charged particles will be determined from their curvature in the 2T magnetic field provided by a 2.8m long solenoid magnet. The superconducting (SC) solenoid, a two layer coil with mean radius of 60cm, has a stored energy of 5MJ (for reference the CDF coil is 30MJ). Inside the tracking volume the value of sin$\theta \times \int {\rm B}_z{\rm d}l$ along the trajectory of any particle reaching the solenoid is uniform to within 0.5%. This uniformity is achieved in the the absence of a field-shaping iron return yoke by using two grades of conductor with higher current density near the ends of the coil. From the value of the field integral and the space point precision provided by the silicon and fiber tracking system, $\Delta p_T/{p_T}^2\approx0.002$. The SC coil plus cryostat is about 1.1 radiation lengths thick.
The solenoid is being built by Toshiba Corp. in Yokohama, Japan. They are under contract to provide the solenoid as specified by Fermilab. The contract was awarded in January 1995 and Toshiba presented its preliminary design to Fermilab in mid-March. Delivery of the magnet to Fermilab will occur in late 1996 after complete testing in Japan.
The cryogenic plant that supplies LHe for both the solenoid and the visible light photon counter (VLPC) readout devices is in the final stages of design and the initial stages of construction. After appropriate modifications to the D0 cryogenic services building, the Accelerator Division (AD) will provide the experiment with warm high pressure He. The expansion engines and heat exchangers, etc. used in making LHe, will be located in the D0 Assembly Hall. Modifications to the AD0 cryo building will begin in the 1995 summer shutdown.
The tracking system was designed to meet several goals: momentum measurement with the introduction of a solenoidal field; good electron identification and $e/\pi$ rejection (to compensate for the loss of the TRD); tracking over a large range in pseudorapidity ($\eta$~$\approx$~$\pm$3); secondary vertex measurement for identification of b-jets from top and for b physics; second level tracking trigger; fast detector response to enable operation with a bunch crossing time of 132 ns; and radiation hardness. The silicon tracker is the high resolution part of the tracking system and is the first set of detectors encoutered by particles from the collision.
Several of the Run II Collider machine parameters have an effect on the silicon design. The luminosity sets a scale for the radiation damage expected over the life of the detector, which in turn dictates the operating temperature (<10 degrees centigrade). The long luminous region length sets the length scale, and motivates our hybrid disk and barrel design. The crossing interval sets the design parameters for the electronics and readout.
Since the Collider interaction point is extended, with a $\sigma_z$ of 25cm, it is difficult to deploy detectors such that the tracks are generally perpendicular to detector surfaces for all $\eta$. This forced us to a hybrid system, with barrel detectors measuring primarily the r-$\phi$ coordinate and disk detectors which measure r-z as well as r-$\phi$. Thus vertices for high $\eta$ particles are reconstructed in three dimensions by the disks, and vertices of particles at small $\eta$ are determined by the barrels.
The interspersed disk and barrel design is shown in Fig. 2. In such a system, the disk separation must be kept small to minimize extrapolation errors. However, each plane of disks also introduces a dead region between the barrels which lowers the overall efficiency of the detector. Thus there is a compromise between vertex resolution at large $\eta$ (~1/disk spacing) and efficiency at small values of $\eta$. This design clearly puts a premium on minimizing the gap between barrel sections.
We have chosen the following overall geometry:
Table 1 shows the numbers of detectors and the geometric parameters of the tracker. The 12cm long barrel segments are separated by 8mm gaps containing F~disks at z=6.4cm, 19.2cm and 32.0cm. Three more F disks are located at each end of the barrel at |z|=44.8cm, 49.8cm and 54.8cm. The H disks are located at |z|=110cm and 120cm.
The barrels and the F disks are based on 50 $\mu$m pitch silicon microstrip detectors, 300$\mu$m thick, providing a spatial resolution of approximately 10 $\mu$m. The small angle stereo design provides good pattern recognition with a resolution in r-z at the vertex of 0.5--1.0~mm, allowing separation of primary vertices due to multiple interactions. The detectors are AC coupled; each strip has an integrated coupling capacitor and a polysilicon bias resistor. This technology has been shown to be sufficiently radiation hard.[2] F disks are made from 12 double-sided detectors which have $\pm15^\circ$ stereo strips.
Results from tests of prototype F disks fabricated at Micron Semiconductor (UK) have shown that these detectors are suitable for use in D0.[3] Leakage currents were <120 nA/cm$^2$; coupling capacitors had a capacitance of >20 pF/cm and a breakdown of >80 V; and the polysilicon resistance was >3.5 M$\Omega$ with strip-to-strip non-uniformity of <3%.
Orders for all barrel detectors have been placed with Micron Semiconductor. The first pilot run (10 detectors) was completed on April 1, 1995 and we are currently evaluating these detectors. In a test beam run planned for April 1995 at TRIUMF we expect to re-verify radiation damage properties of the prototype F disk and barrel detectors.
The SVX II front end readout chips are mounted on a kapton high density circuit (HDI) which is laminated onto a 300 $\mu$m thick beryllium plate and glued to the surface of the detector. The end of the HDI consists of a kapton strip cable which carries signals and bias voltages to the outer radius of the detector (~18 cm) where a connection to a long (~8 m) low mass microstrip cable is made. These cables carry the signals to the port cards located on the D0 support platform.
A prototype HDI has been designed and is in fabrication. However, due to delays in the manufacturing process we are actively seeking an alternate vendor and expect to procure and test the prototype by August 1995.
The mechanical structure must provide a precise and stable support for the individual barrel and disk detectors, provide cooling for the heat generated in the SVX II chips and allow for the necessary cable paths for external connections. In the barrels, the basic mechanical unit is the ladder. Each ladder supports two detectors wire-bonded together, forming a 12 cm long unit with the SVX II readout at one end. A rohacell-carbon fiber support provides extra rigidity to the ladder. The ladders are mounted on beryllium bulkheads, which serve as a support at both ends of the ladder and provide cooling at the readout end by means of an integrated coolant channel.
The F disks are mounted in the 8 mm gap between the barrel segments. In analagy to ladders, disk modules consist of a single F disk detector with SVX II readout at the outer radius. Water cooling is via a beryllium cooling channel which also supports the modules at the outer radius.
The barrels and disks are mounted in a double-walled carbon half-cylinder which acts as a support with zero thermal expansion. The half-cylinder has a length of 2.2 m and an outer radius of 15.3 cm. As detector elements are installed, compensation will be made for the predicted 100 $\mu$m half-cylinder gravitational deflection.
Designs completed in the mechanical systems include: the beryllium bulkhead and fabrication of prototype straight sections to test the cooling channel for leaks; the single-sided ladders, and the cooling system, including finite element analysis. Fabrication of a 10-inch long prototype half-cylinder support is complete.
Barrel and disk modules will undergo thorough testing to avoid using modules with unacceptible fractions of dead channels. This will be done in stages, starting with probe testing of the individual silicon detectors and SVX II chips. Due to the large numbers involved, three sites have been set up with probe station facilities: Fermilab, UC Riverside and Oklahoma University. Further testing of HDI's instrumented with SVX II chips will be performed and, finally, the fully assembled ladders and disk modules will be checked with a 1064 nm wavelength laser, which simulates a charged particle track traversing the silicon.
Design and implementation of all assembly sequences and fixtures has been started. A detailed ladder fabrication sequence has been defined and a prototype ladder assembly fixture has been designed and fabricated.
Some important milestones for the silicon tracker R&D and construction are given below.
A scintillating fiber tracker (CFT) surrounds the silicon vertex detector and covers the central pseudorapidity region. The fiber tracker serves two main functions. First, with the silicon vertex detector, the tracker enables track reconstruction and momentum measurement for all charged particles within the range ${\eta} = \pm 1.7$. Second, the fiber tracker provides fast, ``Level 1" track triggering. Combining information from the tracker with the muon and preshower detectors, triggers for both single muons and electrons will be formed at Level 1. These triggers will be critical to take full advantage of the physics opportunities available with the Main Injector.
The scintillating fiber tracker is shown in Fig. 2. A total of about 80,000 scintillating fibers are mounted on four concentric cylinders with average radii of 20,30,40 and 50 cm. Each of the four cylinders supports a ``superlayer" of four doublet fiber layers. Two of the doublet layers contain fibers oriented in the axial direction, parallel to the beam line. These two axial doublets are separated by the 1.5 cm thickness of the support cylinder. The other two fiber layers are oriented at $\pm1.5^\circ-3^\circ$ stereo angles. All fiber doublet layers are constructed so that one layer is offset by one half of the $\sim 900 \mu$m fiber spacing with respect to its partner. This configuration removes all gaps and improves the doublet position resolution. We have studied the tracking and triggering with Monte Carlo simulations and the expected performance is excellent.
The basic detection element is the multiclad scintillating fiber. The inner polystyrene core is surrounded by a thin acrylic cladding, which in turn is covered by a thin flouro-acrylic cladding. These three materials have indices of refraction of 1.59, 1.49 and 1.42, respectively. The addition of the second cladding increases the light trapping by about 70% with respect to single-clad fibers, and improves the mechanical robustness of the fibers. The nominal fiber diameter is $830 \mu$m; the core diameter is $770 \mu$m and each cladding is $15 \mu$m thick. The polystyrene core is doped with 1% p-terphenyl (PTP) and 1500 ppm of 3-hydroxyflavone (3HF). The fiber scintillates in the yellow-green part of the visible spectrum, with a peak emission wavelength of 530 nm.
Eight meter clear multiclad fiber waveguides conduct the light to the photodetectors and are mated to the scintillating fibers by plastic, diamond-polished optical connectors. The photodetectors for the fiber tracker are placed in the platform under the central calorimeter.
The photodetector must be capable of detecting single photons with high efficiency at high rates and with large gain. We will use the Visible Light Photon Counter (VLPC), a variant of the solid-state photomultiplier. Much research and development in collaboration with Rockwell International Corp. has led to a device with the characteristics: greater than 70% quantum efficiency for the wavelength of interest, gain of roughly 20,000, and a rate capability of at least 10 MHz. The VLPC can be operated at full effficiency with a noise rate of 0.1% or less. The VLPC's are manufactured in arrays containing 8 circular pixels each 1 mm in diameter, well-matched to the clear waveguide fibers. The photodetector operates with a bias voltage of about 7.5 volts. The VLPC's operate at a temperature of 6-8 K, requiring the detectors to be maintained in a cryogenic environment. Cryogenic ``cassettes" are being designed which will house 128 arrays for a total of 1024 channels, and maintain them at a stable operating temperature.
The R&D0 program to develop the fiber system culminated in the operation of a large-scale scintillating fiber cosmic ray test stand at Fermilab. The test stand contained three 128-fiber-wide superlayers (a total of 3072 fibers). Superlayers were mounted at the top and bottom of a carbon-fiber support cylinder, with a third on a flat board on the cylinder axis. Muons with momenta greater than 2.5 GeV/c were selected using a steel filter. The cosmic ray setup was designed to test the major components of the fiber tracker in a configuration as close as possible to the final tracker design. The scintillating fibers were three meters in length, and were optically coupled to eight-meter-long clear waveguides. A cryostat was built to house test cassettes containing 128 VLPC channels each. The test stand was in operation from May through December 1994.
The results [4] of the cosmic ray test were excellent. The VLPC cryostat operated stably and the temperatures of individual cassettes were controlled to better than $\pm 15$ mK, easily good enough for stable VLPC operation. The gain of each of the 3072 channels was monitored by an LED-based calibration system, and overall gains were found to vary less than 1% over the length of the run. The noise rate, which was fixed to 0.1% by setting thresholds on each VLPC channel, also remained constant over the entire run. The light yield and tracking resolution are consistent with expectations. Figure 3 shows the light yield spectrum in photoelectrons for all fibers found on tracks. The most likely value of 8.5 photoelectrons is about a factor of four more than the minimum required for efficient tracking. There was no evidence of any degradation in light yield over the duration of the run. The doublet hit efficiency for cosmic ray tracks is better than 99.9%. The doublet position resolution, plotted in Fig. 4, is found to be 136 $\mu$m.
Currently, a variety of R&D tasks are being completed before construction of the fiber tracker begins. A new set of doublet ribbons with more accurate layer-to-layer registration has been manufactured and installed in the cosmic ray test stand and are expected to improve the position resolution to the theoretical limit of $120 \mu$m. Designs for the fiber ribbon manufacture and for the optical connectors are being finalized. A joint Fermilab-Rockwell study to optimize the VLPC cassette design is near completion. Prototypes for the final calibration system are being tested, and alternative scintillating dyes which may prove superior to 3HF in speed and environmental stability are under investigation.
The readout for both the silicon vertex detector and the fiber tracker is based on the 128 channel SVX II chip developed by Fermilab and LBL. Each channel contains a double-correlated sampling charge sensitive preamp, 32 stages of analog pipeline delay, a Wilkinson 8 bit analog to digital converter and a sparse data readout system ($<5\mu$s for ~3% occupancy). The readout system employs a common digital threshold for all 128 channels. It also implements a `nearest neighbor' readout scheme in which two channels below threshold neighboring a channel above threshold will also be read out. The chip is designed to accept data every 132 ns. At this crossing frequency, the 32 channel delay stage provides 4.2$\mu$s for the Level 1 trigger decision. For the silicon tracking system, this chip is mounted directly on the detector. Two iterations of the SVX II chip have been prototyped in standard CMOS and tested. All parts of the chip are fully functional and the noise has been measured to be $\sigma = 450 e + 65 e$/pF for a risetime of 105 ns.
Figure 5 is a block diagram of the silicon readout system. The SVX II's are controlled by a readout card, called a port card, which is mounted in the detector platform. The silicon detector is connected to the port card by a 28 foot long metallic data path. The port card downloads the parameters to the SVX II chips, controls the chip during data taking, reads out the data after a Level 1 trigger and converts the data to optical signals and sends these signals over fiber optic cables to the moving counting house. It also provides temperature, voltage and current monitoring and some level of diagnostics. It generates the test pulse signal for the SVX II.
The optical signals are received in the Moving Counting House by the Silicon Acquisition and Readout board (SAR), a VME board that acts as a buffer for transferring data to the Level 3 system. Events are held in one of eight local buffers until a valid Level 2 accept is received, when they are sent to the existing VME buffer driver (VBD) for transmission to the Level 3 system. Level 2 rejects are discarded.
The front-end electronics for the fiber tracker must provide a prompt Level 1 trigger pickoff, necessitating the development of a special `precursor' chip between the VLPC and the SVX II. Each channel of this chip has a charge sensitive amplifier, a discriminator with TTL output and a buffer amplifier to put charge onto an output capacitor which is read by the SVX II chip. In order to prevent the chip from oscillating, the trigger is picked off with a different clock cycle than both the input and the transfer to the SVX II. The readout after the SVX II is nearly identical to that for the silicon system.
The trigger scheme for the fiber tracker is based upon the r-$phi$ hit patterns in 4.5 degree sectors and allows four distinct momentum thresholds. The output from the trigger pickoff chip is fed into a series of large Field Programmable Gate Arrays which are preloaded with the appropriate logic to form hits from the eight trigger layers. The FPGA's are static RAM, so thresholds are software settable. Trigger data is combined with the preshower detector on the trigger board and sent serially at 424 Mhz to other trigger systems.
The central preshower detector will aid electron identification and triggering and correct electromagnetic energy for effects of the solenoid. The detector functions as a calorimeter by early energy sampling and as a tracker by providing precise position measurements. The cylindrical detector will be placed in the 51 mm gap between the solenoid coil and the central calorimeter cryostat at a radius of 71~cm, and covers the region $-1.3<\eta< 1.3$. The detector will consist of several layers of axial and stereo scintillator strips with a wavelength-shifting~(WLS) fiber readout. An end-view of the detector for one of the designs described below is shown in Fig. 6. A lead absorber before the detector will be tapered in z so the coil plus lead yield two radiation lengths of material for all particle trajectories. The scintillation light from the preshower detector will be transported via clear light-guide fibers to the VLPCs. The readout is again based on the SVX II.
The fast energy and position measurements enable use of preshower information at the trigger level to aid electron identification. The axial layers of the preshower will be used in the Level 1 electron trigger. Monte Carlo studies show that the information from the preshower detector will provide a factor of five reduction in the low-energy trigger rate by applying a pulse height cut and requiring coarse position-matching with tracks. Off-line, the early sampling of the showers and the good position resolution of the detector will provide additional means for identifying electrons.
Two approaches for the scintillator strips are being considered. The first approach involves machining square scintillator strips, shown in Fig. 7 (a), from large scintillator sheets. Six layers of strips will be arranged into z-z, u-u and v-v views with a uv stereo angle of $\pm 20$ degrees. The two layers in the same view will be staggered by a half-strip to reduce detector inefficiency. Each strip will be 5 mm thick and 5 mm wide. There will be 5,760 channels in this design. The WLS fiber will be placed in a `key-hole' groove at the center of the strip. A prototype module of this design was made and tested using the VLPC readout at the fiber tracker cosmic ray test facility at Fermilab. The module is about one meter long and consists of 128 channels. With over 15,000 cosmic rays recorded per strip, the light yields, signal uniformity, attenuation length, edge effects and crosstalk were studied. The effective attenuation length is measured to be around 11 meter and a light yield of 4.5 photoelectrons per millimeter of scintillator traversed has been achieved for cosmic ray muons. The detailed prototype test results for this design are described in Ref. 5. The results show that there is a considerable margin in achieving the desired performance for the preshower detector.
The second option uses extruded triangular strips. Figure 7 (b) shows the end view of this design. Triangular cells eliminate the need for two staggered layers of a square cross section, and therefore, make the detector more compact. In addition, this design improves the position resolution for minimum ionizing particles (MIP) as a result of light-sharing between two neighboring strips. We are considering a two-layer design of triangular strips. Each layer is 5 mm thick with a 5 mm fiber-to-fiber spacing. The inner layer will be arranged along the z direction. The outer layer will be arranged with the strips orthogonal to those of the inner layer. The channel count of this design is about 4,000. A 128 channel module made of triangular strips is being tested at lab 6 for its optical viability.
Calibration of the detector will be done in two steps. Light emitting diodes (LED) will be utilized to provide a quick calibration on-line. Each non-readout end of the WLS fibers will be equipped with an LED. By comparing the one and two photoelectron peaks from the LED light, the relative calibration of the detector elements can be performed. An absolute calibration will be provided by MIP responses of the detector using the data. The Monte Carlo studies show that the MIP peak can easily be identified.
The optimization of the detector design is progressing well. The R&D effort is expected to be completed this summer and a final design will follow.
Two Forward Preshower Detectors (FPS) will cover the pseudorapidity regions $1.4<|\eta|<2.5$. They will consist of wedges joined together to form conical detectors mounted on the faces of the End Calorimeter (EC) cryostats. Each wedge will be composed of an inactive absorber material (two radiation lengths of lead) sandwiched between two finely-segmented active layers. The active layer before the absorber (layer 1) will serve as a track stub detector and the active layer behind the absorber (layer 2) will sample the showers initiated in the absorber.
The FPS will be part of the electron trigger in the forward region. At trigger Level 1, a spatial match between any hit in layer 1 and a hit with energy deposition larger than expected from a minimum ionizing particle in layer 2 can be formed. At trigger Level 2, such a pair of hits can be matched with a tower in the electromagnetic (EM) calorimeter with energy above a threshold. At Level 3, more sophisticated algorithms will match clusters in the FPS and EM calorimeter.
Particle level studies show that a matching resolution of 5 mm or better is required between FPS layers 1 and 2 to achieve a substantial reduction of the rate due to the accidental overlap of charged hadrons and photon showers. More detailed GEANT simulations are in progress to estimate the backgrounds due to early hadronic showers and to determine the optimal segmentation.
In the offline analysis, the FPS will provide additional rejection against backgrounds for electrons by providing a more precise spatial match between CFT or H disk tracks, FPS track-stubs and showers than the calorimeter could provide. This will be especially helpful in identifying electrons in busy environments, e.g. from b decays. The fine segmentation of layer 2 will allow resolution of showers initiated by a single photon and by two closely spaced photons from the decay of high $p_T$ $\pi^0$s.
The detector technology for the active layers has not been selected. We are considering two options: grooved scintillator sheets with optical fiber/VLPC readout or interpolating pad chambers. The scintillator solution would use the same technology as the Central Preshower Detector and will require no additional R&D effort, but would provide some constraints on the possible segmentation. A pad chamber could be segmented in any way but would require mor bulky frames. The choice will be made on the basis of detailed GEANT studies, which will determine what segmentation is required to achieve the goals outlined above.
The Calorimeter system upgrade is driven by the need to preserve the excellent performance of the calorimeter for Run II running conditions. For practical reasons, we have chosen not to make any modifications to the uranium liquid-argon calorimeter itself, but to restrict all the changes to the front-end electronics. One component of the calorimeter system that will need modification due to the effect of the solenoid field is the Intercryostat Detector (ICD), which uses phototube readout located in what is presently a field free region.
An upgrade of the calorimeter front-end electronics is required because the minimum bunch crossing time will be reduced and the luminosity will be increased in Run II. The electronics upgrade will preserve as much of the existing infrastructure as possible.
To minimize the effects of pile-up in the calorimeter, we have re-optimized the shaping times. The present peak sampling time of $2.2\mu$s will be reduced to 400ns (matching both the charge drift times in the calorimeter and the expected minimum bunch-spacing at the start of Run II). Since this shorter shaping time increases the sensitivity to noise and reflections on the signal cables, we will replace the present cables from the calorimeter cryostat with cables that are impedence matched ($30\Omega$) to both the cables inside the cryostat and the new preamp input impedance. The calorimeter signals are transported to preamps housed on the calorimeter. The present preamp hybrids will be replaced with new hybrids which will have better noise performance through use of dual low-noise FETs, and increased output drive capability in order to drive the long terminated-cable run from the preamp to shaper circuitry (baseline subtractor or BLS). The new preamps require new preamp motherboards and power supplies, but otherwise use the existing mechanical structure.
The shaper circuitry incorporates significant new elements in the design to provide the necessary pipelining of the calorimeter signals in order to provide time for a trigger decision to be made. The analog pipeline will use a switched capacitor array (SCA) originally developed for SSC work by LBL [6], and modified to match D0 specifications. The SCA will sample the peak of the integrated charge signal from a preamp, shaped to provide a symmetric unipolar signal. The calorimeter signals require 12-bit accuracy with 15-bit dynamic range. This range requirement exceeds that which can be achieved with the SCAs, so a dual-pipeline will be used to accommodate the full dynamic range. To minimize the deadtime at high luminosity, the signals will be toggled between two dual-pipelines. Limiting the trigger to only one in a ``superbunch'' (a superbunch consists of 11 (or 33) bunches at 396 (or 132) ns spacing with about a $2\mu$s gap between superbunches) provides a deadtimeless system in which one of the dual-pipelines is reading out while the other is being filled with data. The gap between superbunches will provide a single baseline sample which will be used to remove long term drifts. The present BLS system (60,000 channels) will be fully replaced (including new power supplies, timing system, and pulser system) except for the mechanical infrastructure (crates, cabling, cooling, shielding).
The performance of the system with regard to pile-up has been simulated, and we find that the capability of the upgrade detector at high luminosity is comparable to that of the present detector at our present lower luminosities. Its performance at 132ns has also been checked and found satisfactory up to luminosities approaching $10^{33}{\rm cm}^{-2}{\rm s}^{-1}$.
The preamp design is finished and we anticipate completing the SCA R&D this year. A modest amount of R&D for the new timing and pulser systems should be completed by the begining of next calendar year.
The Intercryostat Detector provides a single energy sample in the region between the Central and End Calorimeter cryostats. This sample serves to improve the detector performance significantly in the overlap region $1.1<|\eta|<1.4$. The solenoidal field will render the present ICD phototube readout inoperable. Thus the upgrade of this detector system consists of modifying the present scintillator-tile system with waveshifter and fiber readout by moving the phototubes to a lower magnetic field environment. This will preserve the present capabilities of this system.
The advent of the Main Injector requires a major redesign of the D0 muon system triggering and front-end electronics systems. Bunch crossing times of 396 and 132 nsec are smaller than the drift time of the wide angle muon system (WAMUS) and the 132 ns time is less than that of the small angle system (SAMUS). The chambers will be preserved but the existing readout electronics will be replaced. In addition, it is necessary to improve the rejection power of the muon trigger systems considerably in order to handle the factor of ten increase in luminosity.
The WAMUS chambers are extruded aluminum tubes with a rectangular cross section. The wires are 50 $\mu$m gold clad, nickel struck tungsten. The drift distance is 5 cm with field shaping electrodes made of ``glassteel" in a pattern such that the relative charge on two adjacent triangular pads yields a coordinate along the wire direction. The SAMUS chambers are stainless steel tubes with a diameter of 3.2 cm. There is no field shaping, and only drift time is recorded. A three-dimensional readout is achieved by using crossed $x\times y\times u$ tubes. WAMUS covers $|\eta|<2.4$ while SAMUS covers $2.4<|\eta|<3.4$. The detectors are organized in three layers; the A layer for WAMUS is located between the calorimeter and the toroid iron, the B and C layers are located outside the toroids, with the C layer outermost. This notation is built into the acronyms in the discussion that follows.
We will use the existing high voltage, low voltage, gas systems, and coaxial-cable runs to the moving counting house (MCH) and the L2 trigger system. The L1 trigger system will be moved to the detector platform and use new serial data links, greatly simplifying the trigger cable plant. As much of the existing infrastructure will be reused as is possible.
The front-end electronics for the muon system must be replaced to accommodate the shorter bunch-crossing times. Improvements in available electronics make it possible to replace the present discrete-component systems with commercially available integrated cicuits.
The chamber wire signals will be processed by new amplifiers with transformer coupling at the inputs to avoid low impedance ground paths between the chambers and the preamplifier inputs. These commercial-IC based preamplifiers will provide equivalent noise performance to that achieved in the present run. The trigger information will be provided by wire signal latches, with the ambiguities, caused by the ``ganging'' of pairs of tubes, resolved by inserting a lumped delay between the adjacent tubes.
The time digitizers use a KEK-developed four-channel pipelined digital TDC chip called a TMC [7] which matches our needs very closely. The bin width when run at 26 MHz is 1.2 ns and the maximum delay is 4.8 $\mu$s, enough to cover the first level trigger (L1) latency of D0 of 4.1 $\mu$s. The time-of-arrival difference between each end of the tube pair signals is used to break the ambiguities of the repetitive pad pattern. We propose to calculate $\Delta t$ by subtracting time values from adjacent tubes.
The charge will be measured using a commercial 500 ns gated integrator, and digitized by a 10-bit 15-Msps ADC. The ADC signals are pipelined to provide deadtimeless L1 operation. This will allow us to maintain 2 mm pad resolution from 1% measurements of the individual pad pulse heights.
In order to form a trigger using the complete hit map for a particular collision, the signals cannot be used until the maximum drift time has elapsed. Thus the wire signal must be stretched to the maximum drift time, which is greater than the crossing interval. The problem of correlating hits with their corresponding collision can only be accomplished with the previously mentioned auxiliary detectors.
The rate capabilities of the data acquisition system (DAQ) dictate a hierarchal trigger scheme. In the muon system, which uses the TMC chip, the L1 pipeline is integral to the chip. The trigger logic generates a new candidate list of hits at a 53/7 MHz rate. The need to send trigger bits to the L1 processor at this speed defines its bandwidth requirements, which can be handled by serial links with rates as high as 1 Gbit/s developed for telecommunications. These serial links will be used to transmit signals to the L1 trigger and to the L2/DAQ in the counting house.
For the ADCs, we must supply an external pipeline. The present plan is to use a FIFO as a pipeline to provide at least eight-deep buffers. The outputs of the L1 FIFOs are attached to a data bus controlled by a digital signal processor (DSP). This high level processor will perform data formating, including pedestal subtraction, and time to distance conversion, and more extensive processing depending on the high level processing algorithm.
Since the bunch crossing time for Run II will be smaller than the WAMUS (SAMUS) drift time of 700 (200) ns, subsidiary detectors are required to remove the ambiguities and to provide the ``time stamp" for the event. The muon trigger detectors consist of:
The cosmic ray shield was completed and brought into operation during Run Ib. The readout for all these detectors is designed. They will be part of the muon system data stream, reporting to L1 and L2/DAQ.
The experience gained from data taken in Run Ia was invaluable for learning how to operate a muon trigger in the Tevatron environment. In particular, the forward region is very susceptible to albedo from the End Calorimeters and the low-beta aperture stops. Prototype scintillators and pad-pixel chambers have been installed in the present experiment yielding the rates, pulse heights and timing spectra which are required for the upgrade design. Shielding studies have led to reduced rates which are of critical importance to triggers based on a three-element coincidence.
We have found that the effective rate can be reduced in both EFB and EFC by a factor of 50 by using 20 nsec timing gates to reject albedo. Similiar factors are measured for CFA scintillator. We are also measuring the reduction of rates due to soft photons, obtained by using pulse height cuts on the scintillator signals. Clearly, scintillator is the detector of choice, but its use is constrained by the total system cost. We are optimizing the cost/benefit to establish the transition point between scintillator and PWC detectors on the basis of measurements taken in Run Ib. The time to freeze the design is the summer of 1995.
The present D0 triggering system includes two hardware triggers Level 0 (L0) and Level 1 (L1), and a Level 3 (L3) software trigger (we have replaced the existing D0 nomenclature of L0, L1, L1.5, and L2.) Interactions with coincident hits in the small angle counters on both sides of the interaction region give a L0 trigger. After a L0 accept, the L1 trigger requires either a minimum $E_{T}$ deposition in the calorimeter or primitive tracks of minimum $p_T$ in the muon chambers. Once an event passes the L1 trigger, the entire detector is read out and the event assembled in a farm of VAX computers. This farm, the L3 trigger, performs a nearly complete reconstruction of the event. If the event includes objects of sufficient interest, it is written to tape. Between L1 and L3, a third hardware trigger (L2) refines the calorimeter-based trigger for electron candidates by examining the shape of the energy deposition, and refines the muon trigger by using finer granularity hardware information. L2 presently interrogates only a subset of the L1 accepts and inhibits data taking while examining the event. Typically, this limits the current L1/L2 accept rate to 150 Hz. The maximum L3 accept rate is 4 Hz.
A typical Run I trigger menu includes high $p_T$ jet, electron, muon and large missing $E_T$ triggers. The L1/L2 and L3 cross-sections, at a luminosity of $2 \times 10^{32} {\rm cm}^{-2} {\rm s}^{-1}$, are 10 and 0.05 $\mu$barns, respectively. These correspond to trigger rates of 2000 Hz for L1/L2 and 10 Hz for L3 at $2 \times 10^{32} {\rm cm}^{-2} {\rm s}^{-1}$, which are beyond our present capabilities. In order to deal with these high rates, the D0 triggering system requires significant enhancement. The upgrade must include a new trigger framework to deal with the increased rates and several new triggering elements, including the fiber track trigger (CFT), the central preshower trigger (CPS), the forward preshower trigger (FPS), and muon trigger detectors, to provide the necessary rejection of background.
The present L0 provides a trigger and luminosity measurement. The magnetic field requires that the L0 system be replaced because of its conventional phototube readout. The luminosity monitor functions will be replaced by the new L0 system.
There are two distinguishing characteristics of the new framework. First, all events will be examined by L2 hardware engines -- not just a subset of events. Second, there will be event buffers between L1 and L2 and between L2 and L3. The addition of eight buffers between each trigger stage de-randomizes the Poisson distributed arrival times of the events, decreasing deadtime due to pileup. The buffers also eliminate the present L1 disable during L2 operation. These two improvements alone increase the L1 accept rate to 5-10 kHz and the L2 accept rate to 800 Hz. Since the Run II event sizes will be half that of Run I, an 800 Hz event transfer rate to L3 and a 10 Hz rate to tape are feasible. In summary, the expected Run II trigger accept rate limits are 10 kHz, 800 Hz, and 10 Hz at L1, L2, and L3, respectively.
The L1 high-transverse-momentum electron trigger will be upgraded from the simple threshold in $E_T$ to include a CFT track and CPS deposition for $|\eta| < $ 1.2 . The CFT momentum threshold gives good rejection against minimum bias (QCD) background. Requiring an energy deposition in the CPS which matches spatially with the track further improves rejection against isolated charged pions. Since the calorimeter L1 trigger does not contain spatial information, it cannot be used to further improve the background rejection. The forward electron trigger will use the calorimeter, as before, and the new forward preshower. The L1 high $p_T$ trigger rate will be $\sim$ 1000 Hz at $2\cdot 10^{32} {\rm cm}^{-2} {\rm s}^{-1}$.
Both the forward and central L2 electron triggers will retain the present electron isolation and shape requirements. In addition, a rejection factor of approximately two should be possible by requiring a coincidence among the calorimeter, CFT, and CPS in the central region and between the calorimeter and the FPS in the forward region. With these new elements, the L2 electron trigger rate at $2\cdot 10^{32} {\rm cm}^{-2} {\rm s}^{-1}$ will be $\sim$ 200 Hz for $E_T>15$ GeV. A L3 electron rejection factor of 100 can be achieved by importing current off-line shape cuts into the software farm; the high-momentum electron rate at $2\cdot 10^{32} {\rm cm}^{-2} {\rm s}^{-1}$ will be $\sim$ 2 Hz for $E_T>20$ GeV.
The L1 high $p_T$ central muon triggers ($|\eta | <$ 1.6) will also incorporate the CFT. Coincidences between the CFT and the inner-layer muon scintillators, and/or the muon chambers themselves, will provide substantial background rejection. The forward (2 $< |\eta | <$ 3) muon triggers will rely on the A, B, and C pixel layers and SAMUS coincidences. Additional shielding, multiplicity cuts, multiple interaction cuts and pulse height discrimination will reduce the rates further. At the lowest $p_T$ (~1.5 GeV), a pair of CFT $\bullet$ A-layer $\phi$ coincidences will serve as a di-muon or J/$\psi$ trigger. A Level 2 di-muon mass trigger could provide rejection factors of five for a J/$\psi$ trigger. The high $p_T$ L1 muon rate at $2\cdot 10^{32} {\rm cm}^{-2} {\rm s}^{-1}$ will be $\sim$ 200 Hz. The goal is to trigger on one muon with $p_T^\mu> 8$ GeV unprescaled, and on two muons for $p_T^\mu > 2-3$ GeV. Measurements and extrapolations based on data and Monte Carlo calculations indicate that this is possible.
Changes in the trigger framework and the addition of triggering elements will meet the high rate demands of Run II. Table 2 is a summary of the L1, L2 and L3 rates for various generic triggers at $2\cdot 10^{32} cm^{-2} s^{-1}$. For completeness, high $p_T$ jet, photon, and missing $E_T$ triggers have been included. Note that there is sufficient bandwidth for more specific low rate top, di-lepton and tri-lepton search triggers.
The data acquisition architecture for D0 in Run II, as illustrated in Fig. 8, will be largely unchanged from the current system. The basic components will remain: VME Buffer Drivers (VBDs) in each front end electronics crate driving one of eight parallel high-speed Data Cables feeding Multi-Port Memories (MPMs) accessed by a farm of event-building and software-filtering Level 3 processor nodes with another VBD -- Data Cable system feeding event data-logging and monitoring host cluster.
The VBD -- Data Cable -- MPM path will handle a rate of 160 Mbytes per second (~ 800 Hz) to the Level 3 processor farm using existing components. The 48 Level 3 nodes will be replaced with ones running an open operating system, i.e. one which will provide the necessary real-time functions yet will allow great flexibility in choice of commercial hardware. An example Level 3 system in current terms might be a PCI-based PC running Windows NT. The processor farm must possess sufficient compute capacity to provide a software trigger rejection factor of approximately eighty by executing a substantial subset of the current offline algorithms.
The VBD0 -- Data Cable -- MPM path from Level 3 to the host system will be designed to accommodate a rate of 5 Mbytes per second (~ 20 Hz). The host system is also to be assembled from commercial components. Incoming event data will be written to local disk buffers for later spooling to local or remote serial media. The host system will also provide the platform for monitoring of the data stream and will act as the interface to the hardware monitoring and control system.
Events will be reconstructed on the FNAL processor farm system, with that portion dedicated to D0 (an estimated 20,000 MIPs) capable of matching the 10 -- 20 Hz data acquisition rate. Following reconstruction, data will be stored on a tightly coupled disk and robotic tape system, and made available for analysis on a centralized analysis processor. We expect ~250 million events per year to be accumulated, requiring 3 Tb disk-resident and 160 Tb tape-resident storage.