TRIGGER REVIEW COMMITTEE REPORT ========================================= (D. BRETON, D. BRIGGS, D. NELSON, M. WITHERELL, G. WORMSER) The committee would like to congratulate the Trigger group for the very complete and thorough work done so far and for the quality of the documentation and presentations given to the Committee. The overall impression of the committee is that the Trigger group has proposed a very sound basis for BABAR trigger and encourages the group to proceed to the Preliminary design phase. The detailed comments are organized in 4 sections : Physics requirements and trigger simulations, Global Trigger architecture, DC trigger, Calorimeter trigger. I/ Physics Requirements and Trigger Simulations. The primary physics requirement for the BABAR trigger should be to obtain 99% efficiency for events which have a reconstructible example of a B meson decaying to a CP final state, at background levels 10 times nominal. This efficiency should be close to 100% for nominal background conditions. These numbers are for conditions of good performance (but not unrealistically good) of the detector elements that are used in the trigger, and do not take into account down time for the trigger/DAQ system. It should also be possible to monitor the efficiencies easily and to measure them accurately. There are other physics requirements for the trigger that are also important, but are secondary in the sense that nothing should be done to achieve them that endangers satisfying the primary requirement. The primary physics requirement we suggest above is significantly more stringent than the "efficiency goal" set forth in the Trigger Preliminary Design Requirements. We believe that the Trigger group is actually working to a higher standard in this respect than the goal they stated implies. GEANT simulations presently show that the experiment can be run at nominal backgrounds with a C2+D2 trigger, which is 100% efficient for the two CP modes studied. At 10 times nominal backgrounds, the C2'+D2' trigger could be used, which adds an additional momentum cut on the A track. In present simulations, these are 99% efficient for the CP modes shown. The two modes used, B0 -> pi+pi- and B0 -> D+D-, do represent a wide spread in the multiplicity of the CP decay mode, and therefore represent reasonable benchmark modes for these studies. We recommend that the Trigger group should revise their requirements on the CP modes, since a higher standard is both desirable and realizable. The present preliminary design does appear to meet that standard. The simulation work that has been done to establish the efficiencies for physics modes and the rates for backgrounds is generally of very high quality. Of course, the present simulations do not yet take into account full reality of the detector. For example, they need to be repeated with realistic cell efficiencies for the drift chamber, including reasonable rates of non-functioning cells. The robustness of the trigger efficiencies to possible detector degradations such as a loss of a axial superlayer due to a broken wire should be studied. Thresholds may need to be applied on pulse heights from individual towers before energy sums are taken for the trigger. It is important that the simulation be made as realistic as possible, including some conservatism about detector performance. We have a few suggestions for how to improve the information coming from trigger simulations in the future. 1)As we have already said, one should always define efficiency as a fraction of reconstructible events. A fiducial "volume" relative to geometrical acceptance but also to momentum cuts,. should be defined for each reaction. For example, a B0 -> D+ D- mode should only be included in the denominator if both D+s decay into a detectable mode with all of the decay products in the fiducial volume of the detectors. This is done in some cases, but not in others. 2)In the particular case of common 1x1 tau modes used for measurements of one-prong branching ratios, the overall efficiency is less important than the ability to measure the efficiency with an accuracy of a fraction of a percent. The trigger group makes many plausible arguments about why it should be possible to do this. However, it is really only possible if the trigger efficiency is very high (>95%) in some restricted fiducial region that can be well defined. It would be useful to define some such benchmark figure so that both the group and reviewers could get some feeling for whether this goal can be met. 3)A very difficult rare B decay should be added to the benchmarks. We would recommend B+ -> tau+ nu. 4)It is important to follow up on the recent CLEO experience to make sure that whatever is causing their present background triggers are represented in the BABAR background simulation and to check whether BABAR trigger strategies could handle their problem. II/ Global trigger design The committee would like to stress that it is essential for BABAR to to have a well understood trigger system operating at day 1. This implies that the system should be kept as simple as possible. Concerning the potential future upgrade of the trigger with the level 2 implementation, the committee recommends that the minimal checks be made to ensure that the dataflow architecture is indeed compatible with a Level 2 trigger and that all detector subsystems would be able to sustain an higher Level 1 rate (10 kHz) which could become the desired rate in presence of the level 2. II.2 Trigger latency The committee noticed that there is at present a 1.3 us latency contigency and would like to know whether it is a sufficient margin at that early stage of the design. A larger contingency might be used to try to reduce the trigger window (see below). II.3 Trigger jitter It is not known whether the calorimeter trigger can meet the 1 us jitter requirement at 10Xnominal background. The committee recommends a comprehensive rethinking of the approach to the problem of extracting event times from the calorimeter trigger data stream (see some further comments on this calorimter section) II.4 The committee encourages the trigger group to stay in close contact with the group in charge of BABAR commissionning in order to extract as much information as possible relevant to the trigger from the 1998 run.. II.5 The committe recommends that the "no man's land" domain between the trigger group and the Fast Control System be clarified in a two or three month time scale. II.6 The proposed trigger system architecture should be consolidated in as few crates as possible consistent with BABAR standards and partitionning requirements. The data rates and links among the crates require more study. II.7 As a general remark, the committee recommends that a great deal of attention is put into the testability issues of all aspects of the trigger system. The possibility to feed every part of the system with test patterns should be carefully studied, especially in such a system where the control of the environment is very difficult. Moreover, a complete standalone test configuration could be of important help to understand the system. III. Drift Chamber Trigger III. 1 Cell efficiency and pt measurement efficiency In conjunction with the DC group, the trigger group should define requirements for single cell efficiency for its use in the trigger logic and for the pt measurement. In the present design, a pt measurement almost requires 15 hits out of 16, placing a high constraint on the cell efficiency. The robustness of the trigger versus various hardware problems suh as broken wires, high background or large cross talk needs to be documented. The committee would therfore welcome more information on the precision in pt provided by the 4/4 requirement, compared with the 3/4 arrangement. The granularity of the HV supplies in the trigger layers should also be studied. III.2 Usefulness of the time-based segment finder The committee was not shown a substantial benefit from the added cost and complexity incurred by segment finder logic that uses drift time information to improve the trigger logic. III.3 Location of the supercell segment finder The committee endorse the deployment of segment finder logic outside the confines of the Drift Chamber backward endplate region. This choice alleviates demands on space and cooling in a critical region. IV. Calorimeter trigger IV.1 Trigger jitter As mentioned above, improvment is required in defining the trigger window. It seems that this point has not yet been completel understood. The time jitter seems to be correlated to both peaking time uncertainty due to signal/noise level and dispersion in the electronics consta Moreover, the effect of a high background level on this jitter is not yet well known. The width of the trigger window will directly depend on the capability to compute the T0 at low energy level and high background rate. The comittee wonders if it would be possible to use faster clocks and/or more parallelism in trigger chain, so to allow to reduce the latency of a part of this chain. The extra time could be used to refine algorithms to compute the event time. IV.2 Trigger implementation The choice of computing the tower sums along the phi strips fits of course well with the DC wire configuration. This configuration seems to fit the physics goals by providing a back-to back capabilities as well as better background rejection than using theta alone. The cost/benefits of a finer segmentation (ie 2D) should however be studied. IV.3 It seems that the need to study in more detail individual thresholds per crystal before the summation is obvious in relation to the potentially effects of background and coherent noise at the calorimeter electronics level. IV.4 Spying The comittee notes that the possibility to use a spying system in parallel to the standard data flow has been studied. The benefits provided by such a system should be studied further by comparing the added complexity vs the added potentiality with respect to the standard diagnostics provided in the DAQ system. Such studies should perhaps be made for the DC trigger also. V Conclusion Concerning the trigger requirements, the committee thinks that the trigger group has done a good job of building in the necessary redundancy and flexibility into the present design. For example, the fact that the good efficiency for CP decay modes of both the C2 and D2 triggers makes it possible to measure the efficiencies accurately. Having the multiplicity of tracks, clusters, and matches available for cuts at the final stage of the L1 trigger makes it possible to react to unforeseen background problems. The proposed implementation is a sound basis for further design. A particular attention has to be paid to the time measurement in the calorimeter in presence of x10 backgrounds. Requirements have to be defined in concert with the DC group for the trigger cell efficiency.