DZero Silicon Track Trigger Preprocessor Conceptual Design Review _________________________________________________________________ 12th March 1998 1 pm - 6 pm Review Panel: Dave Buchholz, Dan Edmunds, Ann Heinson (chair), Patrick LeDu, Petros Rapidis, Paul Slattery, Daria Zieminska STTpp Presentations by: Horst Wahl (Introduction and physics benefits) Ulrich Heintz (Link to CFT L1 and trigger card) John Hobbs (Digital signal processing, & algorithms) Hal Evans (Link to Level 2) Meenakshi Narain (z-vertex finding) Shekhar Mishra (Beam position stability, control, & monitoring) Ulrich Heintz (Cost, manpower, interested parties, & timescale) Horst Wahl (Summary, and answers to questions) Review Panel Report ___________________ The panel thank everyone who has worked on the design of the Silicon Track Trigger preprocessor (STTpp) for their thoughtful presentations and detailed documentation on the conceptual design and predicted performance of the STTpp. We are extremely impressed with the progress shown since the first review in November 1997. ______________________________________________________________________ Charge Item 1. Reexamine the physics case for the STTpp. Is it coherent and defensible? Does it warrant inclusion in the Run II detector? _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ Based on the material presented to us by the STTpp group during this review and the previous one four months ago, the panel believes that the physics case for including the STTpp in the DZero Run II upgrade has been made. There are two facets to the improvements the STTpp will bring to the DZero physics program: (i) It will provide additional tools to help control trigger rates, by reducing the rates from fakes in many triggers, and by enabling more effective use of information at Level 2 and Level 3. These tools include: (a) adding SMT hits to the CFT tracks, which improves the pT resolution for tracks with |eta_det| < 1.6. The sharper pT thresholds will improve ET to pT matching, leading to better central electron identification at Level 2; (b) finding primary z vertex positions within |z| < 25 cm with their associated track multiplicities. These can be used in Level 2 triggers for physics measurements which are particularly sensitive to background, to reject events with too many primary interactions or SMT tracks; (c) using the primary z positions in Level 3 to speed up track finding, and to sharpen electron and jet trigger turn-on curves; (d) providing clustered hit coordinates to Level 3 for use in full tracking and vertexing; (e) providing the fitted track parameters to act as seeds in the Level 3 tracking. (ii) It will enable DZero to trigger on events containing displaced secondary vertices from b or c hadrons decaying in jets or from hadronically decaying tau leptons. This will allow the physics menu to be extended in several ways: (a) it will be possible to select all-hadronic events containing b jets from the QCD background. Some of the interesting channels which will become accessible include WH,ZH->qqbb, tt->qqbqqb, qt(b)->qqqb(b), bt->bqqb, plus direct b-physics production, and many processes beyond the standard model; (b) tau-tagging may enable the all-hadronic decay modes of the tau from H+->taunu to be selected; (c) it may be possible to trigger on a single electron, muon, photon, etc. at lower pT, when combined with the presence of a displaced secondary vertex tagged jet; (d) it will be possible to obtain a sufficiently large Z->bb calibration sample to fix the jet energy scale at mZ for b jets, which will be critical for making a competitive top quark mass measurement. The advantages of being able to use the SMT readout information in the DZero Run II triggers are many, and it seems essential that we have these capabilities if the collaboration is to carry out a competitive physics program at that time. Therefore the review panel unanimously recommends that the STTpp becomes an integral part of the DZero Run II Level 2 trigger. ______________________________________________________________________ Charge Item 2. Examine the conceptual design of the STTpp. Is the design sufficiently developed that the STTpp will work? Is there a credible cost estimate (taking into account the stage of the design)? Is there a reasonable estimate of effort and other resources needed? _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ We have examined the conceptual design of the STTpp as presented in the draft "Technical Design Report, version 1.2", in the review talks, and in other related documentation. We find no major problems with the design that might interfere with its successful operation. We offer the following comments on some details of the design. Algorithms __________ The sequence of algorithms - cluster the strips into hits, find hits in roads, fit tracks, find vertices and impact parameters - is in the correct order and probably the only possible solution for the trigger. Hardware ________ A lot of effort has gone into understanding the cable matching between the CFT and the SMT. Division of the functions between the CFT receiver board and the SMT trigger board looks right, since decoupling the CFT input receivers from the later processing elements will simplify maintenance and make a future processor upgrade easier. The current design has 10 trigger cards in each crate, with each card processing the hits and track roads from two 30 degree sectors of the SMT. This leads to the challenge of designing a VME board containing 32 FPGAs. The review panel recommends that the STTpp group examines the option of putting only one 30 degree SMT sector on each trigger card. This has the advantage that the cards will be easier to design and lay out, and the processing will have more parallelism, leading to a reduction in the time required to find hits in track roads and to fit tracks. The disadvantages are that twice as many receiver boards, trigger boards, and processor boards will have to be produced (but containing about the same number of components as before), and there may be a small loss of efficiency for tracks with large impact parameters which cross the boundary between two neighboring 30 degree sectors. There may be some additional hardware needed which is not explicitly described in the draft TDR yet. This includes support cards to get the Serial Command Link information into the fiber road card, to send data to the Level 2 global processor, and to control the VBDs. Also, at least two custom backplanes seem to be needed and will have to be designed. All of the necessary hardware must be described in the final TDR for a full technical review before detailed estimates of manpower and costs can be completed. Timing ______ Besides the time for performing the track fitting algorithm on the processor board, time will also be needed to: o scatter the data into the processors o retrieve the data from the processors o synchronize at the beginning or end of each event o check for and recover from errors o gather and manage monitoring and alarm data o manage the readout to the Level 2 global processor o manage the readout via the VBD to Level 3 o (and on the z vertex finding board, time for scaling from one radius to another) The time required for all these housekeeping chores is likely to be greater than the time required to execute the fitting algorithm. Software ________ There was not much discussion during the review of the software needed to implement a successful STTpp. The list of packages thought of by the review panel is as follows: o "trgsim" package, a full bit level simulator o "examine" package to compare online operation with simulator o hardware exerciser, debugger, and download packages o package to get calibration and geometry data into running hardware and full simulator o package to monitor online operation of the STTpp o package to program hit finding and hit filtering FPGAs on the trigger cards o package to control data movement, event synchronization, error-checking, -logging, and -recovery for the DSPs on the processor cards o physics studies to complete the design of all parts of the STTpp. Beam Alignment and Survey, and SMT Alignment ____________________________________________ >From physics studies and looking into what CDF plan to do, it is clear that the beam position has to be known and controlled to very high precision. This is a critical element of the design, and failure to incorporate it in every aspect of planning now could very well lead to failure of the trigger. We encourage the STTpp group to work closely with the SMT detector and algorithms groups on SMT alignment issues, so that these groups know what is needed, and with the Beams Division and CDF on beam positioning and surveying, so that we can measure the position of the beam before DZero is rolled in, and control it sufficiently well afterwards. The issues of signal dilution and change in efficiency from beam spread or tilt have been satisfactorily addressed. We would like to know what the effects of misalignment of the individual barrels is on the performance of the STTpp. That is, if each barrel has a ~50 mrad (say) random misalignment with the beam, is the trigger compromised, or can offline alignment corrections be fed back into the trigger to correct hit strip positions to clustered hit locations, compensating for the misalignments? Other Recommendations _____________________ The review panel urges the STTpp group to work closely with the main DZero Level 2 trigger group to thoroughly examine the FIC/SLIC/Alpha approach to triggering being adopted by some parts of the Level 2 trigger system. The advantages of standardizing hardware and software include: o engineering effort can be shared and minimized o reduced numbers of types of spares and simplified repairs o fewer software packages to design, implement, document, and maintain o more people will understand the system during the run. However, there could be disadvantages (based on the brief overview of another design we have seen so far), such as: o a less functional STTpp (including no z vertex measurement) o greater cost o increase in the cost, complexity, and time to design and build the FIC/SLIC/Alpha systems for the rest of Level 2, if they become more complicated in order to accomodate the STTpp The review panel recommends the STTpp group and the DZero Level 2 group come to a decision on whether to use dedicated devices for the STTpp, or standard components, as soon as possible, and whichever choice is made, to then merge and the STTpp group to become a subgroup of the main Level 2 team, regardless of the status of funding, in order that future directions are worked on together before decisions are made. Cost Estimate and Schedule __________________________ The cost estimate is credible, given the conceptual status of the design. So far, insufficient financial resources have been identified to fund the STTpp, even if the NSF Major Research Infrastructure (MRI) proposal is successful. The review panel believes that the STTpp will be a critical part of the DZero upgrade trigger, and as such, it should be included in the baseline design despite this problem. This decision, if adopted, will lead to one of several routes being taken: o NSF MRI is funded, additional funds are found (unspent contingency, additional money from Fermilab or collaborating institutions) -> both DZero and the STTpp start at the beginning of Run II o NSF MRI is funded, additional funding is not found until later -> the STTpp becomes operational at some point after the start of Run II o NSF MRI is funded or not funded, money is taken from some other part(s) of the upgrade to build it -> STTpp starts at the beginning of Run II o NSF MRI is funded or not funded, less money is taken from other part(s) of the upgrade or from elsewhere -> STTpp starts some time after the beginning of Run II. The panel recommends that the decisions whether to take money from some other part of the upgrade to pay for the STTpp in the instance that the MRI and/or other sources of funding are not forthcoming, and when to have the STTpp become operational, should be taken only after explicit discussion and agreement within the collaboration. We recommend, however, that the hit clustering part of the STTpp should be implemented from the very beginning of Run II, even if other parts of it are delayed. This is because the hit information will be essential to Level 3 tracking and vertexing. If the STTpp were implemented some months after the start of the run, say, then temporary Level 3 algorithms which did software clustering and then (faster and less efficient) tracking would have to be implemented for just this short period. It would be a more efficient use of Level 3 manpower to develop only one set of tracking and vertexing algorithms that use the already-clustered hits. We note that all strips, including those in 2 degree and 90 degree stereo ladders and from disks should be clustered and the hit positions sent to Level 3. Regarding the overall project schedule, the panel believes that although the advantages of the STTpp in helping to control Level 2 trigger rates and in allowing effective use of Level 2 and Level 3 with the z vertex position will not be so important during the expected lower luminosity running at the beginning of the run, there are two critical reasons why the whole STTpp should be implemented from the beginning of Run II: o we need to be able to measure the mean collision position of the beams in (x,y,z), and using the STTpp rates will lead to a delay of only a few minutes to do this. The information can be sent to the accelerator people so they can steer the beams to center the collision region at (0,0,0) of the SMT; o we must be able to trigger on displaced vertices to find b-jet-enriched all-hadronic event samples. These goals are so important to DZero that the entire STTpp system should be installed and running from Day 1 of Run II if at all possible. Effort Needed _____________ A reasonable estimate of the engineering and technical effort needed to design and implement the STTpp has been made, for the hardware parts that have been considered so far. System integration will take additional engineering expertise, which should be found outside of Marvin Johnson's overstretched group. The Nevis Labs' facility could be considered for this task. The STTpp is an extremely complex project, with many aspects to its design and operation in both hardware and software. Therefore, the strength of the team proposing to build it is critical to its prospects for success. The review panel notes that the team has members with previous experience designing triggers (DZero L1.5 Calorimeter trigger), and that they have a track record within DZero and elsewhere of successfully completing projects of similar magnitude. We note the following: (i) Based on the parts of hardware identified above that have not yet been incorporated in the TDR, and on the long list of essential software not discussed by the participants, the STTpp project team will have to expand in order to complete the project on time. (ii) The people working on the STTpp would have invested this effort in some other part of the DZero upgrade. This diversion of effort from very capable people will hurt the upgrade, even though it has not been explicitly identified in which areas it will be. Despite this, the panel believes that the STTpp is of sufficient importance to DZero to warrant inclusion in the upgrade at the earliest possible time, depending on the availability of funding and the conscious division of that funding between the various parts of the upgrade. Miscellaneous _____________ (a) Improving the SMT The panel notes that although we have discussed above a possible cost-benefit analysis of building the STTpp, on time or late, versus some other part of the upgrade and its schedule, we have not discussed other ways to improve the same physics capabilities of DZero in Run II which could be included in such a cost-benefit analysis. There are changes to the design of the SMT itself which could work in tandem with the STTpp to improve the high pT physics program. These possible changes include adding two barrels to the current six barrel design to increase the acceptance for high pT physics processes, moving two or more F-disks out from between the barrels to reduce the amount of central material for low eta tracks (or removing them, to pay for one or more of the other changes), and making the single-sided ladders in the outer two barrels into double-sided 90 degree stereo layers so that the z vertex position finding can be extended from the central +-25 cm region to cover the full luminous region. The panel strongly endorses the physics program supported by the STTpp, and therefore encourages the collaboration to urgently examine these proposals for improving the design of the SMT to maximize the Run II physics prospects for DZero. Since the eight barrel option is a straightforward and popular extension to the current SMT design, the STTpp group should complete their design for eight barrels in r-phi and six barrels in r-z, and then scale it back only if it is not possible (financially and/or technically) to modify the SMT design. (b) Using the SMT Disks in a Level 2 Trigger It would be useful for the STTpp group to briefly examine whether it could be possible in the future to use hits from the H-disks and F-disks in another (new) part of the STTpp. This information may come in handy if the device is presented to the PAC and a question is asked on the subject. (c) Using the STTpp in Run III A Run III tracking system is not yet specified. However, we encourage the STTpp group to consider how their trigger will perform when the instantaneous luminosity increases from ~2x10^31 cm^-2s^-1 to 5x10^32 cm^-2s^1, or higher if there were no luminosity leveling. We also encourage the group to consider whether it is expandable to allow hits from pixels to be used instead of from strips. That is, should the hit clustering hardware be physically separated from the hardware finding hits in track roads. This issue also relates to implementation of the hit clustering hardware from the very beginning of Run II even if the rest of the STTpp has to be completed late. While we strongly endorse consideration at this early conceptual design stage of how the STTpp might operate in Run III conditions, building in extra flexibility to allow such putative operation must not compromise the timely completion of the project for Run II. ______________________________________________________________________ Charge Item 3. Examine the present status with respect to the issues raised by the previous committee. _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ The panel is satisfied that the issues raised by the previous committee have been satisfactorily addressed during this review. ______________________________________________________________________