Trigger Studies Group (TSG)

The Trigger Studies Group has just concluded its third 2013 workshop, where all POGs presented the improvements to the physics object reconstruction, and all PAGs have shown their plans for Trigger development aimed at the 2015 High Level Trigger (HLT) menu.

The Strategy for Trigger Evolution And Monitoring (STEAM) group is responsible for Trigger menu development, path timing, Trigger performance studies coordination, HLT offline DQM as well as HLT release, menu and conditions validation – this last task in collaboration with PdmV (Physics Data and Monte Carlo Validation group).

In the last months the group has delivered several HLT rate estimates and comparisons, using the available data and Monte Carlo samples. The studies were presented at the Trigger workshops in September and December, and STEAM has contacted POGs and PAGs to understand the origin of the discrepancies observed between 8 TeV data and Monte Carlo simulations. The most recent results show what the rates of the existing triggers would be at 13 TeV.

Together with PPD, the group has coordinated a new Monte Carlo production campaign, addressing the requests from all POGs and PAGs, in view of the detailed Trigger studies foreseen for next year. The production is starting using CMSSW 6.2.x and taking into account the post-LS1 configuration of the detector.

The "half-rate" toy menu, designed with the OpenHLT workflow to halve the individual rate of each 8 TeV trigger, has been shown, when taken as a whole, to indeed yield half of the total rate of the original menu.

This shows how the HLT menu could look like in 2015 in the absence of substantial improvements in objects reconstruction and selection criteria, and POGs and PAGs are using it as a starting point for the development of the HLT paths for Run 2.

On the validation side, the group has followed the HLT validation of the CMSSW 7.0.x release cycle, using both an ad-hoc event-by-event validation tool and the DQM monitoring from the POGs and PAGs.

STEAM is currently starting a dedicated campaign aimed at developing, improving and rationalising the DQM modules, where all paths and sensitive quantities should be monitored throughout release cycles and HLT menu changes. In view of this activity the group has also produced various data skims of good events, as selected by the PAGs, to be used in future validation cycles.

During LS1, the Software, Tools, Online Releases and Menus group (STORM) has followed the activities of the DAQ, Computing and Offline sectors of CMS, in order to make sure that HLT menus and software can cope with the modifications in the underlying systems. The "half-rate" menu has been integrated in the current CMSSW releases, and will be included together with the 8E33 HLT menu in the upcoming Monte Carlo production.

Besides helping the development and integration of the paths themselves, STORM is also actively working together with the Offline experts for the optimisation of the algorithms, in order to use the computing power of the online farm as efficiently as possible, and in that way increase the physics capabilities of the HLT menu.

Another big chunk of the STORM-related activity deals with the improvements foreseen for the ConfDB database back-end, the web-based browser and the GUI. The back-end has been redesigned for faster access, and the tools that deal with the CMSSW releases are being adapted for the migration to Git.

Since the end of Run 1, the Field Operations Group (FOG) has been preparing for data taking in Run 2, with particular focus on the transition to the file-based HLT and on analysis of the CPU capacity of the HLT farm. For the file-based HLT, our main task has been the creation of a test stand that can be used to emulate the new online environment. To determine the CPU capacity we have been running the 2012 8E33 HLT menu over very high pile-up data on three of the newest HLT farm machines.

We have also been participating in monitoring upgrade task force and provided on-call HLT experts during the Global Run in November. FOG has also been documenting and preparing tools in preparation for training new HLT experts required for data taking in Run 2.

Level-1 (L1) Trigger

In July, integration testing for the upgrade to the Level-1 trigger commenced with tests of the components of the calorimeter trigger upgrade at the CMS integration centre at Prevesin’s Building 904. The optical mezzanine cards necessary to duplicate ECAL trigger data for parallel commissioning of the new Trigger were tested, alongside prototypes of the processing electronics planned for use in the upgrade. The successful testing of the mezzanine cards has allowed orders for the full production of these items to be placed. In late September, testing of a slice of the calorimeter trigger began with a demonstration of the time-multiplexed trigger concept. During the test the 10 Gb/s optical links required for the upgrade were run without errors for long periods and the latency of representative algorithms and links was measured, verifying the estimates made in the Technical Design Report. Testing will continue throughout 2014, building up a complete slice of the trigger, including the calorimeter inputs, the global trigger and later the muon trigger systems.

Studies of Trigger algorithms have continued, with a workshop in mid-September. Particular emphasis has been on clustering and isolation for eγ and tau triggers, pile-up estimation and subtraction in hadronic triggers and improvements in the momentum estimation and isolation of muons. Studies indicate significant improvements are expected with the new trigger systems and initial versions of the algorithms are planned to be available in CMSSW in early 2014. Development and optimisation will continue throughout the lifetime of the Trigger as has been the case with the existing Trigger.

by Roberta Arcidiacono, Aram Avetisyan, Andrea Bocci, Grant Christopher, Andrea Perrotta, Muriel Vander Donckt, Alex Tapper