CMS Bulletin

TRACKER

Pixels Tracker

With the 2012 proton-proton run almost complete, the pixel detector continues to operate well in an environment with large pile-up and high L1 rate. During this period, the pixel detector has shown excellent stability, with the number of current active channels from each the BPIX and FPIX the same as from the first month of 2012 running, resulting in 96.3% of the detector active. This total includes the recovery of six FPIX channels, temporarily disabled due to an unexpected dependence on the magnetic field. From a dedicated study that identified a small crack in an optical cable connector, a repair was made which restored 120 ROCs in the FPIX.

During 2012 there has been a close collaboration of the online operations with the offline studies, resulting in the first dedicated HV bias scans used for the pixel Lorentz Angle measurement. These scans help to better understand this important parameter that changes with temperature, irradiation, and bias voltage. This is in addition to all other scans taken throughout 2012, namely the timing and HV depletion-depth scans, the latter performed at monthly intervals during the entire year. These measurements are essential to study changes in the effective space-charge distribution in irradiated pixel silicon sensors. They are important for the further operation of the present pixel detector and for the design of the planned upgraded detector.

With peak instantaneous luminosities exceeding 7E33, reducing the impact of radiation effects that affect individual hardware components (SEUs) has been a key activity in 2012. Through improvements in the readout software and firmware, the pixels are now able to automatically reconfigure whenever an SEU causes the pixels to go into a permanent BUSY state. In the rare cases when the reconfiguration fails to fix the problem the affected channels are disabled on the fly by operating in a RunningDegraded state. In addition, the maximum time spent in BUSY has been reduced from 2 seconds down to 0.25 seconds, decreasing the total dead-time from the pixel detector to below 1%. On the online monitoring side the pixel DQM can now show the fraction of ROCs that become silent due to SEUs as a function of time in a run.

The work on the pixel lab at P5 continues to make significant progress, with the installation of the floor, electrical supplies, and pixel chillers complete. The current work on cooling and the dehumidifier is progressing well, and the entire project is on schedule and should be ready in time for the upcoming LS1.

Strip Tracker

During 2012, the number of active channels has been quite stable and are now at 97.5%. Few modules, which were already misbehaving before, were eventually masked from the DAQ resulting in a loss of 0.3%.

Over a delivered luminosity of 20.7 fb^–1 (up to 17 November), the strip tracker caused data losses of 213 pb^–1 (downtime), 70 pb^–1 (bad quality) and an additional but unavoidable 156 pb^–1 at the beginning of fills (raising the high voltage). The overall losses correspond to 2.1% of total delivered luminosity, and are equivalent to 13% of CMS losses. A lot of effort was made throughout the year to improve the strip tracker’s stability and high level of efficiency.

A semaphore implemented in DCS now continuously checks the beam conditions, allowing the raising of high voltages automatically at the declaration of “Stable Beams”. This simplified procedure is in place since three months, and results in an extra 0.5 pb^–1 being recorded per physics fill.

The strips suffered from a few hardware failures: one VME crate PSU, a VME link card and a disk controller on one DCS PC. The tank level sensor stopped functioning for the SS1 cooling plant, causing detector downtime; the same happened recently for SS2, but luckily without consequences. In both cases the sensors were recovered with a reset.

A continuous effort was made to fight DAQ instability, identifying sources of downtime and looking for fixes or quick recovery. The use of soft-error recovery was successfully commissioned to the most frequently affected component (TIB-2.8.1, aka FED101), determining a relevant downtime saving. Experts are now looking to extend this mechanism to other less frequent problems. Several improvements were made on FED firmware, in order to make FEDs robust to any non-standard situation, such as data not being sent occasionally by the detector front-end or extra data received without L1 accept associated. Several smaller problems in the current implementation of the resync mechanism have also been found and fixed. The situation is continuously improving but the scarceness of the different event types makes debugging difficult and time consuming.

Calibration runs were made to monitor detector stability and bias scans were performed to provide voltage depletion measurements, crucial for understanding the detector evolution with radiation.

Three years of running are close to the end and the strips, despite the dramatic increase of the instantaneous luminosity, performed in an excellent and stable way thanks also to the daily activity of the operation team. DQM shows that the noise occupancy continues to be at a negligible level and the hit efficiency is stable at a very high value.

The strip tracker is now looking forward to the 25 ns pilot run and the proton-lead collisions. The preparation for LS1 is now well established, thanks also to several comprehensive reviews that have been held. A full report was presented at the 3^rd Technical Coordination workshop.