Cosmic shakedown

2 June 2009

In the counting room (courtesy of Claudia Marcelloni)



For the last seven weeks, ATLAS has been going through the motions, practicing and problem-solving in preparation for beam.

Like a choir rehearsing ahead of a concert, the first five weeks – so called ‘slice weeks’ – each focused on running different combinations of parts of the detector, to inspect how each performed and cooperated.

Week 1, commencing April 13th, was all about the SCT, Pixels, and Beam Condition Monitor. Week 2 dealt with the Tile and Liquid Argon calorimeters, the L1 Calo trigger system, and some High Level Trigger (HLT) algorithms specific to the calorimeters. The third week concentrated on the suite of muon sub-systems.

Week 4 took things up a notch, combining all of the above and focusing on testing HLT algorithms as much as possible with cosmics. Finally, week 5 was the first time that the forward detectors – LUCID and the Zero Degree Calorimeter – had been run together and integrated into the ATLAS data stream. Look out for articles reporting in more detail on these final two weeks in the next issue of e-News (June 15th).

Due to ongoing work on the cooling for the silicon detectors, the SCT and Pixel detectors could not be switched on, and so week 1 involved only the off-detector part of Pixel and SCT electronics, with Monte Carlo data being 'plugged' into their readout devices. The rest of the slice weeks were full cosmic runs.

The muon detectors with the largest surface areas can detect 1000 cosmic particles per second, but in LHC conditions, up to 100,000 ‘interesting’ events will be sent to the HLT per second. To simulate the stress on the HLT, ‘fake’ so-called random triggers were layered on top of the cosmic ones.

“We’ve tested so far up to 80 kilohertz. The bulk of this is fake triggers, with about 1 kilohertz of real cosmic muons,” says Run Coordinator Christophe Clement. “The HLT can filter out the fake triggers, run algorithms on the real cosmic ones, and then write maybe 200 interesting ones per second to Tier 0.”

According to Christophe, the slice weeks have been pretty successful, particularly considering how much the landscape has changed since the detectors were last run together, in Autumn 2008; the Detector Control System (DCS) which monitors the hardware has been upgraded, bits of the detectors have been replaced and repaired, the online software and HLT software have both been upgraded, and the whole detector has been opened and is now almost closed.

“I think we can say now that we have upgraded all software for Data Acquisition and Trigger as well as the Detector Control System, and we’re more or less at the same level that we were last year in terms of stability,” says Christophe “and then we start to push more at the trigger rates.”

Stability tests, where the system is left to run unhampered for extended periods, were performed on the weekends of the slice weeks. “It’s like a test-program for a plane,” explains Christophe, “they’re going to do all possible things: fly it into a storm, try to land it when it’s snowing. That’s what we do during the week. Then on the Friday evening, we say ‘OK, now we’ll just try to fly straight for the weekend, without touching anything, and see if it works for a long flight’.”

Most of the sub-systems were able to run well for extended periods, although there were some unexpected instabilities at high rate. This week and next, experts from the calorimeters and L1 Calo will meet in Geneva to try to get to the bottom of those problems.

Weekend tests were also run using a simulated beam schedule, to give groups a better sense of how they will need to work during beam time – stopping and starting and reconfiguring between LHC fills. “We were quite positively surprised at how well we were able to do this,” says Christophe, reporting crude data-taking efficiency calculations of 91 per cent during the long muon weekend. Ignoring a glitch on the general power grid on a certain Sunday morning at 3 a.m., the figure for the calorimeter weekend would have been 97 per cent.

Since the slice weeks, more tests have been done with the muon system and the TRT, as ATLAS works towards a two-week combined magnet run, due to begin on June 22nd.

“For these two weeks, we hope to run cosmics with less debugging,” says Christophe. “The weeks we had so far were really to try to address technical issues. Hopefully [the combined run] will be much smoother, and we can calculate our data taking efficiency much better.”

The data taken so far will be analysed for weeks to come, and used to perfect the calibration, alignment, and synchronisation of ATLAS. Over 40 million cosmic events were triggered in the muon slice week alone, and there is now a big push to bring trigger timings for different parts of the detector and different types of triggers into alignment. When real collision particles start shooting through the different layers of the detector, all the electronics must be synchronised and shouting their findings in unison, a choir hitting its notes in time.

“We’re fine-tuning this now,” says Christophe. “If we can get everything within 25 to 50 nanoseconds on the cosmics before the beam it would be a great success. And then we can improve with the collisions.”

 

 

Ceri Perkins

ATLAS e-News