Arnd Meyer (RWTH Aachen) Oct 8, 2003Page 1 Data Taking Status and Plans Achievements Concerns Plans Arnd Meyer, RWTH Aachen D Ø Collaboration Meeting October.

Arnd Meyer (RWTH Aachen) Oct 8, 2003 Page 1

Data Taking Status and Plans

AchievementsConcerns

Plans

Arnd Meyer, RWTH AachenDØ Collaboration Meeting

October 8, 2003


Data Taking Status

Total datasample ontape withcompletedetector> 200pb-1


Data Taking Status cont.

● Lab reached its goal of delivering 225pb-1 in FY03

Also for DØ: BD delivered

∫L dt = 227.7pb-1 in FY03

26pb-1 per month since May

But had to run 6 weeks longer than hoped

● Cuts into our running time next year

Startup November 17 – difficult to anticipate rapid startup

Six week shutdown next summer / fall

● We do not expect to get significantly more ∫L dt in FY04 than we got this year

25% pbar tax (Recycler commissioning), studies ⇒ 233-328pb-1 in FY04

See Dave Mc Ginnis' presentation on Oct 3 ADM

FY02 FY03


FY 04 Luminosity Profile

More (2x/week) andshorter (8 hours)accelerator study periods

Studies only if >140 hrsof store time in theprevious 14 days

Higher deliveredluminosity throughimproved stacking rate

Improve stacking ratethrough shorter pbarproduction cycle time(2.4sec 1.7sec) 11.3mA/hr (FY03) 18 mA/hr (FY04)

≃ 3 months turn-onafter shutdown

pessimistic∫L dt ≃ 233pb-1

design∫L dt ≃ 328pb-1


... and Plans / Dreams

You are here

3⋅1032cm-2s-1

2004

● STT fully commissioned● Missing pieces of CTT

fully commissioned● Taking data with rates of

2.5kHz / 1kHz / 50 Hz after L1 / L2 / L3 and 90% average efficiency

● CPS / FPS used in the trigger and for physics

● Most data quality problems are caught online (Global Monitoring et al.)

● 1-2 fewer people on shift

● Taking shifts and improving the detector is not considered a necessary evil

● We have 0.5fb-1 of good data on tape● Reco takes 1sec/event on my 2-year old desktop


Data Taking Efficiency

Shutdown

Winter '03Shutdown

The LuckyWeek

Pre-shutdownspecial runs/

studies


Data Taking Efficiency cont.

● Average data taking efficiency for 2003 is 86%● Current upper limit is ≃ 95%

3-4% global front end busy

1% begin and end store transitions

<1% run transitions

● Typically in the upper 80%'s for the last six months

Since Beaune, “lost” 83.5 hours of store time (12.6% by time)

12 hours for special runs

Largest single failure (4 hrs): low airflow trips of L1CAL on July 27/28. Fan belt replaced. Symptomatic: some of the largest downtimes are one-time occurences

Failures by component (without special runs):

– SMT: 12 hours

– Muon/L1Muon: 12 hours (trips, readout errors/crashes, trigger problems)

– CAL: 9 hours (mostly BLS power supplies and hot trigger); + 5 hours L1CAL

– CFT/CTT/PS: 5 hours

– L3/DAQ/Online: 4 hours

Tracking crates readoutcollaborations' decision: L1A vs. FEB

fairly optimized, contiuouseffort to keep this low


Data Taking Efficiency cont.

● Much of the time running close to our desired efficiency to tape – 90%

● At the same time, data quality improves

Number of conditions that causes us to stop data taking (automagically or manually) is continually increasing

● Credit to many (few) dedicated people!● Last weeks of data taking had large number of special runs,

increased number of accelerator studies – still ran over 80%● Large downtimes are discussed at the weekly operations meeting● Several systems marginal in terms of expert coverage: one

or no resident expert – no manpower for proactive improvements

http://www-d0.fnal.gov/runcoor/


Run II Bests

Regularly updated:

Best days by data taking efficiency

– 95.0% on June 22nd; so far 8 days with 93% or better efficiency

Best runs and days by recorded luminosity (Aug 10, 488nb-1; May 4, 1.68pb-

1)

Best stores by initial DØ luminosity (Aug 10, 4.55⋅1031cm-2s-1)

http://www-d0.fnal.gov/runcoor/


A “Typical” Store

Wobbly L1 rates (CAL)Initial Lum. 3.9⋅1031cm-2s-1

Store lost (quench)5-3% L1 Busy

Present max. rate guidelines:Level 1 1.4 kHz FEB < 5% (5-10% headroom to accountLevel 2 800 Hz Muon r/o for rate fluctuations)Level 3 50 Hz Offline (30% room)


Control Room Shifts

● It is a burden on the collaboration to fill that many shifts (and schedule them!)

The shifter duty is 7 shifts / 6 months per person on masthead

Only about 1 shift per month per person (on average!)

● Calorimeter and Muon shifts consolidated into CalMuon shifts since June

Rocky at times, but overall OK

Cost of additional training offset by savings in total number of shifts

There is more training involved – took some time to be realized by “old” shifters

● Next natural choice for merging is SMT / CFT – will require initiative from detector groups (clear instructions, simplify, automate)

● More than a third of the collaboration have not yet taken a single shift in 2003

● Request for new DAQ shifters: response was mediocre

Have a very good crew currently, but could use more volunteers – we need >26 new DAQ shifters per year. Need one more DAQ shifter this year!

I think I am too naïve for this job

The fact that we are collecting data with high efficiency is to a large partdue to the presence of 5 – 6 well trained people in the control room


Data Quality & Global Monitoring

● Online data quality monitoring consists of three (four) parts

Significant Event System (“Slow Control”)

– Catches an increasing number of hard- and software failures

– In many cases pauses the run to ensure consistent data quality

– Working very well, could use additional experts guidance

DAQ Artificial Intelligence

– Notifies shifter of abnormal conditions (global rate fluctuations, BOT trigger rate, ...)

– Automagically fixes certain problems (SCLinit), e.g. sync. problems in L2

Sub-detector monitoring examines

– Many expert-level plots (but experts are not generally on shift)

Global Monitoring

– Trigger rates, Trigger Examine, Vertex Examine, Physics Examine

● Global Monitoring has great potential, but there are many issues – examples follow


– If we can't fill all shifts, need to think about merging with Captain's and other shifters' duties

Global Monitoring cont.

● Technical Issues

During the transition from trigger list v11 to v12, ran for weeks with wrong/bad reference plots (different triggers, then rapidly changing prescales)

PhysEx uses random sample – should be based on certain triggers

Low statistics (slow reconstruction)

● Psychological Issues

Lack of interaction between detector shifters and GM

– GM detects feature in Gtrack phi distribution – SMT shifter cannot correlate with his occupancy plots

– Need effort from all detector groups to bring their expertise into GM plots (only Muon group has done this so far)

● Organizational Issues

Shifts not being filled (e.g. 8 in August)


DQ & GM cont.

● LmTrigger urgently needs improvement

For example averaging over different time periods, uncertainties, luminosity dependence of trigger cross sections, ...

Extremely important tool to identify problems quickly

● Overall, somewhat slow progress (remember Beaune?)● If we want to continue reducing the number of shifters in the control

room, GM needs a major effort (time, people, attention) From the collaboration – great task for groups new to DØ

Automation should be the goal

Need to catch all major problems online● Up to one third of the data is thrown away in the analysis stage –

sad!

Everybody who discards data through “bad” or “good” run lists should

make that extra step and think about how to catch the problems earlier!


Shutdown Plans & Progress

● In the middle of 10 week shutdown (Sep 8 – Nov 17)

7-8 weeks for experiments, 2-3 additional weeks with limited access

● Four scheduled power outages – recovered well from the first two (and unscheduled outage on Sep 26), two more on October 13 and 18

● 24x7 DAQ shifts and day shift Captains – thankless task! (“Good God, please, someone, if you see me in the parking lot, run me over. Kill me. I am so bored to death.” - anonymous DAQ shifter)

Keep as much as possible of the detector in working order

● Major goals for the shutdown

Improve reliability: reduce access time, periods with incomplete detector etc.

Improve quality of the data: reduce calorimeter noise, repair Silicon HDI's

● Detector groups together with mechanical and electrical support groups have developed detailed job lists including detector opening and closing, allocation of manpower resources from detector groups and support teams, survey as needed during major moves

● Access went smoothly so far – no accidents, on schedule, great support● See Friday's talks and parallel sessions for details


Silicon Status

● Bias scans (August): measure depletion one layer at a time

Use tracks from CFT and other SMT layers to determine cluster charge as a function of HV

Runs with HV varied between 0% and 100% in 10% steps

SMT in full readout (no sparsification)

Average over ladders (statistics)

0

50

100

150

200

250

0.0E+00 5.0E+12 1.0E+13 1.5E+13 2.0E+13 2.5E+13 3.0E+13 3.5E+13 4.0E+13 4.5E+13

Fluence of 1 MeV n / cm 2̂

DS ladder I DS Ladder IIDS wedge SS Ladder ISS Ladder II DSDM LadderHamburg Model (n) Current Dose

● Shutdown: Repairs of failed HDI's / electronics

Before shutdown: 136 disabled (up from “irreducible” 84 just after Jan shutdown)

12 are repaired (2 weeks of 2 shifts/day with 4 people); ≃100 are presently disabled

59 are unstable – study in the coming two weeks. Some recoverable for partial readout

● Cancellation of Silicon replacement means we must plan to operate current detector for the life of the experiment

Have to evaluate what steps can be taken to increase chances of the detector's reliable operation long term

CurrentDose


CFT and Preshower

● Maintenance of LVPS's and installing upgraded LVPS (better connectors) – this week

● Maintenance of the VLPC He cooling system

● Modifications of AFE boards to remove unused SVX inputs from the readout

Reduce data size and DAQ deadtime

Pedestal RMS's across all channels are higher than before the shutdown, under investigation

Continue testing AFE modifications over the next week.


Calorimeter

● Calorimeter Workshop, plenty of presentations during the week

● Replaced all large cooling fans for preamplifier cooling

● Studies of calorimeter noise – access priority given to “noise task force”

Controlled power-up after power outages

Investigating how SMT noise gets into the Calorimeter

Source of 10MHz noise

– Went around with a small antenna to identify sources

Grounding test planned for October 13

– Detailed plan in place

– Disconnect AC, safety ground, telephone, etc.

– Attach current controlled power supply, slowly increase current up to 100A, and look for heat sources

Preliminary findings this week, final report later


Muons, Lumi, FPD

● Forward Muon First-time access to A layer forward muon tracker (8-12 hours for

opening+closing) completed: replacement of preamplifiers, gas leaks, gas monitors; C layer repairs in progress

Number of non-working channels now 0.15% (trigger counters), 0.5% (drift tubes)

● Central Muon Installation of extra trigger counters under way – running into a few snags

(tight clearance on east side) but no show-stoppers

Installation of 144 remote power cycle relays for front-end electronics and all relevant cabling – on schedule, first driver board tested

Pulled a couple of wires drawing moderate to high currents for investigation

Installed Power PC's in the remaining muon readout crates that had 68k's

● Luminosity system – Cable work in the gaps last Thursday. Come out of the shutdown with complete readout electronics

● Forward proton detector – Maintenance and installation of electronics for full system operation on schedule. Waiting for parts


Trigger, DAQ, Online, General

● Firmware upgrades on L1CTT and L1Muon, maintenance on L1CAL: progressing well

● All Level 2 Alphas have been replaced with Betas! Running smoothly so far, updated shifter instructions

Now progressing on “worker” and “administrator” code to run with 2 Betas per crate

● Online & Controls Replaced 8 disks that have died over the last 2 years (4 disks for the data

logger)

Offline switch upgrade planned for Oct 13 – required electrical work means extended downtime for all online systems as well

Major online software upgrades this week – Python, Epics, VxWorks

● Still some downtimes ahead for TFW (new TCC etc.) and Online (IP shuffling etc.) upgrades/maintenance

● General detector maintenance – Air handlers, hydraulic systems, vacuum jackets, cooling water systems, ODH heads, etc.

● Old Cryo UPS replaced – Switched to commercial power source last Tuesday, then to new UPS on Thursday


Conclusions

● Detector is running well (better than some people want to make us believe)

DØ was in good shape for physics data taking until the last store

Dmitri, detector group leaders and crews, support groups, daily operations have been doing a great job in operating the detector and have done a lot of planning for a successful shutdown

Data taking efficiency 86% for the year – “physics analysis efficiency”??

214pb-1 integrated luminosity on hands with full detector in readout

Progress in online data quality monitoring not as good as hoped for

● Shutdown progressing well and on schedule

A lot of work scheduled

Brace yourself for 2 more power outages: Oct 13 (½ hour) and Oct 18 (8 hours)

● Bring the detector back into shape by the end of October

Assume access ends on November 17th, close detector the week before

Plan to finish all major jobs in 7 weeks, only limited access during weeks 7 – 10

Arnd Meyer (RWTH Aachen) Oct 8, 2003Page 1 Data Taking Status and Plans Achievements Concerns Plans Arnd Meyer, RWTH Aachen D Ø Collaboration Meeting October.

Documents