2. Turnaround: from LHC to FCC?
David Nisbet, with input from K. Fuchsberger

3. Outline
LHC turnaround in 2016
LHC v FCC
Conclusions

4. Efficiency: turnaround and beam cycle
Turnaround Time
Beam Cycle Time
Turnaround time
Stable beams to Stable Beams
A total of 157 fills in 2016
Beam cycle time
Only physics fills that arrived to Stable Beams
A total of 179 fills in 2016
General
Only proton fills are considered
Beam mode declaration time is used (from logging DB)
CRS to 3m
Squeeze to 40cm
TOTEM bump
Standard, BCMS
RB.A12
RQ5.L1

5. Turnaround See Evian presentation for more details 2015 2016 Min 2.5 h
p-p physics, ignoring faults > 24 h
Fills following accelerator mode change have no associated turnaround
Following the Restart, Technical Stops, Special Physics Commissioning, Machine Development
2015
2016
>4h = faults & measurements
2016
2015
Min
2.5 h
2.2 h
Median
5.2 h
6.3 h
Mean
7.1 h
6.6 h
K. Fuchsberger, M. Solfaroli + AWG (CERN-ACC-NOTE )

6. Clean turnarounds 14 Turnarounds No gap in Fill-number No fault
No Precycle
No EOF MD
14 Turnarounds
2016
Min
2.50 h
Median
2.67 h
Mean
2.74 h

7. Injection (physics beams)
Injection in 2016 limited to 96b/injection
46 injections for 2076b Nominal (t_fill_min = ~26mins)
48 injections for 2220b BCMS (t_fill_min = ~28mins)
For fills reaching STABLE BEAMS
In 2015: 122 fills, with an average of 761b up to 144b/inj
In 2016: 179 fills, with an average of 1688b up to 96b/inj
127 fills with >2000b
2016
(>2000b)
2015
Beam Quality
Dumped while injecting
average for 2016 (<2000b) = 32min/579b
Aug-16 21:08: SPS RF cavity
Aug-16 15:13: ADT adjustments
Aug-16 19:30: Normal
16-Aug-16 09:53: min SPS no beam then normal fill
Several factors impact the injection process:
Train length, TL steering, SPS super cycle length, BQM setup, Injector beam quality + availability, intensity (vac, cryo, etc)
Faults and machine setup
2016 >2000b
2016
2015
Median
37.8 min
37.0 min
44.1 min
Mean
42.1 min
39.2 min
48.2 min

8. Adjust
ADJUST in 2016 = TOTEM dispersion bump, collide IPs 1+5, collide IPs 2+8
Optimisation after finding collisions requires operator interaction
2015
2016
15x EoF MDs removed from data
Checks and problem solving
Totem bump played at end of squeeze.
20 – 25 mins = regeneration, alternative check (ADT, BSRT, etc)
25 – 30 = mixed. Some long adjusts. Regeneration + lumi publication.
53min (fill 5370) = CMS lumi very difficult to find. MD but normal collisions (6 Oct) – effect of bump in 5R.
78min (fill 5330) = BPM faults prevent declaring SB.
2016
2015
Min
9.2 min
4.9 min
Median
14.1 min
12.5 min
Mean
16.2 min
13.7 min

9. Turnaround in one go Beam Mode LHC Theory 2015 LHC Median
2017 LHC Median (estimate)
2016 FCC Theory (estimate)
2017 FCC Median (estimate)
SETUP
10 min
222.7 min
158.5 min
147.0 min
10 + ? min
INJECTION
38 min
58.1 min
51.6 min
40 min
40 + ? min
PRERAMP
4 min
5.4 min
4.2 min
3 min
5 min
RAMP
20 min
20.4 min
19 min
FLATTOP
4.8 min
0 min
SQUEEZE
18 min
13.1 min
18.0 min
11 min
ADJUST
12.5 min
14.1 min
5min
RAMP DOWN
31 min
41.0 min
40.0 min
TURN-AROUND
2.2 hrs
6.3 hrs
5.2 hrs
4.5 hrs
1.8 hrs
1.8 + ? hrs

10. Observations
Clearly the greater the size of the machine the more systems are required
One must ensure design for availability
Otherwise it is unreasonable to use the LHC figures directly without mitigation
In my limited experience, the median value is a useful indicator of efficiency
The large outliers tend to push the average up disproportionately
Turnaround estimates for a realistic machine should include both ‘theoretical minimum’ and ‘realistic forecast’
Method for realistic forecast?
Turnaround (min) + t_fault (median) = 1.8hrs + 2 hrs?
The FCC relies heavily on the availability and injection time being close to forecast
In the LHC scenario, can it really cycle 4x in ~40 mins?
In my opinion, more data should be provided to justify why this is achievable

11. Conclusions
The LHC does a little more in a little less time in 2016 compared to 2015
Ratio of median / theory was 2.8 in 2015, 2.3 in 2016
Could trend to 2.0 in 2017 (if fault statistics remain similar)
BUT
Data from 2015 and 2016 shows long tails
As clearly stated in 2016 paper, must assume some distribution in estimate
Mitigating arguments should be provided if the theoretical value is retained as baseline (applies both for LHC and FCC) , for example design for availability
The estimate made in 2016 is a reasonable ‘theoretical minimum’
Theoretical minimum for FCC = 1.8 hrs
Assumes circuits are designed to ramp down inside time of main dipole ramp (not the case in the LHC today)
However turnaround estimates for a realistic machine should include both ‘theoretical minimum’ and ‘realistic forecast’
Method for determining realistic forecast needs to be evaluated and agreed…
In particular the times for ‘setup’ and ‘injection’
Apply same methods to both and FCC…

12. References
FCC-hh turn-around cycle, R Alemany Fernandez et al, CERN-ACC
Concepts for magnet circuit powering and protection, M. Prioli, FCC Week Rome 2016