On reproducibility (again)

1
On reproducibility (again)

• Based on several inputs of N. Sammut, S.
  Sanfilippo
• Presented by L. Bottura to the LHCCWG
• 11.9.2007

2
Components reproducibility

• Geometric
• Change in the cable positions
• Saturation
• No issue detected
• Persistent currents
• Pre-cycle current
• Modeling

• Decay Snapback
• Pre-cycle current and time
• Residual magnetization
• It packs everything we did not understand properly

3
GeometrySummary of uncertainty

• uncertainty estimated as 3 ? of multipoles
  repeatedly measured on the same magnet (few
  magnets tested)
• after powering
• after training
• u(b1)2.8 units
• u(b3)0.3 units _at_ 17 mm

4
Persistent currentsSummary of uncertainty

• The effects are large (of the order of 10 units)
• The variability associated with powering cycles
  is very large

• MB (IFT 2 kA vs. nominal)
• u(b1) 1.5 units
• u(b3) 1 units _at_ 17 mm
• MQY (Imin 50 vs. 200 A)
• u(b2) 10 units _at_ 17 mm

These values are relevant only if the pre-cycle
is changed from run to run
5
DecaySummary of uncertainty

• Although we have seen (much) better, we maintain
  that the empirical model (data fits) has a
  typical error that can amount to up to 20 of
  the effect
• Main source of uncertainty is from the modeling
  of powering history, all other effects (aperture
  differences, cycle details, ageing) are small and
  have negligible systematic
• Why so cautious ?
• The sample of magnets used for the data-fitting
  is limited (10 magnets)
• The biased sample adds uncertainty in the
  projection of the average

6
Uncertainty after correction
Values estimated for MBs in July 2004, RMS review
NOTE variations of pre-cycle from the nominal
one (e.g. due to limitations during commissioning
or changes in optics) will cause an additional
uncertainty that can be much larger than the
above values
7
New elements and a reminder

• Tracking test, July 2007
• Effects of hysteresis on correctors
• Other mysteries and miseries

8
Tracking - what ?

• The plan was (and still is) to have 2 MBs and 1
  MQ running simultaneously to verify
• B1/B1const
• B2/B1const
• integral B30 and integral B50 MBs
• We experienced difficulties in maintaining stable
  operating conditions on the SM-18 benches (mainly
  on cryogenics), which made running two benches
  simultaneously practically impossible
• A first test campaign was nonetheless performed,
  testing B3 and B5 corrections on a single MB
  powered using a customized version of LSA from
  the SM-18 control room

9
Main result

• After considerable tweaking, and many, many
  measurements(1) we could compensate b3 to
  approximately 0.5 units (0.25 units peak-peak)

(1) The measurements taken during the tracking
test amount to about 10 of the whole
measurement raw database (ovation for Nicholas
and Marek)
10
Why so deplorable ?

• Measurement issues
• Instrument calibrations (gains, coils 0.1 units)
• Completeness of the harmonics integral (small ?)
• Modeling issues
• Persistent current model
• Relevance of the data source (50 A/s vs. 10 A/s
  0.5 units)
• Accuracy of the model (0.1 units)
• Residual magnetization model (?)
• Control issues
• Translation of nominal optical parameters to
  currents for the power supplies (checked on-line,
  further check in progress)
• Timing (small)
• ?

11
10 A/s vs 50 A/s precycle
Difference in pre-cycles produce substantial
variations in the apparent persistent currents,
affecting the model. We have to check the
accuracy of the extrapolation.
12
Other miseriesEffect of precycle - MQT
Data courtesy of W. Venturini
The effect low current cycling can be massive
55
13
and misteriesa2 anomaly in Ansaldo-2 (2002)

• The shape of the a2(I) has a strong anomaly in
  one aperture of on Ansaldo-2 (2002) reassembled
• This data is real !
• not a cable hysteresis
• measurements are OK as far as we can tell
• a magnetic piece (protection layer, shim,) in
  the collared coils?
• Observed in few other magnets
• Depends linearly on maximum current reached

14
A conclusion for this round

• For the moment, we maintain the original
  estimate, with the following caveats
• The whole series tests were performed using 50
  A/s pre-cycles (or proportional for magnets
  working at nominal current other than 11850 A).
  The 50-10 extrapolation may have troubles
• We still cannot explain why the tracking (a
  single harmonic, a single magnet) is so
  miserable. A new test campaign is being set-up
  for mid October. Volunteers are welcome