1. 05.03.2003Chamonix 03 / Presentation 5.5 / J. Wenninger1 Orbit control for machine operation and protection Orbit control requirements Feedback performance & limitations Feedback architecture Summary & Outlook J. Wenninger AB/OP Main persons involved in orbit FB (past & present) : L. Jensen, R. Jones AB/BDI J. Andersson, S. Chtcherbakov, K. Kostro, T. Wijnands AB/CO M. Lamont, R. Steinhagen, J. Wenninger AB/OP Q. King AB/PO

2. 05.03.2003Chamonix 03 / Presentation 5.5 / J. Wenninger2 Stabilization requirements I Collimation (see also R. Assmann):  Cleaning section : <  0.3    m  TCDQ absorber in IR6 : <  0.5    m @ 7 TEV  On day 1 + some   @ injection and during the ramp.  For larger  * in physics. Collimation inefficiency versus position error …for nominal performance in physics and for  * = 0.5 m !! Stabilization to  200  m is sufficient :

3. 05.03.2003Chamonix 03 / Presentation 5.5 / J. Wenninger3 Stabilization requirements II Beam dumping system (see also B. Goddard) :  CO stabilized to  1 mm (peak) @ kicker & septa in IR6 – H plane only !  in the shadow of the collimation requirements / TCDQ. Injection :  CO stabilized to 0.2 mm rms at the TDI. Machine protection :  Stabilize CO around the WHOLE ring to ensure that the aperture limits are always in the collimation section. Very important for the triplets. Machine performance & operation :  Minimize beam excursions with respect to reference CO to help control feed-downs from multipoles (injection & snapback).  Stabilize the orbit during the squeeze.  Minimize beam movement at the IRs in physics.  Make life (much) easier for operation !

4. 05.03.2003Chamonix 03 / Presentation 5.5 / J. Wenninger4 Observed orbit drifts : ~ 200-500  m rms over a few hours ~ 20-50  m rms over ~ minute(s) LEP/LHC tunnel is a quiet place. Ground motion spectrum ~ f -3 Ground motion @ LEP orbit rms    ground movement  Uncorrelated motion :   35  Waves (E. Keil): f < 5 Hz   1 f > 5 Hz 1 <  < 100 CO movements at f > 0.1 Hz are in or below the few  m range ! 1  m 1 nm OPAL cavern IP4 (S. Redaelli) Ocean waves @  = 100 m @   100-150 m

5. 05.03.2003Chamonix 03 / Presentation 5.5 / J. Wenninger5 Vertical low-  quadrupoles @ LEP moved vertically ~ 100  m during the machine cycle : Orbit drifts of 2-5 mm rms dominant effect on LEP orbit Not entirely reproducible Related to temperature Lot’s of problems in the ramp due to the absence of a real-time feedback. Magnet girders @ LEP We must watch out for : Triplet movements Vibrations (cryo…)  = kick due to low-  movement @ one IP 1  rad  40  m rms @  =100 m

6. 05.03.2003Chamonix 03 / Presentation 5.5 / J. Wenninger6 Orbit movements during Snapback and decay  Random b1 errors (~ 0.75 units)   1 mm rms in the horizontal plane (with a large spread).  Random a1 errors (~ 2.6 units)   3-4 mm rms in the vertical plane.  Feed-down from b2 errors   0.2 mm rms in both planes !

7. 05.03.2003Chamonix 03 / Presentation 5.5 / J. Wenninger7 Ramp, squeeze, collisions Ramp :  “Experience” shows that drifts of few mm rms have to be expected.  Magnetic centre of the warm quads expected to move by ~ 100  m. (should be Ok !) Squeeze :  Large drifts – up to 20 mm rms (IR1 & IR5  * : 18 m  0.5 m)  Effects are very sensitive to the input conditions : orbit offset,  -fct and strength change in IR quads. Collisions :  Ground motion …  (Parasitic) beam-beam kicks. @ LEP the inability to control the orbit in real-time during ramp & squeeze probably cost us ~ 5% overall efficiency ! and was responsible for > 30% of the lost ramps.

8. 05.03.2003Chamonix 03 / Presentation 5.5 / J. Wenninger8 Orbit drifts & requirements in short  Most drifts occur / build up on time scales of few seconds to minutes.  need a good feedback gain at and above ~ 0.1 Hz.  The squeeze could be the most delicate phase for the orbit FB.  The most critical requirement apply during collisions where slow ground motion is hopefully the main ‘enemy’…  During the initial operation, requirements are not as stringent – 200  m rms tolerance is probably OK.  Most perturbations produce ~ REPRODUCIBLE drifts (except ground motion)  80% (?) or more of those effects can be anticipated and feed-forward.  reduces load on FB.

9. 05.03.2003Chamonix 03 / Presentation 5.5 / J. Wenninger9 Power converters & magnets Cold orbit correctors :  Circuit time constants   10 to 200 s (arc correctors ~ 200 s).  For small signals the PC is limited to frequencies of ~ 1 Hz. Warm orbit correctors :  Circuit time constants  ~ 1 s.  PC could run well beyond 10 Hz !  Too few of them in the cleaning section to build a closed correction !  would need warm (or super-power cold) correctors in the cold section of the machine !  Cannot profit from their speed – we could consider slowing them down to remove this source of fast orbit movements ! Controls :  All PCs accept real-time input @ up to 100 Hz.  Each PC can only be controlled by a SINGLE feedback loop !

10. 05.03.2003Chamonix 03 / Presentation 5.5 / J. Wenninger10 Orbit acquisition Per ring and plane : 500 orbit measurements ~ @ every quadrupole. The real-time orbit acquisition will run at 10 Hz. For a good FB performance : sampling frequency ≥ 20 x (fastest perturbation to stabilize)  FB limited ~ 0.5 Hz ! SPS tests in 2002 on 4 BPMs equipped with LHC readout:  Transmission delays over standard SPS network are OK for 10Hz CO.  Very good electronics performance.  CO resolution < 20  m for nominal intensity. Extr. flat top 10 Hz sampling of the LHC beam cycle in the SPS averaged over 2 hours Start of ramp

11. 05.03.2003Chamonix 03 / Presentation 5.5 / J. Wenninger11 Feedback performance To improve the performance towards higher frequencies  orbit sampling of 20 Hz or more !  Delay of 1 period (100 ms).  Limitations due to the correction strategy not included !  2 period delay (200 ms) may be more conservative for initial operation…  reduced gain. Gain = 10 @ 0.1 Hz Gain = 1 @ 1 Hz Feedback gain (not ultimate performance !)

12. 05.03.2003Chamonix 03 / Presentation 5.5 / J. Wenninger12 Complications, complications… Ramp : Energy tracking. Squeeze :  Orbit response matrix must be updated to track optics changes.  Reference orbit must be updated (crossing scheme…). LHC energy stabilization at injection with horizontal orbit correctors :  The same correctors are also used by orbit FB.  FB also responsible for energy ?  Energy trims not via real-time inputs since very slow changes ! Ring 1 – Ring 2 coupling in IRs 1,2,5 & 8 :  Handle rings individually or in common ?  Individual ring handling will NOT work well for the squeeze. Trims : Must allow some form of manual corrections (bumps, Xing angles …). Post-mortem diagnostics

13. 05.03.2003Chamonix 03 / Presentation 5.5 / J. Wenninger13 Feedback Strategies Global correction / feedback :  By definition such a FB affects the orbit in (at least) one entire ring. Local correction / feedback :  Uses a subset of monitors and correctors.  Provides a LOCAL correction, i.e. does not affect the orbit outside its ‘working’ range. Requires a buffer region to enforce the closure. Collimation IR This is NOT really what we want (for protection…) !

14. 05.03.2003Chamonix 03 / Presentation 5.5 / J. Wenninger14 Marrying local & global FB loops The classical approach (Light sources) : frequency de-coupling  Very fast local loops (> 50 Hz), sampling rate ~ kHz.  One slow global loop (0.1 Hz).  Does not work (well) @ LHC due to the ‘slow’ sampling and large perturbations during snapback and squeeze. A single global loop with chained corrections :  Can apply both global & local corrections – complete info available !  Very flexible & easy to (re)configure.  Avoids correction weighting – tricky to tune.  Total correction =  corrections Input Orbit Predicted Orbit Predicted Orbit Corrected Orbit … Global Corr. Local Corr. # 1 Local Corr. # 2 Local Corr. # n

15. 05.03.2003Chamonix 03 / Presentation 5.5 / J. Wenninger15 Centralized feedback architecture Global correction as “workhorse” – good to satisfy most requirements entire CO information available. can be made rather insensitive to bad monitors. can be easily configured and adapted.  numerical problems are more complex.  large amount of network connections to front-ends. Local corrections ensure tight constraints in local sections…  (very) sensitive to faulty monitors. FB Data transfer first tests  OK ! lightweight ‘protocols’ please !

16. 05.03.2003Chamonix 03 / Presentation 5.5 / J. Wenninger16 Ground motion correction in collision Simple global correction :  “Conservative” correction strategy – insensitive to isolated faulty BPMs.  Decouple rings (i.e. common beam pipe elements not used). IP1Primary Coll. Residual orbit shifts after ~ few hours of coast / 1 beam  =10  m  = 17  m Note the very large residual drift @ IP1 despite a 100 x smaller   correction strategy !

17. 05.03.2003Chamonix 03 / Presentation 5.5 / J. Wenninger17 Entirely local feedback architecture FB reduced # of network connections. numerical processing simpler.  less flexibility.  not ideal for global corrections.  coupling/X-talk between loops is an issue.  problem with boundary areas to ensure closures. Example of an aggressive solution… the Swiss Light Source…

18. 05.03.2003Chamonix 03 / Presentation 5.5 / J. Wenninger18 SLS : global correction with local loops ! One can cast the solution of the orbit problem in the form of a matrix multiplication (  = kicks, y = input orbit) All non-zero elements are very close to the diagonal Each local FB loop receives a piece of the matrix to perform a global orbit correction (+ needs to talk to its neighbor !). Equivalent to a MICADO correction using ALL AVAILABLE orbit correctors of the machine – every “bad” monitor kills you ! A =  LHC matrix

19. 05.03.2003Chamonix 03 / Presentation 5.5 / J. Wenninger19 BPM reliability in critical areas Cleaning Section :  Stabilization to the required accuracy with a local correction can only be achieved throughout the cleaning sections if the BPMs are reliable at the level of  50  m or better.  To detect systematic errors at the level of 100  m or less is not simple !  Those BPMs are installed in a very difficult area (radiation). Triplets – inner IR region :  The directional couplers in the common beam tube have a tough job to separate the beams.  This is a critical region with  * = 0.5 m – aperture !  Experience will show how much we can trust them.  Fortunately we start with 75 ns bunch spacing  OK !

20. 05.03.2003Chamonix 03 / Presentation 5.5 / J. Wenninger20 Summary & outlook Stabilization requirements for protection & collimation Tough @ 7 TeV + squeezed – but no show-stoppers. The squeeze is likely to be the most delicate phase. Architecture & correction strategies More systematic simulations & tests required to :  choose implementation – local / global…  check ring decoupling and strategies. Fast orbit movements or failures cannot be avoided by any orbit feedback  interlocks on beam movement / beam position. SPS tests in 2004 Test of a closed local orbit FB based on 6 BPMs equipped with standard LHC electronics  good test bed & milestone.  end 2003