Presentations Heard very good presentation by Li-Hua Yu about orbit feedback simulations (at DC) –Established that general number of BPMs and fast orbit feedback correctors is large enough to just about meet 10% of vertical beamsize stability goal (assuming just quadrupole jitter and BPM noise) Other presentations (Bob Hettel, Christoph Steier, Michael Boege, Igor Pinayev, Roland Mueller, Om Singh) also addressed orbit feedback issues, design considerations and solutions (for NSLS-II and other sources). –Multitude of solutions, all have advantages and trade-offs. –General performance at state-of-art facilities is approaching NSLS-II requirements, but not quite there, yet – especially for longer time periods.
Observations, Recommendations, … General guidelines: –Orbit feedback should be flexible, reliable, upgradeable, error tolerant (BPMs will fail because of large number), operator friendly One should concentrate right now on issues that need to be decided now (number, location, type of BPMs, correctors) – postpone to look in too much detail into specific BPM receiver technology – will be obsolete once you need them Simulation that should be done soon –Do quantitative comparison of multiple corrector magnet and BPM configurations looking at how well conditioned SVD inverted matrix is (spread of singular values) –Do time resolved feedback simulations (probably need to be numerical, not analytical because of nonlinearities and time discretization of system): Study sensitivity to all performance parameters, like update rate, latency times, noise, DAC/ADC resolution, algorithm choices (controller, different PID parameters for different singular vectors, …) to evaluate trade-offs in diagnostic choices. Might be beneficial to have person with fast feedback expertise working on system optimization.
O, R (ctd.) More on BPM hardware and challenges in Om Singhs summary! –RF BPMs – state of the art not fulfilling NSLS-II specs consider 2 RF BPM types (optimized signal and button geometry in straight) (Digital) RF BPMs will improve substantially in next years – but you have to evaluate the risk of commercial systems not meeting all your performance goals when you will need them, if you do not undertake a (joint) development yourself We recommend linear encoders (best on invar rods to ground) at least for straight section BPMs, maybe arc BPMs as well Optimally incorporating BPMs into feedback system requires experts and substantial software development –X-ray BPMs are essential (2 per beamline) Common hand over point between machine and users Electronics should have full bandwidth and be incorporated homogenously into fast orbit feedback (even if hey might not be used in feedback algorithm initially) Noise performance is generally better than X-ray BPMs – might need to make integral part of global feedback for NSLS-II Need intelligence (calibration constants are gap dependent for IDs) Can be fully incorporated into global feedback (need provision to zero gain if ID open) Transition maybe easier by using X-ray BPMs to modify golden orbit setpoints for RF BPMs
O, R (ctd.) Corrector magnets (again more in Oms summary): –Definitely need additional fast correctors right next to IDs. –Should attempt to have at least 2 fast correctors between any beamline source points. –All correctors for fast feedback need to be the same –Corrector choice affects feedback system layout If all correctors are fast it would allow to have fast feedback only – could have some performance advantage over fast+slow and makes system simpler Caveats: bigger noise put onto beam by correctors, need resistive windows in aluminum chamber, risk of window and bellow transfer function not being equal
O, R (Ctd.) Algorithm choices –System needs to be distributed (no long analog cables) but could have some central processing –Slow+fast feedback interference problem is easy to solve (APS, ALS, …) and is not an argument against using a mix of slow+fast feedback –Different time constants for different singular vectors seems attractive (simple version of it is slow+fast feedback) –Consider more advanced control algorithms than straight PID (the 60 Hz notch filter of NSLS-I is one example) –Recommend against local feedbacks –System should have automated BPM error detection (and if possible automated reconfiguration); MTBF of libera system and # BPMS suggests failure every couple of weeks –Consider (in your detailed numeric simulation studies) update rates of system well beyond 5 kHz which is just about state of the art now and will be easy to achieve in the future.
O, R (ctd) It is very beneficial for accelerator experts to understand beamlines and beamline experts to understand accelerator and to closely collaborate. –The photon source should be made as stable as reasonable, but accelerator and beamline system needs to be approached as one. Very important for troubleshooting, commissioning and to figure out where it is easiest to achieve a certain stability goal (in a specific frequency band) Beamsize stability is equally important to orbit stability! –Local skew quadrupole correction with multiple skew quadrupoles per sector is state of the art. Need to study what is a reasonable number and distribution of skew quadrupoles. –Correction is typically by feed forward (probably sufficient). Could be advantageous to be incorporated into feedback system algorithm. –Insertion devices need to be corrected locally. –Potential issue is eddy current time constant. Skew quadrupoles in arc will be on thick aluminum chamber which limits their use (for all purposes NSLS-II is discussing so far that is OK). –IDs need local correction in straight – those skew correctors should have high bandwidth (if possible at least for EPUs)