Choose timezone
Your profile timezone:
FCC-ee optics tuning WG meeting
→
Europe/Zurich
6/R-012 - conference room (CERN)
FCC-ee optics tuning WG meeting
Satya, GHC v25 tuning (latest version of GHC, L*=2.4m) with sextupole field errors.
1832 BPMs and orbit corrs (@ quads). 616 skew quads (@ sexts). Strength errors of 2 units for arc quads and arc sextupoles.
90 seeds tune successfully out of 100 (most seeds fail at the first stages of tuning). Linear optics are really well corrected. Tunes are within 0.001 from design (tune is part of the full response matrix). Three iterations are applied. All arc quads are used.
DA is more deteriorated than for the v23 studies. Maybe to be clarified, deterioration comes from longer L* =2.4m or from sextupole strengths errors. Slight impact on DA after knobs is also seen. Vertical emittance is targeted as 1.48 pm. Analytical computation of emittance gives results close to design. The knobs deteriorate slightly the vertical emittance (D*y correction identified as source as it requires skew quad fields, plus maybe there is leakage).
Lifetime calculation for 512 turns without beam-beam gives a lifetime, above 85% of design lifetime for all seeds. However, target should be very close to 100%, hence granularity in this plot might not be enough.
Christian rises the point of dispersion measurement at 10 um, which seems challenging. The IP Dx/y tolerance is recalled to be 1um from EPOL studies. This will be studied by EPFL.
Kevin, LCC tuning
Work on-going, trying to understand the large DA loss after adding BPM misalignment errors. Orbit at FD is suspected with excursions of up to 50 um, however this would be hard to beat as BPM misalignment is 10um. The global status should be given in the next meeting in 2 weeks.
Emittance from tracking and analytical estimate is compared. In tracking there is a large transient that required 10000 turns to converge, the tracking value reaches values well below analytical estimate in one seed. Further studies needed.
Kyriacos, Beam Dynamics observations
The emittance computed from tracking in corrected seeds from Satya disagrees with analytical estimate and it is much worse than nominal. Xsuite team is researching on more accurate emittance estimates. It is recalled that DA with beam-beam for a tuned lattice further deteriorated DA.
GHC and LCC nominal comparison is given for Z. DA is a bit better for LCC. For MA values are similar.
FMA studies are performed displaying key resonance affecting the beam. Simular diffusion between GHC and LCC.
Q. Bruant, AI algorithm for BPM faulty detection and noise reduction
DBSCAN is a clustering algorithm using a distance-based approach.
Anoumalies BPM detection works only for high amplitude kicks using normalized features. Raw features perform worse. Increasing number of turns does not change significantly the result.
Bad BPMs are identified with this technique in SuperKEKB HER data. There are about 6 BPMs known to be bad that are well identified by the unnormalized analysis. The results are less clear for the normalized analysis, hence this is not suitable. More statistics are needed (visit to KEK in 2 weeks).
Simulations were performed in SAD using codes from the CERN team. Xsuite SuperKEKB lattice is now available. Comparisons between SAD and Xsuite would be useful. Mael will be visiting KEK in one week for 2 weeks.
Charles, Dual-Decoder CNN Autoencoder for Unsupervised Anomaly Detection & Denoising in TbT-BPM Data
Similar work as in previous presentation using autoencoders.
Analysing HER data, it starts with SOMA identifying 3 bad BPMs out of the 6 known bad BPMs. The autoencoder finds new bad BPMs. Tune distribution in denoised data is inspected. Important differences in the tune amplitude is found between SOMA and denoised data. Signal-to-noise ratio improves by factor 3 without shifting tune.
Compared to SVD clean this denoising has the advantage of not needing any input (SVD clean needs number of modes to keep).
Maybe actual tune distributions before and after cleaning could be looked at. Future FCC simulations to be performed.
There are minutes attached to this event.
Show them.
