5 months of experience with 8 units of Libera Brilliance - generally good performance (stability & resolution) also high quality Sum signal (resolution & stability) --> charge measurements 1 unit (c14) broke down due to faulty board /connections resolution disturbed by (real) 22KHz vertical beam motion ? resolution disturbed by non-equal phased RF inputs (different cable lengths) resolution much disturbed by partial fill (i.e. 30%) and single (4 & 16) bunch fills But : offset-tuning offers a solution to all these disturbances (or not ?) also : the fine-tuning of certain (FA) filters can also help to suppress aberations Status now : 1) performance under 1, 4 & 16 bunch (and 30% fill) need close investigations 2) this with various values of offset-tune & cleverly designed filters 3) while analyzing the (spectrum of) FA (10KHz) data
Once optimum timing adjusted inject 1/3 fill in SR and kick it out after 1 single Turn : One single Turn should be seen on the SUM signal, without smearing Conclusion : the modified DDC filter works well
Situation with standard DDC filter ‘smearing’
22KHz component of ~15um pk-p in Vert. Plane, + strong DSC Libera-switching (at 27 TbT-intervals) (C8)
DSC OFF (0) DSC ON (2) 2 um
Tuned (MC=SR) PMO-Tuned (MC~SR) 1 um DSC ON (2) all the time
DSC ON DSC OFF DSC ON DSC OFF DSC ON 500nm DSC ON : ~ 12 nm rms ‘perfect-stable’ beam DSC OFF : ~110 nm rms
Tuned (MC=SR) Poor Man’s Offset-Tuning (MC~SR, only few (milli-)Hertz offset)
For a certain time we believed some aliasing effect between the 13 KHz DSC mechanism & the 22KHz real beam motion was responsible for the resolution degradation. Seemed plausible In fact, later it was found that simply non-equal phased input signals (because of different cable lengths) are enough for a Libera Brilliance resolution degradation. (when tuned) Seemed surprising . . . . this is the ‘Brilliance’ that does not even need offset-tuning according to official IT documents . . . .
tuned ~350nm DSC off off-tuned (20) DSC on CW RF signals from splitter (equal ampl. & phase) into the 4 Libera inputs
tuned DSC on DSC off off-tuned (20) ~2um CW RF signals from splitter into the 4 Libera inputs, but with 11cm extra cable for D
Degradation with partial fill and 1, 4, 16 bunch fills output after the 40MHz wide BPF that the ESRF puts just in front of the Libera RF inputs
tuned off-tuned (20) DSC on DSC off Libera fed with button & beam signals in 16 bunch fill, but going through Sum & Split. (i.e. stable beam) after 40MHz BPFs the max. ADC values are strong but well below saturation (i.e. <30K)
TUNED ~7 um DD-decimated, 10Ksa=1.8sec
OFF-TUNED (20) ~7 um DD-decimated, 10Ksa=1.8sec
Transition : from off-tuned (20) to tuned DD-decimated, 10Ksa=11sec
Spectrum from FA output 16 bunch, offset-tune=20 Beam. Hor. Stab.Hor. Stab.Vert. Beam. Vert. Spectrum from FA output 16 bunch, offset-tune=20
Beam. Hor. Stab.Hor. Stab.Vert. Beam. Vert. Spectrum from FA output 16 bunch, offset-tune=20
Beam. Hor. Off-tune=30 Beam.Hor. Off-tune=20 Beam.Vert. Off-tune=20 Beam. Vert. Off-tune=30 Spectrum from FA output 16 bunch, offset-tune=20 & 30
When tuned, what is causing the poor performance of the Liberas ?? 40MHz 5.7MHz SAW-Libera 15MHz ADC signals with 16 bunch fill i.e. side bands at 5.7MHz 5MHz Blue : with 40 MHz BPF (and the internal SAW 15MHz BPF) Red : with 5 MHz BPF
* ? 40MHz 5.7MHz ‘Leaking SAW’ ‘Leaking SAW’ ADC signals with 16 bunch fill i.e. side bands at 5.7MHz 5MHz Blue : with 40 MHz BPF (and the internal SAW 15MHz BPF) Red : with 5 MHz BPF
Blue : 16 Bunch with 40 MHz BPF & the internal SAW 15MHz BPF SAW-Libera 15MHz 7 X fsa – 2 X frf created by Libera itself Blue : 16 Bunch with 40 MHz BPF & the internal SAW 15MHz BPF Black : CW RF with 40 MHz BPF & the internal SAW 15MHz BPF
offset=20units offset=10units Another Libera artefact appears in SA data with offset-tuning : a slow 200nm pk-pk oscillation, the period is determined by the value of the offset data taken under 16 bunch, ‘perfect stable’ beam (Sum & Split)
The slow 200nm pk-pk oscillation in SA data can be minimized by playing with the switching delay parameter, data taken under 16 bunch, ‘perfect stable’ beam (Sum & Split)
offset=20units offset=10units offset=40units Another Libera artefact appears in DD t-b-t data with offset-tuning : ~ 12um pk-pk oscillation in vert. plane, the period is determined by the value of the offset data taken under 16 bunch, with ‘perfect stable’ beam (Sum & Split) offset=40units
offset=20units offset=10units offset=40units Another Libera artefact appears in DD t-b-t data with offset-tuning : ~ 25um pk-pk oscillation in vert. plane, the period is determined by the value of the offset data taken under 16 bunch, with real beam (Cell 8, BPM 3) i.e. including the ESRF’s 15um 22KHz vertical beam motion, (just about visible in 1rst graph) offset=40units
Numbers in the ESRF Libera [ in units of fsr ] ESRF harmonic : 992 352,202,600 Hz ADC sample freq. : 304 107,933,055 Hz T-b-T : 1 355,043 Hz DSC mechanism : 1 / 27 13,150 Hz T-b-T decimated : 1 / 64 5,548 Hz FA output : 1 / 36 9,862 Hz SA output : 1 / 36 x 1024 9.6 Hz Just note : 304 = 19 x 16 4 x 27 = 108 = 3 x 36 I-Tech, being confronted with our findings, is re-thinking the 304 and 36 design
What’s next : At the ESRF : - close investigations under 1, 4 & 16 bunch (and 30% fill) - this with various values of offset-tune & cleverly designed (FIR-FA) filters - while analyzing the (spectrum of) FA (10KHz) data At I-Tech : re-produce the ESRF findings (1, 4 & 16 bunch) with ultra-short pulse generator find the optimized offset-tune values and FIR filters to satisfy all ESRF operation study the impact of different numeric designs ( 304 305, 36 35 )