1. CLIC Drive Beam Stabilisation
Author: Alexander Gerbershagen Supervisors: Prof Philip Burrows (University of Oxford)
Dr Daniel Schulte (CERN)
What is CLIC Drive Beam?
The Compact Linear Collider (CLIC) is a proposed multi-TeV e+e- collider being developed at CERN, which is based on two-beam technology. One of the beams, so called Drive Beam, with low energy but high current, is used to deliver RF power for the accelerating structures of the high energy, low current beam, called Main Beam. This scheme has been chosen, since the CLIC main beam frequency (~12 GHz) is too high for any conventional RF source.
Fig. 1: Layout of the CLIC collider at 3 TeV
How does Drive Beam recombination scheme work?
The Drive Beam is produced in 240ns long trains with 180° phase shifted bunches at 0.5 GHz frequency. Acceleration power is provided by standard 1GHz RF source, so that all bunches are accelerated equally (Fig. 4). The frequency is then increased from 0.5 GHz to 12 GHz by recombining bunches in the delay loop
And the combiner rings. The trains overlap during
this recombination process and the errors
from different trains come together
in one train (Fig. 5).
Why do we need to stabilize it?
First, bunches pass the recombination scheme
of the Drive Beam. Then the RF power is extracted
from the Drive Beam in so called Power Extraction
and Transfer Structures (PETS) and the RF waves
are led from there to the accelerating structures
of the Main Beam.
The stability of the Drive Beam is a critical issue,
since its jitters and instabilities would lead
to unstable power supply to the main beam.
Some tolerances (e.g. phase, gradient
and energy error tolerance) are very strict.
The filling time of the RF structures is ~ 60ns,
hence we integrate the errors
over intervals of same order
of magnitude (10ns)
to receive following distribution
of error as function of frequency (Fig. 3).
Fig. 2: Two beams scheme of CLIC
Fig. 4: Layout of the CLIC RF complex
Fig. 3: Charge error as a function of frequency
Fig. 5: Train errors overlap
How can we stabilise it?
In order to avoid the main beam RF jitter one has to compensate
the errors of the beam before the PETS e.g. with help of a feedback system.
Dependent on jitter frequency, the feedback system can reduce or increase the error (Fig. 6). To optimise the total impact of feedback, we have to integrate error over frequencies and minimise the value as function of feedback latency and gain (Fig. 7).
We realize that if t(latency) + t(gain) ≈ 240ns (which corresponds to the train length) the feedback is optimal.
Conclusion:
Cool, we can even reduce the white noise errors with a feedback system!
This is in general not possible, but it becomes possible in this case due to recombination scheme.
What is a feedback system?
Ideal feedback system:
1. Measures error value
2. Compensates error on the next bunch
For CLIC the bunch spacing (0.08 ns) is much smaller then the latency time of the feedback
and gain time of the correction kicker (~ 10 ns). Hence, the correction will look
like on Fig. 9.
But if the jitter frequency
is anti-resonant
to feedback latency,
the feedback
system corrects
the measured value
too slowly, and only
increases the error
(Fig. 10). The standard
feedback system used in
our simulations, is called
proportional-integral-
derivative controller
(PID controller).
Fig. 10:
Resonant (top)
and anti-resonant
(bottom)
feedback
Fig. 6: Charge error as a function of frequency
(without and with feedback)
Fig. 9: Example of feedback correction
with long latency and gain time
and small bunch spacing
What will we do in the future?
Include more realistic effects in the simulation
Perform the simulations for possible future feedforward systems
Use the results as input for the main beam studies
Make predictions and test them at CLIC Test Facility (CTF3)
Simulate usage of Feedback On Nanosecond Timescale (FONT) system
Fig. 7: Current error integrated over frequencies
as a function of latency and gain length
Applying the feedback before
the recombination and integrating
over 10ns thereafter puts the error
at the blue bunch and its correction
on the red bunch in one interval, hence
reducing the total error (Fig. 8).
Special thanks for providing support and material:
Prof Philip Burrows
Dr Daniel Schulte
Dr Franz Tecker
Dr Guido Sterbini
Dr Oleksiy Kononenko
Fig. 8: Scheme of bunch recombination and
integration over 10ns intervals
For more information see: