Review of the e-p feedback
experiments
Rod McCrady
Los Alamos National Lab
IU e-Cloud Feedback Workshop March 13, 2007
LA-UR-07-1613
Overview
• Pickup, process v, feedback 4 turns later
– Q = 2.1875, 4×Q = 8.75
– Cables and LLRF require >3 turns
RF amp
Signal
Processing
Kicker
Pickup
IU e-Cloud Feedback Workshop March 13, 2007
LA-UR-07-1613
Beam
Low-Level RF System

Monitor

Filter
Input Level
Control

Fiber Optic
Delay
Variable
Delay
Variable
Attenuator
Variable
Attenuator

Gain
Control
• We have plenty of signal strength
• Fiber optic link compresses at -14dBm
IU e-Cloud Feedback Workshop March 13, 2007
LA-UR-07-1613

RF
switch
Setting the timing
Use kicker as “BPM”
Mark time of arrival of 1µpulse on 5th
traversal
LLRF

Observe time of arrival of pulse from PAs
(This will be from the 1st traversal)
Adjust delay so that damper pulse from 1st
traversal arrives when beam arrives on
5th traversal
Oscilloscope
Beam
Pickup
LLRF

Kicker
LLRF

Oscilloscope
Oscilloscope
Beam
Beam
Pickup
Kicker
Pickup
IU e-Cloud Feedback Workshop March 13, 2007
LA-UR-07-1613
Kicker
Complicating factors
• Short store time
– Complicates measurements and system diagnosis
• Long bunch
– A few complexities introduced by this
• v signal from BPM
– (dy/dt)×I(t)
• Broad band
• Rapid growth
IU e-Cloud Feedback Workshop March 13, 2007
LA-UR-07-1613
Factors Limiting Performance
• System gain
• System bandwidth
– Power amplifiers
– Kicker
• Signal fidelity
– Especially phase
•
•
•
•
Optimization of betatron phase advance
Beam in the gap
Longitudinal “noise”
Onset of horizontal instability
IU e-Cloud Feedback Workshop March 13, 2007
LA-UR-07-1613
Long bunch & Short store time
• Short store: difficult to use spectrum analyzer, etc.
– Very little frequency information on-line
• Frequencies change:
Injected
f 201.25 MHz
IU e-Cloud Feedback Workshop March 13, 2007
LA-UR-07-1613
Bunched
f 201.25 MHz
Long Bunch
Coherent
...After synchrotron
transverse oscillation
motion
IU e-Cloud Feedback Workshop March 13, 2007
LA-UR-07-1613
BPM v signal
• Need beam position quickly (<1s) with wide bandwidth
(10 to 300MHz)
• v(t) = Vtop(t) – Vbottom(t)
• v  intensity
• Looks like derivative of position in bandwidth of this
system
• 90 phase shift at all frequencies
– Cannot compensate with a delay
IU e-Cloud Feedback Workshop March 13, 2007
LA-UR-07-1613
BPM v signal
Signal at upstream end of stripline electrode:
VU (t )  C I b (t ) F  y (t )   I b (t   ) F  y (t   )  where  
L 1
1 
  
c  b  s 
Difference of top and bottom electrodes (v):
VU (t )  VU ,top (t )  VU ,bottom(t )  2aCIb (t ) y(t )  Ib (t   ) y(t   )
ybeam (t )  y0 sin t
For an oscillating beam:
VU (t )  2aCI 0 y0 sin( t )  sin( t   )
 
  
VU (t )  4aCI0 y0 sin 
 cos  (t  ) 
2 
 2  
Pickup Response
Note 90 phase shift at all frequencies.
Looks like derivative of position.
for f 300MHz
  and sin  cos
d
ybeam (t )  y0 cost
dt
IU e-Cloud Feedback Workshop March 13, 2007
LA-UR-07-1613
200
400
600
800
BPM v signal
• One could integrate the v signal
– We tried a passive integrator
Vin
R
• 1/ response was unpalatable
• Reduced signal level
– In retrospect, maybe not a big deal
Vout
C
• Other ideas
C
– Another differentiator: sin t   2 sin t
Vin
Vout
R
– Comb filter also gives 90 phase shift
• We haven’t seen any benefit from comb filters
– Different pickup type
• Buttons
• Slotted coupler
IU e-Cloud Feedback Workshop March 13, 2007
LA-UR-07-1613
Betatron Sidebands
• Why are they present in the v signal:
– Beam pulse traverses BPM at fR=2.8MHz (revolution frequency)
• Revolution harmonics n × fR
– Position changes turn-to-turn due to betatron motion
• f = Q × fR = (k+q) × fR
• A BPM only knows about q, the fractional tune
– fR is modulated by q × fR
• Betatron sidebands: (nq)×fR (upper and lower sidebands)
• Lower sidebands are associated with instabilities
Beam Position
q 2
IU e-Cloud Feedback Workshop March 13, 2007
LA-UR-07-1613
Experiments
• Explore limitations of the system
• Elucidate complicating factors
• Improve performance of the system !
•
•
•
•
•
•
•
•
Drive / damp
Noise-driven beam
Tests of system fidelity
Investigate effects of saturation in the LLRF system
Tests of comb filters
Effects of longitudinal noise
Compare Qthr with/without damping
Grow / damp
IU e-Cloud Feedback Workshop March 13, 2007
LA-UR-07-1613
Drive - Damp
• Signals are complicated by synchrotron motion of beam
• Hoped to compare passive vs. active damping rates
• Next time use coasting beam
IU e-Cloud Feedback Workshop March 13, 2007
LA-UR-07-1613
Noise-Driven Beam
• Does it “damp” as well as feedback does?
– One of my darkest fears
• Does it initiate instability?
• Does it interfere with coherence?
• Makes the beam more unstable.
IU e-Cloud Feedback Workshop March 13, 2007
LA-UR-07-1613
Effects of saturation
• Re-configured system
• Monitor input
• Operating in
compression is better
• What’s the benefit?
150mVp-p  no compression
• Attenuator for input level
• Attenuator for gain
WM41 top
– Damping early?
– Compression is OK?

Monitor
1
-8dB
WM41 bot


300MHz
LPF
2
-8dB
8.5dB gain
F.O. Tx
F.O.
Delay
F.O. Rx
17dB
IU e-Cloud Feedback Workshop March 13, 2007
LA-UR-07-1613

Variable
Attenuator
Input signal
level control
Variable
Attenuator
Gain
Control.

1
2
RF
switch
PM44 bot
PM44 top
Beam in the gap
• Compare conditions at low Vbuncher to intentional BIG
• Explore both axes of threshold curve
IU e-Cloud Feedback Workshop March 13, 2007
LA-UR-07-1613
Longitudinal noise
•
•
•
•
Problem: v signal has intensity information
PSR fR = 72.00×flinac  micropulse stacking
2006: changed to fR = 72.07×flinac
Longitudinal noise was reduced
– 402.5MHz is ~USB of mode 144 when using 72.07
72.00
72.07
• But no improvement in damper performance
IU e-Cloud Feedback Workshop March 13, 2007
LA-UR-07-1613
Less longitudinal noise, but…
• 402.5MHz is ~USB of mode 144 when using 72.07
=2×linac frequency
• Vertical oscillations at 402.5MHz
IU e-Cloud Feedback Workshop March 13, 2007
LA-UR-07-1613
Vary the vertical tune
• How perfect does the betatron phase advance need to
be?
• Can give some indication of what frequencies matter
• Found that several 1/100ths units on vertical tune made
little difference.
– 3.18 to 3.20
IU e-Cloud Feedback Workshop March 13, 2007
LA-UR-07-1613
Vary the Timing
• Increase & decrease LLRF system delay till damping is clearly
worse
• How perfect does the betatron phase advance need to be?
• Can give some indication of what frequencies matter
• ~90  ~2ns  100 to 150MHz
Damping
4ns
t
IU e-Cloud Feedback Workshop March 13, 2007
LA-UR-07-1613
Signal Fidelity – Phase Errors
• Phase errors in power amplifiers and cables
50
Power Amplifier (0dBm)
0
Phase (deg)
Phase (deg)
40
30
20
10
0
2 Cables
-5
-10
-15
-20
-10
0
50 100 150 200 250 300
f (MHz)
0
IU e-Cloud Feedback Workshop March 13, 2007
LA-UR-07-1613
50 100 150 200 250 300
f (MHz)
• To filter out revolution harmonics
– Wasted power
– Closed orbit offset
Response dB
Comb Filters
0
10
20
30
40
50
60
0
1
2
• Subtract signal from time-delayed signal (t=Rev)
FO
xmitter
Optic fiber
– Similar to stripline BPM
• 90 phase shift at all frequencies
• ? Might help mitigate dy/dt from v signal ?
– 180 phase shift from one passband to the next
0
20
40
180 °
60
LSB
40.5
41
41.5
42
42.5
43
4
Mode #
coax

IN
3
43.5
IU e-Cloud Feedback Workshop March 13, 2007
LA-UR-07-1613
5
6
7

FO
rcver
sin t  sin(...) cos(t )
OUT
Comb Filters
•
•
180 phase shift from one passband to the next
Damping in one passband means driving in the next
– Two ways to deal with it:
1) Twice as many passbands
Only LSBs matter anyway
2) Two comb filters in series
Lose 90 phase shift
•
0
20
40
360 °
60
LSB
40.5
41
41.5
42
42.5
43
43.5
1
2
6
Time domain picture
– Which “turns” to feed back
– One positive, one negative
2 q
0
3
5
4
IU e-Cloud Feedback Workshop March 13, 2007
LA-UR-07-1613
Results of Comb Filters
• Revolution harmonics reduced
– Signals to kicker:
• Ultimately, no better damping achieved
IU e-Cloud Feedback Workshop March 13, 2007
LA-UR-07-1613
Instability in the Horizontal Plane
• If we control the vertical motion, will the intability show
up in the horizontal?
– Some predictions of instability tune
– In PSR: Qh / Qv = 3.2 / 2.2
IU e-Cloud Feedback Workshop March 13, 2007
LA-UR-07-1613
Experiments: To Do
• Understand mechanisms for frequency spread
– Coasting beam
• Why does system perform better in compression
– Damp early, then turn off damper
– Turn on damper late, without early damping
• Can we get a better input signal? (other than v)
• What frequencies really matter?
IU e-Cloud Feedback Workshop March 13, 2007
LA-UR-07-1613