Transcript Document 7363062

Stabilization Projects at SLAC
Eric Doyle, Leif Eriksson, Josef Frisch, Linda
Hendrickson, Thomas Himel, Thomas Markiewicz
Richard Partridge
NLC Project, SLAC
1
Beam Stabilization
●
●
●
●
Goal: Stabilize beams to ~1nm at a Linear
Collider IP
Slow Beam Based Stabilization (luminosity)
Fast Beam Based Stabilization (IP deflection)
Magnet position Stabilization:
–
●
●
Interferometer, Inertial Sensor based.
Very fast Beam Based Stabilization: Feather /
Font
Nanometer BPMs
2
Ground Motion
3
Beam Based Stabilization
●
Beam based measurements are the only long term
measurement of beam positions
–
●
●
●
Mechanical objects are not stable to nanometers!
For Timescales > 10 minutes, Luminosity
Optimization feedback
120 Hz Feedback (for NLC) based on deflection
scans.
Note that 120Hz feedback has unity gain at
~10Hz.
4
Calculated Gain for 120Hz Beam
feedback
5
6
Magnet Position Stabilization
●
●
●
Interferometer based feedback
–
Measures magnet position relative to ground
–
Work ongoing at UBC (Tom Mattison).
Accelerometer based feedback
–
Measures magnet position relative to "fixed stars"
–
Work ongoing at SLAC (this talk).
Ground referenced (Interferometer) and inertial
feedback both work in simulation. Effectiveness
depends on ground motion spectrum.
7
Commercial Interferometer Technology
• Heterodyne system provides immunity to ambient
light, and high resolution phase measurement.
8
Interferometer
Measurement Limits
• Zygo company ZMI-4004
Measurement resolution
1/2048 Fringe
– 0.31 Nanometer single pass
• 4 axis / VME module
• Data rate 10MHz.
• Zygo #7712 Laser Head
– 0.5ppb Stability 1 Hour
– OK for 1nm to > 1Meter
9
Environmental Effects - Air
●
●
●
Air tpemerature and Pressure:
–
1ppm/°C
–
1ppm/2.8mm Hg pressure,
–
1ppm/90% Humidity
Compensation
–
0.1ppm to 1ppm from calculation
–
< 0.1ppm from refractometer compensation
Difficult to get 1nm over 1M in Air.
10
Other Environmental Effects
●
●
●
Even Vacuum not ideal - windows
–
Fused Silica has small temperature coeficient, but
index variation with temperature is large ~10ppm/°C
–
For 1 cm path in fused silica, need .01°C
May be difficult to provide vacuum paths for
interferometers.
Assuming 10cm between reflector and center of
magnet / BPM, need .001°C short term stability.
11
Interferometer Overall
●
●
●
Performance typically limited by environmental
issues.
Commercial heterodyne systems available from
Zygo, Agilent, probably other companies
Provide stabilization to the GROUND
–
Cannot do better than a perfectly rigid mechanical
support.
–
Need to decide how to evaluate performance
12
Inertial Stabilization Work at SLAC
1. Stabilize a simple block using low sensitivity
commercial seismometers (done)
2. Stabilize an “extended object” with mechanical
properties similar to a final focus magnet using
low sensitivity commercial seismometers.
3. Stabilize an “extended object” with high
sensitivity seismometers
4. Construct a high sensitivity non-magnetic
seismometer suitable for use in a detector.
13
Magnet Suspension
Hard Support
Soft Support
Small motion without
feedback
Couples high frequencies:
will excite internal modes
Requires high actuator
forces: 10 N
Large motion at support
resonance without
feedback
Attenuates high frequencies,
minimal excitation of
higher modes
Low actuator forces: .01 N
Used for this project
14
Actuator, Sensor
●
●
With soft supports, actuator strengths can be low
~.01 N (100Kg, 100nm, 5Hz Resonance)
Use “electrostatic Actuators”
–
–
–
●
Capacitive gap, ~100cm2, 1mm, gap, 1KV
Low stiffness, Fast response time
Force proportional to V2, not dependant on position
(if motion << 1mm).
Sensor: Low cost, low sensitivity geophones for
now
15
Data Acquisition System
●
DSP (Old TI TMS320C40), for closed loop feedback
–
–
●
May upgrade to modern DSP if needed (C6000 series)
So far not a performance limit
24 channel A-D, D-A.
–
16 bit, 250KHz hardware, Typically operated at a few KHz
●
Variable gain input amplifiers
●
Variable frequency input filters for anti-alias.
●
Hardware: MIX bus / VME / Ethernet / Sun
●
Software: DSP C, VxWorks, (EPICS), Solaris, Matlab
16
Feedback Algorithm
●
Characterize system
–
–
Drive all actuators, measure all sensors, all
frequencies
●
Find normal modes
●
Find sensor resonances
●
Find couplings
~96 parameter fit (works!)
●
Six independent feedback loops
●
State-space type feedback.
17
Single Block Stabilization System
Note: frequencies below 2 Hz
filtered out
18
19
Spectrum, Feedback On / Off
20
21
Integrated spectrum with simulated beam /
beam feedback
22
“Extended Object”
●
●
●
●
●
Designed for same resonant frequencies and
masses as a real magnet support.
Magnet support tube replaced by support beam
under magnet for convenience
Use “Soft” supports ~ 3-7 Hz.
Use 8 sensors, 6 for solid body modes, 2 for first
higher modes
Use 8 electrostatic actuators
23
Extended Object Drawings
24
Extended Object
Actuator
Sensor
Support Spring
25
Extended Object
26
Extended Object
27
Characterization of Extended Object
28
29
30
Extended Object Status
●
Sensors, actuators, DAQ operating
●
6 solid body, and 2 internal modes identified
●
●
Feedback software requires minor modifications
from single block system
–
6 to 8 sensors and actuators
–
“Code Rot” since single block tests
Attempt to close loop soon
31
Possible Technical Issues
●
Extended object is far from symmetric – expect wide
range of couplings to sensors, actuators and modes.
–
●
●
Very weak control over “roll” mode
Internal modes are high frequency (75, 120Hz), probably
not excited.
Sensor tilt sensitivity: Tilt indistinguishable from
transverse acceleration
–
Orthogonalization now frequency dependant.
–
May need to solve fully coupled problem (more computation)
32
Why Build Our Own Sensor?
●
●
●
●
Want ~3x10-9M/s2/sqrt(Hz) noise at F > 0.1Hz.
Compact sensors for machinery vibration
measurements (used for single block test) have
noise ~300X larger
Geo Science seismometers have good noise
< 10-9M/s2/sqrt(Hz), but are magnetically
sensitive and physically large
Could not find commercial sensors which met our
requirements
33
General Seismometer Design
• Thermal mechanical
noise sets ultimate limit
• Readout noise can be
low
• Thermal noise limited
acceleration given by
4kbT0
A
mQ0
34
Vertical Sensors Difficult
●
●
●
Need to measure 3x10-9M/s2/sqrt(Hz) on top of
Earth's gravity 9.8M/s2.
Spring "sag" under gravity is large for low
frequency suspension
Small changes in suspension spring length or
spring constant will appear as acceleration signals
Thermal changes typically limit low frequency
performance - typically operate in vacuum
– Material creep can be a serious issue
–
35
Suspension Design
●
●
Want low fundamental resonance frequency in a
compact geometry.
Simple mass on spring frequency goes as
f=(1/2p)sqrt(g/L): f = 1.5Hz (our design) L = 11cm
●
Pre-bent spring gives high second order mode f.
36
Feedback Seismometers
●
●
●
High suspension mechanical Q improves
sensitivity - but results in large amplitude motion
at resonance
Below resonance sensitivity decreases as 2 leads to dynamic range problems
Use feedback to keep suspended mass
motionless relative to sensor housing.
(Standard technique)
–
Can use feedback force as acceleration signal
–
Optionally use force and residual error as signal
37
Suspension flexure spring (pre-bent
to be flat under gravity load)
Electrostatic
feedback pusher
~50V, 500um
Cantilever
Signal null
at center
Mass
Slow adjust
flexure
Adjust motor
Sensor Housing
DAC
position
feedback
Signal
split
DAC
phase match
feedback
VCO
Cable delay
1 nanosecond
Signal combine
I/Q
ADC - get position
and phase mismatch
~20dBm information
38
Sensor Parameters
●
●
Suspended mass 40 grams
Resonant frequency 1.46Hz
–
●
●
●
Next mode ~96Hz, ANSYS simulation (not seen)
Mechanical Q ~50
Theoretical Thermal Noise 2.5x10-10
M/s2/sqrt(Hz)
– 10X better than needed
Theoretical electrical noise X2 smaller than
mechanical thermal noise
39
Electrodes on PCB
Spring
Cantilever
40
Mechanical Design Issues
●
●
BeCu spring (high tensile strength, non magnetic)
–
Pre-bent, operated at high stress to increase higher
mode frequencies
–
Extensive creep measurements done at SLAC
Thermal effects very large!!
–
~10-8Co corresponds to (0.1Hz) noise limit
–
Use multiple "thermal filters", Gold plating to reduce
temperature variations. Operate in < 1 um vacuum.
–
Expected to be ultimate low frequency noise limit
41
Spring
Cantilever
RF Out
RF IN
Electrodes,
Test Mass
42
43
Sensor Status
●
●
●
●
●
Construction of prototype sensor complete
RF system operational, but with kludged control
of out of phase signal.
Sensor mounted on 30 Ton Shielding block on
elastomer supports.
Two Streckheisen STS-2 Seismometers mounted
on block to provide reference signals.
Data very very preliminary!!!
44
Sensor Testing
●
●
Do not have a location sufficiently quiet to
measure sensor noise
Compare sensor with STS-2 seismometer
–
●
Look for correlation with STS-2
–
●
STS-2 noise much better than we need in this
frequency range
Compare with correlation between two STS-2s.
Data analysis very preliminary
45
46
47
48
49
50
Data Interpretation
●
●
●
●
All noise issues expected to be at low frequencies
Expect sensor noise to be flat in acceleration
frequency down to some frequency. Then expect
1/f noise to cut in (unknown frequency).
Expect STS-2 noise to be flat in acceleration
down to 0.01 Hz.
Compact Geophone (used for single block test),
expect noise to be 1/f in acceleration (velocity
sensor).
51
Sensor Noise Estimate From Correlations
●
STS-2 to STS-2 Correlation good to ~10-8M/s2 to
.025Hz.
–
●
●
Actual sensor limit probably 10x better, but indicates
measurement limits in this setup
Compact geo-sensor to STS-2 correlation good to
~7x10-7M/s2 at 0.25Hz
New sensor to STS-2 correlation good to
~4x10-8M/s2 to 0.05Hz.
52
Noise Estimates
●
●
●
●
●
Use correlation and assumed frequency spectrum.
STS-2, measured: <0.25nm at 1Hz, 25nm at 0.1Hz.
Probably measurement limit.
Compact Geosensor (used for block tests). 5nm at 1Hz.
5000nm at 0.1Hz (This is a velocity sensor, below
resonance, noise ~1/F3).
New Sensor: 1nm at 1Hz, 100nm at 0.1Hz.
With “NLC” style beam-beam feedback, demonstrated
sensor noise is OK down to < .01Hz.
53
Sensor Noise Limits
●
●
●
Sensor operating at low RF power. Results in x10
reduction of ideal sensitivity. (probably not the
limit now)
Some evidence of spring “creak” – small steps
during creep. Investigating
Sensor not magnetic immune – contains low
resistance current loop on cantilever. Being
replaced with insulating cantilever.
54
Sensor Upgrades
●
Non-conducting cantilever Aluminum Oxide.
●
Non-conducting mass Hafnium Oxide (dense).
●
●
RF splitting on PC board (probably ceramic), to
replace kludged connector.
Various detailed mechanical changes to reduce
size, improve manufacturability
55
56
Stabilization for ATF Nano-BPM
●
●
●
●
Inertial and / or interferometer stabilization
Beam rate 1-6 Hz (compare with 120Hz for
NLC), Need low frequency system.
Need good stability at least to <1Hz, probably to
<0.1Hz.
Need to understand how to use beam to evaluate
system performance
57
Inertial Stabilization Issues
●
●
●
Inertial sensor: Low beam rate (< 6 Hz, vs. 120Hz for
NLC) requires very low frequency sensor.
–
Sensor noise scales as 1/F2
–
Present performance of SLAC sensor not good enough.
–
May want to use 3 Streckheisen STS-2 sensors.
Can probably measure 1nm down to ~0.25 Hz.
–
1nm at 0.1 Hz very difficult
–
Only interesting if beam rate ~ few Hz.
At best, performance is somewhat marginal
58
Interferometer Stabilization Issues
●
Interferometers should be good to <1nm for timescales
of seconds
●
Not pushing state of the art!
●
Ground motion at single point >>1nm at 0.1Hz.
–
●
Need to measure 2 point relative ground motion.
–
●
At SLAC see ~300nm at > 0.1Hz
Use STS-2 or similar – best measurement
Quandary: Need inertial sensor to measure ground
motion to evaluate interferometer performance!
59
Beam Issues
●
●
Need to make 2 point comparison – compare line fit to
one (3 BPM) structure with next structure.
Magnetic fields – need ~micro-Gauss-M field variation
for nm motion.
–
–
–
Need to measure. Typically see mill-gauss at 50Hz in
laboratory.
Phase shifts relative to power line can be a problem!
Must turn off all magnets between BPMs.
●
May need to build magnetic field feedback system.
●
Lever arm: 3 BPMs projecting to more distant point.
60
Ignoring problems:
●
Place inertial sensors on LLNL support frame
–
●
●
●
Space for 3 Streckheisens, or 3 pairs of SLAC sensors.
Place 6 interferometer beam lines (in vacuum) to
ground).
Replace LLNL support frame supports with springs, and
electrostatic actuators.
Use SLAC DAQ system to close loop based on both
seismic sensors and interferometers
–
Adjust frequency roll-off between inertial and ground
61
62
Comments on ATF stabilization
●
●
●
System is complex, and requires complex
mechanical integration
Light paths through support table are required for
interferometers.
Need to integrate LLNL support / feedback
system with LLNL support / feedback system
63
Short Term Plan
●
Stabilize extended object with commercial low
noise (but magnetic sensitive) sensors.
–
●
●
Hope to meet NLC performance
Construct an updated non-magnetic seismometer
which meets NLC requirements.
Work on stabilization of ATF NanoBPM system
64