a d v a n c e d l i g h t s o u r c e 1 april 18-20, 2007christoph steier, nsls-ii workshop...

A d v a n c e d L i g h t S o u r c e

1April 18-20, 2007 Christoph Steier, NSLS-II workshop

Stability Issues at the ALS

Stability Issues at the ALSChristoph Steier

ALS Accelerator Physics Group

• The Advanced Light Source• Introduction

• Slow/Fast orbit feedback•Energy Stability•Beamsize Stability•Top-off preparation•RF phase noise



Aerial view of the Advanced Light Source



1/10 Electron Beam Size

Sorce Location Horizontal Vertical Straight Section 30 m 2.3 (0.8) m Bend Magnet #2 10.3 m 1.3 (0.5) m

Nominal Energy

1.5-1.9 GeV

Circumference 196.8 m RF frequency 499.642 MHz Harmonic number

328

Beam current 400 mA multibunch (future 500 mA)

65 mA two-bunch Nat. emittance 6.3 nm

at 1.9 GeV Emittance Coupling

Typical about 2% (future 0.4%)

Nat. energy spread

0.097%

Refill period 3 daily fills multibunch 12 two-bunch

(future top-off about every 30 s)

ALS Parameters and Beamlines



• 12 nearly identical arcs – TBA; aluminum vacuum chamber• 96+52 beam position monitors in each plane (original+Bergoz)• 8 horizontal, 6 vertical corrector magnets per arc (94/70 total+chicanes)• Beam based alignment capability in all quadrupoles • 22 corrector magnets in each plane on thinner vacuum chamber pieces - FOFB

BPM, Corrector locations



What has been done at the ALS to maximize stability

FEED FORWARD

• Insertion device compensation (10 Hz for most IDs, 200 Hz for EPUs)

• Beta-beating, tune and coupling feed-forward (presents additional challenges to orbit stability!)

“PASSIVE”

(i.e. remove the sources)

• Temperature stability (air below 0.1, water below 0.3 degree peak-to-peak; 0.1 for RF)

• Minimized water induced vibrations

• Power supply stability (no switched mode supplies, thick aluminum vacuum chamber in most magnets)

• Vibration - reduce the effects by mechanical design (ALS has big girders and moderate amplification factors) and remove the source (cryo-coolers).

• Reduce RF-phase noise (mode-0 noise for IR users)

FEEDBACK

• Local orbit feedback is not routinely used at ALS

• Global orbit feedback (1 Hz update rate slow, 1.1 kHz fast)



Instrumentation at the ALS

I. Beam position monitors (BPMs)

Old in-house design (96) plus J. Hinkson/J. Bergoz multiplexed BPMs (currently 50); Bergoz BPMs used in feedback: noise level is about 0.3 – 0.5 microns at 200 Hz bandwidth and 200-400 mA; current dependence less than 1-2 micron for 200-400 mA

II. Photon beam position monitors (PBPMs)

Several very diverse designs; most are not integrated in accelerator control system; some beam-lines use them for local feedback (time-scales of feedback range from hours to ms); installed some

new PBPMs recently (plan to install on most bend magnets)

III. Power supplies

All power supplies at ALS SR are SCR or linear; no switched mode. Noise level is typically less than 10-4 integrated over all frequencies (some main supplies 10-5). 16-20 Bit control (all corrector magnets are 20 Bit); corrector bandwidth >500 Hz.



Instrumentation at the ALS II

IV. Control system

High level control system has throughput of about 100 Hz and delays of less than 10 ms after upgrade. Low level (fast feedback – distributed cPCI crates) runs at 1.1 kHz with standard computer and network equipment, network synchronized timing. Extermely reliable

V. Other

Tested some simple methods to measure BPM and magnet motion; plan to incorporate measurement of BPM position relative to common accelerator-experiment ground plate into feedback

Two synchrotron light monitors (emittance, energy spread); Streak camera (bunch length); LFB, TFB record + grow/damp; Fill pattern (high bandwidth BPM); CSR; fluctuations; nonlinear crystals; …



User requirements

Three examples of experiments that currently are the most sensitive: Micro focusing beamlines on bending magnets (e.g. Micro XAS, especially in combination with molecular environmental science samples, i.e. dirt); problem is that sample is very inhomogenous and small source motion causes the spectrum to change significantly. I0 normalization does not help! Dichroism experiments (i.e. on EPUs) measuring very small polarization asymmetries; orbit motion can cause small shifts of the photon energy out of the monochromator, resulting in fake asymmetries. STXM – bend magnet beamline has very high bandwidth beamline feedback – not very sensitive to orbit (but very sensitive to beamsize) variations; undulator beamline does not have fast feedback (heavier mirror) – needs very good fast orbit stability (and beamsize)

After upgrades to the slow orbit feedback (arc sector, chicanes) and the EPU dipole, quadrupole and skew quadrupole feed-forward, and implementation of fast orbit feedback, even our most sensitive experiments are currently happy with the orbit stability.



Achieved orbit stability at ALS

Frequency Magnitude Dominant Cause

1 day – 4 weeks

3 m Horizontal 5 m Vertical

1. BPM chamber motion 2. BPM electronics drift and

systematic errors 3. Limited number of

BPMs/correctors

1 h – 24 h

1 m Horizontal 2 m Vertical

1. BPM chamber motion 2. BPM electronics drift and

systematic errors 3. Limited number of

BPMs/correctors Minutes << 1 m 1. BPM noise and beam

vibration (aliasing) 2. Corrector resolution

(digitization)

.2 to 300 Hz

<2 m Horizontal <1 m Vertical

1. Ground vibrations 2. Cooling water vibrations 3. Power supply ripple 4. Feed forward errors

Beam Stability in straight sections w/ Orbit Feedback and w/ Insertion Device Feed-Forward

• Improved fast jitter with fast global feedback (2004) • Improved 60 Hz noise with conversion of fast analog FF to digital (ground loop)• Improved insertion device FF compensation with better chicane magnets (important in two bunch)• Improved slow orbit stability with continuous addition of stable BPMs (ongoing)



Feed-forward example: EPU COMPENSATION

Without compensation the EPU would distort the electron beam orbit by ±200 m vertically and ±100 m horizontally. Using corrector magnets on either side of the EPU, 2-dimensional feed forward correction tables are used to reduce the orbit distortion to the 2-3 m level. Update rate of feed-forward is 200 Hz.

Mechanically the EPU can move from left to right circular polarization mode in ~1.6 seconds

Apple-II type elliptically polarizing undulators are more complex than other IDs The jaws can move in two directions (vertically and longitudinally) The motion in the longitudinal direction is fast (up to 17 mm/s at ALS)

This makes orbit compensation more difficult



EPU FEED FORWARD ORBIT CORRECTION

Orbit Error without Feed Forward Correction 200 Hertz Feed Forward Correction



Orbit Correction

Fast (200 Hz) or slow (10 Hz) local feed forward for all insertion devices (2-d tables for EPUs)

Fast global orbit feedback (1111 Hz, up to 80 Hz closed loop bandwidth (3 dB))

Slow global orbit feedback (1 Hz) No frequency deadband between feedbacks Complete (more correctors) global orbit correction plus local orbit

correction at all IDs every 8h after refill. Photon beam position monitors at ALS are not used to correct

beam orbit – instead they feed back on beamline optics. Bandwidth from h to about 10 kHz (IR beamline)



Software used for slow orbit feedback

All ALS high level controls accelerator physics routines are implememted in Matlab

Orbit feedback is controlled using a GUI which allows to ramp for injection, do single orbit corrections, standardize the lattice, etc.

Matlab includes all Matrix manipulation tools necessary and has proven to be very reliable

Code is very flexible (algorithm development is simple and can if urgent need arises even be done during user operation)

Based on Matlab Middle Layer – now widely used at many (most) light sources



RF-Frequency Feedback

• Largest long term effect is rain season (plus outside temperature)• Short term the fill cycle has a strong effect (heating), but insertion device gap changes are equally important and in an FFT also tidal effects show up



Energy calibration (resonant depolarization)

• High precision measurement of beam energy is relatively simple at low energy light sources like ALS• Allows some conclusions about long term orbit/magnet/ground plate stability• Implemented rf-frequency feedback at ALS and verified it with energy measurements

1.28

1.26

1.24

1.22

1.20

1.18

Count

rate

(M

Hz)

1400120010008006004002000

Time (sec)

7.7

7.6

7.5

7.4

7.3

7.2

7.1

Lif

eti

me (

hours

)

a)

b)

25.6

25.4

25.2

25.0

24.8

1.0451.0441.0431.0421.041

1.9022 1.9020 1.9018 1.9016 1.9014 1.9012 1.9010 1.9008

Beam energy (GeV)

b) resonance at lower sideband

26.0

25.8

25.6

25.4

0.4820.4810.4800.4790.478

Excitation frequency (MHz)

1.90221.90201.90181.90161.90141.90121.90101.9008

a) resonance at upper sideband

Norm

alize

d c

ount

rate

(H

z/m

A2 )

-3

-2

-1

0

1

2

3

E/E

(x103 )

-4 -2 0 2 4

frf/f rf (x106)

=1.628±0.004 x10-3

-10

-5

0

5

10

E/E

(x105 )

86420

Time (hrs)



Fast Feedback Layout

•Design choices:

• Distances at ALS are relatively large -> distributed system

• Wanted to avoid expensive specialized hardware (like reflective memory, DSPs)

•Multiplexed (Bergoz) BPMs provide enough bandwidth and low enough noise

• D/A converter resolution for corrector magnets was upgraded from 16 to 20 Bit.

• Update rate of system is currently 1.11kHz.

• Motivation: Fast Orbit stability with passive measures already very good (2-4 microns rms). Improvement into <m range required active/fast feedback



Computer Hardware of ALS FOFB

Use network timing (network is 100 Mbit/s, full duplex, switched), normal PowerMAC/cPCI hardware used in control system upgrade



Feedback Implementation Details

• Combination of fast and slow global orbit feedbacks in both planes – no frequency deadband

• Fast Feedback currently 24 BPMs in each plane and 22 correctors in each plane. 1.11 kHz update rate, bandwidth DC-60 Hz. Only ½ of singular values used.

• Slow Feedback 52 BPMs in each plane, 26 horizontal correctors, 50 vertical correctors, RF frequency correction. 1 Hz update rate, about 60% single step gain, bandwidth DC-0.1 Hz. Typically all SVs used.

• Slow feedback communicates with fast feedback to avoid interference in frequency overlap range. Setpoints/golden orbit used by fast feedback is updated at rate of slow feedback.



Orbit feedback performance

• Fast feedback routinely used in user operation since spring’04 with very positive user response.• Extremely reliable. One beam dump and total of 4 (minute long) feedback outages in first 2 years.• With slow and fast orbit feedback the ALS achieves submicron stability in the vertical plane:

• Integrated rms motion 0.01 to 500 Hz in the vertical plane is below 0.5 micron (projected to the 2.25 m beta function, 18 micron vertical beamsize at center of straight)• Horizontally the integrated rms motion is now reduced to below 2 microns (at 13.5 m beta function and 300 micron horizontal beamsize).

• Long term stability (week) is of the order of 3 microns.



Beam spectra with feedback

• Beam motion with feedback in open (red) and closed loop (blue) at out of loop BPM.• Feedback is very effective for moderate frequencies. Right now closed loop bandwidth (3 dB) is about 80 Hz. • Correction at low frequencies below the individual BPM noise floor (only ½ of SVs used).• System is set up conservatively at the moment – no excitation at higher frequencies.



Simulink model of FOFB system

• Comparison of simulated (Simulink) and measured step response of feedback system in closed loop

•PID parameters were intentionally set to create some overshoot (demonstrating that time constants and performance of system are well understood).



Frequency Overlap – Master/Slave

ALS needs slow and fast feedback (do not have enough fast correctors) Avoided frequency dead band – fast system not DC blocked Synchronization by SOFB updating FOFB golden orbit



Fast feedback magnets can be noise source

Strong corrector magnets with high vacuum chamber cut off frequencies can be significant sources of orbit noise

Observed at several light sources Feedback of course will (partially)

correct this, but it is much better to avoid effect in the first place

In case of ALS, problem was not power supply noise, but ground loops which we introduced by analog summing junctions for fast local feedforward (2+2 correctors)

Switching to a digital feedforward (with same update rate) and eliminating ground loops reduced 60 Hz noise (w/o fast feedback) substantially



Summary of user input for top-off

I/I of 0.3% is small enough

Injection every 30 s is OK, but should not be much more frequent

No burst mode (several injections just after each other – 1 Hz)

Bunch cleaning for two bunch needs to be incorporated in Top-off

Most experiments do not see injection transients

• Some (especially microscopes with short integration times) do see them and will make use of provided gating signals

500 mA is good compromise (minimum upgrade to beamline optics)

20 – 30 pm vertical emittance is close to limit for best beamline optics (sagittal focusing)



Impact of injection transients

Incoming beam is only small fraction of total intensity• Its unavoidable oscillations are no problem

Injection elements also perturb stored beam1. Non-closure of fast bump2. Stray field of pulsed septum magnets3. Potential influence of booster

— Conducted experiments with users and measured transients using BPMs (fast and turn-by-turn)• Results:

— Most experiments insensitive to any distortion (protein crystallography, PEEM, most spectroscopy beamlines)

— Very few experiments (STXM, IR) see no-closure of bump and will require gating (multibunch feedbacks help)

— All experiments that see transients can use gating— Some examples on the following slides



Effect of the Bumps

RMS Beam sizes are300 by 23 (later 8) microns

Transverse feedback system reduces the duration of the transients



• With full sine current pulse show that slowly decaying eddy currents from first and second half sines mostly cancel. •Delayed stray field using ‘full sine’ excitation reduced by factor of 10!

Septum Stray Field Reduction



EPU effects

1. Variation of on axis field integrals with EPU phase (causing orbit distortions).

2. Variations of the (mostly vertical) beamsize (both with gap and with phase):• Due to focusing changes (systematic focusing

terms from the bulk of the undulator). • Due to coupling terms (skew quadrupole like or

solenoid like). 3. Higher order effects impacting the dynamic (or

momentum) aperture, for example due to the field roll-off, which is quite significant and systematic in circular polarization mode.



Beamsize Stability

Receive more inquiries about beamsize than orbit stability!

• Low beam energy, already pretty good orbit stability

Vertical beamsize variations due to EPU motion were big problem.

Is caused by skew quadrupole (both gap and row phase dependent)

Search for root cause still underway.

Installed correction coils for feedforward based



RF phase noise

Mode 0 motion nowadays is very small – 0.03 degrees rms Dominated by noise from master oscillator, rf distribution system, rf

frequency correction … not HVPS• Fast RF amplitude feedback reduces effect of HVPS to this

level Use improved master oscillator + filtering at several points in low

level RF frequency distribution system



Magnitude of mode zero motion

‘Bad’ case corresponded to an energy oscillation of 3*10-5 resulting in position oscillations of about 1.5 m and angle changes of 4 rad at source point

Problem periodically reappeared and needed to be fixed again• Harmonic cavities lowering mode zero frequency further – into

range of amplification due to LFB• RF frequency feedback introducing DAC noise



Summary

Users are very happy with current orbit stability at ALS and handle feedback based on photon beam monitors themselves

Fast orbit feedback brought significant improvement for frequencies between 0.1 and 80 Hz.

Preparing for top-off

• Studied and minimized transients with users• Users helped define scope of upgrade

For ALS, beamsize stability often is bigger issue than orbit

Are continuing to improve stability

• Short Term (this year): • More BPMs for feedbacks• Faster update rate for fast feedback (goal is 4 kHz)

• Medium Term:• Slowly installing more photon BPMs (so far not in orbit feedback)• Started to think about path beyond Bergoz BPMs• Tested some means of measuring physical BPM positions relative to

ground slab – plan to eventually include in feedback

a d v a n c e d l i g h t s o u r c e 1 april 18-20, 2007christoph steier, nsls-ii workshop...

Documents

n c e d

i g h t s o u r c e

orbit stability

als global orbit feedback

workshop stability issues

als sr

feedback range

workshop instrumentation