1. nanoBPM Progress
January 12, 2005
Steve Smith

2. nanoBPMs at ATF
Steve Smith
ATF2 Workshop
January 6, 2005
SLAC

3. Cavity BPMs C-Band Cavities Livermore Spaceframe Dual Downconversion:
Hexapods
flexure legs
Dual Downconversion:
First IF at 476 MHz
Second IF at 25 MHz
Digitize at 100 MSamples/sec

4. Algorithm Digital Downconversion:
Multiply digital waveform by complex “local oscillator” eiwt
Low-pass filter (currently 2.5 MHz B/W)
Sample complex amplitude of position cavity at “peak”
Divide by complex amplitude from reference cavity
Scale by calibration constants
Refine calibration with linear least-squares fit to other BPM measurements
Removes rotations, calibration errors.

5. Data: Raw & Demodulated

6. Calibration Move BPM more than beam jitters Estimate scale, phase
Doesn’t use information from other BPMs

7. Does calibration work in presence of beam jitter?
Look at same calibration data
Use other BPMs to remove beam jitter

8. Fits to Calibration Data, All BPMs

9. Measurement
Predict Y2
Linear least-squares fit to (x, y, x’, y’) at BPMs 1&3

10. Preliminary Resolution
s ~ 20 nm
Individual BPM resolution is better, this is measurement – prediction from 2 other BPMs
Calibration scale is clearly off by ~20%

11. Move BPM in 1 mm Steps

12. X Resolution

13. Anticipated Improvements
Analysis improvements
Adjust cavity parameters
Frequency
Decay constant
Optimize algorithm
Filter bandwidth may be reduced => improve statistical power
Optimize measurement sample time
Investigate handling of saturated pulses
Is saturation handled properly in this algorithm?
Are there non-linearities apparent at large amplitudes?
Potential Physical improvements
Lock Local Oscillators to accelerator RF
Improve understanding, operation of movers

14. Exercise BPM Movers Y2 is a little low Move Y2 up 1 mm Oops, wrong way
Go up 2 mm
What happens during the move?
40 mm excursions(!) in process of making 1 mm move
Why is resolution worse after the move?
s=40 nm, was 20nm
BPM rotated during move? No

15. Conclusions Resolution is not difficult
~20 nm, should get better with optimized analysis
Demonstrating resolution is hard
Must beat beam jitter, drifts
Stability looks good, but is poorly studied so far.
Implemented system should be able to prove BPM capabilities
Redundancy?
Movers?

16. End ATF2 Talk

17. Progress This Week
Parameters
Saturation
Stability

18. Anticipated Improvements
Analysis improvements
Adjust cavity parameters
Frequency Done, but makes no difference
Decay constant
Optimize algorithm
Filter bandwidth may be reduced => improve statistical power
Optimize measurement sample time
Investigate handling of saturated pulses No improvement, essentially no saturation in runs I’ve been evaluating.
Is saturation handled properly in this algorithm? Can’t tell yet
Are there non-linearities apparent at large amplitudes?
Potential Physical improvements
Lock Local Oscillators to accelerator RF
Improve understanding, operation of movers

19. Stability Calibrate using Dec-16 calibration runs
Examine all runs in \14-Dec-2004_19_55\
Refine calibration using a single run (Run9)
Regressed y2 against (y1, y3, x1, x3, y1’, y3’) i.e. left out x2, y2’, x’s
Plot mean & rms of (y2-y2est)
~ 2min /run -- for ~20 minutes stable to nm!
Sometimes things move!

20. Stability (cont) Now examine all runs in \14-Dec-2004_21_4\
Using same calibration-regression
Plot mean & rms of (y2-y2est)
Only run1 looks like previous runs!!!
Is all lost?

21. Stability (cont) Look at earlier data \14-Dec-2004_17_30\
Using same calibration-regression \14-Dec-2004_19_55\Run9
<100 nm motion in 1 hr!

22. To Do Investigate handling of saturation Understand Calibration
Calibrate tilts
Understand scales
Understand regression
What part is helping? How much?
What’s it hiding?
How stable are coefficients? What do they mean?
Should we establish DST files?
what format?

23. Minimal Regression Regress against Y1, X2, Y3 Y2 = 10mm +
Different choices of regression variables yield different direction of motion beyond data set regressed!

24. Drift < 50 nm over 1 hr !!

25. Stability Must be careful about regression variables Drifts look small
(at least for a data set selected for small drifts)