Cavity BPMs for Happex and G0
John Musson
Triplet Configuration…X, Y, and I
• TM010 Mode for I
• TM110 Mode for X & Y– Slugs provide proper
excitation, reducing TM010 x-talk
– Nominal output: 54nV-uA/um (-132 dBm)…per MAFIA simulations
Cavity Response
___________________Courtesy Jürgen Schreiber, ECFA/DESY LC workshop, Amsterdam, April 1-4, 2003
I & Q Demodulation
I Q DE- MOD
I @ 28 Msps
ADC
70 MHz LO
90 Degree I & Q
56 Msps ADC
0
23
1
+
1
COUNTER
REG
ADC70 MHz LO
56 Msps System Clock 14-Bit 2’s
Complement
0
23
1
+
1
REG
Q @ 28 Msps
=
+I +Q -I -Q +I +Q -I -Q +I +Q -I -Q +I +Q -I -Q +I +Q -I -Q +I +Q
Receiver
Functional Description
Receiver Parameters
• Noise Floor: -91 dBm (200 uV)– 3690 uA-um per shot (per cavity simulation)
• Bandwidth = 100 kHz• Processing Gain = 6MSPS/100 kHz = 18 dB • Best-case Resolution at 50 uA ~ 9 um (at full
BW)• Additional Integration Gain (16ms): 32 dB• DAC Output BW ~ 75 kHz (200 ksps)• Calculated resolution for Eff = 50 %: 0.7 um
– Hoping for 1 um
EPICS Interface
Happex Run
BCM Crosstalk
Helicity-correlated position differences, vs stripline,1nm
ResolutionBCM DD
Glitches
BCM Linearity
Helicity-correlated position difference. xtalk
Same plot, better data!
Happex Run
Bad News…..• Limited Dynamic Range
– Required external amps and filters• Crosstalk…~45 dB of C-C isolation
– BCM signal would corrupt X & Y• Software Problems
– Register overflow resulted in glitching and “Bedposts”• Synchronous Detection => Dedicated MO
– Phase noise and distribution issues (“LOL”)
Data COURTESY l. Kaufman, K. Paschke, R. Michaels
In Addition
• Setup is a learned behavior!– We devised a procedure, which proved to be
more difficult than expected with actual beam.
• Hall personnel eagerly participated…..– More eyes– Technical understanding of benefits and
limitations– Fantastic model for future systems
G0 Improvements
• Hardware– Crosstalk path identified. IF “traps” installed on Local Oscillator
lines => > 60 dB– Amplifier removed from BCM (I) channel– Additional bench testing to understand
• Software– Register rollover identified, corrected, and tested.
• MO– Try asynchronous operation, due to large Phase Noise in Halls
• Hall personnel also system-savvy!– Data courtesy R. Suleiman
Non-Linear Behavior
“Double Bounce” at Zero X-ing
Glitching
Known Improvements• 50 nA Sensitivity
– Currently have 74 mm resolution! Need additional 37 dB to achieve 1 mm. LNA?
– I-cavity is not a problem…plenty of signal• Shore up all hardware fixes (ie. LO-IF traps)• Firmware and Additional tests. • Setup Procedure (EPICS) and All-Save• Return to Synchronous MO
– Must improve MO Distribution to Hall(s)• Cavity Investigation on Downstream X&Y
– How can we duplicate beam+cavity behavior in the lab?
• Thank you to all for data, feedback, and especially patience!
CORDIC Algorithm
• COordinate Rotation DIgital Computer– Jack E. Volder, The CORDIC Trigonometric Computing Technique, IRE Transactions on
Electronic Computers, September 1959 – Ray Andraka, A Survey of CORDIC Algorithms for FPGA Based Computers, FPGA '98.
Proceedings of the 1998 ACM/SIGDA sixth international symposium on Field programmable gate arrays, Feb. 22-24, 1998, Monterey, CA. pp191-200.
• Iterative method for determining magnitude and phase angle– Avoids multiplication and division
• Nbits+1 clock cycles per sample• Can also be used for vectoring and linear
functions (eg. y = mx + b)
Concept
• Exploits the similarity between 45o, 22.5o, 11.125o, etc. and Arctan of 0.5, 0.25, 0.125, etc.
• Multiplies are reduced to shift-and-add operations
Angle Tan ( ) Nearest 2-N
Atan ( )
45 1.0 1 45
22.5 0.414 0.5 26.6
11.25 0.199 0.25 14.04
5.625 0.095 0.125 7.13
2.8125 0.049 0.0625 3.58
1.406125 0.0246 0.03125 1.79
0.703125 0.0123 0.01563 0.90
cossin
sincos,',' yxyx
iiiiii
iiiiii
dxyKy
dyxKx
2
2
1
1
Y
X
Binary search, linked to sgn(Y)
Successively add angles to produce unique angle vector
Resultant lies on X (real) axis
0,1
0,1
i
ii yif
yifd
)2arctan(1i
iii dzz
i
iid )2arctan(
Functionally.....
with a residual gain of 1.6