AP2 line BPM systemBill Ashmanskas, Sten Hansen, Terry Kiper, Dave Peterson, 2005-09-21

Background

AP2 line 275 m long, travels from target to DebuncherDp/p 4% (at end of line)Particles per pulse ~ 1010 at end of line, ~ 1011 earlier in line~ 108 antiprotons reach Debuncher every ~ 2 seconds 1.5 ms bunch train, bunched at 53.1 MHzbefore amplification, signals while stacking range from about -19 dBm (30 mV) upstream to about -35 dBm (4 mV) dnstreamso we amplify downstream signals 20dB in tunnelReverse proton signals range from -50 dBm (1 mV) upstream to -23 dBm (20 mV) downstream (before downstream amplifiers)maybe we should have amplified upstream, too?

Motivation

Simulations by V. Lebedev et al. indicate that ~ 50% more flux into Debuncher may be possible with well understood optics

Would like to use AP2 line BPMsto measure optics vs. Dp/p for reverse protonsto correct orbit using reverse protonsto monitor orbit drift and intensities while stacking

History

Legacy AP2 BPM readout did not provide useful data:signals much smaller than other beamlines 53 MHz signalscrosstalk from Debuncher injection kickerKicker crosstalk looked pretty severe on oscilloscopesee http://www-bd.fnal.gov/cgi-mach/machlog.pl?nb=pbar03&page=318

But scope data looked OK offline with narrowband processingAdded BPF before scope to reduce dynamic range needed(Re)connected existing 20 dB amplifiers to make interfering kicker signal less importantGot ~ 1 mm resolution with scopes good enough to be usefulDecided to build boards that do signal processing equivalent to what we did with scopes less clutter, easier readout, less costly

Blue = envelope read from new BPM boardRed = raw BPM signal (scope)Green = BPM signal after bandpass filter (scope)

We built something different from the echotek solution because

Modest demands only 53 MHz neededWe wanted to control the details, since we werent sure how much tinkering would be needed to make the system workDidnt have legacy system already working, to help specify upgrade

I spent enough time with CDF trigger to learn to hate VxWorksSubstantially reduced infrastructure cost appealed to usGives us a toolkit for other ad-hoc projects

It was much more fun to do it this way

Anyway, the resolution is 100 mm in the lab.We think, some rainy day, we ought to be able to push it to < 50 mm

65 MHz LPFBPMsignal53.1 MHz quadraturedemodulatorRFLOIQ5 MHz LPF5 MHz LPF10 bit dual ADC26.5 MHzFPGATI MSP430TCP/IPJava OACSimplified block diagram

Misc featuresNIM module, 4 type N inputs, 2 BPMsCan lock to external RF or 10 MHz or not10 bit sampling of IF I and Q waveforms44dB adjustable gain in demodulator chipDiagnostic DAC can drive (via analog switches) each input up to about full scale (at maximum input gain); also drives FP lemo32 MB SDRAM available (e.g. data capture), but not used nowDebug/test via USB console or telnetMSP430F149, programmed in C, provides command interfaceWIZnet 1 x 2 daughtercard provides TCP/IP stackNearly all processing done in Altera Cyclone 1C6Q240 FPGARemote FPGA + CPU update has been demonstrated (but not yet fully implemented in the field)Single network connection reads out ~ 10 boardsSimple I/O protocol for remote register access (see next page)

Java program by Jim Budlong (analogous to existing Debuncher BPM program)

N.B. chnl 721 is broken in tunnel

Resolution beam motion

Resolution beam motion

Response

Board has 44 dB variable attenuation in IQ demodulator chip

At full scale, max gain, signal 5 mV peak (-36 dBm)(84/2) x 384 x 2 32000 counts intensity reading(sample 26.5 MHz, +511 counts FS, A+B chnls)

A+B rms/mean 0.09% (using boards own DAC as source)AB / A+B rms 0.09% 90 (70) mm for 100 (75) mm BPM

Using min gain, FS, signal 0.5 V peak (+7 dBm)A+B rms/mean 0.11%, AB / A+B rms 0.11%

Response

Using AWG in instrumentation area, 24 dB atten, 83% FS,A+B rms/mean = 0.65%, A-B / A+B rms = 0.092%69 mm for 75 mm BPM (92 mm for 100 mm BPM)

Using another BPM boards DAC as source (clocks not synchronized), still at FS, A+B rms/mean 0.4%, AB / A+B rms 0.13%Would be interesting to understand why this does worse than AWG

These plots are 1 entry per channel per board

Relative gain settings (ask for -6dB, what do you get) vary 1.5% (rms/mean)

Absolute channel gains vary 10% (rms/mean)

CPU corrects each B signal for gain ratio between A and B

This sort of effect I vs. Q gains, pedestals, etc. is likely part of our excess resolution

We already subtract separate I,Q pedestals, but perhaps we could do better

To-do listRemote software update needs to be finishedBetter debug handles, e.g. waveform readout (partially exists)Make lots of debug info accessible without creating a dozen new ACNET devices per BPMStreamlined (automatic) handling of stacking vs. studies settingsSimple application to manage calibration constants, gain settings, timing offsets, etc.

More comprehensive set of bench measurements may be nice

Try mixing down with 52 MHz, not 53.1, followed by digital downconversion in FPGA? Better resolution??Implement fast fast time plots (as done for damper board)??Use SDRAM for circular bufferDecode TCLK, MIBS directly

Other applications?MI SBD trigger module (Nathan Eddy, Bob Flora): doneDRF2 AWG: in place (but needs ACNET hooks)Very similar boards do various Debuncher LLRF functions

D to A line (boards in place; need better tunnel electronics)Debuncher 53 MHz BPM orbits while stacking (likely)Already use one BPM as intensity monitor: switch SA for BPM board?

Downconvert 75 / 79 MHz diagnostic schottky signals?

Phase meter for MI injection??Could do bunch length, too, if modified for 106 or 159 MHz operation

Could potentially replace many turn-by-turn oscilloscope setups (pickup + RF + mixer + LPF + scope + console app)

http://pbardebuncher.fnal.gov/wja/docs/ap2bpm/