rf system improvements for performance and reliability dan van winkle kirk bertsche, john fox,...

RF System Improvements for Performance and Reliability

Dan Van WinkleKirk Bertsche, John Fox, Themis Mastorides,

Claudio Rivetta, Heinz Schwarz

Brief Outline

• Current performance and future plans

• Brief RF System Review

• What keeps us up at night– Longitudinal Growth Rates– Aborts

• General ongoing global issues

• Conclusions and Outlook

Current and Planed Performance

Current performance and future plans

• Run 6 Parameters (max achieved)

LER HER

Beam Current 3.026 (1.4X) 1.96 A (2X)

RF Voltage 4.05 16.5 MV

# Cavities 8 28

# Klystrons 4 11 (1.8X)

Voltage/Cav 506 590 kV

Klystron Pwr 755.1 845.9kW

Current performance and future plans

• Run 7 Parameters (max planned)

LER HER

Beam Current 4.00 (1.9X) 2.2 A (2.2X)

RF Voltage 5 17.5 MV

# Cavities 8 28

# Klystrons 4 11 (1.8X)

Voltage/Cav 625 625 kV

Klystron Pwr 1005 980 kW

Brief Overview of PEP-II RF System

The PEP-II LLRF

Gap Loop

SAT LOOP

HVPS

Klystron120W Driver

IQ MOD

IQ DEMOD

DIRECTLOOP

COMB LOOP

RFCAV

476 MHzREF

STATION REF (EPICS)

+

- -

BEAM

TUNER LOOP

Slow Loop (EPICS)

Fast Loop (Electronics)

476 MHz

Baseband

STATION REF (EPICS)

•15 Stations•4 LER•11 HER

•8 Cavities LER•28 Cavities HER

The PEP-II LLRF

• LLRF Station

1.2 MW Klystron

Temperature Controlled LLRF “Blue Box”

LLRF VXI Crate

Fast Interlock Chassis

120W Klystron Pre-Amplifier

Insomnia Producing Problems

What Keeps us up at Night

• Longitudinal Growth Rates– Grow Damp measurements in 2003 showed

5-10X greater growth rates than predicted by a linear model.

Longitudinal Growth Rates

• Previous MAC talks have addressed various ideas we’ve tried which include:– Non-Linear Modeling Effort (Claudio Rivetta

Talk)– Klystron Linearizer (MAC Oct ’06 Talk)– Klystron Pre-Amplifiers (MAC Oct ‘06 Talk –

Preliminary)

Solution Flow Diagram

High Growth Rates Observed

Linearizer Project Non-linear Model

Pre-amp EvaluationComb Rotation

New Pre-Amps

Reduced Growth Rates Observed

Longitudinal Growth Rates

• Today, I’ll Discuss:– Progress since last MAC on:

• Identifying key parameters and measurements for replacement amplifiers

• Progress on purchasing and replacement• Results of new installations

Pre-Amplifier Specification and Measurements

Klystron Pre-Amplifiers

Full Power Carrier

Swept Low Level CarrierSmall (network analyzer swept) signal injected along with large carrier to simulate small signal modulation on CW carrier

Klystron Pre-Amplifiers

Unusual (distorted) response seen when carrier is present

Since we use the amplifier with carrier and small signal modulation, the modulation “sees” this response rather than the flat (desirable) response.

Klystron Pre-Amplifiers

• Old Data showed LR4-2, HR12-2 and HR12-6 were especially bad in small signal response

Klystron Pre-Amplifiers• LR4-2 Distortion affected ability to implement comb rotation

LR42 was nearly unstable with 20 degrees of comb rotation. Simulations show this is due to non-linear pre-amplifier response.

Klystron Pre-AmplifiersNon-linear distortion is also a key parameter of these amplifiers. Rather than use two large tones as is typically done in a TOI measurement, we decided to try a new technique similar to our network analyzer technique. Namely, a small signal in the presence of a large signal.

Klystron Pre-Amplifiers

Klystron Pre-Amplifiers

Klystron Pre-Amplifiers

In this case, AmpC was a class A amplifier powered off 240V AC.

Amp B was a class AB amplifier powered of 120V.

We chose amp B based upon “good enough” performance and much less expensive price

New Pre-Amp Performance In Station

Klystron Pre-Amplifiers

Gap Loop

Klystron120W Driver

IQ MOD

IQ DEMOD

DIRECTLOOP

COMB LOOP

RFCAV

476 MHzREF

STATION REF (EPICS)

+

-

-

BEAM

+

IQ DEMOD

DAC

ADC

Built in stimulus driver allows for unique in-situ measurements

Klystron Pre-Amplifiers

• New Amps installed in all stations

Klystron Pre-Amplifiers

Klystron Pre-Ampliers

• Since Last October– Significant Time spent characterizing and

specifying amplifiers in a “new” way.– Found Vendor who met specs– Bought 17 and installed 15 new pre-amplifiers– Amplifiers show much improved response and

allow for 20 degrees of comb filter rotation

Longitudinal Growth Rates

Conclusions

• We now feel system is prepared to move to higher current realms in terms of longitudinal growth rates

• Further work may include:– Asymmetric Combs– RFP asymmetry calibrations

What else keeps us up?

• Aborts– Biggest Contributors are cavity arcs

during startup after down (Heinz S.)

– We also had many issues with the HVPS systems (things breaking)

– Things we plan to work on are:• PG&E Power Dips (55 Aborts during run 6)

• LR4-4 Drive Glitches (41 Aborts during run 6)

HVPS Dips

Pout vs. Pin

600

650

700

750

800

850

900

950

1000

1050

1100

0 10 20 30 40 50 60 70

Pin (W)

Po

(k

W)

77 kV

77.5 kV

76.5 kV

What’s Going On?

For 77 kV @ 900 kW Output, ~24 W InputFor 77.5 kV @ 900 kW Output, ~22 W InputFor 76.5 kV @ 900 kW Output, ~27 W Input

Swing Required of 5W to keep output constant at 900 kW

Pout vs. Pin

600

650

700

750

800

850

900

950

1000

1050

1100

0 10 20 30 40 50 60 70

Pin (W)

Po

(k

W)

75.5 kV

76 kV

75 kV

Constant Running

Power can not reach 900 kW and goes over the top

HVPS Dips

• What to do?– Lower drive power on Klystron

• Upside:– Allows for greater head room

– More linear running

• Downside– Can’t reach as high power

– Higher Collector Power

HVPS Dips

• Since these dips appear to be coming from PG&E, there is very little we can do locally to mitigate them.– Constant monitoring of drive set-points will be

required– For peak currents, we may need to live with

the occasional power dip abort

LR4-4 Drive Dips

Ongoing problem since 2005

•AIM HVPS monitor shows noise burst

•Followed by Klystron Forward “wiggle”

•Followed by very short dropout in drive signal

•Followed by beam abort…

LR4-4 Drive Drop

• Drive mysteriously drops (or rises)

• Cavity responds after delay

LR4-4 Drive Dips

• Things Tried so far:– Replaced several modules (not well controlled study)– Moved drive set point– Low trip rate and missing measurement points makes

this difficult to diagnose• Plans

– New klystron being installed. Will let run for some time to see if this makes any difference (not likely)

– Begin plan for swapping various modules. Likely culprits are gap module and RFP module. Will start with one then wait 2-3 weeks, then swap another.

– Extra Monitoring on HVPS signals to attempt to understand mysterious HVPS noise burst

– More plans to be developed in LLRF ongoing meetings. This will become high priority.

General Issues

General Ongoing Issues

• Cavity Tuning Polynomials– Polynomial fits for cavity de-tuning vary with

temperature– Occasionally cavity temperatures have been varied

without concern for these polynomials (RF experts not notified)

– Net result is constant tuning of RF stations as current is pushed.

– Machine seems to run better after stations are given “tune cavities” and “make polynomials” tune ups.

– This process take time without beam so administration is reluctant to do except when necessary

General Ongoing Issues

• Cavity Tuning Polynomials– We are working on scheme to make this

process run-able by operators. – Process must be “bullet proof”– Will remind operators to run at opportunistic

times when we are without beam for 30 minutes

Summary and Conclusions

Summary and Conclusions

• Much progress has been made over the last 3-5 years in improving the reliability and performance of the PEP-II RF systems. Some highlights are:

– Re-designed RFP modules– Re-designed IQ&A modules– Fixed “Stuck Tuner” Problem– Better temperature control on blue boxes– Designed new R2 COMB Module (reduced two VXI modules to one)– New pre-amplifiers– Comb rotation for better beam stability– Filters on RE signals– Non-linear modeling for low order model longitudinal damping improvement– Low Group Delay Woofer for increase low order mode damping– Gage Board diagnostics for transverse and longitudinal troubleshooting– Matlab GUIs for fault files– Matlab GUIs for “tune cavities” and “make polynomials” – RF training for operators– Model based RF station tuning– AIM module diagnostics of HVPS signals– SLAC Klystrons in most stations– Klystron Linearizer Development– Re-designed VXI Clock module– Re-designed VXI AIM Module– Found & Fixed old VXI COMB stuck overflow problem– Upgraded VXI CPU from 40MHz 68040 to 350MHz PPC (and re-designed VXI interface on all old VXI modules)– Re-wrote RFP DSP Ripple loop code to reduce phase ripple– Added RFP analog ripple loop– Found & fixed AC power wiring issue with VXI Crate power supplies, which was causing intermittent problems– Upgraded 476Mhz RF Phase Reference System to reduce phase drift and improve stability as well as added more diagnostics– Developed full-fledged RF station mock-up in lab for development & troubleshooting– Insulated Blue VXI Racks and added heaters with closed-loop control for better temperature regulation– Improved RFP Calibration routine– Improved IQA calibration method

Summary and Conclusions

• The PEP-II RF system is running relatively well (2.5 aborts per day) considering the complexity of the overall system.

• This rate is still too high, but there will be difficulties in reducing this rate due to the lack of “big ticket” aborts to go after.

• We will (of course) continue working on these issues to the last day of running.

Summary and Conclusions

• Finally– Running at the highest currents will require

constant vigilance• If we lose a station, we WILL NOT be able to

continue running at full current

Acknowledgements

Technical Review and Discussions:Dmitry Teytelman, Mike Browne, John Dusatko, Jim

Sebek, Ron Akre, Vojtech Pacak, Alan Hill, Kirk Bertsche

Original Concept and DesignPaul Corredoura, Rich Tighe and Flemming Pedersen

Support and PermissionUli Wienands, John Seeman, Mike Sullivan

All this work was done under contract #DE-AC02-76SF00515 from the U.S. Department of Energy

The PEP-II RF TEAMHigh Power RF

Alan Hill, Heinz Schwarz, Vojtech Pack, Al Owens, Ron Akre

Accelerator Research Department

John Fox, Claudio Rivetta,

Controls Department

Mike Browne, John Dusatko, David Brown, Bill Ross

High Voltage

Marc Larrus, Dick Cassel, Paul Bellomo, Serge Ratkovsky

Control Software

Mike Laznovsky

Accelerator Dept

Mike Sullivan, Uli Wienands, William Colocho, Franz-Josef Decker, Alan Fisher, Stan Ecklund, Mat Boyes, Kirk Bertsche

Management

John Seeman, Ray Larsen, Sami Tantawi

References

• Dan Van Winkle – MAC ’06 Talk– http://www.slac.stanford.edu/~dandvan/mac_1006_dvwR3.ppt

• Dan Van Winkle – MAC ’04 Talk– http://www.slac.stanford.edu/~dandvan/MAC_12_04.ppt

• Dan Van Winkle – Internal Linearizer Review– http://www.slac.stanford.edu/~dandvan/project_review_3_08_06.ppt

• Claudio Rivetta et al – PRST Longitudinal Simulation Paper– http://prst-ab.aps.org/pdf/PRSTAB/v10/i2/e022801

• Dan Van Winkle - EPAC ’06 Klystron Linearizer– http://www.slac.stanford.edu/pubs/slacpubs/11750/slac-pub-11945.pdf

• John Fox – RF Amplifier Selection PAC ’07– http://www.slac.stanford.edu/pubs/slacpubs/12500/slac-pub-12636.pdf

• Dan Van Winkle – LLRF Workshop 2007 Invited Talk– http://www/~dandvan/llrf07.ppt