JLAB-TN-09-041

RF and CHL capacity tradeos; Q measurement

J. Benesch, M. Bickley, J. Creel, M. Drury, A. Freyberger,M. Wright, C. Reece, D. Turner

24 July 2009

AbstractCryogenic capacity is a concern for the PVDIS run in late 2009 and the Qweak run 2010-

2012. In both cases ∼20 g/s or more of liquid Helium (LHe) must be transferred from theCHL to the end station for target cooling. With the CHL running full out, ∼215 g/s wouldbe available to CEBAF and the FEL. On April 21 a test was run with the machine set up inQweak conditions. RF fault rate was ∼12/hour with cavities clamped to reduce CEBAF/FELow to 205 g/s. On June 15-17 there was a PVDIS test. 100 µA of 6068 MeV 5-pass beamwas delivered to Hall-A with a RF trip rate of ∼12 RF faults/hour, a CEBAF heat load of 216g/s and 17 gradient clamps on low Q cavities. This demonstrated that PVDIS may be run inlate 2009 however the estimated combined load of the NP program and the FEL program willexceed the CHL capacity. The April test showed that Qweak will be possible at acceptable faultrate at with a CEBAF+FEL load of 210 g/s on the CHL (excluding target cooling) after twomore refurbished cryomodules are installed in January/February 2010.

1 CEBAF Heat Load and the Cavity Q values

The cavity Q is the ratio of stored to lost energy of a cavity at a specied frequency. Cavities withlow Q produce more heat (loss) per MV of acceleration and therefore it is important to know theQ of each cavity in order to optimize the heat load of the ensemble of cavities in CEBAF.

The CEBAF ve cell cavities have a shunt impedance of 960 Ω/m. The RF power transferred tothe liquid helium in a ve cell (0.5m) cavity is E2×0.5/(960 × Q). For SL21, with seven cell cavities,E2*0.7/(968 × Q). Knowing Q at operating gradient E is critical for balancing heat load and faultrate. (E is known at ∼5% accuracy due to slow thermal oscillations in RF control modules whichaect the beam-based gradient calibration. See Section 1.1.1 )

1.1 RF Power to liquid helium level rate of change

A method to measure the cavity Q which put a known amount of electric heat into a cryomoduleto calibrate rate of change of LHe in the cryomodule with a known insertion of heat. Then therate of change in the LHe level is measured due to the presence of a known cavity gradient. Thecalibration measurement is used to convert the LHe rate of change due to the cavity gradient intoWatts, and combined with the known gradient this provides a measure of the cavity Q. Two mainapplications, acquisition and analysis, were written to support this activity by automating as muchof the process as possible. Two of the authors [JB and CR] looked at most of the Q data. Thegures are taken from JB's examination of the data.

1

Calibration of the rate of change in the liquid helium level for a zone (cryomodule) is the rststep in the process. The distribution of the measured calibration constant for each zone is shown inFigure 1. Units are Watts/(liquid level rate of change). The mean of 47500 is used for Q calculationin the swing heater zones as those cannot easily be isolated for zone calibration.

Figure 1: Distribution of zone calibration factors, JT change with heat input.

Using the zone calibration for the 31 zones measured and the mean of these 31 measurements forthe six swing heat zones, data was acquired and analyzed successfully for all but a dozen linac cavitiesand the eight cavities in zone SL21. The results are shown in Figure 2. There is a non-physicalhigh-Q tail. For use in machine setup, all Qs above 8.25E9 were set to that value, 2.7 standarddeviations above the mean of the truncated data set (Figure 3). The original commissioning Qspeaked at 6.8E9 for ve cell cavities. Faster transition through the Q disease range 150 to 50 Kin subsequent cooldowns may account for the increase in the high end. Better magnetic shieldingwithin some of the refurbished cryomodules may also help.

SL21 has small helium vessels encompassing six cells per seven cell cavity and inadequate con-ductive cooling of the ends groups. Cavity gradients are limited by end group heating. Small JTvalve position change when RF is turned on or o suggests the value used for all the cavities, 7.2E9,works well.

1.1.1 RF Gradient Calibration

The Q measurement relies on knowledge of the cavity gradient. The cavity gradient is determined bya calibration process that uses a reference cavity. The reference cavity gradient has been determinedby a cross calibration against the Arc energy measurement.

Figure 4 illustrates the accuracy of the RF gradient calibration process: Seven measurementsof one cavity with one reference cavity and a spread of 17%. The lem fudge factor is 0.7% for theNL and 1.6% for the SL, so the cavity GMES on average is not too far o reality. For individual

2

Figure 2: Measured Qs using zone calibrations from gure 1. Values above 1E10 are non-physical.

cavities JB believes the accuracy is about ±5%; a span equal to the dierence between the upperand lower 95% means in Figure 4.

Figure 5 shows that the error in Q is comparable to or larger than the gradient error illustratedin Figure 4. For each of the cavities with more than one valid Q measurement we plot the standarddeviation of the Q measurements divided by their mean, even if there are only two measurements.The mean of this distribution is 0.2, making the error twice that in E2 suggested above, there isalso a very long tail. The inaccuracy in Q is the largest contribution to the error in the powercalculation for more than 75% of the cavities.

The thermal oscillation which causes the gradient calibration uncertainty has a period of abouttwenty minutes. It may be possible to reduce the eect on the measurement by changing themeasurement, making sure that one is at the same phase of the oscillation, but that would multiplyby ten the time required to perhaps two weeks of beam time - unacceptable. The method must bereworked, changing to a parasitic one. Perhaps the energy lock cavities could be calibrated usingthe long method and then used as the standard for other cavity measurements.

1.2 Q measurement results

Accuracy of the Q values may be checked, at least integrated over cryostats, by shutting the RF oand watching the JT valve position change. The Qs and gradients are used by the linac energy model(lem) to calculate total RF heat for each module, plus a control allowance. The Autoheat programapplies resistive heat to control at the target value. If the Q values integrated over the cryomoduleare wrong, the electric heat is not equivalent to the RF heat and the JT valve moves when RF isturned on or o. The Q measurements accomplished since September 2008 have reduced the JTvalve changes. Table 1 captures four RF on/o transitions. The number of JT valves that move more

3

Figure 3: Measured Qs with top 22 removed and tted normal distribution. Near-normal.

than 3% has decreased substantially since Oct08 as well as the average change in JT valve position1.

Table 1: Results from several RF on/o test which measured the number of JT valves that movedmore than 3% and the average amount of JT valve motion across the RF on/o transition.

Date N(JT Moved) <|JT Change|>(%)

Oct08 24 4.7Apr09 22 4.3May09 14 3.0Jun09 12 2.5

Accurate Qs allow one to manually limit gradient in extreme cavities, trading o heat loadfor fault rate in a less blunt manner than simply restricting all cavities to, say, 9 MV/m. Theoptimization algorithm chosen in the original lem included both cryo load and RF faults in theobjective. Unfortunately it found unstable solutions. When the code was rewritten optimizationwas restricted to RF faults. Alternate optimization schemes are possible and should be developednow that the Q values appear robust.

1 JT Valve Elog: http://opweb.acc.jlab.org/CSUEApps/elog02/elog_item.php?elog_id=1474772

4

Figure 4: Seven beam-based calibrations of cavity NL03-2 spanning 0.98-1.15. All were taken withNL07-1 as the reference cavity. October 2005-July 2006. Three measurements on the same day were1.034, 1.039, 1.146. The data for the last looks as good as that of the other two.

2 Beam Tests

During the Spring09 Physics run two beam studies were performed. The goal of these studieswas to stress the RF systems and heat load on the CHL in order to understand the interactionsbetween systems (FEL, CEBAF, CHL, ESR) and to provide benchmarks for the future experimentalprogram. In addition the CEBAF and FEL heat loads and CHL capacity were monitored duringthe 5.97 GeV operation of CEBAF providing additional information on RF and CHL performance.The results are summarized in Table 2 and the details of the tests are in the following sections.

2.1 April 21, 2009 test: Qweak conditions

Linacs were lemmed to 550 MeV, the value for Qweak. The FEL heat load was imposed. Thetransfer lines to the end station were cooled and a small amount of ow established. The linacswere relemmed thrice with progressively lower limits on maximum cavity gradient: 11.5, 10, 9. Thelast reduced the heat load from the linacs suciently to allow 25 g/s to be sent to the end sta-tion refrigerator within an overall total of 210 g/s. Unfortunately, the RF fault rate was 12/hour,much higher than desired by Qweak. Fortunately, two more refurbished cyromodules will be in-stalled before Qweak runs. These should allow the RF fault rate to be reduced below 8/hour withCEBAF+FEL ow ∼210 g/s and target ow as needed 2.

2.2 June 15-17 test: PVDIS conditions

This was a systems test in preparation for the PVDIS experiment in Hall A scheduled for November-December 2009. The experiment requires 100 µA of 6068 MeV beam, an energy 50% above theoriginal CEBAF specication. The physics run beginning April 23, 2009, was 100 MeV lower and

2MatWright's elog can found here:http://opweb.acc.jlab.org/CSUEApps/elog02/elog_item.php?elog_id=1466618.

5

Figure 5: The distribution of the standard deviation/mean for cavities that have more than one Qmeasurement.

at ∼10 µA, the highest energy physics run in CEBAF history. A similar 6068 MeV test was runAugust 1-7, 2000. It is documented in TN00-028.

Steering was less dicult this time than in 2000 because the even-YR power supply was up-graded. Arc 3 bus was increased to get the second common dipole to the required eld without ashunt adder. About 2/3 of the capacity of the arc 3 horizontal correctors was used to null the eectof the bus increase. It took less than a shift to get CW to the BSY dump. No optics tuning wasneeded due to the small (1.7%) scaling from the previous physics run.

Increasing the current from 15 µA to 100 µA was dicult due to the lack of fast RF diagnostics.There was a fast, likely millisecond scale, RF instability which was detected only by the beam lossaccounting system. The linacs were lemmed at 600, 750 and 400 µA to redistribute gradient in thehope of lowering the stress on the unknown weak klystron. The last setting was the best. It waseconomized, mod anodes set to reduce forward power margin, in another attempt to nd the weakelement3. lists cavities which were turned down in the course of increasing 6 GeV current to 100µA ve pass. Figure 6 plots the Hall-A beam current for the 2.5 hours when reasonable stable highcurrent 6.07 GeV beam delivery to Hall-A was achieved.

The small changes were not saved in the cavity history les. They are available in the archiver.We can't be sure that it was one of these cavities without using a fast data acquisition system tomonitor a similar test. Since there is no opportunity for such a test before the 6068 MeV physicsrun, we'll have to roll a system around looking for problems while delivering for physics, as we didduring G0, or continue to rely on indirect indicators. Identiable RF fault rate during the entireperiod at 6068 MeV was ∼11/hour. Fault rate with high current was >20/hour due to the beamloss (fast RF) trips. This is not acceptable for physics, but it is not uncommon for initial setups to

3 Economizer eorts elog: http://opweb.acc.jlab.org/CSUEApps/elog02/elog_item.php?elog_id=1474667

6

Table 2: Recent (2009) measurements of CHL loads during Physics operations and special beamstudies tests.

QWeak Test May@5.97 June@5.97 PVDIS TestDate of test 2009-Apr-21 2009-May-26 2009-Jun-13 2009-Jun-16Energy (GeV) 5.60 5.97 5.97 6.07Linac Current (µA) 0 50 50 500Cap. Heater Margin(W/module)

7 9 9 8

CEBAF Flow (g/s) 184 201.7 199 200FEL Flow (g/s) 26 6.3 17 16CHL-ESR (g/s) 25 0 0 0Total Flow (g/s) 235 208 216 216Excess CHL Capacity(g/s) 0 27 19 19Total (Excess+Flow) 235 235 235 235RF trip rate (trips/hour) 12 7 8 10

suer high trip rates until the setup has been optimized and stabilized 4, has current plot at 100µA. The operator remarked on the over 3 minutes before tripping. Returning to Q and heat load,seventeen cavities limited for heat load during the test are listed in the elog entry 5. The cavityMV/W, MV/W vs. gradient and power to LHe distributions at the time of this test is shown inFigures 7, 8 and 9 respectively. Three are modestly limited because of a design aw in the autoheatsoftware. The others are limited because of low Q. Of the latter, three in SL08 will be remeasuredwith the commissioning cart during the summer down. JT valve motion in this zone suggests thethree cavities have high Q at 5 MV/m and low Q at higher gradients due to eld emission. If thesegradient restrictions do not provide sucient heat reduction for the PVDIS run, others may be cutusing MV/W as a gure of merit. Fault rate will increase.

Figure 6: 2.5 hours at 80-100 µA to hall A, 400-500 µA total in South Linac. Note that the beamhas many fewer trips at 400 µA total than at 500 µA, in line with an RF instability.

racr, a routine for extracting statistics about RF variables from the archiver 6, was used to lookat two periods. For 6/17 0050-0145 zero beam current data was extracted. For 6/17 0210-0340,

46 GeV trip rate elog: http://opweb.acc.jlab.org/CSUEApps/elog02/elog_item.php?elog_id=14746685Cavity Gradient Capping Elog:http://opweb.acc.jlab.org/CSUEApps/elog02/elog_item.php?elog_id=1475103.6Documentation:http://devweb.acc.jlab.org/controls_web/certied/racrs/racrs.pdf.

7

Figure 7: Distribution of MV/W at gradients from the last 6068 MeV allsave June 17, 2009. 17values below 0.5. Calculated using Qs graphed in Figure 3 plus the 22 assumed 8.25E9.

linac current 460-511 µA data was extracted (mean 494, 9 µA). Two cavities with large variation inGASK were found, R1C8 and R2L5. Their drive systems should be checked by engineering duringthe summer down. Eighteen cavities with high PASK variation were found: R157, R185, R188,R195, R1C6, R247, R2C1, R2D5, R2G8, R2H3, R2J2, R2J3, R2J4, R2L2, R2L3, R2L4, R2L7,R2L8. If resources are available, these should also be checked.

2.3 Cryogenic loads

During the June PVDIS test the CHL was running at 216 g/s at 2.07K (CFI14323N). The FEL RFsystem heat load was 300W, equivalent to 16 g/s at 2K (R4XXHTPA). Zero g/s at 4K 3 atmosphereswas being transferred to the end stations (CFI11139). CEBAF was therefore using 200 g/s at 6.07GeV and 607 kW beam power. This compares favorably with the 188 g/s used in late 2004 at 5.5GeV. Between the April Qweak and June PVDIS tests gradient reductions guided by measured Qsallowed CHL ow reduction of 10 g/s and ended with higher heater margin than in early April,meaning that additional ow reduction was possible. The ows during the two day June test areshown in Figures 10 and 11.

The FEL takes as much as 26 g/s. This would increase the CHL 2K ow to 226 G/s, leavingonly 9 g/s for the end station. This is insucient for PVDIS. For Qweak, CEBAF load should dropby more 300W and 15 g/s due to the lower energy and additional cryomodules to be installed earlyin 2010.

Table 2 captures the loads during the PVDIS and QWeak tests as well as for two periods duringthe Spring09 5.97 GeV operations.

8

Figure 8: MV/W vs 6 GeV gradient. A quadratic t is no better than the linear, surprisingly. Same17 points highlighted as in Figure 7.

3 Estimated required CHL load for the remaining 6 GeV NP pro-gram

Using the CEBAF load data from PVDIS-test and the nal two weeks of 5.97 GeV operationprojections for the loads during the remaining 6 GeV program are estimated. These estimates arecaptured in Table 3. From this data the LEM tool is used to estimate the CEBAF Linac heat loadfor the future congurations. The ratio of the estimated heat load for the future congurations tothe heat load during the June 2009 running is used to scale the CEBAF 2K load to provide anestimate of the required 2K ow from the CHL.

Table 3: The estimated CHL load for the future NP program. The FEL load of 25 g/s is notcaptured in this table.

Happex-III PVDIS Prex QWeakEnergy (GeV) 5.76 6.07 5.06 5.56Linac Current ( µA) 400 500 150 150Cap. Heater Margin (W/module)CEBAF Flow (g/s) 190 200 170 185CHL-ESR (g/s) 0 18 0 20Total Flow (g/s) 190 218 170 205Excess CHL Capacity(g/s) 45 17 65 30CHL Capacity(Excess+Flow) 235 235 235 235

9

Figure 9: Power to LHe, Watts, for 297 cavities. Gradients from 6 GeV test 6/17/09. Qs as discussedabove. The distribution is not LogNormal but the curve guides the eye nicely.

Figure 10: CAPHTRMGN is control margin for CEBAF linac heaters; multiply by 36 for totalwatts. CFI11139 is transfer from CHL to ESR, zero at this time. CFI5601 is 4K subcooled liquidto FEL. CFI14323N is total 2K ow to cold box. IBC1H04CRCUR2 is CW current to hall A. TheFEL and ESR ow meters do not give quantitative data because the helium is supercritical.

The FEL load plus additional experimental target loads need to be less than the expected excessCHL capacity in order for the Physics program to be executed as scheduled. The addition of C50-9before PVDIS and C50-10 before PRex will help optimize the machine conguration (CEBAF loadand RF trip rate) but the overall benet has large error bars so it is not included in these estimates.A ballpark estimate for the addition of the C50-9 module may reduce the heat load by up to 100 W(or 5 g/s) and the trip rate by 2 trips/hr. If this is true than the FEL limitation of 17 g/s duringPVDIS may be relaxed, but even if 5 g/s is achieved the FEL operations will need to be cappedduring the PVDIS run.

3.1 PVDIS and the FEL

It is clear that the PVDIS run in Nov/Dec09 represents the largest load on the CHL by the NPprogram. The excess CHL load during this period will only support FEL operations that produceless than 325W of RF Heat.

3.1.1 SBR Option

There is another option that will allow the FEL to run at full heat load during the PVDIS. Thisoption involves running the Standby Refrigerator (SBR) to feed the CHL-ESR transfer line. This

10

Figure 11: Same as Figure 10 with large amplitude signals hidden. Again, the two CFI signals onthis graph do not provide quantitative data due to the supercritical helium. The zero to the ESR,CFI11139, is accurate.

option has a cost of about $100k/month, requires a month's notice to prepare and places additionalstang load on the Cryo group. In addition it involves using a spare device as a productionmachine which leaves the CHL exposed to failure modes for which there are no spares. Recoverytime would be on order of months during which the CHL would be operating without a spare.Operations considers this a very high risk scenario.

3.2 QWeak, FEL and Hall-A cryo-targets

The 2004 Qweak agreement (without FEL signature) capped the FEL at 20 g/s or 380W. Butthe FEL typically runs at 25 g/s (475 W). The estimated excess CHL capacity during the Qweakprogram is 30 g/s, enough for 5 g/s supplemental ow for Hall-A targets and 25 g/s for the FEL.The expected fault rate for QWeak is expected to be less than the measured 12 trips/hour duringthe QWeak cryo load test due to the addition of two C50 cryomodules (C50-9 and C50-10), betteroptimized heat load and gradient distribution. The addition of the two C50 modules is estimatedto reduce the trip rate to less than 8 trips/hour. It should also be noted that these tests were ofshort duration and there was no time to optimize the RF conguration to minimize the trip rateduring the test.

The 185 g/s load in Table 3 is a new estimate of the CEBAF at QWeak energies. The fact thatit agrees with the 2004 estimate of 188 g/s suggests that the CEBAF heat load has been restoredto its historical levels from the abnormally high levels of Spring08 through March09.

3.3 FY11 and beyond

The C-100 prototype known as Renasence will be installed in the FEL in October 2010 after refur-bishment. Beam energy will be raised from 110 MeV now to 150 MeV for the UV program. This willincrease the FEL load an unkown amount. Q and x-ray measurements should be made on FEL02and FEL04. If heat load and low Q are driven by eld emission, backing o as little as 0.5 MV/mmay have much more than a quadratic eect on heat load.

4 Summary

Figure 12 graphically summarizes the presented measured and estimated cryogenic loads. Mea-surements from 2009 beam studies and operations agree well with QWeak load estimates made in

11

0

50

100

150

200

250

HapIII

est.

PVDIS

est.

PRexest.

Qweak

est.

Qweak

test

May09

5.97

Jun09

5.97

PVDIS

test

CHL2K

Loads(

g/s)

CEBAF loadCHL-ESR load

FEL loadExcess Cap.

CHL Max.Cap.

Figure 12: The estimated future and measured CHL loads for dierent congurations. The FELload is assumed to 25 g/s in all future estimates if such a load can be supported. No FEL load isshown during PVDIS.

2004, suggesting that the CEBAF load has been restored to historical levels. Q measurements havebeen completed on the majority of the installed SRF cavities in CEBAF. Q measurements of the24 cavities in the FEL would be benecial in understanding the FEL load.

With our present understanding of the cryogenic loads, simultaneous operation of the FEL,CEBAF and QWeak target cooling can be supported along with 5 g/s of CHL capacity for Hall-A target load available. During the PVDIS run in Fall09, the CHL capacity is exceeded by thecombined loads of CEBAF, FEL and Hall target loads.

A Appendix - microphonics (Jay Benesch)

One of the columns in lem.dat is related to microphonics. Larry Doolittle used PASKR/5 for thevalue. There's been a debate among cognoscenti on the best variable to use. C. Hovater suggestsPASK-peak, much larger than any of the others associated with the phase loop. That would be quiteconservative and limit many cavities due to available power. In 2001 I used an old data acquisitionprogram by Kurt Brown to acquire PASKR statistics during a high energy, high beam loading run. Iupdated lem with (PASKRmean+PASKRsigma)/5 and haven't touched that column in eight years.Since there's a static detuning allowance of 18 degrees and the old prescription worked well, I'velooked only at PASKRmean+PASKRsigma and PASKsigma, with and without beam current, thistime around. Figures 13-15 show the PASK distributions during these recent test and comparison tohistorical values. The racrs output mentioned at the end of section 2.2 was used for this evaluation.

Results:

• PASKsigma ∼ (PASKRmean+PASKRsigma), so one might as well use the former

12

• Values are larger without beam loading than with it, so use loaded values

• There's little correlation between the 2001 values and the ones from June 17, 2009

• The 2009 mean is twice the 2001 mean

• The higher microphonics allowance is predicted to increase fault rate at 6068 MeV and 500µA by less than 2%, so it's worthwhile to use the new numbers. Cryo load drops slightly.

• The lem.dat le lemdat30jun09.txt uses PASKsigma/5 with beam loading at 6 GeV for mi-crophonics. The values used are displayed below. If the cavity was turned o I used 0.35, alittle under the mean and over the median.

Figure 13: Distribution of PASKsigma during period 6/17/09 0210-0340 with 460-510 µA in thelinacs, i.e..beam loaded.

13

Figure 14: Comparison of new PASKsigma values with 2001 values used in lem for microphonics.Correlation is poor. If one excludes 10% of the points, correlation is still low, 0.11.

Figure 15: PASKsigma with beam loading (Y) versus without beam loading (X). A mediocre t dueto the extremes of PASKsigma with zero current (zeroI). Note that the horizontal axis is twice thevertical. Discarding the zero current points > 1 would improve the t, but there's no justicationfor this.

14