CEBAF cryo requirements
J. Benesch
for Accelerator Operations
2K flow during 6 GeV run
~225 g/s with FEL dynamic load (R4XX) on, 215 g/s with FEL dynamic load off
Using R. Walker’s 665 W static+dynamic and 19 W/(g/s) for FEL, CEBAF ~190 g/s.
Preconditions for 6 GeV run
• in situ Q0 measurement with JT valves to improve
accuracy of heat load calculations
• accept higher trip rate; reduce gradient of or turn off low Q
cavities with high gradient capability
• accept higher trip rate; reduce gradients of C-50 cavities
which would otherwise require more than 168 W/zone
(MOPS limit)
Changes since 6 GeV run
• Ten cryomodules cycled to 300K. Resulting Q0s are still
being measured, but many are lower and none improved.
• Admiral replaced first 7-cell cryomodule in slot SL21.
– Three of five cavities measured so far are at half last-measured Q,
~2e9 now.
– Gradient lowered in two of these three and one other to prevent
LHe boiling.
– CM at 300K during move from FEL.
• For C25 modules, gradients at which field emission causes
arc fault every two days dropped 7% after 300K cycle.
More field emission = more heat; not yet quantified.
TN-10-008 http://www1.jlab.org/Ul/Publications/view_pub.cfm?pub_id=9288
Basic equation
P = (E2l)/(Q(r/Q))
(r/Q) shunt impedance from cavity design
l cavity length
E gradient, V/m
Q cavity quality factor Q0
example: E= 6.9 MV/m, l=0.5m, (r/Q)=960 ohms/m, Q=4.5e9
P= (6.9e6)2*0.5/(960*4.5e9)=2.38e13/4.32e12=5.5W
C100 cavities have r/Q = 1241 ohms/m, lower loss at fixed E
Cavity E and Q are distributions which must be taken into
account if any estimate is to be realistic.
Q0 distribution – C25 modules
-2.33
-1.64
-1.28
-0.67
0.0
0.67
1.28
1.64
2.33
0.5
0.8
0.2
0.05
0.01
0.95
0.99
Norm
al Q
ua
ntil
e P
lot
1e+9 3e+9 5e+9 7e+9
100.0%
99.5%
97.5%
90.0%
75.0%
50.0%
25.0%
10.0%
2.5%
0.5%
0.0%
maximum
quartile
median
quartile
minimum
8.25e+9
8.25e+9
8.25e+9
8.25e+9
5.9e+9
4.6e+9
3.4e+9
2.8e+9
1.68e+9
6.66e+8
6e+8
Quantiles
Mean
Std Dev
Std Err Mean
Upper 95% Mean
Lower 95% Mean
N
4.8042e+9
1.8277e+9
119992166
5.0406e+9
4.5678e+9
232
Moments
Q0
Distributions
Includes 29 suspect high-Q measurements capped at 8.25e9. There were only a
few this high during initial commissioning, not 10%.
Q0 distribution – C25 modules (2)
-2.33
-1.64
-1.28
-0.67
0.0
0.67
1.28
1.64
2.33
0.5
0.8
0.2
0.05
0.01
0.95
0.99
Norm
al Q
ua
ntil
e P
lot
10
20
Cou
nt
1e+9 3e+9 5e+9 7e+9
Normal(4.39e+9,1.41e+9)
Mean
Std Dev
Std Err Mean
Upper 95% Mean
Lower 95% Mean
N
4.3947e+9
1.4131e+9
99425202
4.5907e+9
4.1986e+9
202
Moments
Location
Dispersion
Type
µ
s
Parameter
4.3947e+9
1.4131e+9
Estimate
4.1986e+9
1.2874e+9
Lower 95%
4.5907e+9
1.5662e+9
Upper 95%
-2log(Likelihood) = 8948.94091039153
Parameter Estimates
Shapiro-Wilk W Test
0.991162
W
0.2558
Prob<W
Note: Ho = The data is from the Normal distribution. Small
p-values reject Ho.
Goodness-of-Fit Test
Fitted Normal
Q0
Distributions
C25 measurements less suspect high-Q values. Normal distribution.
For simulation, use mean 4.5e9, 1.4e9 partially accounting for 8.25e9 points removed
Q0 distribution – C50 modules
-2.33
-1.64
-1.28
-0.67
0.0
0.67
1.28
1.64
2.33
0.5
0.8
0.2
0.05
0.01
0.95
0.99
Norm
al Q
ua
ntil
e P
lot
1e+9 3e+9 5e+9 7e+9
100.0%
99.5%
97.5%
90.0%
75.0%
50.0%
25.0%
10.0%
2.5%
0.5%
0.0%
maximum
quartile
median
quartile
minimum
8.25e+9
8.25e+9
8.25e+9
6.18e+9
5e+9
3.75e+9
3.1e+9
2.5e+9
1.8e+9
1.4e+9
1.4e+9
Quantiles
Mean
Std Dev
Std Err Mean
Upper 95% Mean
Lower 95% Mean
N
4.1651e+9
1.537e+9
171838347
4.5072e+9
3.8231e+9
80
Moments
Q0
Distributions
Includes 4 suspect high-Q measurements capped at 8.25e9.
None were seen above 7e9 in initial commissioning.
Q0 distribution – C50 modules (2)
-2.33
-1.64
-1.28
-0.67
0.0
0.67
1.28
1.64
2.33
0.5
0.8
0.2
0.05
0.01
0.95
0.99
Norm
al Q
ua
ntil
e P
lot
5
10
15
Cou
nt
1e+9 3e+9 5e+9 7e+9
Normal(3.95e+9,1.25e+9)
Mean
Std Dev
Std Err Mean
Upper 95% Mean
Lower 95% Mean
N
3.9501e+9
1.2456e+9
142878858
4.2348e+9
3.6655e+9
76
Moments
Location
Dispersion
Type
µ
s
Parameter
3.9501e+9
1.2456e+9
Estimate
3.6655e+9
1.0742e+9
Lower 95%
4.2348e+9
1.4825e+9
Upper 95%
-2log(Likelihood) = 3397.99555909662
Parameter Estimates
Shapiro-Wilk W Test
0.977144
W
0.1862
Prob<W
Note: Ho = The data is from the Normal distribution. Small
p-values reject Ho.
Goodness-of-Fit Test
Fitted Normal
Q0
Distributions
C50 measurements less suspect high-Q values. Normal distribution.
For simulation, use mean 4.0e9, 1.3e9 partially accounting for 8.25e9 points removed
Gradients – C25 modules
-2.33
-1.64
-1.28
-0.67
0.0
0.67
1.28
1.64
2.33
0.5
0.8
0.2
0.05
0.01
0.95
0.99
Norm
al Q
ua
ntil
e P
lot
10
20
30
40
Cou
nt
3 4 5 6 7 8 9 10 11
Normal(6.90982,1.48076)
Mean
Std Dev
Std Err Mean
Upper 95% Mean
Lower 95% Mean
N
6.9098225
1.480759
0.0980657
7.1030579
6.7165872
228
Moments
Location
Dispersion
Type
µ
s
Parameter
6.9098225
1.480759
Estimate
6.7165872
1.3561741
Lower 95%
7.1030579
1.630746
Upper 95%
-2log(Likelihood) = 825.040944114034
Parameter Estimates
Shapiro-Wilk W Test
0.984320
W
0.0128*
Prob<W
Note: Ho = The data is from the Normal distribution. Small
p-values reject Ho.
Goodness-of-Fit Test
Fitted Normal
lem_gradient
Distributions
For simulation, use mean 6.9 MV/m, 1.5
Gradients – C50 modules
-2.33
-1.64
-1.28
-0.67
0.0
0.67
1.28
1.64
2.33
0.5
0.8
0.2
0.05
0.01
0.95
0.99
Norm
al Q
ua
ntil
e P
lot
5
10
Cou
nt
4 5 6 7 8 9 10 11 12 13 14
Normal(10.905,2.01554)
Mean
Std Dev
Std Err Mean
Upper 95% Mean
Lower 95% Mean
N
10.905006
2.0155435
0.2375341
11.378635
10.431376
72
Moments
Location
Dispersion
Type
µ
s
Parameter
10.905006
2.0155435
Estimate
10.431376
1.7316468
Lower 95%
11.378635
2.4116532
Upper 95%
-2log(Likelihood) = 304.255151928576
Parameter Estimates
Shapiro-Wilk W Test
0.936218
W
0.0012*
Prob<W
Note: Ho = The data is from the Normal distribution. Small
p-values reject Ho.
Goodness-of-Fit Test
Fitted Normal
lem_gradient
Distributions
Not normal, but for Monte Carlo purposes assume it is with mean 10.9 MV/m, 1.5 as for C25
Monte Carlo simulation inputs
Type G mean G Q0 mean Q0
C25 6.9 MV/m 1.5 MV/m 4.5e9 1.4e9
C50 10.9 1.5 4.0e9 1.3e9
C100 A 18 1.8 8.2e9 1.2e9
C100 B 18 1.8 7.0e9 2.0e9
• 16W/module static heat C25/50
• 50 W/module C100 includes transfer line changes but not bayonets,
15W/module or 75W/linac, per R. Ganni presentation
not included: 100W/linac as
• 2W/module electric heat allowance
• 50W/linac electric heat in three swing heat zones
Monte Carlo simulation
• Written by Arne Freyberger
• Inputs from experimental values, cryo group estimates,
cavity specifications (e.g. R/Q)
• Gradient and Q values for C25 and C50 from lem and
measurement, respectively.
• C100 scenarios agreed with SRF May 18, 2010:
– 18 MV/m, 1.8 for cases with all modules on
– 22 MV/m max
– case A: Q0 8.2e9, 1.2e9
– case B: Q0 7.0e9, 2e9
– case: one C100 off; adjust all gradient means in that
linac up to compensate
Nominal configurations
• 6 Gev injector: two C25
• 12 GeV injector: one C25, one C100
• ignore quarter
Module type 6 GeV 12 GeV
C25 32 31
C50 10 10
C100 0 11
Monte Carlo simulation – 6 GeV run
1200+67.5+2 (energy) = 1297.5 MeV
Monte Carlo simulation – 6 GeV
Monte Carlo – tweak 6 GeV
“Optimize G dist” pairs high Qs with high gradients, as we did by hand for 6 GeV run.
Experimental heat load comparable to this one. 190 g/s * 20W/g/s = 3800 W
Monte Carlo – 6 GeV tweak
Single module heat simulations
• C25 53 +/- 5 W/module
• C50 159 +/- 12 W/module
• C100 253 +/- 13 W/module (case B)
12 GeV – case A (index 1)
Includes injector CMs so 2303 MeV energy gain needed. Add 2 as before = 2341 MeV.
C50 mean gradient reduced as MeV/W lowest for these.
12 GeV – case A
12 GeV – case B (index 2)
Lower Q mean and larger for C100. 7E9 is Q spec for C100 at 19 MV/m. 2K heat from
waveguide, ~1W/kW in C50, will drive effective Q down. Experiment needed.
12 GeV – case B
12 GeV – NL case A (index 4)
NL+Injector 1203 MeV, so again 2 high. Optimistic Qs.
12 GeV – NL case B (index 5)
~300W more heat than case A due to lower, specified Q
12 GeV NL – case B less one C100 (index 6)
Heat up ~200W. Add another 50W for static load of non-functional C100 if left in place.
Fault rate doubles as C25 gradient up 10%. C100 mean gradient up 6.7%, which may not be
available. C50 gradient 11 MV/m as lem places these at max possible. 260W per C100
12 GeV – SL case B (index 8)
12 GeV – SL case B, one C100 down (index 9)
Fault rate doubles due to C25 push. C100 mean gradient up 10% to 19.8 MV/m.
Heat up ~200W. Add another 50W for static load of non-functional C100 unless pulled.
280W per active C100.
SL case Ba: JLAMP (index 10)
3281W vs 3714W, 430W savings vs SL case B. mean+2
Ten more C50s (index 3)
Fault rate falls by ~40% due to reduction in C25 module count and their lower mean
gradient. Availability improvement ~3%. Heat load about the same. Higher Q at lower
C50 gradient per commissioning measurements.
NL: ten C50, one C100 down (index 7)
Fault rate remains the same instead of doubling as before as C25 gradients unchanged.
C50 module gradients increased so heat up ~300W vs case without extra C50s. If C50 Qs
at high gradient improve in next lot of ten, heat needn’t rise.
Heat summary – 12 GeV
Index Case Linac Heat mean Sigma Mean+2
1 Optimistic Q Both 6641 120 6881
2 Specified Q Both 7176 158 7492
3* Specified Q, ten more C50 Both 7160 150 7460
4 Optimistic Q NL 3505 85 3675
5 Specified Q NL 3795 117 4029
6 Specified Q, one C100 down NL 3993 124 4241+50
7* Spec Q, ten more C50, one C100 down NL 4109 126 4361+50
8 Specified Q SL 3480 117 3714
9 Specified Q, one C100 down SL 3660 120 3900+50
10 SL JLAMP (four more C100, four fewer C25) SL 3107 87 3281
6+8 Spec Q, one NL C100 down Both 7471 165 7801
* fault rate about 60% of case above. Electric heaters not included above, 100W/linac.
Not included
• 100W/linac electric control heaters - one may view the
numbers shown as mean plus one sigma plus this heat.
• Superconducting RF gun
• Superconducting fifth pass RF separator: 50W static and
dynamic in cryounit; 50W??? 2K transfer line from SL.
CASA design assumed a conventional RF separator so
beam line is available if SRF/cryo is not cost effective.
• Error: 530W for transfer line was allocated 265W/linac
instead of 302/228 W NL/SL per Rao’s presentation. 37W
too low for NL, 37W too high for SL.
Conclusions
• NL OK in all cases
• SL OK in all cases with present FEL requirement of 665W
• SL not OK with any JLAMP case – maximum available
with four additional C100s ~1100W, 430W increase over 6
GeV run value.
• Additional C50s will reduce fault rate.
– Application of improved surface processing developed
for C100 to increase C50 Q(G) will reduce heat load.
• Additional C50s will keep fault rate from increasing 50%
with a C100 down. (Fault rate will double in linac with C100
down. Faults about equal in linacs, so overall rate up 50%.)
SL case Ba: JLAMP (index 10)
Replace nine C25 with four C100 per JLAMP proposal, leaving five SL slots empty. Heat load
260W below SL case B (mean+2 ). If this is chosen early enough, the five empty slots in the
SL need not be equipped for cryomodules, removing the 75W bayonet load too. Net savings
335W then.