TETRODES ARE EFFICIENT En Efficient RF Sources workshop

TETRODES ARE EFFICIENT ! En. Efficient RF Sources workshop 2014, Cockcroft Institute Eric Montesinos, CERN-RF

Outline Frequency & Power range of tetrodes & diacrodes Theoretical and measured efficiency Availability vs Efficiency, an example with a YL 1530 tetrode operated in the CERN SPS Imposed parameters and their impact to efficiency Overhead Circulators High Voltage Power Supply Fundamental Power Coupler Granularity Conclusion: Tetrodes are very efficient En. Efficient RF Sources workshop, 03 -04 June 2014, Cockcroft Institue Eric Montesinos, CERN-RF 2

Frequency & Power range of tetrodes Tetrodes & Diacrodes available from industry Power k. W per single tube 10000 CW 1000 100 En. Efficient RF Sources workshop, 03 -04 June 2014, Cockcroft Institue 200 300 Frequency MHz 400 Eric Montesinos, CERN-RF 500 3

Frequency & Power range of tetrodes Tetrodes & Diacrodes available from industry Power k. W per single tube 10000 peak < 1 ms 1000 100 En. Efficient RF Sources workshop, 03 -04 June 2014, Cockcroft Institue 200 300 Frequency MHz 400 Eric Montesinos, CERN-RF 500 4

Frequency & Power range of tetrodes Tetrodes & Diacrodes available from industry Power k. W per single tube 10000 peak < 1 ms CW 1000 100 En. Efficient RF Sources workshop, 03 -04 June 2014, Cockcroft Institue 200 300 Frequency MHz 400 Eric Montesinos, CERN-RF 500 5

Theoretical efficiency Heat out RF power in DC power in En. Efficient RF Sources workshop, 03 -04 June 2014, Cockcroft Institue Eric Montesinos, CERN-RF 6

Amplifier class Class A Class B Operative curve Class C Operative curve Output Signal Unsused area Input Signal C B AB A 100% 75% 25% 0% A AB B Input Signal Efficiency 50% Output Signal Less than 180⁰ C Amplifier Class Description Class-A Full cycle 360⁰ of conduction Class-AB More than 180⁰ of conduction Class-B Half cycle 180⁰ of conduction Class-C Less than 180⁰ of conduction 90⁰ 0⁰ Conduction Angle 0 En. Efficient RF Sources workshop, 03 -04 June 2014, Cockcroft Institue Eric Montesinos, CERN-RF 7

Theoretical Class B efficiency DC power is Pdc = Vdc Idc Assuming the tube is linear whilst it is conducting, the dc anode current is found by Fourier analysis of the current waveform and is Idc = Ipk/π Irf = Ipk/2 = Idc π/2 And ideal class B, Vrf = Vdc So, RF power is Prf = ½ Vrf Irf Prf = ½ Vdc Idc π/2 = π/4 Vdc Ipk Class B Theoretical efficiency η = Prf/Pdc = ¼ Vdc Ipk / Vdc Idc η = 78. 5 % En. Efficient RF Sources workshop, 03 -04 June 2014, Cockcroft Institue Vdc Eric Montesinos, CERN-RF 8

Class B efficiency in practice En. Efficient RF Sources workshop, 03 -04 June 2014, Cockcroft Institue Ipk UG 2 Two reasons for not achieving this impressive number 1. tube is not fully linear whilst it is conducting 2. Anode voltage must be higher than G 2 voltage, VG 2 being ~ 10% Vdc This leads into Pdc = Vdc Idc = Vdc 1. 05 Ipk/π Prf = ½ Vrf Irf = ¼ 0. 9 Vdc Ipk Theoretical efficiency in practice η = Prf/Pdc = ¼ 0. 9 Vdc Ipk / 1. 05 Vdc Ipk/π η = 67 % Class B Vdc Eric Montesinos, CERN-RF 9

Measurement on a YL 1530 tube YL 1530 @ 200 MHz (Filament – 7. 5%) 70 Efficiency RF/DC 60 Operating point RF/DC efficiency = 66. 4 % 50 40 30 20 10 0 0 10 20 30 CW output power [k. W] En. Efficient RF Sources workshop, 03 -04 June 2014, Cockcroft Institue Eric Montesinos, CERN-RF 40 10

Overall YL 1530 efficiency @ 35 k. W UG 2 Ipk Class B Vdc En. Efficient RF Sources workshop, 03 -04 June 2014, Cockcroft Institue Eric Montesinos, CERN-RF 11

CERN SPS 1976 (7 km) En. Efficient RF Sources workshop, 03 -04 June 2014, Cockcroft Institue Eric Montesinos, CERN-RF 12

CERN SPS YL 1530 Power plants Two amplifiers, 64 YL 1530 tubes Frequency 200. 2 MHz One amplifier delivers 650 k. W with 32 YL 1530 -0. 2 d. B for combining plant (-4. 7 % power) Each single tube is operated 21. 2 k. W Overall Efficiency = 48 % En. Efficient RF Sources workshop, 03 -04 June 2014, Cockcroft Institue Eric Montesinos, CERN-RF 13

Overall YL 1530 efficiency @ 21 k. W Ipk 35 UG 2 Ipk 21 Class B Vdc En. Efficient RF Sources workshop, 03 -04 June 2014, Cockcroft Institue Eric Montesinos, CERN-RF 14

SPS Supercycle LHC 2008 (27 km) North Area SPS 1976 (7 km) CNGS 2006 En. Efficient RF Sources workshop, 03 -04 June 2014, Cockcroft Institue Eric Montesinos, CERN-RF 15

SPS Supercycle LHC 2008 (27 km) During 2012 North Area SPS 1976 (7 km) CNGS Supercycle 48 seconds 2006 North Area CNGS LHC 19. 25 seconds / 650 k. W 9. 6 seconds / 488 k. W 19. 15 seconds / 0 k. W For 7000 hours of operation 2. 5 GWh En. Efficient RF Sources workshop, 03 -04 June 2014, Cockcroft Institue Eric Montesinos, CERN-RF 16

Operating with ‘OFF Line’ Tubes In normal operation, one tube delivers a maximum power of 650 k. W + 4. 7 % / 32 = 21. 2 k. W Systems have been designed (1980) to deliver full voltage into the cavity even with only 24/32 tubes 32 P 18 P In case of 8 tubes ‘OFF Line’, remaining tubes shall deliver 23 k. W x (32/24)^2 = 37. 8 k. W 32 x 21. 2 k. W = 18 x 37. 8 k. W = 680 k. W YL 1530 tubes are rated 37. 5 k. W En. Efficient RF Sources workshop, 03 -04 June 2014, Cockcroft Institue Eric Montesinos, CERN-RF 17

Tubes statistics 200 180 160 # tubes 140 120 100 80 60 40 YL 1530 SPS 200 MHz, 2 x 32 tubes in parallel (Filament – 7. 5 % & 21 k. W instead of 35 k. W) Total 1980 -2013 : 254 tubes Average lifetime : 32’ 000 hours Tubes with > 28’ 000 hours : 78 % (also true for tubes stored 20 years !) 20 0 1 y 2 y 3 y # years of operation before failure of the tube En. Efficient RF Sources workshop, 03 -04 June 2014, Cockcroft Institue Eric Montesinos, CERN-RF >= 4 y 18

Amplifiers statistics YL 1530 SPS 200 MHz, 2 x 32 tubes in parallel Over temp 8% Unknown 8% Arcing 7% Loss of emission 39% Input reflections 38% Reason of fault # of faults % of faults Loss of emission 99 39 Input reflections 97 38 Unknown 20 8 Over Temp 20 8 Arcing 18 7 We lose 14 -16 tubes per year, that we exchange during Technical Stops 3 -4 times a year Unpredictable : 23 % (equivalent to 4 tubes) Graceful degradation : 77 % (possible to monitor and predict) En. Efficient RF Sources workshop, 03 -04 June 2014, Cockcroft Institue Eric Montesinos, CERN-RF 19

Availability of the CERN SPS • - 7. 5 % Ufilament → +37. 5 % lifetime • Off Line tubes for the unpredictable faults • Overall efficiency 48 % during cavity maximum voltage • Supercycle average power 390 k. W Over 7000 hours of operation in 2012, these two RF Power plants have been responsible for only 16 hours of down time ! En. Efficient RF Sources workshop, 03 -04 June 2014, Cockcroft Institue % SPS down time With these operating conditions 2 1 0 2003 2008 Eric Montesinos, CERN-RF 2013 20

Availability of the CERN SPS Availability of an Injector is the key parameter CERN’s total annual electricity consumption is 1. 3 TWh, for an operation of almost 7000 hours per year When SPS is down, electrical bill without physics, costs 4500 CHF / h (minimum, only electricity cost based on 60 CHF / MWh) 20 % for LHC cryogenics + cooling and ventilation 10 % for LHC experiments 7 % for the North area experiments (http: //cds. cern. ch/record/1324541? ln=en) En. Efficient RF Sources workshop, 03 -04 June 2014, Cockcroft Institue Eric Montesinos, CERN-RF 21

Availability of the CERN SPS Savings We can improve the efficiency from 48 % to 61 % of the RF Power source by • Increasing the filament to the nominal value (- 37. 5% lifetime) • Reducing the number of tubes in operation (2 x 20 tubes operated 34 k. W) • no additional down time with ‘off line’ operation BUT reduced beam intensity(OR Three hours stop for a tube exchange En. Efficient RF Sources workshop, 03 -04 June 2014, Cockcroft Institue 66, 000 CHF Money (electrical bill savings) 2. 5 GWh at 60 CHF/MWh with 48 % efficiency costs 312, 000 CHF 61 % efficiency costs 246, 000 CHF Costs 273, 000 CHF (money, not taking into account impact to physics) 14 additional tube exchanges, 42 hours physics lost & electrical bill due to LHC & experiments infrastructure costs Eric Montesinos, CERN-RF 22

Availability of the CERN SPS Savings We can improve the efficiency from 48 % to 61 % of the RF Power source by • Increasing the filament to the nominal value (- 37. 5% lifetime) • Reducing the number of tubes in operation (2 x 20 tubes operated 34 k. W) • no additional down time with ‘off line’ operation BUT reduced beam intensity(OR Three hours stop for a tube exchange En. Efficient RF Sources workshop, 03 -04 June 2014, Cockcroft Institue 66, 000 CHF Money (electrical bill savings) 2. 5 GWh at 60 CHF/MWh with 48 % efficiency costs 312, 000 CHF 61 % efficiency costs 246, 000 CHF Costs 186, 000 CHF (money, not taking into account impact to physics) 14 additional tube exchanges, 42 hours physics lost & electrical bill due to LHC & experiments infrastructure costs Eric Montesinos, CERN-RF 23

Availability of the CERN SPS With respect to this CERN SPS example, applicable to all RF plants with smaller energy consumption compare to ‘other’ energy consumers in the machine, being more efficient with RF power plant with an overall efficiency increased from 48 % → 61 %, and no margin in the RF power plant, for the same beam quality Will induce more accelerator stops Cost more electricity at the end Even worse if the k. Wh cost increases in the coming years… En. Efficient RF Sources workshop, 03 -04 June 2014, Cockcroft Institue Eric Montesinos, CERN-RF 24

Overhead YL 1530, 35 k. W @ 200 MHz 70 10 µs (x 1. 7) Output power k. W 60 100 µs (x 1. 5) 1 s (x 1. 4) 50 CW (x 1. 2) 40 35 k. W nominal 30 20 10 0 Courtesy Wolfgang Höfle 0 1000 2000 3000 4000 5000 Input power W En. Efficient RF Sources workshop, 03 -04 June 2014, Cockcroft Institue Eric Montesinos, CERN-RF 25

Output power / nominal Output power Overhead Tetrodes, Klystrons, SSPA 1. 8 1. 6 1. 4 1. 2 klystrons 1 Tetrode Overhead needed for LLRF operation 0. 8 Klystron 0. 6 SSPA 0. 4 RF Pulse 0. 2 0 0 1 2 3 4 Input power / nominal Input power En. Efficient RF Sources workshop, 03 -04 June 2014, Cockcroft Institue Eric Montesinos, CERN-RF 26

HVPS Gridded tubes Pdc klystrons HV 10 -20 % tolerant For tetrodes (and gridded tubes more generally) HVPS is very simple No RF -> idle current (can be zero in class B or class C) RF Pulse RS 2004 tetrode 100 90 80 70 60 50 40 30 20 10 0 Unominal (8. 5 k. V) - 10 % U nominal -20 % Unominal 0 En. Efficient RF Sources workshop, 03 -04 June 2014, Cockcroft Institue RF Pulse time Pout k. W Even if HV is drooping, the LLRF will impose output power, and tetrode remains able to deliver requested Power HV less tolerant 5 Pin k. W Eric Montesinos, CERN-RF 10 15 27

Circulator If the reflected power is not too long, Tetrodes do NOT need a circulator With a good reflected power protection, Tetrodes can sustain full reflection We operate the SPS since 1976 -1980 without a single failure due to reflected power Our protection is made with old relays, few milliseconds En. Efficient RF Sources workshop, 03 -04 June 2014, Cockcroft Institue Eric Montesinos, CERN-RF 28

Fundamental power coupler Even if we can imagine cavities able to receive a large amount of power In the frequency range, Fundamental Power Coupler will limit the maximum power that can be delivered to the cavity The limit will be ~ 1. 0 MW peak power ~ 0. 5 MW average power (for a stable operation 10 -20 years lifetime of a coupler) En. Efficient RF Sources workshop, 03 -04 June 2014, Cockcroft Institue Eric Montesinos, CERN-RF 29

Granularity In the frequency range, Tetrodes power limit is close to FPC power limit FPC One (or few) tube per cavity Redundancy ensured by a small additional number of tubes/cavities FPC This will ensure to operate with a high efficiency, still having availability En. Efficient RF Sources workshop, 03 -04 June 2014, Cockcroft Institue FPC Eric Montesinos, CERN-RF 30

Granularity Single higher power source Need a circulator to protect the tube FPC More losses in circulator & splitting system FPC No individual control of the cavities FPC Limited to the weakest cavity Will lose a large number of cavities in case of source failure FPC Not easy to have a few spares En. Efficient RF Sources workshop, 03 -04 June 2014, Cockcroft Institue Eric Montesinos, CERN-RF 31

Granularity Multi lower power sources Need a circulator to protect the sources More losses due to combining system - 0. 1 d. B is - 2. 5 % in Power At 352 MHz or 400 MHz A simple N connector is - 0. 01 d. B If redundancy is made by increased number of modules, not anymore the best operating point, efficiency is reduced If at the efficient operating point, same number of few additional cavities as with tetrodes En. Efficient RF Sources workshop, 03 -04 June 2014, Cockcroft Institue FPC FPC FPC Eric Montesinos, CERN-RF 32

Granularity ‘New’ way of making the storage of spares 2 MW One tube is in ‘stand-by’ mode and can possibly replace any faulty tube within few minutes with a ‘OFF line – ON line’ matrix All four tubes in operation will be operated at their best efficiency point Still have robustness thanks to the fifth tube En. Efficient RF Sources workshop, 03 -04 June 2014, Cockcroft Institue 2 MW New SPS power plants: 2 x 2 MW @200 MHz Four + one diacrodes option scheme Eric Montesinos, CERN-RF 33

Conclusion (I/II) Regarding the frequency range & regarding the power range If AVAILABILITY of the machine is the key criteria Tetrodes are a good choice Single (few) tubes per cavity with good efficiency & redundancy with few additional cavities (on hold) Multi tubes solution for redundancy operated at a lower power, lower efficiency with nevertheless a very correct final overall cost If EFFICIENCY is the key criteria Regarding the arrangement choices, Tetrodes can be operated 60 -70 % overall efficiency in operation En. Efficient RF Sources workshop, 03 -04 June 2014, Cockcroft Institue Eric Montesinos, CERN-RF 34

Conclusion (II/II) If within the frequency & power range Never neglect a tetrode (diacrode) option Carefully look at it, and you will see they are very good ! En. Efficient RF Sources workshop, 03 -04 June 2014, Cockcroft Institue Eric Montesinos, CERN-RF 35

Thank you for your attention En. Efficient RF Sources workshop, 03 -04 June 2014, Cockcroft Institue Eric Montesinos, CERN-RF

Theoretical Class B efficiency DC power is Pdc = Vdc Idc Assuming the tube is linear whilst it is conducting, the dc anode current is found by Fourier analysis of the current waveform and is Idc = Ipk/π Irf = Ipk/2 = Idc π/2 And ideal class B, Vrf = Vdc So, RF power is Prf = ½ Vrf Irf Prf = ½ Vdc Idc π/2 = π/4 Vdc Ipk Class B Class A Class C Theoretical efficiency η = Prf/Pdc = ¼ Vdc Ipk / Vdc Idc η = 78. 5 % En. Efficient RF Sources workshop, 03 -04 June 2014, Cockcroft Institue Vdc Eric Montesinos, CERN-RF 37

Tubes costs in the CERN SPS 7000 hours operation per year 2 x 32 tubes operation 2 x 20 tubes operation Filament, nominal -7. 5 % Lifetime 32000 hours Reduced thermal losses Broken tubes per year 14 Cost 350, 000 CHF Filament, nominal Lifetime 20000 hours (-37. 5%) Increased Thermal losses Broken tubes per year 14 Cost 350, 000 CHF Overall efficiency 48 % Electricity 2. 5 GWh, 0. 125 CHF/k. Wh Cost = 0. 06 2. 5/0. 48 Cost = 312, 000 CHF Overall efficiency 61 % Electricity 2. 5 GWh, 0. 06 CHF/k. Wh Cost = 0. 06 2. 5/0. 61 Cost = 246, 000 CHF En. Efficient RF Sources workshop, 03 -04 June 2014, Cockcroft Institue Eric Montesinos, CERN-RF 38

Pulsed machine Operating frequency 352 MHz Pulse duration 3. 5 ms Repetition rate 14 Hz Number of cavities 26 RF Power 400 k. W Total RF peak power 10. 4 MW Total RF average power 510 k. W Redundancy by number of cavities (operation still possible with one/two cavity OFF) Operated at maximum power Same tetrode being driver Pre-driver SSA ~ 600 W En. Efficient RF Sources workshop, 03 -04 June 2014, Cockcroft Institue Eric Montesinos, CERN-RF 39

Pulsed machine Operating frequency 352 MHz Pulse duration 3. 5 ms Repetition rate 14 Hz Number of cavities 26 RF Power 400 k. W Total RF peak power 10. 4 MW Total RF average power 510 k. W HVPS Redundancy by number of cavities (operation still possible with one/two cavity OFF) Operated at maximum power Same tetrode being driver Pre-driver SSA ~ 600 W En. Efficient RF Sources workshop, 03 -04 June 2014, Cockcroft Institue Eric Montesinos, CERN-RF 40