Design and Manufacture of the RF Power Supply and RF Transmission Line for SANAEM Project Prometheus
G. Turemen, S. Ogur, F. Ahiska, B. Yasatekin, E. Cicek, A. Ozbey, I. Kilic, G. Unel, A. Alacakir
PPrepared for submission to JINST
Design and Manufacture of the RF Power Supply and RFTransmission Line for SANAEM Project Prometheus
G. Turemen, a , b , S. Ogur c F. Ahiska d B. Yasatekin a , b E. Cicek e A. Ozbey f I. Kilic b G. Unel g A. Alacakir b a Ankara University, Department of Physics, Ankara, TURKEY b TAEK, Saraykoy Nuclear Research and Training Center, Ankara, TURKEY c Bogazici University, Department of Physics, Istanbul, TURKEY d EPROM Electronic Project & Microwave Ind. and Trade Ltd. Co., Ankara, TURKEY e Gazi University, Department of Physics, Ankara, TURKEY f Istanbul University, Aircraft Technology Program, Istanbul, TURKEY g University of California at Irvine, Department of Physics and Astronomy, Irvine, USA
E-mail: [email protected]
Abstract: A 1-5 MeV proton beamline is being built by the Turkish Atomic Energy Authorityin collaboration with a number of graduate students from different universities. The primary goalof the project, is to acquire the design ability and manufacturing capability of all the componentslocally. SPP will be an accelerator and beam diagnostics test facility and it will also serve thedetector development community with its low beam current. This paper discusses the designand construction of the RF power supply and the RF transmission line components such as itswaveguide converters and its circulator. Additionally low and high power RF test results arepresented to compare the performances of the locally produced components to the commerciallyavailable ones.Keywords: Accelerator subsystems and technologies, waveguides, modeling of microwave sys-tems, passive components for microwavesArXiv ePrint: 1504.01576 Corresponding author. a r X i v : . [ phy s i c s . acc - ph ] J u l ontents The SANAEM Project Prometheus (SPP) at the Turkish Atomic Energy Authority (TAEK)’sSaraykoy Nuclear Research and Training Center (SANAEM), aims to gain the necessary knowl-edge and experience to construct a proton beamline needed for future necessities [1]. A Proof ofPrinciple (PoP) accelerator with modest requirements of achieving at least 1 MeV proton energy,with a peak beam current of few tens of µ A’s is under development. This PoP project has also thechallenging goal of having the design and construction of the entire setup using local resourcesranging from its ion source to the final diagnostic station, including its RF power supply and theRF power transmission line which are the foci of this paper. There are also two secondary goalsfor this project: 1) To train the next generation accelerator physicists and RF engineers on the job;2) To encourage and develop skills of the local industry in the accelerator component construction.The ion source is a pulsed 1 kW 13.56 MHz RF driven inductively coupled plasma chamber where20 kV potential is used to extract 10 ms H + ions with a repetition frequency of 1 Hz [2]. The lowenergy proton beam’s current, profile and emittance are measured using a compact measurementstation at the solenoid based LEBT line, which was previously described elsewhere [3]. A "4-vane"Radio Frequency Quadrupole (RFQ) operating at 352.21 MHz is used as the low beta acceleratingcavity [4], [5]. This operating frequency ( f ) was selected to be compatible with similar machinesin Europe and therefore to benefit from the already available RF power supply market. The designrequirements of the accelerating cavity is given in Table 1. Other parameters such as the inter-vane– 1 –oltage (V), input energy ( E in ), output energy ( E out ), beam current acceptance ( I ), power dissi-pation at maximum Q factor ( P d ) and the vane length ( L ) were chosen to be adequate for a firsttime machine. To compensate for any possible reduction in Q factor of the manufactured cavity,a safety margin is considered and the power requirement of the RFQ is estimated as 120 kW. Theaccelerating cavity will be followed by a beam diagnostics section and finally by a beam dump. Table 1 . SPP RFQ design parameters
Parameter Value E in (MeV) 0.02 E out (MeV) 1.3 f (MHz) 352.21 V (kV) 60 I (mA) <15 P d (kW) 65.5 L (m) 1.2 The overview of the RF power transmission line for the SPP project is depicted in Fig.1. The racksshown in lower left part of the figure represent the RF power supply unit (PSU). The RF powersupply is required to be pulsed at a maximum duty factor of 3% and to have a peak output powerof 120 kW at 352.21 MHz. The master oscillator’s frequency precision is required to be at least1 kHz in order to fine-tune the accelerator cavity (at this level, no LLRF system is used to followthe frequency of the cavity). A precision of 10 kW in power is considered to be enough to set theproper voltage level in the RFQ. The acceptable beam transmission limit of the RFQ cavity is setto be 70% for the project. The cavity’s inter-vane voltage should not fall below 56 kV compared tothe design value of 60 kV to ensure such a transmission rate. Since 4 kV potential difference wouldcorrespond to about 16 kW difference from the PSU, its stability requirement can be written as ± igure 1 . The overview of the SPP RF transmission line. The SPP power supply is required to operate at 352.21 MHz as a narrow bandwidth device, and todeliver about 120 kW peak power in pulsed mode at a maximum duty factor of 3%. It is designedby the project members and manufactured by a local RF company [6]. It is envisaged as a twostaged amplifier, as shown in Fig.2. The first stage was implemented with a solid state amplifierand the second stage with a vacuum tube amplifier. The whole system was planned to allow bothremote operation over IP network and local control and monitoring over the touch screen panel onthe control rack.
Figure 2 . The two stage hybrid PSU design for SPP. Note that the solid state stage is only water cooledwhereas the second stage with tetrode tube requires both water and forced air cooling.
The first amplification stage uses the BLF-578XR high power transistors [7] to achieve acombined output of 6 kW in continuous and over 8 kW in pulsed modes. Each solid state amplifier(SSA) circuit provides 1.1 kW per amplifier board in pulsed mode for pulses shorter than 500 ms.The output of stage one is, thus, obtained by combining the outputs of eight such boards using homebuilt combiner modules (Fig.3). – 3 – igure 3 . Block diagram of the combiner module.
The low power RF input of the splitter/SSA/combiner component is provided by a low noisePLL (phase-locked loop) controlled oscillator and amplified with a 1 W pre-amplifier. The driveramplifier’s output power is controlled with a PLC (programmable logic controller) based gaincontroller to adjust the input power of the SSA modules. All error signals are evaluated with thePLC and triggering the fast-interrupts to protect the RF system, if needed. The low power 3dBcoupler splits the output of the driver amplifier into two and channels these into two identicalsplitter/SSA/combiner boards. On these boards, a microstrip Wilkinson 1/4 splitter was designedwith a teflon substrate and a 35 µ m silver plated copper strip to further split the received signal andchannel it into 4 identical SSA modules. Each SSA module amplifies the RF signal up 1.1 kW withthe water cooled high power transistor modules. The output signals of the transistor modules arecombined with a microstrip Wilkinson 1/4 combiner for both boards. Finally, the combined outputof both boards are coupled up with a high power 3 dB coupler to provide 8 kW RF signal.The assembled combiner and a number of amplifier cards can be seen in Fig.4. The powerand the rise time test results of the combined pulse from the SSA combiner are shown in Fig.5 andFig.6, respectively. The PSU provided about 42 ns of rise-time for an RF pulse of about 200 µ s,fast enough for filling the SPP RFQ cavity. The power tests of the SSA pulse are performed with-70 dB attenuated signal and a 7.1 kW pulse power was reached in these tests.The second amplification stage uses a tetrode amplifier, both water and air cooled, to achievethe final peak power of 120 kW. The selected tetrode tube is the TH595 [8] which provides up to200 kW peak power in short pulse operation mode, for frequencies up to 450 MHz. Fig.7 containsa view of the tetrode amplifier, the amplification cavity and the forced air cooling system during– 4 – igure 4 . SPP PSU solid state amplifier combiner. Figure 5 . Power test results of the pulse. – 5 – igure 6 . Rise time of the RF pulse is measured as 42 ns on a dummy load. assembly. For the SPP setup, the tetrode gate voltages are adjusted as G1= -200 V and G2= 900 V,the filament voltage as 7.3 V, whereas the anode voltage is set to 13.5 kV. The completed RF PSUincluding all the controls, power supplies and cooling as installed in the laboratory consists of fiveracks. The DC power supply, control unit and the first amplification stage are as shown in Fig.8during operation.
Figure 7 . SPP PSU Tetrode amplifier rack during assembly in the lab. Left section contains the RF amplifiercavity which would later house the tetrode amplifier and on the right side one can see the air intake for forcedair cooling of the tetrode. – 6 – igure 8 . SPP PSU in use. The left rack has the DC power supply, the central rack has the controller and themaster signal generator, and the right rack has the power supplies for the tetrode gates, the first amplificationstage and the output to the Tetrode amplifier rack.
As a PoP project, SPP transmission line contains a large variety of the available components,where 3 1/8" rigid coaxial lines, a waveguide circulator, full height and half height rectangularwaveguides (WR2300), a flexible rectangular waveguide and their converters and adapters can becited as examples (Fig.9). The selected rigid coaxial line has 50 Ω impedance and can carry a peakpower of 440 kW, up to a maximum frequency of 855 MHz [9]. The WR2300 series rectangularwaveguides operate at a lower maximum frequency, up to 450 MHz and these can carry a maximumpower of about 700 MW. Figure 9 . Block diagram of the SPP RF transmission line.
In the whole transmission line only one component, a flexible rectangular waveguide fromMega [10], was purchased to ease alignment of the coaxial coupler to the RFQ. All the other RF– 7 –omponents were first designed using Computer Simulation Technology (CST) Microwave Studio’s(MWS) Frequency Domain Solver [12] for the operating frequency of 352.21 MHz and producedlocally by the manufacturers in Middle East Industry and Trade Center (OSTIM) in Ankara. Withthe relatively modest power requirements, the SPP RF transmission line operates in normal air anddoesn’t have any special isolation gas requirements. The maximum power capability requirementof the components is pulsed 120 kW with 0.01% duty factor and 1 Hz repetition frequency. Upondelivery, reflection ( S
11) and transmission ( S
21) characteristics of the components were measuredat the operating frequency using low power output from a vector network analyzer (VNA) [13].Lastly, the high power tests were performed with the RF PSU and the whole RF transmission line.
The half height RF dump is made from a triangularly ending waveguide loaded with silicon carbide.Its length is about 50 cm and it can be air cooled for high power operations. The two E-bends areto be used with the circulator setup, whereas the H-bend is for the section of the transmission linejust before the RFQ. Therefore the bends are designed as HH and FH modules as shown in Fig.10.These components are optimized to perform below −
85 dB for S
11 (with PEC assumption) andproduced as seen in Fig.11 and Fig.16, respectively.
Figure 10 . Full height H-bend(left) and half height E-bend (right) designs for SPP.
Figure 11 . Half height E-bend attached to RF dump.
Fig.11 contains the manufactured and assembled HH E-bend attached to the RF dump to tunethe coaxial-rectangular waveguide converters just before integration to the RF transmission line(Fig.13). As the RF transmission line contains both FH and HH sections, a transition module– 8 –as necessary. The design of the HH to FH transition was performed by using CST MWS whileoptimizing the total length, the transmission efficiency and the producibility (Fig.12). Consideringthese factors, the S
11 of the HH to FH transition is optimized to −
61 dB, similarly with PECassumption.
Figure 12 . Left : HH to FH transition design.
Right : The finished product.
The S
11 measurements were performed using the VNA, which provided a low power RF signalto the HH coaxial to rectangular (TEM to TE) waveguide converter with a home-built 3 1/8" toN type reducer. The 50 cm converter was attached to a 50 cm HH WR2300 waveguide, the 90degree E-bend and finally to the RF dump as seen in Fig. 13. The low power pulse of the VNA,eventually absorbed by the RF dump, was used to tune the converters. Before the production phase,the S
11 parameters of the reducer and the TEM to TE waveguide converter are optimized to −
69 dBand −
62 dB, respectively, with the above mentioned simulation setup. The maximum CW powercapability of the TEM to TE waveguide converter was found as 400 kW without any cooling. Fig.14contains the S
11 measurement results for the previously described setup. After tuning the TEM toTE waveguide converter, at the operating frequency of 352.21 MHz, the attained S
11 value was −
60 dB.As previously discussed, the RF transmission line also requires a FH rectangular to coaxial(TE to TEM) waveguide converter. For the TE to TEM waveguide converter the S
11 parameterwas optimized to −
63 dB with computer simulations. The CW power handling capacity of the TEto TEM waveguide converter was calculated as 275 kW without any cooling. After the productionof the TE to TEM waveguide converter the second measurement employs the same setup, exceptthe HH TEM to TE waveguide converter was replaced by the mentioned HH to FH adapter and theFH TE to TEM waveguide converter. The S
11 for this case read as −
59 dB after tuning, whichcan be found in the same figure, second screenshot. The HH TEM to TE and FH TE to TEMwaveguide converters were designed as narrow-band devices due to the RFQ cavity’s requirements.For a comparison, an RF-lambda [11] product RFWA2300 has 1.3 VSWR in a bandwidth of 370MHz and our converters have bandwidths of 4 MHz. By means of this narrow-bandwidth, designedconverters have a sharper S11 and are less lossy at the operating frequency of the cavity.– 9 – igure 13 . The S
11 measurement setup for HH TEM to TE converter.
Figure 14 . Top : S
11 for TEM to TE conversion.
Bottom : S
11 for TE to TEM conversion. Both measurementsare made by using a low power RF signal from the VNA. – 10 –nsertion loss measurement setup is given in Fig.15. Basically a low power pulse is emitted bythe VNA which is converted to TE mode and propagated from left to right through HH waveguide,HH to FH adapter and finally FH TE to TEM waveguide converter, all being previously discussed.The pulse arriving to the right side is received by the same VNA which can calculate the S − .
17 dB.
Figure 15 . The S
21 measurement setup of the converters.
Following the low power tests, the most of the waveguide components of the RF transmissionline were assembled as in Fig.16. To measure the S
21 parameter of the whole transmission line,the same procedure was followed. For the SPP operating frequency, S
21 of the entire line wasmeasured as − . Figure 16 . Top : The S
21 measurement setup of the transmission line. A low power RF pulse from the VNAgoes from left to right and finally back the VNA.
Bottom : The S
21 results of the RF transmission line.
The reducers were also tested with the VNA pulse and an insertion loss of -0.04 dB was– 11 –ecorded (Fig.17).
Figure 17 . The S
21 measurement setup for the reducers.
A circulator is a non-reciprocal passive RF transformer, used to deliver the RF power to a targetdevice, e.g. an accelerator cavity, while protecting the RF power supply with an RF load in theevent of an impedance mismatch between the supply and the load. An H-plane waveguide junctioncirculator, to be operated at the frequency of 352.21 MHz, has been designed using CST MWS [12]for the SPP project. This RF device is intended to deliver the RF power generated by the RF powersupply unit, with minimal loss, while protecting the power supply against any power reflections.The maximum power handling requirement for the project is about 120 kW which will be suppliedfrom port-1. The circulator is designed to transfer this power to port-2, as seen in Fig.18.
Figure 18 . Electric field inside the circulator for 352.21 MHz. – 12 –hat port is connected to the RF transmission line, delivering RF power to the RFQ by themeans of waveguides and rigid coaxial cables, discussed previously. The air cooled RF dump thatwas discussed in the previous section 4.1 is connected to port-3 to absorb any reflections fromport-2.The designed circulator consists of two aluminum disks and two disks of ferrites which areattached inside the HH WR2300 waveguide. Following CST MWS simulations, the ferrite materialis determined as NG-1600 from the catalogue of Magnetics Group [14]. NG-1600 has a saturationmagnetization 4 π M s at 1600 Gauss, its maximum line width is 10 Oersted, and its dielectric constantis 14.6 while the loss tangent is below 2 × − , where all values are given by the supplier at 9.4 GHz.The Curie temperature is given as 220 ◦ C which was considered in the thermal simulations. Thenecessary setup to keep the ferrites at room temperature during the operation was made availablevia water cooling. However due to low RF duty factor (0.01%), so far no such need has arisen.The relation between the saturation magnetic field of the ferrite (cid:174) M s and applied field (cid:174) H both havingtime dependence of ω can be formulated as below: M x M y M z = X xx X xy X yx X yy
00 0 0 H x H y H z , (4.1)where elements of X correspond to; X xx = ω m ω ( ω m − ω ) = X yy , (4.2) X xy = j ωω m ( ω m − ω ) = − X yx . (4.3)Thus, magnetic field becomes; (cid:174) B = µ (cid:169)(cid:173)(cid:173)(cid:171) + X xx X xy X yx X yy
00 0 0 (cid:170)(cid:174)(cid:174)(cid:172) (cid:174) H = [ µ ] (cid:174) H , (4.4)where the Larmor and magnetization frequencies are respectively; ω = µ o γ H , (4.5) ω m = µ o γ M s , (4.6)where, µ is vacuum permeability and γ is the gyromagnetic ratio. The permeability of the ferrite µ has been discussed in detail by Wu and Du [15]. Comparing these two frequencies, it can beseen that this circulator is working above the resonant frequency. An advantage of this operationregime is to have lower field losses as pointed out by Bosma [16]. The fields inside the circulatorare so-called the acceleration mode which is TM (or in European notation E ) and the view ofthe magnetic field can be seen from Fig.19, while the electric field is already presented in Fig.18.The performance of the designed circulator, obtained after lengthy optimization studies, canbe seen in Fig.20. At the SPP operating frequency of 352.21 MHz the return loss ( S S
21) and the isolation loss ( S
31) are computed as -55 dB, -0.26 dB and -30 dB, respectively.– 13 – igure 19 . Magnetic field inside the circulator for 352.21 MHz.
VSWR is calculated to be about 1.005:1. The power handling capacity of the circulator is about30 kW CW without needing any water cooling. As can be seen in Fig.20 the bandwidth of thedesigned circulator is 12 MHz for a maximum VSWR of 1.2. For a comparison, an RF-lambdaproduct RFWC2300A having 1.2 VSWR in a bandwidth of 17 MHz, has -0.3 dB maximum insertionloss and -20 dB minimum isolation loss. Therefore the performance of this locally designed andmanufactured circulator is comparable (-0.26 dB vs -0.3 dB and -30 dB vs -20 dB) to an industrialone. Finally, the biasing magnetic field of the circulator is supplied by a locally designed and builtdipole electromagnet to fine tune of the ferrites. The dipole is able to produce an external magneticfield of 2.3 kG to the ferrites without any cooling requirements.
Figure 20 . Performance of the circulator design at 352.2 MHz.
Most of the circulator components have been bought or produced locally, except the two ferritedisks purchased from TSI ceramics [14]. The non-magnetic components were constructed from– 14 –061 series aluminum and attached to a support frame made from 2 cm thick iron layers that alsoserved as a return yoke for the magnetic field. For field tunability, the biasing magnets were madesolely from two solenoids forming a dipole. The coils of the dipole are designed to carry a currentof about 17 A each. The total weight of the circulator including the coils and yokes is about 1 Ton.The 3D drawing of the whole circulator setup is given in Fig.21.
Figure 21 . The SPP HH waveguide circulator together with its TEM to TE waveguide converter on port-1(left side), its RF dump on port-3 (lower right) and its RF transmission line connection on port-2 (upperright).
The individual components were assembled in the laboratory by the team members and installedin the RF transmission line as seen in Fig.22. The biasing field estimated in CST as 1910 G hasbeen optimized as 1880 G after the installation using a VNA.After all the components were assembled, the full transmission line (with circulator) was fedwith low power RF signals to measure the S-parameters. The insertion loss for the full line ismeasured as -0.34 dB which is corresponds to 7.5% power loss (Fig.23). Simulated electric field– 15 – igure 22 . 352.21 MHz waveguide circulator as it is installed in the RF transmission line. Note that theconnection to the RF PSU (port-1 one left side) was made with a 3 1/8" rigid coaxial line. The electromagnet,PSU and the gauss-meter are right below the circulator. distribution of the whole RF transmission line is shown in Fig.24.
Figure 23 . The insertion loss measurement of the whole RF transmission line.
Figure 24 . The electric field pattern of the full RF transmission line. – 16 –
High Power Measurements
The high power tests of the RF system were performed within three phases. First, at the SSA output,second at the tetrode output, and third at the end of the RF transmission line. The performanceof the RF PSU was measured with a -60 dB RF load and a oscilloscope (Fig.25). All high powermeasurements were done with calibrated coaxial cables and attenuators. First, high power testswere performed at the SSA output with the previously mentioned setup. A 5.8 kW RF power wasmeasured with a pulse length of 100 µ s. Second, the same setup was used to calculate the tetrode’samplification gain through feeding the RF tube with SSA power. A 15.9 kW output RF power wasmeasured with an input power of 780.3 W. Therefore the gain of the tetrode tube was calculated as13.1 dB. With this gain the tetrode is able to produce a 118.9 kW RF power with 5.8 kW RF inputpower. Unfortunately tetrode’s output RF power at this power level could not measured due to therestriction of maximum power capability of the existing RF load. After the gain measurements, thetetrode output was connected to the RF transmission line. As the third phase, the high power testswere performed at the end of the RF transmission line. An SSA power of 2.4 kW was fed to thetetrode and the amplified power was measured at the end of the RF transmission line as 47.4 kW.Regarding a power loss of 7.5% for the whole transmission line, the output power of the tetrode wascalculated as 50.9 kW and the gain as 13.3 dB. Accordingly with this gain the RF PSU is enoughto provide the required RF power to the RFQ cavity with a safety factor of 1.8 in case of reductionin the quality factor (Q) of the cavity. Figure 25 . High power test setup at the end of the RF transmission line.
The RF system is protected by predefined interlock conditions (Fig.26). On the power supplyside, the temperature of the SSA is continuously monitored, as well as the availability of the– 17 –ater cooling. For the tetrode amplifier, both water and air cooling requirements are imposed foractivating the second stage. Tube’s anode, G1 and G2 voltages and currents are monitored andin case of unwanted level change the fast-interlocks (100 ns) protect the system. Additionally, thePSU finite state machine requires the "ON" signal from the independent measurement of the waterflow rate (should be higher than 15 l/m) and from the circulator biasing electromagnets. In caseof any failure, the PSU initiates an automated shutdown procedure. The cooling water pressureand temperature are also continuously displayed and saved to a file. The reflected RF power iscontinuously monitored with a directional coupler just before the RFQ to provide a fast shutdownsignal to the RF PSU in case of a value above a user defined threshold.
Figure 26 . Fast-interlock diagram of the RF system.
The RF power supply and the RF transmission line for the SPP proton beamline are designed, builtand installed as a combined effort between the SPP team and local manufacturers. The measurementresults (power and S-parameters, etc.) on the RF transmission line correspond well to the designexpectations.The educational and training outcomes are immense since the accelerator physicists and engi-neers were able to design, to help the manufacturers to produce their design, and finally to comparethe finished product to the concept they had initially. Moreover, the manufacturers understand theexpectations concerning the level of precision and details for accelerator component production.And finally the fact that, in Turkey, there is now a set of physicists, engineers and manufacturerswho can speak the same language, work on a complex product from its design to its delivery couldbe a big regional asset.
Acknowledgments
The authors are grateful to A. Tanrikut for useful comments. This project is supported by theTAEK, under project No. A4.H4.P1. – 18 – eferences [1] G. Turemen et al.,
Project PROMETHEUS: Design and Construction of a Radio FrequencyQuadrupole at TAEK , Proceedings of the International Beam Instrumentation Conference 2013(IBIC’13) (2013), Oxford, UK.[2] G. Turemen et al.,
SPP beamline design and beam dynamics , Proceedings of the InternationalParticle Accelerator Conference 2014 (IPAC’14) (2013), Dresden, Germany.[3] H. Yildiz et al.,
Compact Measurement Station for Low Energy Proton Beams , Journal ofInstrumentation vol (2017), DOI:10.1088/1748-0221/12/02/T02006.[4] I.M. Kapchinsky, V.A. Teplyakov, A Linear Ion Accelerator with Spatially Uniform Hard Focusing ,ITEP-673 (1969).[5] T. Wangler,
RF Linear Accelerators , 2nd edition, Wiley Books (2008).[6] EPROM Electronic Project & Microwave Ind. and Trade Ltd. Co. Ankara, Turkey.[7] NXP Semiconductors Elektronik Ticaret, Istanbul, Turkey;
Field Structures of Waveguide Junction Circulators with Irregular ShapeFerrite Simulated Based on Exact Treatment , JPIER vol (2006) 33-54.[16] Henk Bosma, A General Model for Junction Circulators; Choice of Magnetization and Bias Field , IEEE Transactions on Magnetics