High efficiency pulsed power converters for the ESS accelerator Carlos Martins European Spallation Source Davide Castronovo, Roberto Visintini Elettra Sincrotrone Trieste

ESS Klystron modulator – Stacked Multi-Level (SML) topology Improved efficiency (~92%), due to minimal number of conversion stages; Good AC grid power quality (constant power absorption = flicker-free operation, sinusoidal current absorption, unitary power factor); 3.5ms pulse Magenta: HV output pulse; Yellow: LV input voltage; Green: LV input current Part I – LV power converter stage (@ 1kV) Part II – HV oil tank assembly Pulse (1kV) AC line voltage AC line current Unitary power factor; Sinusoidal line current, constant amplitude -> No flicker

SML modulator – Integration of PhotoVoltaic energy sources A yearly energy of 3.4 MW-Year could be collected from such a field.

Magnets & PC System Optimization Linac Warm Units (LWU) and M&PC “Single Pass Machine”  is “DC mode” really needed? Is “Pulsed Mode” a feasible alternative? Advantages? Disadvantages? Magnets and their associated PC are the two sides of a system Optimization involves magnets AND power converters together DC vs. Pulsed Complexity/costs vs. simplification/spares Let’s consider the Linac Warm Units of ESS. The accelerator is a single pass one. Particles are emitted at 14 Hz rate, therefore, is DC mode really needed? The consequence is to verify if a “pulsed solution” is possible and which are the constraints on the pulse shape and period. And, from this considerations analyse and compare pros and cons. Quite obvious, indeed. For the specific case, magnets and power converters are the two halves of a single system. Any optimization process – technical and economical – involve both at the same time. DC and pulsed modes act in opposition on magnets and power converters. DC mode bring some complications with the magnets, for example the water cooling and simplify the structure of the power converters. Conversely the pulsed mode could simplify the magnets (e.g. air cooling) and complicate the power converters. Optimization is a matter of weighting the complexity and costs increase on one dish with the simplification and the economical and maintenance spares on the other.

Plosses = ρ*Volcu*Jrms2 Generalities on magnet design Resistance of a magnet: R = ρ*l/S ; where: ρ is the resistivity; l is the coil average length; S is the coil cross section Inductance of a magnet: L=k*N2 ; where: k is a constant dependent on the geometry, magnetic core properties, air gap; N is the number of turns; Magnetic field in the air gap: Bg= k’*N*I ; where k’ is a constant, I is the current Voltage at the terminals of a magnet: V=L*dI/dt+R*I - DC magnet (dI/dt=0) -> V=R*I - Pulsed magnet -> V≈L*dI/dt (R*I term is often negligible) V≈k*N/(k’)*Bg/Tr, assuming linear rise of Bg where: Tr is the rise time Power losses on a magnet: Plosses = R*Irms2 = ρ*l/S*Irms2 = ρ*Volcu*Jrms2 where Volcu is the volume of copper and Jrms is the current density, Jrms=Irms/S (Jrms=1..3 Arms/mm2) Bg ~ k’*N*I V ~ N*Bg/Tr + R*I Plosses = ρ*Volcu*Jrms2 This is the list of magnets in the LWU. The quadrupole types Q5, Q6 and Q7 are the more numerous and are involved in the study while the others can be DC.

Plosses = ρ * Volcu * Jrms2 Generalities on magnet design   DC magnet Pulsed magnet DC pulsed J t k: integer Plosses = ρ * Volcu * Jrms2 In a DC magnet (Jrms= Jpeak): - Plosses = ρ * Volcu * Jpeak2 In a PULSED magnet (Jrms= √δ *Jpeak): - Plosses = ρ * Volcu * Jpeak2 * δ Where δ is the duty-cycle of the pulse - In the case of ESS, δ = 8 to 10% (to accommodate for margins in ramp time and flat top stabilization) -> The losses on the magnet for the same copper volume (Volcu, i.e. same magnet size) and same peak current density (Jpeak, i.e. same field strength and same ) are about 10 times lower if a pulsed magnet is used; -> Alternatively, if a factor of 10 is not needed and can be lowered, a PULSED magnet becomes both more compact, more efficient and water cooling free. This is the list of magnets in the LWU. The quadrupole types Q5, Q6 and Q7 are the more numerous and are involved in the study while the others can be DC.

Magnets & Power Converters (PC) Some (quite obvious) considerations: Power is always paid twice: for generating it, and removing it Keep the peak voltage on PC and magnets <1 kV  avoid MV-rated components, cables, rules,… Minimize the number of different types of PC (operation, maintenance, spares,…) Unify interfaces PC-RCS and PC-PSS+MPS(+) among PC types/families Minimize use of water cooling both on magnets and PC Plant  de-ionized water, radiation resistant rubber pipes for magnets,… Operation  integration into MPS,… Reliability  risk of leakages, clogging of coils,… Maintenance  operate close or over delicate equipment,… (+)RCS = Remote Control System; PSS = Personnel Safety System; MSS = Machine Protection System I want to share with you some – I admit, quite obvious – considerations (from our experience with Elettra and FERMI). First of all, power is always paid twice. Think to the excitation current of a warm magnet, for example. You pay once for supplying the magnet with the power converter and than you have to dissipate the heat generated in the coils. And the latter one is even worse, due to the power needed by the cooling plant (pumps, refrigerators,…). Then, 1 kV is a border line between two worlds of components, cables, rules,… To operate at Mid-Voltage is more expensive and complicated. Minimization of power supplies types and the unification of control and interlock interfaces are self-evident benefits. Finally, minimizing the water-cooling brings to consistent savings both economical – plant, maintenance – and in time – reliability, complexity of the machine protection system, and so on.

Operating Mode, Cooling LWU* Magnets and Power Converters Magnet Type Description Operating Mode, Cooling Quantity N. of PC Q5 Quadrupole magnet Pulsed, Air-cooled 26 C5 Dual-plane corrector magnet DC, Air-cooled 13 Q6 95 C6 55 110 Q7 12 D1 Vertical dipole magnet DC, Water-cooled 2 1 Q8 6 C8 4 8 This is the list of magnets in the LWU. The quadrupole types Q5, Q6 and Q7 are the more numerous and are involved in the study while the others can be DC. *LWU = Linac Warm Unit

How to merge the two quadrupole families into one? Basic parameters and specifications Parameters   Q6 Q7 unit Number of required quads n 95 12 # Maximum integrated gradient IntGMax 2.3 2.9 T Range of integrated gradient IntGrange 1.20 - 2.20 0.85 - 2.70 Minimum magnetic length Leff 230 275 mm Minimum bore diameter Ø 112 Maximum overall length LOverall 350 Due to the magnets characteristics we decided to focus on two quadrupole families, Q6 and Q7. They have the same bore diameter and overall length. In addition the integrated gradient range of Q7 covers the range of Q6. The key question is that the majority of magnets are Q6: the question that arose was how to optimise the design, merging the two families into a single magnet type or at least two types with somewhat in common (lamination geometry). Q6 and Q7 have the same dimension but different nominal ranges Q7 have a wider range of use (lower and upper values), but… …there are many more Q6 than Q7 (95 vs. 12) How to merge the two quadrupole families into one?

Possible pulse waveform The transit time of the particles is 2.8 ms every 71.4 ms a 4 ms-long flat-top is sufficient For the pulsed excitation, two waveforms have been considered: Trapezoidal, 4.5 ms rise/fall, 4 ms flat-top, 14 Hz repetition rate (71.4 ms period) Trapezoidal, 8 ms rise/fall, 4 ms flat-top, 14 Hz repetition rate (71.4 ms period) Current Pulse [4.5 – 4 – 4.5] ms Here it is the studied waveform. We extended the study also to a little more relaxed one, 8 ms of rise time instead of 4.5 ms. We will explain the reason for this later.

Magnet Cooling: Water or Air Coil dimensioning: Solid / hollow copper Maximum water cooled current density = 4 A/mm2 Maximum air cooled current density = 1.1 A/mm2 (at max Current, DC or RMS if pulsed) According to the cooling mode, the maximum current density in the coils is 4 A/mm2 with water or 1.1 A/mm2 with air. This is a comparison among four possible solutions air- or water-cooled for both cases, DC and pulsed for the Q6 and Q7.

Q6 Q7 Magnet: DC vs. Pulsed (Air Cooled) Turns per pole = 24 I@GMax = 435A I@GMax = 439 A IRMS@GMax = 137A IRMS@GMax = 138 A PMag = 0.21kW PMag = 0.27 kW LMag = 8.2 mH LMag = 10.3 mH VPK = 0.79 kV VPK = 1kV Pulsed (*) DC Turns per pole = 78 Turns per pole = 96 I@GMax = 134 A I@GMax = 137 A IRMS@GMax = 134 A IRMS@GMax = 137 A PMag = 2.62 kW PMag = 3.44 kW VMAX = 20 V VMAX = 25 V Q6 Q7 Taking into account the two most interesting solutions for both Q6 and Q7 (DC means water-cooled), we evaluate here the differences among the models. It is clear how the pulsed solution has a very low power consumption compared to the DC one. While Q6 peak voltage in the pulsed case is well below 1 kV, Q7 is very closed to this limit. In order to reduce the peak voltage, a possible solution is to relax the pulse, increasing the rise and fall times. In this way, since we are less sensitive to magnet inductance, it is possible to increase for both models the number of turns: the required gradient is obtained with a lower peak current (*) Rise time 4.5 ms, flat top time 4 ms and fall time 4.5 ms

Q7 (with plexiglass cover) Magnet: Q5 and Q7 Optimization Q6 (with iron shield) Q7 (with plexiglass cover) This is the current optimized design – almost ready for a prototype construction and test – where both families Q6 and Q7 share the same iron and coils and the difference is the iron shield on the Q6 type instead of the plexigalss one on the Q7. The current optimized design: two identical magnet types for families Q6 and Q7.

Power Converters Considerations DC Required stability: ±100 ppm vs. nominal DC Unify the PC type for both Q6 & Q7: 35 V / 150 A Unify the remote control interface FOR ALL PC (not only Q6 & Q7) Large number of PC for Q6 & Q7: 120 PC (incl. 10% spare) Stability higher than “average commercial” Custom-made solution could be economically comparable to “Commercial” AND tailored to actual needs Pulsed Required stability: ±100 ppm vs. nominal on Flat-Top Additional requirement: Output voltage and internal DC-Link voltage <1 kV Possible unification of the PC for both Q6 & Q7: 700 VPK / 400 APK Unify the remote control interface FOR ALL PC (not only Q6 & Q7) Large number of PS for Q6 & Q7: 120 PC (incl. 10% spare) Various topologies are possible and available, to be further investigated Custom-made solution is needed to meet the actual needs Let’s move our focus on the power supplies, starting with the DC case. It is possible to define a common PS type for both quadrupole types. If we consider these other aspects: the common interface, a relatively large number and stability requirements beyond what is usually available from manufactures as standard product. A custom solution could be taken into account also for DC, with the advantage of been tailored for our needs. Let’s now consider the pulsed option. In order to remain under 1 kV, we changed the pulse shape: with 8 ms of rise and fall time, the magnet design allows the reduction of the peak current to 400 A, and the peak voltage is significantly reduced as well. It is again possible/convenient to unify the PS types into one only. While the DC solution is somewhat quite straightforward, for the pulsed one various topologies are possible and this aspect need further investigations. In any case a custom solution is needed.

Conclusions ●○ Compared several design for magnets Q6 and Q7, both in DC and Pulsed DC is a standard, well-known solution: Power consumption  2.4 kW for each QC6; 3.5 kW for each QC7 Water cooling of magnets  de-ionized water plant,… “Low Power” but stable power supplies (~5 kW) Pulsed excitation is a less common solution: Power consumption  significantly more efficient than DC Air cooling of magnets  no piping, etc. but heat to environment High peak output voltage  risk of exceeding 1 kV (design & operations) Shape of the pulse is important (e.g. rise time: 8.0 ms vs. 4.5 ms) Our conclusions can be summarized as follow. Adopting the DC solution, means following a well known standard, at the cost of significantly higher power dissipation in the magnets and the need of water cooling with the drawbacks I’ve just mention. The power supplies are low power ones but require a stability higher than commercial standard. The pulsed excitation is a less common solution that brings some remarkable positive effects, in terms of power consumption and air cooling requirements. For this, one has to take into account higher peak voltages and pay attention on the shape of the pulse. Since these issues are controllable I’ve not marked them as real drawbacks.

Conclusions ●● Costs considerations for magnets: Q6 and Q7 could be the same model with and without the shields Both DC and Pulsed could have the same yoke geometry Pulsed mode yoke must be realized with Fe-Si 3-3.5%, lamination 0.35 mm Both DC and Pulsed have the same coils dimension DC coils are water-cooled The total cost of pulsed magnets are slightly lower than of DC magnets Costs considerations for power supplies: Custom PS (either DC or Pulsed) Pulsed PS are more powerful (peak) than DC, more expensive Potential savings with Pulsed Solution: No need of de-ionized water (plant installation and running) Reduced electrical power consumption (mains and dissipation) Reliability of operations and reduced maintenance Concerning the costs, we can make the following considerations. For the magnets themselves, the major cost of the yoke material and construction in the pulsed solution is counterbalanced by cheaper coils and air cooling. And vice versa for the DC, where the major cost is in the water-cooled coils than in the yoke. From the pre-design, it could be possible to unify the quadrupole types and have the same geometry for the poles and size of the coils. For the power supplies, a custom production is possible for both solutions, DC and pulsed, but – it is clear that the cost of pulsed PS is higher than the DC one. This could seem a disadvantage for the pulsed solution but widening the view range, and considering the plants behind the magnet-power supply systems, there are significant potential savings both in construction and operations. Just to mention three. There is no need of de-ionized cooling water, no plant (or reduced size plant), no pipes. The electrical power consumption is significantly reduced, and the dissipation as well. Operations benefit from increased reliability and reduced maintenance

Thank You!