(2) across the load through a step-up nano-crystalline core based transformer with 𝑛 turns ratio. Accordingly, the generated pulses have a peak voltage of ±𝑛𝑁𝑉𝑠 as a result of the double magnification. The validity of the proposed topology is assessed via Matlab/Simulink simulations and scaled-down experimentation. The obtained results validate the proposed topology for water treatment applications. The proposed PG topology is introduced in section II, along with its operation principle. Simulation and experimental results are present in sections III and IV, respectively. Finally, section V outlines the conclusions. SM2 Vs SMN Arm 2 SM1 Arm 1 i Nano-Crystalline Transformer r Vt - SM1 SM2 SMN + 1:n II. PROPOSED TOPOLOGY PRINCIPLE OF OPERATION L R S1 - VL S2 + Tx Vs+- Cc SM A VAB Tm B SM Switching States Tm ON OFF OFF Tx OFF ON OFF State Bypassed Inserted Idled Fig. 1. Proposed PG topology. Ts V tpp tpc tnp II III tnc Vpp I IV The SMs of Arm2 are charged sequentially and the load is subjected to Null Voltage (NV) Negative pulse generation via Arm2 SMs The SMs of Arm1 are charged sequentially and the load is subjected to Null Voltage (NV) t Positive pulse generation via Arm1 SMs The proposed PG topology is shown in Fig. 1. The upper arms, Arm1 and Arm2, are formed of 𝑁 series-connected HBMMC SMs which are charged sequentially from the LVDC supply 𝑉𝑠 through an 𝑟𝐿 branch via reverse blocking switches 𝑆1 and 𝑆2 respectively. During charging of a particular arm SM capacitors, the other arm SMs are idled. The SM capacitors of the charging arm are charged sequentially by bypassing the other series SMs in the arm while inserting the charging SM capacitor. The operating sequence of the topology to generate bipolar rectangular pulses, shown in Fig. 2, is illustrated in Table I. The pulse generation is formed of four stages, consecutively, the positive pulse, the positive SM capacitors charging, the negative pulse, and the negative SM capacitors charging. The formed voltage pulse (𝑉𝑡 ) across the primary winding of the transformer is stepped up by transformer turns ratio 𝑛 before being applied across the load. Hence, the generated voltage pulse (𝑉𝐿 ) is magnified by the number of 𝑁 sequentially charged SMs as well as the transformer turns ratio 𝑛 yielding to a voltage pulse of peak ±𝑛𝑁𝑉𝑠 across the load. By varying the number of MMC cells in-circuit, up to N cells, output pulse shaping is possible. Reverse blocking semi-conductor switches are required for 𝑆1 and 𝑆2 which can comprise an IGBT in series with a diode, as shown in Fig. 1. This is because during pulse generation, 𝑆1 and 𝑆2 arms are OFF and a reverse voltage of (𝑁 − 1)𝑉𝑠 is applied across them. The rating of the IGBT switch is 𝑉𝑠 , and since an LVDC input source is used, a single IGBT should be sufficient. Additionally, Zero Voltage Switching (ZVS) is assured during IGBT turn ON/OFF, thus, series connection of IGBTs (if required) or the diode(s) should not present a sharing issue. The nano-crystalline core material is preferred (over ferrite) for high-frequency operation due to its high core permeability, hence high magnetizing inductance, high flux density, and near square hysteresis loop . The reduced transformer volume, due to high-frequency operation, enhances the modularity of the proposed topology. Bipolar rectangular voltage pulses are advantageous over unipolar rectangular pulses in terms of applying mechanical stresses to the sample under treatment in addition to electrical stresses . Using a transformer in the proposed PG supports bipolar voltage pulse generation, however, two aspects should be considered namely: the leakage inductance of the transformer and the voltage-second balance. Vnp Fig. 2. Generated bipolar rectangular voltage pulse across the transformer primary.
(3) TABLE I SEQUENCE OF OPERATION AND CIRCUIT CONFIGURATION DURING THE GENERATION OF A COMPLETE BIPOLAR PULSE CYCLE - vt + vt L + 0 vt Vs S2 Tm Req r - + vt L S1 + -Cc Req r - - Tx Vs NVs Arm2 Arm1 Arm2 Vs L S2 IV + Cc /N Req r + Arm1 Arm2 Vs Req III Tm - 0 L S2 Vs -Cc Negative arm SMs charging S1 NVs + Negative Pulse S2 Vs Arm1 Tx II + r Positive arm SMs charging S1 Cc /N Arm2 Arm1 I S1 Circuit configuration Positive Pulse S1 is ON (while charging SM S1 and S2 switches are OFF. S1 is OFF and S2 is ON (while capacitors of Arm1) and S2 is Arm1 SMs are bypassed to charging SM capacitors of OFF. Arm2). provide a path for Arm2 Arm1 SM capacitors are discharging SM capacitors. Arm2 SM capacitors are inserted sequentially to re- Negative pulse 𝑉𝑛𝑝 is formed inserted sequentially to recharge Arm1 SM capacitors, charge the individual Arm2 SM across the transformer primary each to Vs. capacitors to Vs. for a duration of 𝑡𝑛𝑝 . All Arm2 SMs are idle*. All Arm1 SMs are idle*. To charge SMs of Arm1, the To allow charging the SMs of first charging SM is inserted, Arm2, the first charging SM is then S1 is turned ON (hence, it inserted, then S2 is turned ON has ZVS). (hence, it has ZVS). After charging the last SM, S1 After charging the last SM, S2 is turned OFF safely since the turns OFF safely since the charging current is zero. charging current is zero. During Arm1 SMs charging During the Arm2 SMs charging duration 𝑡𝑝𝑐 the transformer duration 𝑡𝑛𝑐 the transformer primary voltage is nullified. primary voltage is nullified. *The idled arms eventually act as an open circuit since no current passes, as represented in the circuits outline. Sequence of operation S1 and S2 switches are OFF. Arm2 SMs are bypassed to provide a path for Arm1 discharging SM capacitors. A Positive pulse of voltage peak 𝑉𝑝𝑝 is formed across the transformer primary for a duration of 𝑡𝑝𝑝 . The transformer leakage inductance limits the generated duration of the pulses, thus, the leakage should be measured in order to determine the allowable pulse duration range. Voltagesecond balance is assured for symmetrical bipolar pulses, however, asymmetrical pulses must maintain the voltagesecond balance (otherwise, the transformer core will accumulate flux and saturate). Thus, the following equation should be applied to determine the suitable pulse polarity magnitude and duration while assuring transformer voltagesecond balance 𝑉𝑝𝑝 𝑡𝑝𝑝 = 𝑉𝑛𝑝 𝑡𝑛𝑝 (1) where, 𝑉𝑝𝑝 and 𝑉𝑛𝑝 are the peak of the positive and negative pulse polarities while 𝑡𝑝𝑝 and 𝑡𝑛𝑝 are the corresponding pulse polarity durations, respectively (assuming rectangular pulses). Effectively, the generated pulses should be in the kilo-volt magnitude range (1-100 kV) with pulse durations between nanoseconds and milliseconds ; generally decreasing in duration as voltage increases. The water sample under treatment is modelled as an 𝑅 load when considering pulse durations of micro-seconds and above . III. selection of 𝑟𝐿𝐶𝑐 values is made such that the capacitor voltage has a smaller drop  while the input charging current has an 4𝐿 underdamped response, that is 𝐶𝑐 < 2 . Accordingly, the SM 𝑟 capacitor size is calculated from , 𝐶𝑐 = (2) where 𝛽 is the percentage remaining voltage on the SM capacitor after pulse generation and 𝑡𝑝𝑙 is the longest pulse polarity duration (i.e. the longest among 𝑡𝑝𝑝 and 𝑡𝑛𝑝 ). While the charging current is expressed as 𝛽𝑉𝑠 𝑖(𝑡) = 𝑟 𝐿 𝑟2 − ) 𝐶𝑐 4 𝑒 −2𝐿𝑡 𝑠𝑖𝑛√ (√ 1 𝑟2 − 2𝑡 𝐿𝐶𝑐 4𝐿 (3) After re-charging the SM capacitor, the charging current is reduced to zero, 𝑖(𝑡𝑐 ) = 0, solving (3) for the charging time 𝑡𝑐 yields SIMULATION RESULTS The charging of the individual SM capacitors through the 𝑟𝐿 branch is based on a slightly underdamped response of the 𝑟𝐿𝐶𝑐 circuit, hence, the SM capacitors have fast charging. Thus, the 2𝑁𝑡𝑝𝑙 (1 − 𝛽 2 )𝑅 𝑡𝑐 = 𝜋 1 𝑟2 √ − 𝐿𝐶𝑐 4𝐿2 (4)
(4) The SM charging time, 𝑡𝑐 , is the control variable to determine the maximum possible repetition time provided by the PG, where 𝑇𝑠 = 2𝑁𝑡𝑐 + 𝑡𝑝𝑝 + 𝑡𝑛𝑝 (5) Voltage, kV Parameter Simulation LVDC input voltage Input inductance Number of SMs/arm Transformer turns ratio Load resistance SM capacitance SM charging time Percent remaining voltage 10 Primary voltage Load voltage 5 TABLE II SIMULATION AND EXPERIMENTAL SPECIFICATIONS 0 𝑉𝑠 𝑟𝐿 𝑁 𝑛 𝑅 𝐶𝑐 𝑡𝑐 𝛽 500 V 0.1Ω and 2µH 5 4 2 kΩ 10 µF 14 µs Experimental 30 V 0.5Ω and 5µH 3 3 1 kΩ 15 µF 30 µs > 0.95 -5 5 -10 0 0.5 1 Cycles 1.5 2 0 Voltage, kV (a) 510 505 -5 Voltage, V 500 495 -10 490 0 0.5 1 Cycles 1.5 2 1.5 2 (a) 485 10 480 0 0.5 1 Cycles 1.5 2 8 6 (b) 4 Voltage, kV 510 505 Voltage, V 500 2 0 -2 -4 -6 495 -8 490 -10 0 485 480 0 0.5 1 Cycles 1.5 2 (c) 20 0.5 1 cycles (b) Fig. 4. Generation of different bipolar pulse shapes whilst ensuring the transformer voltage-second balance constraint. (a) Combined null-load voltage duration bipolar pulses of 10µs pulse durations and 10 kV peak. (b) Asymmetric bipolar pulses of 10µs positive-pulse duration and 4 kV peak and 5µs negative-pulse duration with 8 kV peak. Current, A 15 10 5 0 -5 0 0.5 1 Cycles 1.5 2 (d) Fig. 3. Generation of 10 kV peak, 10µs bipolar pulses. (a) Voltage pulses across the transformer primary and the load. (b) Five SM capacitor voltages of the negative pulse, Arm1. (c) Five SM capacitor voltages of the positive pulse, Arm2. (d) Input charging current of the SM capacitors. Matlab/Simulink simulations are used to validate the proposed topology, with the specifications given in Table II. Bipolar rectangular pulses with positive and negative durations of 10µs, voltage pulse peak of 10kV and repetition rate of 5 kHz are shown in Fig. 3a. In Fig. 3a, the primary voltage of the transformer is 2.5kV, which is the sum of the sequentially charged five SM capacitors, while the voltage across the load is 10kV, since the transformer has 𝑛 = 4. The capacitor voltages of Arm1 and Arm2 are shown in Figs. 3b and 3c, respectively, where each capacitor fluctuates around 500 V, with a voltage ripple lesser than 5%.
(5) The current flows from the LVDC supply to charge the individual SM capacitors during the charging period is shown in Fig. 3d. It can be seen that the current drops to zero after the SM capacitor re-charged to 500 V with charging time of 14µs. The flexibility of the proposed PG is explored by generating asymmetrical bipolar pulses and combined null-load voltage durations pulses as in Fig. 4. Fig. 4a shows asymmetric bipolar pulse with a positive polarity peak of 4 kV and 10µs duration, while the negative pulse polarity has an 8 kV peak and 5µs duration. Moreover, combining the null-load voltage durations is explored in Fig. 4b with 10µs pulse durations and 10 kV peak voltage. Zoomed view Ref Ref Primary voltage Secondary voltage Time: 50 µs/div. Voltage: 100 V/div. Time: 50 µs/div. Voltage: 40 V/div. (a) (b) Ref Ref Time: 250 µs/div. Voltage: 10 V/div. Time: 50 µs/div. Voltage: 40 V/div. (c) (d) Ref Time: 50 µs/div. Current: 100 mA/div. (e) Fig. 5. Scaled-down experimental results of the proposed topology. (a) Primary and secondary voltage pulses of non-combined null-load voltage durations bipolar pulses. (b) Primary voltage pulses of non-combined null-load voltage durations bipolar pulses. (c) Primary voltage pulses of combined null-load voltage durations bipolar pulses. (d) Three individual SM capacitors voltage of Arm2. (e) Input charging current.
(6) IV. EXPERIMENTAL RESULTS Although transformer leakage inductance is ignored in the simulations, its effect is seen in the experimental validation. To minimize transformer leakage inductance, the primary turns are wound over the secondary windings. Its experimentallydeduced value from the secondary side leakage inductance 𝐿𝑙 , which will be connected across the load, is 3.56 μH. Accordingly, for load resistance of 1 kΩ, the generated pulse will require ∆𝑡 = 𝐿𝑙 ⁄𝑅 = 3.56 ns for each polarity to reach the required peak value. Thus, for proper operation, the pulse polarity duration time should be larger than ∆𝑡. Consequently, microsecond pulse durations can be generated safely, which are targeted in this paper. The primary and secondary voltages of bipolar pulses with pulse polarity duration of 10μs and repetition time of 400μs, are shown in Fig. 5a. The voltage-peak of the generated pulses across the load is 300 V, since the transformer turns ratio is 3 and the primary voltage is 100V, as in Fig. 5a. The primary voltage of bipolar pulses with non-combined as well as combined null-load voltage durations and a peak voltage of 90V, are shown in Figs. 5b and 5c, respectively. Accordingly, each of the three individual SM capacitor voltages in the two MMC arms fluctuate around 33V as shown for Arm2 SMcapacitors in Fig. 5d. Finally, the charging input current is shown in Fig. 5e for the pulses in Fig. 5a. V. CONCLUSION This paper presented a new PG topology to generate HV bipolar pulses for disinfection in water treatment applications. The proposed PG is based on HB-MMC SMs which provide modularity and scalability of the topology. The individual SM capacitors are charged sequentially through reverse blocking semiconductor switches and an 𝑟𝐿 branch from a LVDC input supply. The selection of the 𝑟𝐿 branch is such that, during the sequential charging of the SM capacitors, the charging current has an underdamped response, therefore the capacitors charge fast. A step-up nano-crystalline core based transformer, with low leakage inductance, is connected across the load for pulsevoltage magnification. The proposed topology was assessed via simulations and scaled-down experimentation, which established the viability of the topology for water treatment application. ACKNOWLEDGMENT This work was supported by the Qatar National Research Fund (a member of the Qatar Foundation) under NPRP Grant (7203-2-097). The statements made herein are solely the responsibility of the authors. REFERENCES  Water Treatment Manual: Disinfection, Office of Environmental Enforcement, Environmental Protection Agency–EPA, Wexford, Ireland, 2011.  H. Bluhm, Pulsed power system: Principles and applications: Berlin: Springer, 2006.  J. Raso and V. Heinz, Pulsed electric fields technology for the food industry: Fundamentals and applications: New York ; London : Springer, 2006.  L. Lamy Rocha, J. F. Silva, and L. M. Redondo, "Multilevel high-voltage pulse generation based on a new modular solid-state switch," IEEE Trans. Plasma Sci., vol. 42, pp. 2956-2961, Oct. 2014.  A. A. Elserougi, A. M. Massoud, and S. Ahmed, "Modular multilevel converter-based bipolar high-voltage pulse generator with sensorless capacitor voltage balancing technique," IEEE Trans. Plasma Sci., vol. 44, pp. 1187-1194, 2016.  M. A. Elgenedy, A. Darwish, S. Ahmed, and B. W. Williams, "A modular multilevel-based high-voltage pulse generator for water disinfection applications," IEEE Trans. Plasma Sci., vol. 44, pp. 2893-2900, 2016.  A. A. Elserougi, A. M. Massoud, and S. Ahmed, “A modular high-voltage pulse-generator with sequential charging for water treatment applications," IEEE Trans. Ind. Electron., vol. 63, pp. 7898-7907, 2016.  A. A. Elserougi, I. Abdelsalam, A. M. Massoud, and S. Ahmed, “A fullbridge submodule-based modular unipolar/bipolar high-voltage pulse generator with sequential charging of capacitors,” IEEE Trans. Plasma Sci., vol. PP, pp. 1-9, 2016.  B. W. Williams, Power Electronics: Devices, Drivers, Applications, and Passive Components. London, U.K.: Macmillan, 1992.  K. H. Schoenbach, S. Katsuki, R. H. Stark, E. S. Buescher, and S. J. Beebe, "Bioelectrics-new applications for pulsed power technology," IEEE Trans. Plasma Sci., vol. 30, pp. 293-300, 2002.  M. A. Elgenedy, A. Darwish, S. Ahmed, and B. W. Williams, "A transition arm modular multilevel universal pulse-waveform generator for electroporation applications," IEEE Trans. Power Electron., vol. PP, no. 99, pp. 1-1, 2017.