Pulse generator

This patent document claims the benefit of DE 10 2006 060 417.2 filed Dec. 20, 2006, which is hereby incorporated by reference.

The present embodiments relate to generating a voltage pulse with a pulse generator.

Marx generators generate transient pulse voltages of short duration and high amplitude.

Marx generators charge a large number of capacitors in parallel with direct voltage to the value of a stage voltage and then connect the capacitors suddenly in series. As the parallel-connected capacitors are being charged, the charging currents are added, and when subsequently series-connected, the voltages are added, as required, via the capacitors.

A charging voltage is provided, for example, via a transformer and an appropriately rated rectifier. The charge voltage charges a pulse capacitance via a charging resistor and/or a charging coil. The charging process generally occurs relatively slowly over a period in the range of several seconds. The voltage at the pulse capacitor follows an e-function and reaches its quasi-stationary final value after approximately 5τ. A spark gap is adjusted such that at the quasi-stationary final voltage value, the spark gap fails to break down. The spark gap can be brought to flashover after an arbitrarily selectable time period by suitable ignition devices. At this moment, the spark gap constitutes a closed switch. The temporal course of the pulse voltage emitted is determined via the component values of the generator and via the electrical values of the load.

With a single-stage pulse circuit, which may be called a cell, no higher voltage than the charging voltage can be achieved at the load. For the pulse voltages needed, a multi-stage arrangement of several cells in a Marx circuit is used in high-voltage installations.

DE 103 20 425 A1 discloses a Marx generator. The Marx generator has a number n of pulse capacitors. The pulse capacitors may be called stage capacitors. DE 103 20 425 A1 discloses a trigger/ignition device in the Marx generator including n stage capacitors (n is natural and greater than 1), an equal number of switches/spark gaps, and 2(n−1) charging branches. The spark gaps of the Marx generator operate in self-breakdown mode. The trigger/ignition device includes at least one pulse transformer connected to a pulse generator. In at least one of the charging branches of the Marx generator, an overvoltage, which is sufficient for self-breakdown, is generated in a time-determined manner for a short period at the adjacent spark gap. The charging branch with the associated stage capacitor bridges the spark gap, except at the output end. The output winding of the pulse transformer acts during the charging as a charging coil/inductivity. The input winding is connected to the pulse generator. The voltage pulse, which is generated with the pulse transformer upon ignition/triggering of the pulse generator, is added to the charge voltage of the associated stage capacitor and generates for a short period the overvoltage necessary for self-breakdown of the adjacent spark gap.

All n stage capacitors of the cells are charged simultaneously via the charging direct voltage. Coils can limit the charge current. The sparking distances of the spark gaps are chosen such that the paths fail to break down when the maximum charge voltage is reached. When all the pulse capacitors have been charged to their quasi-stationary final voltage value, ignition at the first spark gap takes place. At the next spark gap, a voltage that is now twice as high, namely twice the charge voltage, is applied so that this will also ignite. Within an extremely short time, all the spark gaps ignite in this manner, such that n times the stage voltage declines at the load.

The present embodiments may obviate one or more of the drawbacks or limitations of the related art. For example, one embodiment relates to a method for triggering a pulse generator.

In one embodiment, a method for controlling a pulse generator with multiple cells is provided. The pulse generator may be a Marx generator or other pulse generator with differently connected cells. The pulse generator generates a pulse having a power greater than one megawatt.

Each cell includes a controllable semiconductor switch and a capacitor for generating a voltage pulse through series connection and/or parallel connection of the capacitors by the semiconductor switches. In a Marx generator the capacitors may be stage capacitors. For example, the semiconductor switches may be IGBTs, DMOS-FETs, GTOs, thyristors or other similar switches. The semiconductor switches may be switched off. Each cell may include one or more semiconductor switches.

The capacitors may be connected in series, parallel, or a combination thereof. In the case of a series connection, the capacitor voltages may be added. In the case of a parallel connection, the discharge currents may be added. For the parallel connection, for example, all the discharge currents may flow out of the capacitors into a current node. The discharge currents may be added by adding the magnetic flow density generated by the discharge currents in a pulse transformer.

In one embodiment, the method can, for example, be embodied as software in a wired program routine (ROM) or in a programmable memory (EEPROM).

The method may include measurement of one switch-on time difference for each cell respectively. A time meter, which may be a timer, can measure time. Real time meters have a temporal resolution. The switch-on time difference for each cell is measured between a switch-on signal and a system response. The switch-on signal forms the start point of the time measurement and the arrival of the system response forms the end point of the time measurement. The ‘system response’ may be a pulse response, weight function, or step response. The system response is the output signal of a one-dimensional time-invariant system. A pulse or step is fed, as a switch-on signal, to the input of the one-dimensional time-invariant system.

The switch-on signal switches the respective semiconductor switch of the corresponding cell to the conducting state. The switch-on signal is, for example, a rising edge. Alternatively, a falling edge could also be used as a switch-on signal.

The system response is dependent on the switching to the conducting state. The conducting semiconductor switch connects the charge voltage of the capacitor to the output. At the output, for example, a significant course of the output voltage can be evaluated as a system response.

The semiconductor switches of the cells for the voltage pulse may be controlled. The controlling of the semiconductor switches may be a function of the switch-on time differences. The switch-on signal is generated for the respective cell in such a time-shifted manner that the system response of each cell occurs simultaneously.

Simultaneous is understood here to be a match that is achievable within the framework, for example, of a temporal resolution of the measurement and a temporal resolution of the control.

In one embodiment, a switch-off time difference may be measured. For example, measurement of one switch-off time difference for each cell, respectively, is provided. The switch-off time difference is measured between a switch-off signal for switching the respective semiconductor switch of the cell to the blocking state and a system response dependent on the switching to the blocking state. The switch-off signal is, for example, a falling edge. Alternatively, a pulse on a reset line may be used. The switch-off time difference of a cell may be different from the switch-on time difference. If a voltage pulse with a defined length is generated, the switch-off time differences or a variable derived from the switch-off time differences may be stored.

In one embodiment, the semiconductor switches are controlled. For example, the controlling of the semiconductor switches of the cells for the voltage pulse is a function of the switch-off time differences. The switch-off signal of the respective cell is generated in such a time-shifted manner that the system response of each cell occurs simultaneously. The system response may be dependent on the switching to the blocking state.

In one embodiment, the modules including respectively multiple series-connected cells are triggered for generating the voltage pulse. The modules may be coupled such that the pulse power of the modules is added.

A module-switch-on time difference for each module may be measured. The module-switch-on time difference is measured between a switch-on time for generating a current pulse of the module and a system response dependent on the current pulse. The system response dependent on the current pulse is dependent on an impedance acting at the output of each module. Electrical characteristics of the cells of the module are a factor influencing the system response dependent on the current pulse. The system response dependent on the current pulse can, for example, be determined from the course of the measured output current of the module. The module-switch-on time difference may be measured, for example, by a time meter (timer). The time meter is started at the switch-on time and stopped when the system response dependent on the current pulse is received.

The switch-on time may be determined from the switch-on differences of the individual cells of the module. The switch-on time may be defined in a mean range between the longest runtime of a cell of the module and the shortest runtime of a cell.

The control of the modules for the voltage pulse is a function of the module-switch-on time differences. The switch-on time of the respective module is generated in such a time-shifted manner that the system response of each module occurs simultaneously. The system response is dependent on the current pulse.

In one embodiment, the switch-on time differences and/or the switch-off time differences and/or module-switch-on time differences and/or the switch-on times and/or variables dependent hereon are stored in a table in a non-volatile memory. A non-volatile memory is, for example, an EEPROM. Changes in the values may be stored, so that, for example, a history can be read out from the memory for diagnostic purposes.

Electrical characteristics of the cells can change over the lifetime and via influences such as, for example, temperature fluctuations. The measuring may be carried out cyclically or in a predetermined order, such as during a system warm-up sequence.

In one embodiment, a change of electrical characteristics of a component is determined from the measurement of the system response. A component is any possible structural element, for example, of one of the cells or of the overall system of the pulse generator. Diagnostic information may be output as a function of this change. The diagnostic information enables maintenance of the system and reduces possible downtimes.

In one embodiment, a system generates a voltage pulse with a Marx generator.

The system generates a voltage pulse with a pulse generator. The system includes multiple cells in series connection and/or parallel connection. Each cell has a capacitor and a semiconductor switch. A driver may be integrated for controlling the semiconductor switches.

A control circuit is connected to each cell for controlling the semiconductor switch. The connection may be direct or indirect, for example, via drivers and/or a transmission path.

The control circuit may execute the functions stated below. The control circuit includes, for example, a programmable microcontroller for executing functions. The programmable microcontroller may include integrated subcircuits for executing the functions stated below.

The control circuit may generate a switch-on signal for switching each individual semiconductor switch to a conducting state. For example, switchable power components, such as IGBTs, DMOS-FETs or similar devices, can be used as semiconductor switches. In the conducting state, these have a low switch-on resistance so the stage capacitors are electrically connected via the switched-on semiconductor switch. Depending on the security requirements, the switch-on signal can include, for example, a simple edge or, for a secure transmission, of a binary code.

The control circuit may measure one system response for each individual cell, respectively. The system response is dependent on the switching of the semiconductor switch to the conducting state. For measuring, an analog-digital converter or a threshold switch may be connected to an output for measuring voltage.

The control circuit may measure a time difference between the generation of the switch-on signal and the system response for each cell, respectively. For the time measurement, a time meter (timer) may be started as a function of the switch-on signal and stopped as a function of the system response.

The control circuit may provide time-shifted control of the semiconductor switches of the cells for a series connection and/or for a parallel connection of the capacitors for the voltage pulse. The time-shifted control may be a function of the respective time difference, such that the system response of each cell is simultaneous.

In one embodiment, a plurality of modules may include multiple series-connected cells. The outputs of the modules are coupled for generating the voltage pulse. The coupling may be an electrical connection and/or a capacitive coupling and/or a magnetic coupling by a pulse transformer. As a result of the coupling, the pulse power of the modules may be added. The coupling may produce a series connection and/or a parallel connection of the modules.

In one embodiment, the control circuit may individually control each module. The control circuit may be connected to each module via an interface. For high-speed signal transmission, the interface is a parallel interface. Alternatively, for secure signal transmission, a serial interface or a data bus is used.

In one embodiment, the pulse generator may be a Marx generator. The cells may be connected in a Marx circuit. The capacitors of each Marx cell may be stage capacitors.

A previously explained method or a previously explained system for a magnetron or a klystron may be used, for example, for medical applications. For medical applications, a pulse should have a shape that approximates, as closely as possible, the ideal rectangle. Only the duration of the plateau of the rectangular signal may be used for medical applications. Rising courses, overshoots and falling courses of the voltage pulse may represent losses. Only during the plateau of the rectangular voltage pulse can an electron beam be generated in a stable manner, for example, for a klystron, and used for medical purposes. An adjustability of the temporal plateau width enables the selective generation of a high-frequency wave for the treatment of cancer. The wave characteristics may be adapted via the characteristics of the voltage pulse to the particular type of cancer.

The features of the method, the system and developments and embodiments thereof stated previously may be combined with one another and/or stand-alone. The exemplary embodiment illustrated in the figures illustrates by way of example a possible, but not conclusive, combination of features.

FIG. 1 shows one embodiment of a block diagram of a module including multiple cells in a Marx circuit,

FIG. 2 shows one embodiment of a block diagram of a system including three modules, a control circuit, and a pulse transformer,

FIG. 3 shows a flow diagram of one embodiment of a method for generating a pulse voltage, and

FIG. 4 shows diagrams including switching signals and system responses.

In one embodiment, as shown in FIG. 1 , a module M 1 may include multiple cells Z 11 , Z 12 , Z 13 , Z 14 and Z 15 of a Marx generator. Each cell Z 11 , Z 12 , Z 13 , Z 14 and Z 15 of the Marx generator has a stage capacitor C 1 , C 2 , C 3 , C 4 and C 5 . The module M 1 has a first input I 1 + and a second input I 1 − for applying a charge voltage. When the charge voltage is applied, a charge current flows which charges the stage capacitors C 1 , C 2 , C 3 , C 4 and C 5 . The diodes D 1 , D 2 , D 3 , D 4 , D 5 , D 10 , D 20 , D 30 , D 40 and D 50 connect the stage capacitors C 1 , C 2 , C 3 , C 4 and C 5 to one another in parallel for the charging. The charge current for the stage capacitor C 1 of the cell Z 11 flows through the stage capacitor C 1 , the diodes D 10 , D 20 , D 30 , D 40 and D 50 and through the charging inductor L to the negative power input I 1 −. The charge current for the stage capacitor C 2 of the cell Z 12 flows through the diode D 2 , the stage capacitor C 2 , the diodes D 20 , D 30 , D 40 and D 50 and through the charging inductor L to the negative power input I 1 −. The corresponding charge flow for the stage capacitors C 3 , C 4 and C 5 is illustrated in the same way from FIG. 1 .

Each cell Z 11 , Z 12 , Z 13 , Z 14 and Z 15 has one semiconductor switch S 1 , S 2 , S 3 , S 4 and S 5 , respectively. The semiconductor switches S 1 , S 2 , S 3 , S 4 and S 5 are connected to the stage capacitors C 1 , C 2 , C 3 , C 4 and C 5 and outputs O 1 −, O 1 + of the module M 1 such that, when the semiconductor switches S 1 , S 2 , S 3 , S 4 and S 5 are switched jointly to the conducting state, the stage capacitors C 1 , C 2 , C 3 , C 4 and C 5 of the module M 1 are connected in series. The series connection of the stage capacitors C 1 , C 2 , C 3 , C 4 and C 5 may be connected to the outputs O 1 −, O 1 + of the module M 1 .

The charging process may be complete when all the stage capacitors C 1 to C 5 of the cells Z 11 to Z 15 have the desired charge. When the charging process is complete, the semiconductor switches S 1 , S 2 , S 3 , S 4 and S 5 may jointly be switched at a desired time to the conducting state so that a voltage pulse of, for example, 5 μs is emitted as an output voltage UA 1 at the outputs O 1 −, O 1 +. The peak voltage value may correspond to the sum of the individual capacitor voltages in the charged state.

An individual stage capacitor C 1 , C 2 , C 3 , C 4 or C 5 of the module M 1 or some of the stage capacitors C 1 , C 2 , C 3 , C 4 and/or C 5 of the module M 1 may be connected to the outputs O 1 −, O 1 +. For example, to connect the stage capacitor C 1 of the first cell Z 11 , the semiconductor switch S 1 is switched to the conducting state while the other semiconductor switches S 2 , S 3 , S 4 and S 5 remain in the blocking state. Stage capacitor C 1 is connected by the switching of the first semiconductor switch S 1 via the diodes D 2 , D 3 , D 4 and D 5 to the positive output O 1 + and via the semiconductor switch S 1 to the negative output O 1 −.

The semiconductor switches S 1 , S 2 , S 3 , S 4 and S 5 include, respectively, a control input Is 11 , Is 12 , Is 13 , Is 14 and Is 15 . Depending on the signal applied to this control input Is 11 , Is 12 , Is 13 , Is 14 and Is 15 , the respective semiconductor switch S 1 , S 2 , S 3 , S 4 and S 5 is switched to the conducting or to the blocking state. For triggering, these control inputs Is 11 , Is 12 , Is 13 , Is 14 and Is 15 can be connected to a driver with fiber-optic isolation. One or more drivers may be integrated in the module M 1 .

In one embodiment, as shown in FIG. 1 , the semiconductor switches S 1 , S 2 , S 3 , S 4 , S 5 are IGBTs. Alternatively, other semiconductor switches, such as thyristors, GTOs or DMOS field-effect transistors, can be used as semiconductor switches. The module M 1 may include a current-measurement device A and a voltage measurement device V, which may be connected to the outputs O 1 −, O 1 +. The number (five) of cells Z 11 to Z 15 shown in FIG. 1 is only an example. According to the principle shown in FIG. 1 , two or more cells may be connected.

In one embodiment, as shown in FIG. 2 , three modules M 1 , M 2 and M 3 include five cells Z 11 , Z 12 , Z 13 , Z 14 , Z 15 , Z 21 , Z 22 , Z 23 , Z 24 , Z 25 , Z 31 , Z 32 , Z 33 , Z 34 , Z 35 , respectively, of a Marx generator. The module M 1 , shown in FIG. 2 , corresponds to the module M 1 shown in greater detail in FIG. 1 . The structure of modules M 2 and M 3 corresponds to the structure of module M 1 . The inputs I 1 +, I 1 −, I 2 +, I 2 −, I 3 +, I 3 − of the modules M 1 , M 2 and M 3 can be connected to a supply voltage for charging the stage capacitors. The output voltages UA 1 , UA 2 and UA 3 of the modules M 1 , M 2 , M 3 act upon windings W 1 , W 2 and W 3 of a pulse transformer. Each module M 1 , M 2 , M 3 includes a current-measuring output OI 1 , OI 2 and OI 3 and a voltage-measuring output OU 1 , OU 2 , OU 3 . Measuring outputs and control inputs of the modules M 1 , M 2 , M 3 are connected, respectively, via an interface Int 1 , Int 2 and Int 3 to a control circuit 1 . The interfaces Int 1 , Int 2 , Int 3 may be bidirectional interfaces. The number (three) of modules M 1 to M 3 in FIG. 2 is only an example. One or more modules may be used.

FIG. 3 shows a method that may be implemented using an electric circuit in a Marx configuration, as shown, for example, in FIG. 1 or 2 . The temporal distance between the transition of two semiconductor switches from the blocking to the conducting state and/or from the conducting to the blocking state may be decreased. The semiconductor switches S 1 to S 5 are individually controlled by a switch-on signal and a switch-off signal, respectively. The semiconductor switches S 1 to S 5 may receive the switch-on signal and the switch-off signal at different times.

A switch-on time difference may be measured for each cell Z 11 to Z 15 (Z 21 to Z 25 , Z 31 to Z 35 ), respectively. The switch-on time difference is measured between a switch-on signal for switching the respective semiconductor switch of the cell Z 11 to Z 15 (Z 21 to Z 25 , Z 31 to Z 35 ) to the conducting state and a system response dependent on the switching to the conducting state. The semiconductor switches of the cells Z 11 to Z 15 (Z 21 to Z 25 , Z 31 to Z 35 ) may be controlled for the voltage pulse as a function of each switch-on time difference by the measurement of the individual switch-on time differences. The switch-on signal to the respective cell Z 11 to Z 15 (Z 21 to Z 25 , Z 31 to Z 35 ) is generated in such a time-shifted manner that a system response of each cell Z 11 to Z 15 (Z 21 to Z 25 , Z 31 to Z 35 ) is simultaneous. The system response being dependent on the switching to the conducting state. This applies analogously to switch-off time differences from the conducting to the blocking state with the associated system response.

FIG. 3 represents a part of such a method schematically as a flow diagram. In act 1 , the stage capacitors C 1 to C 5 of the module M 1 are charged. In act 2 , a check is carried out as to whether the charging is complete. If the stage capacitors C 1 to C 5 are not yet adequately charged, the check from act 2 is carried out again. Otherwise, in act 3 the first cell Z 11 is ignited, a time meter (timer) started and a cell counter incremented. The semiconductor switches S 2 to S 5 of the remaining cells Z 12 to Z 15 are in the blocking state. For ignition, a switch-on signal with a rising edge is applied to the semiconductor switch S 1 of the cell Z 11 . The time of the rising edge starts the high-resolution timer (time meter). This timer has a resolution of less than or equal to 5 ns.

In act 4 , a check is carried out as to whether a system response has been received. The system response has a current (I) and/or voltage course (U) in relation to time (t). The current (I) and/or voltage course (U) may be measured at the outputs O 1 +, O 1 − of the module M 1 . If a system response is detected in act 4 , the time that lapses between the edge and the arrival of the system response is measured. The system response stops the timer for this purpose.

In act 5 , a table is opened in a memory of the control circuit 1 and the value of the switch-on time difference and/or the switch-off time difference Δtn for the cell 11 is written. In act 6 , the next cell ( 12 ) is ignited, the time meter (timer) started and the cell counter incremented.

In act 7 , a check is carried out as to whether all the cells Z 11 to Z 15 of the module M 1 have been measured. If not all of the cells of module M 1 been measured, act 4 is carried out again for the cells not measured. When all the runtimes of the cells Z 11 to Z 15 have been measured, after act 7 , the measurement is carried out in act 8 for the cells Z 21 to Z 25 of module M 2 . If a faulty behavior of the system is detected in at 9 , a service advice is output in act 10 . In act 11 , the measuring program is closed and the system operates in a secure mode or shut down completely.

FIG. 4 shows diagrams relating to the measurement acts in FIG. 3 and to the corresponding control. The left-hand section of FIG. 4 displays the measurement results for the individual cells Z 11 to Z 15 .

At time tE 1 the positive edge of the switch-on signal SF 1 (the solid line in the diagram in FIG. 4 ) for the cell Z 11 prompts its ignition. The semiconductor switches S 2 to S 5 (of the other cells Z 12 to Z 15 ) remain in the blocking state. The system response (dashed line) corresponds to a rise in the output voltage UAZ 1 . In the first cell Z 11 , the time difference Δt 1 between the positive edge SF 1 of the switch-on signal and the positive edge of the output voltage UAZ 1 is 30 ns. In the exemplary embodiment shown in FIG. 4 , the system response UAZ 1 is determined from the output voltage course. Alternatively, the power course or the current course may be used to determine the system response.

The time difference Δt 2 for the second cell Z 12 may be determined. The time difference Δt 2 may be the time between the edge SF 2 at time tE 2 and the output voltage UAZ 2 for the second cell Z 12 . In the exemplary embodiment shown in FIG. 4 , the time difference Δt 2 is 100 ns.

The time difference Δt 3 for the third cell Z 13 may be determined. The time difference Δt 3 may be the time between the edge SF 3 at time tE 3 and the output voltage UAZ 3 for the third cell Z 13 . In the exemplary embodiment shown in FIG. 4 , the time difference Δt 3 is 150 ns.

The time difference Δt 4 for the fourth cell Z 14 may be determined. The time difference Δt 4 may be the time between the edge SF 4 at time tE 4 and the output voltage UAZ 4 for the fourth cell Z 14 . In the exemplary embodiment shown in FIG. 4 , the time difference Δt 4 is 80 ns.

The time difference Δt 5 for the fifth cell Z 15 may be determined. The time difference Δt 5 may be the time between the edge SF 5 at time tE 5 and the output voltage UAZ 5 for the fifth cell Z 15 . In the exemplary embodiment shown in FIG. 4 , the time difference Δt 5 is 50 ns.

The right-hand section of FIG. 4 displays the control corresponding to the measurement results. A reference time tM 1 is calculated from the time differences Δt 1 to Δt 5 . The reference time tM 1 may be a mean value calculated from the time differences Δt 1 to Δt 5 . In the example shown in FIG. 4 , this is an average runtime of 82 ns. In an alternative embodiment, the mean between the shortest and the longest runtime could be used as a reference time. For example, the time duration between reference time and system response would be 90 ns (not shown). The determined reference time tM 1 may be the optimum time as it lies between the shortest and the longest runtime. At the reference time tM 1 , the generation of the respective edge SF 1 ′ to SF 5 ′ of the switch-on signal for the respective cell Z 11 to Z 15 is time-shifted forward or backward to the times tES 1 , tES 2 , tES 3 , tES 4 and tES 5 (as times of generation of the edge of the switch-on signal).

The respective runtime measured between the respective edge of the switch-on signal SF 1 to SF 5 and the respective system response UAZ 1 to UAZ 5 may be evaluated, such that the system response UA of each cell Z 11 to Z 15 occurs simultaneously and causes an output voltage UA to rise for each cell simultaneously. The output voltages UA are added in accordance with the series connection of the stage capacitors C 1 to C 5 and generate the desired voltage pulse (not shown).

FIG. 4 shows the switch-on time differences between switch-on signal and system response. The switch-off time differences between a switch-off signal and the system response may differ from the switch-on time differences. It is possible to proceed analogously for the switch-off time differences. The switch-off signals (falling edge) for the cells Z 11 to Z 15 are time-shifted and filed in a table, such that the cells as a result switch off simultaneously. The semiconductor switches S 1 to S 5 block simultaneously.

All other program parts use this table of switch-on time differences and switch-off time differences in a subroutine for a simultaneous system response of all the cells Z 11 to Z 15 . This applies analogously to the identically structured modules M 2 and M 3 comprising the cells Z 21 to Z 25 and Z 31 to Z 35 , as shown in FIG. 2 .

In one embodiment, asymmetrical voltage distribution to the series-connected semiconductor switches S 1 to S 5 may be reduced or even avoided completely. For example, the semiconductor switches, such as IGBTs, have different parameters that relate to the runtime. The simultaneous system response of all the cells Z 11 to Z 15 may prevent the voltage pulse forming in steps as a result of non-simultaneous switching. This exemplary embodiment may consequently increase the magnitude of the gradient of the edge of the system response.

In one embodiment, as shown in FIG. 2 , the outputs of modules M 1 , M 2 and M 3 are connected to windings W 1 , W 2 and W 3 respectively of a pulse transformer PT. The pulse transformer PT includes a secondary winding WL, which is connected to a load RL, for emitting the voltage pulse. To prevent an asymmetrical current distribution within the individual modules M 1 , M 2 , M 3 , which respectively drive a primary winding W 1 , W 2 , W 3 of the pulse transformer PT, the runtime differences between the individual modules M 1 , M 2 , M 3 are measured according to a principle similar to the method explained above.

At time tM 1 of the rising edge for the module M 1 , which has been charged previously, a high-resolution timer is started. The timer may include a resolution less than or equal to 5 ns. The module time difference, which elapses until the system response dependent on the current is received in the control circuit 1 , is determined. The module time difference is filed in a table in a memory. The module time difference is dependent on characteristics of the components of the module and on the impedance of the pulse transformer PT, the impedance of the feed line and optionally on the impedance of the load RL. All the factors influencing the system response may be taken into account together by measuring the runtime.

Module time differences may be determined for the modules M 2 and M 3 and stored. Once the module time differences of all the modules M 1 , M 2 and M 3 have been recorded, a joint switching time is calculated for all the modules M 1 , M 2 and M 3 . The joint switching time may lie at the mean point between the shortest and the longest runtime. For each module M 1 , M 2 , M 3 , a switch-on signal and a switch-off signal are now time-shifted in relation to the joint switching time, such that the current-dependent system response of all the modules M 1 , M 2 , M 3 occurs simultaneously or almost simultaneously. The distribution of current within the individual modules M 1 , M 2 , M 3 may be symmetrized, such that the current flow of the individual modules M 1 , M 2 , M 3 through the pulse transformer PT occurs simultaneously.

The control circuit 1 may have multiple functions, which can be implemented by hardware and a programmable module such as, for example, a microcontroller. As shown in FIG. 2 , in the first function block 2 , the voltage-dependent (OU 1 , OU 2 OU 3 ) system responses of the cells Z 11 to Z 15 , Z 21 -Z 35 , or Z 31 -Z 35 of the respective module M 1 , M 2 , M 3 are synchronized using the time differences by time-shifting the switch-on signal/switch-off signal of the respective cell Z 11 to Z 15 , Z 21 -Z 35 , or Z 31 -Z 35 by the runtime regulation. In the second function block 3 , the modules are synchronized in relation to each other by determining the current-dependent (OI 1 , OI 2 , OI 3 ) system responses of the respective module M 1 , M 2 , M 3 and then by synchronizing by time-shifting the switching points of the modules M 1 , M 2 , M 3 relative to each other.

The third function block 4 includes the higher-level control, such as, for example, the load regulation for generating an electron beam of a klystron, incorporating the additional parameters Kly I, Kly U, tp and PRP. For medical applications, a voltage pulse may have a shape approximating a rectangle. A stable electron beam may be generated during the duration of the plateau of the rectangular signal. An adjustability of the shape of the voltage pulse is effected by the parameters Kly I, Kly U, tp and PRP for the targeted generation of a high-frequency wave for the treatment of cancer. The wave characteristics dependent on the parameters Kly I, Kly U, tp and PRP may be adapted to the particular type of cancer. The power of an electron beam or of a photon beam may be adjusted via the parameters Kly I and Kly U. A pulse width may be adjusted by the parameter tp. The pulse repetition frequency on the other hand can be adjusted by the parameter PRP.

The advantages are inherent in the modular concept in a Marx configuration with n series-connected and m parallel-connected semiconductor switches (IGBTs etc.), including a cascaded active regulation of the voltage and current distribution. In this way the system can also be used with multi-core pulse transformers in particular for magnetron loading or for a klystron loading. In addition a stabilization of the operating parameters is preferably integrated in the control such that, for example, a temperature drift or an ageing symptom, for example of the semiconductor switches, can be determined as operating parameters by means of cyclic sequential measurement intervals during what is termed a pre-scan sequence or a warm-up sequence.

An elaborate and therefore cost-intensive selection, a pairing and matching operations can be avoided by the system shown in FIG. 2 . Constant monitoring of the cells Z 11 to Z 35 may be carried out in synergy by the system shown in FIG. 2 . For example, short-circuits may be detected. Slowly approaching failures may be detected, for example, in that the system response maps a deterioration of electrical parameters of a stage capacitor or of a semiconductor switch. The deterioration of electrical parameters may be determined through measurement of the current/voltage course and/or the time difference, for example, by comparison to threshold values. Diagnostic information may be output. The diagnostic information enables replacement of the component that has deteriorated before a total failure.

Redundant operation may be ensured. If one or more cells fail, the other modules may be adapted in accordance with the number of failed cells, such that an appropriate system response is generated again at the output of all the modules. Operation with reduced power may be possible.

The present embodiments are not restricted to the exemplary embodiment shown in FIGS. 1 to 4 . The basic idea underlying the present embodiments can be applied to all Marx generators with multiple semiconductor switches, such as a programmable control circuit. A single or an arbitrary plurality of modules including two or more cells can be provided. It is also in principle possible to select another reference time which does not lie at the mean point of the runtimes.