CLASS PHI-2 POWER CONVERTER WITH SELF-VOLTATING SWITCHING CIRCUIT

Description

[0001] The invention relates to a DC / AC power converter with high or very high switching frequency. The invention finds its application in particular in the conversion of a direct voltage into an alternating or direct voltage, in HF (from 3 to 30 MHz) and VHF (from 30 to 300 MHz) radio frequency ranges. Switching the converter in the radio frequency ranges makes it possible to reduce the size of the reactive components (inductors, capacitors) of the power conversion circuits, and thus to reduce the overall volume of the power conversion chain, which can be advantageous for applications where compactness and mass are important constraints.

[0002] In DC / AC power converters (also called inverters), switching is performed with a power switch, for example a GaN (Gallium Nitride) transistor or a MOS transistor. The transistor switches from the on state to the off state or vice versa, thanks to a control circuit called a gate driver. Typically, in power converters, one or more resistors can be added to the transistor gate to control voltage or current variations on the transistor when the converter starts (a so-called non-resonant structure). With each switching command transmitted by the gate driver to the transistor, there is energy dissipation in the resistor(s) added to the transistor gate.For switching frequencies of the order of tens or hundreds of kHz, the sum of losses due to dissipation can be considered negligible over a given duration. On the other hand, over the same duration, the sum of losses due to dissipation is much greater for switching frequencies of the order of tens or hundreds of MHz. In order to reduce this type of dissipation, and to subsequently avoid degrading the efficiency of the converter, a so-called resonant structure can be used, based on passive energy storage components (capacitors and inductors), instead of using a non-resonant structure. The resonant structure, unlike the non-resonant structure, makes it possible to store energy during a switching phase of the transistor, and to restore it during the following phase instead of dissipating it in the parasitic elements of the transistor.

[0003] Among the DC / AC converters known from the state of the art, the Phi-2 (Φ2) class converter has many advantages. The Phi-2 class converter is described in particular in the document « A high-frequency resonant inverter topology with low-voltage stress » (Rivas et al., IEEE Transactions on power electronics, vol. 23, No. 4, July 2008) and the document “High-Frequency, High-Power Resonant Inverter With eGaN FET for Wireless Power Transfer” (Choi et al., IEEE Transactions on power electronics, vol. 33, No. 4, March 2018). The PHI-2 class converter, illustrated by figure 1 , includes an input inductance Lf, connected to the voltage source Vin to be converted, and having a value of the same order of magnitude as the inductance Lr of a series resonant network Lr-Cr connected between the drain of the transistor and the load. It also includes an Lmr-Cmr filter, the resonant frequency of the Lmr-Cmr filter being equal to twice the switching frequency of the transistor. The Lmr-Cmr filter is added in parallel with the transistor in order to short-circuit the second harmonic of the drain-source voltage of the transistor and thus reduce the voltage stress on the transistor. Such a structure is easily integrated on a printed circuit, due to the reduced value of the input inductance Lf. The absence of a high-value inductance also makes it possible to obtain a shorter transient than in other types of converters, in particular those of class E, which can be advantageous for achieving rapid power calls in the converter.The different values ​​of the converter components are selected in particular so that, when the transistor switches, the Vds voltage is almost zero, in order to respect the soft switching condition (ZVS, or “Zero Voltage Switching” in Anglo-Saxon terminology).

[0004] In soft switching (ZVS), the values ​​of the resonant network components and the output capacitance Cf (also called shunt capacitance) of the transistor are selected so that the voltage Vds across the transistor is zero each time the transistor switches from the on state to the off state and vice versa. Since losses in a transistor are due to the product of the voltage across its terminals and the current through it, a zero voltage at each switching minimizes losses.

[0005] In a class Phi-2 converter, the drain-source voltage is also equal to approximately twice the input voltage Vin, which limits the voltage stress on the transistor. By reducing the voltage stress, it is then possible either to gain in efficiency, with a lower on-state resistance R DSON than for a class E converter, or to gain in compactness, by reducing the size of the chip on which the various elements of the conversion circuit are located. In addition, the lower voltage stress makes it possible to consider the parasitic capacitance as being more stable in value, which facilitates its modeling.

[0006] The switching control of the transistor can be realized by a gate driver circuit CP designed as an integrated circuit on an electronic chip, as illustrated by the figure 1 . However, this type of gate driver is not suitable for switching frequencies above 30 MHz, and in particular above 35 MHz. Indeed, beyond this frequency, the delays introduced by the internal circuits (logic gates) of the CP gate driver become preponderant, and the CP gate driver is no longer able to lower the driving signal below the threshold voltage Vth of the transistor (1.3 V for a GaN transistor). The transistor can then no longer switch, and the converter no longer works.

[0007] The invention therefore aims to do without active components and logic gates for controlling a transistor in a Phi-2 structure converter.

[0008] An object of the invention making it possible to achieve this aim, partially or totally, is a power converter according to claim 1.

[0009] Advantageously, the converter may comprise a second inductor connected to the output terminal, a first capacitor connected between the output terminal and the reference potential, the second inductor, the first capacitor and the first capacitive divider bridge being configured to form an “L” impedance matching network for a predefined load value at the output of the converter.

[0010] Advantageously, the converter may comprise a resistive divider bridge connected between the input terminal and the reference potential, having a second midpoint connected to the first inductance and to the reactive circuit, the resistive divider bridge being configured to provide a DC component from the input voltage, the DC component being equal to a threshold voltage between the control electrode and the second electrode of the power switch.

[0011] Other characteristics, details and advantages of the invention will emerge from reading the description given with reference to the appended drawings given by way of example: [ Fig.1 ] there figure 1 , already described, represents a Phi-2 class converter according to the state of the art; [ Fig.2 ] there figure 2 represents a Phi-2 class converter according to the state of the art, comprising an impedance matching network; [ Fig.3 ] there figure 3 represents a first embodiment of a Phi-2 class converter according to the invention; [ Fig.4 ] there figure 4 represents an embodiment useful for understanding the invention.

[0012] The invention is described in the case where the power switch is a field effect transistor (for example MOSFET, JFET). The substrate of the transistor can be made of gallium nitride (GaN), silicon carbide (SiC), or any other semiconductor material. The drain, the source and the gate mentioned in the description can be designated more generally respectively by a first electrode, a second electrode and a control electrode. The invention can thus also be applied to other types of power switches, for example an IGBT type transistor, a bipolar transistor or a thyristor.

[0013] A Phi-2 class converter is generally sized to deliver an output power Pout to a desired load impedance Rdes, from a DC input voltage Vin, at a switching frequency f switching . Commercially available power radio frequency loads do not necessarily have a nominal load impedance Rfab equal to the desired load impedance Rdes. In this case, an impedance matching between the desired load impedance Rdes and the nominal load impedance Rfab must therefore be performed. In other cases, for a desired load impedance Rdes, it is not possible to size a Phi-2 class converter that allows a certain output power value to be obtained. To obtain the desired output power, it is therefore necessary to use another load value and adapt the impedance.Impedance matching can even be performed when the desired load Rdes and the nominal load Rfab are theoretically equal, due to the different parasitic elements that the circuit may include.

[0014] It is appropriate in this case, as illustrated by the figure 2 , to insert an “L” impedance matching circuit, comprising an Lmatch impedance matching inductor and a Cmatch impedance matching capacitor.

[0015] The article « Resistance Compression Networks for Radio-Frequency Power Conversion" (Y. Han, O. Leitermann, DA Jackson and DJ Perreault, IEEE Transactions On Power Electronics, Vol. 22, No. 1, January 2007) describes a method for determining an impedance matching circuit.

[0016] There figure 3 represents a first embodiment of a class Phi-2 converter according to the invention. In the class Phi-2 converter according to the invention, the impedance matching capacitor Cmatch of the “L” impedance matching circuit is replaced by an equivalent network of capacitors C1-C3-C4. The first capacitor C1 is in parallel with the capacitor C3 in series with the capacitor C4. A capacitive divider bridge PDC is obtained by taking a voltage between a first intermediate point PM1, located between the capacitors C3 and C4, and a reference potential GND (for example a ground). The manufacturer's load Rfab is located between the output terminal BSO and the reference potential GND. The output voltage Vout can be measured across the manufacturer's load Rfab. The capacitive divider bridge PDC and the manufacturer's load Rfab are connected in parallel. Thus, a fraction of the output voltage Vout is taken across the capacitor C4.The fraction of the output voltage Vout is calculated to provide a voltage level close to the driving voltage of the INT transistor. The output voltage Vout is a sine wave, which must be brought back to an acceptable value by the gate of the INT transistor (approximately 6 V for a GaN transistor), centered around the value of the threshold voltage Vth of the INT transistor.

[0017] A second capacitor C2 is directly connected to the gate G of the INT transistor. The reactive circuit CRE, formed by the second capacitor C2 and the input capacitance Cgs of the INT transistor, forms another capacitive divider bridge, to further reduce the gate drive voltage. The second capacitor C2 also has a function of filtering the DC component of the gate voltage Vgs.

[0018] A first inductor L1 is connected to the intermediate point PM1 of the capacitive divider bridge PDC. The first inductor L1 forms an oscillating network with the capacitor C4, the second capacitor C2 and the gate-source capacitance Cgs of the transistor INT. The value of the first inductor L1 is determined so that the oscillation frequency of the oscillating network is equal to the desired switching frequency of the transistor INT. This value can, however, be corrected to take into account the sizing of the other components. The value of the first inductor L1 is also determined so as to provide a phase shift to the output signal, so as to respect the soft switching condition (ZVS).

[0019] The Vgs voltage is not linear: it does not increase linearly, depending on the total charge of the gate. The Vgs voltage plateau is equal to the Miller voltage (voltage value corresponding to the Miller plateau), and which designates the voltage necessary to recharge the drain-gate capacitance of the transistor Cgd, also called Miller capacitance. Outside this plateau, the Vgs voltage increases, with the same slope before and after the plateau, proportionally to the total charge. Thus, the Vgs voltage obtained on the gate G of the INT transistor may not be centered around the threshold voltage Vth. In the worst case, the Vgs voltage may be constantly lower than the threshold voltage Vth, which would prevent any switching of the INT transistor.

[0020] To compensate for the nonlinearity of the voltage Vgs, a resistive divider bridge PDR is connected between the input terminal BEN and the reference potential GND. The resistive divider bridge PDR comprises a first resistor R1 in series with a second resistor R2. A second midpoint PM2 is arranged between the first resistor R1 and the second resistor R2. A fraction of the DC input voltage Vin is taken at the second midpoint PM2. This fraction is added to the sinusoidal component extracted from the output voltage Vout. Adding a DC component to the sinusoidal component ensures that the gate drive signal oscillates around the threshold voltage Vth of the transistor INT.

[0021] The second resistor R2 can be a variable resistor. Adjusting the value of the second resistor R2 allows the converter's efficiency to be finely varied.

[0022] The INT transistor gate driver circuit is therefore self-oscillating in that it creates and maintains the oscillations necessary for switching the INT transistor.

[0023] There figure 4 illustrates a variant of the converter according to an embodiment which is not included in the subject matter of the claims.

[0024] The converter according to this variant also includes a capacitive divider bridge PDC, and a first inductance L1.

[0025] The CRE reactive circuit is composed of at least one series resonant circuit. On the figure 4 , two series resonant circuits (CRS1, CRS2) are illustrated, but the presence of only one series resonant circuit can be considered. Each of the series resonant circuits is connected between the gate of the transistor INT and the reference potential GND. Each of the series resonant circuits is composed of an inductor (Lh1, Lh2) and a capacitor (Ch1, Ch2). The values ​​of the inductor and the capacitor of a first series resonant circuit CRS1 are determined so that the resonant frequency of the first series resonant circuit is equal to the frequency of the highest amplitude harmonic (after the first harmonic, which corresponds to the fundamental component) of the gate voltage of the transistor INT. Thus, the inductor and the capacitor of the first series resonant circuit short-circuit the highest harmonic. Similarly, a second series resonant circuit CRS2 (of rank 2) can be arranged in parallel with the first series resonant circuit CRS1.The values ​​of the inductor and capacitor of the second series resonant circuit are determined so that the resonant frequency of the second series resonant circuit is equal to the frequency of the second highest harmonic of the gate voltage of the INT transistor. Other series resonant circuits of the same order can be arranged in parallel. The gate signal is thus made more sinusoidal by removing the harmonics that distort it, which causes faster turn-on or turn-off of the INT transistor, which contributes to improving the efficiency of the converter.

[0026] A resistive divider bridge is connected between the input terminal BEN and the reference potential GND, in order to add a DC component and center, if necessary, the sinusoid of the gate control signal around the threshold voltage Vth.

[0027] The driving signal is thus mainly made up of its fundamental component, which is sinusoidal in shape. Driving the gate with a sine wave allows for better switching of the transistor and higher efficiency.

[0028] The converter, according to one or other of the variants, contains only passive components, and without additional DC source. Finally, it does not contain any transistor other than that necessary for the Phi-2 structure, which avoids having to manage dead times. It is recalled that the management of dead times is necessary when there are at least two transistors connected in series, in parallel with a voltage source, in order to avoid the presence of short circuits when two transistors are conducting at the same time. At high and very high frequency, the switching period is so small that it becomes of the same order of magnitude as the dead times necessary for the proper functioning of the circuit, which complicates the management of dead times.

[0029] The converter described above can be part of a DC / DC conversion device, for example by connecting the output of the converter to a transformation stage, which would itself be connected to a rectifier.

[0030] The following section describes an example of dimensioning of the converter which is the subject of the invention, according to the embodiment illustrated by the figure 3 , for an input voltage Vin of 30 V, an output power of 30 W, and a desired switching frequency f switching of 40 MHz.

[0031] The sizing of the input inductance Lf, the series resonant network Lr-Cr and the filter Lmr-Cmr is described in the document « A high-frequency resonant inverter topology with low-voltage stress” (Rivas et al., IEEE Transactions on power electronics, vol. 23, No. 4, July 2008).

[0032] We obtain Lf = 50 nH, Lmr = 35 nH, Cmr = 113 pF, Lr = 50 nH, Cr = 2 nF.

[0033] The dimensioning of the matching circuit {Lmatch-Cmatch} is described in the article « Resistance Compression Networks for Radio-Frequency Power Conversion” (Y. Han, O. Leitermann, DA Jackson and DJ Perreault, IEEE Transactions On Power Electronics, Vol. 22, No. 1, January 2007).

[0034] We obtain Lmatch = 79 nH, Cmatch = 165 pF.

[0035] The Cmatch capacitor is divided into C1 parallel with (C3 series C4), so as to obtain an impedance for the branch (C3 series C4) very large (about ten times) compared to the output load, so as not to deprive the load of the power it should receive.

[0036] The capacitive PDC divider bridge must be sized to bring the Vout voltage back to around 6V (the transistor's drive voltage). We thus obtain the two sizing equations: [Math. 1] 1 2 π . f commutation . C 3 + C 4 ≫ 10 . R fab [Math. 2] C 3 C 3 + C 4 > V div V out

[0037] Where V div is equal to the voltage between the first midpoint PM1 and the reference potential GND, and R fab the load impedance provided by the manufacturer.

[0038] We find C3 = 10pF, C4 = 100 pF.

[0039] To find the value of the first inductance L1, we start by calculating the equivalent capacitance Ceq of the gate capacitance Cgs and the second capacitor C2.

[0040] Then, starting from f commutation = f oscillation (where f oscillation is the resonance frequency of the L1-Ceq circuit), we obtain the equation: [Math. 3] f commutation = 1 2 π . L 1 . Ceq

[0041] We find L1 = 200 nH, this value is provisional and will be modified upon introduction of the second capacitor C2.

[0042] The second capacitor C1 is introduced to realize another voltage divider at the gate of the transistor INT. Cgs is given by the datasheet of the transistor INT and the value of the second capacitor C2 is determined by the equation: [Math. 4] C 2 C 2 + Cgs = V gs V osc

[0043] Where V gs represents the voltage between the gate and the reference potential GND, and V osc represents the voltage between the second midpoint PM2 and the reference potential GND.

[0044] We obtain C2 = 300 pF, and we adjust the value of the first inductance L1 to 230 nH using simulation in order to maintain the oscillation at 40 MHz.

Claims

1. A power converter for converting a DC input voltage (Vin) into an AC output voltage (Vout), the power converter having a structure of type Phi-2, and comprising: - an input terminal (BEN) for the DC input voltage (Vin), - an output terminal (BSO) for the AC output voltage (Vout), - a power switch (INT) equipped with a control electrode (G), a first electrode (D) and a second electrode (S) linked to a reference potential (GND), characterised in that the converter further comprises a self-oscillating circuit, connected between the output terminal (BSO) and the control electrode (G), and configured to supply and maintain a sinusoidal drive signal to the control electrode of the power switch (INT) from the output voltage (Vout); the self-oscillating circuit comprising a first capacitive divider bridge (PDC) connected between the output terminal (BSO) and the reference potential (GND), and further comprising a first inductor (L1) connected on one side to a first midpoint (PM1) of the capacitive divider bridge (PDC) and on the other side to the control electrode (G) via a reactive circuit (CRE), the reactive circuit (CRE) being formed by a second capacitor (C2) connected between the control electrode (G) and the first inductor (L1) and by the stray capacitance (Cgs) between the control electrode (G) and the second electrode (S) of the power switch (INT).

2. The converter according to claim 1, comprising a second inductor (Lmatch) connected to the output terminal (BSO), a first capacitor (C1) connected between the output terminal (BSO) and the reference potential (GND), the second inductor (Lmatch), the first capacitor (C1) and the first capacitive divider bridge (PDC) being configured to form an "L" configuration impedance matching network for a predefined load value at the output of the converter.

3. The converter according to one of claim 1 or 2, comprising a resistive divider bridge (PDR), connected between the input terminal (BEN) and the reference potential (GND), having a second midpoint (PM2) connected to the first inductor (L1) and to the reactive circuit (CRE), the resistive divider bridge (PDR) being configured to supply a DC component derived from the input voltage (Vin), the DC component being equal to a threshold voltage between the control electrode (G) and the second electrode (S) of the power switch (INT).