CROSS-REFERENCE TO RELATED APPLICATION

[0001] This application claims the benefit of U.S. Provisional Application 62/750,896, filed October 26, 2018, the disclosure of which is incorporated by reference in its entirety.

FIELD OF THE INVENTION

[0002] The present invention relates to zero-voltage-switching of a half-bridge converter for on-board chargers and other applications.

BACKGROUND OF THE INVENTION

[0003] A half-bridge is a component of many Switch Mode Power Converter (SMPC) topologies. An isolated Dual-Active-Bridge (DAB), for example, is an SMPC that uses four half bridges and a transformer to deliver power from the primary to the secondary of the transformer. The DAB has one full-bridge (two half-bridges in parallel) driving the primary winding and one full-bridge driving the secondary winding. A current-fed half-bridge, shown in Figure 1 , includes two semiconductor switches in series with a current source or storage element, like an inductor, with a connection to the midpoint node. The nodes at the end of the series combination are referred to as a linkage. In the case of a voltage-fed half-bridge, shown in Figure 2, the linkage is referred to as a DC linkage, since its value must not go negative. Voltage-fed half-bridges may come in any combination of controlled or un-controlled switches. Examples include fully-controlled half bridges (e.g., a bi-directional buck-boost DC/DC converter), half-controlled half-bridges (e.g., simple buck DC/DC converter), and uncontrolled half-bridges (e.g., the front end rectifier of Totem-Pole PFC topologies).

[0004] Treating the half-bridge as one type of switching block, the current-fed half-bridge

(CFHB) and the voltage-fed half-bridge (VFHB) can be used to construct a SMPC. Switching blocks are active power components and are placed in a configuration with passive power components to create the SMPC power path. The voltage-fed half-bridge can additionally be controlled, half-controlled, or uncontrolled. Many or most controlled semiconductor switches used in SPMCs are only controlled when current is flowing through its primary conduction path in one direction and will self-activate (become reverse biased) when current flows in the opposite direction. Two such switches in a half-bridge configuration are considered a controlled switching block despite having uncontrolled states.

[0005] Given a positive emf across the linkage, each of four block states can be commanded by a controller. More particularly, each switch has two states: conducting and non conducting (blocking). Possible control vectors (drive signals) include the following:

The high-side control vector applies the positive rail of the linkage, Vu _nk , to the midpoint, defining the voltage across Sl at v _0Ssi = ~ 0V and across S2 at v _0Ss 2 = Vu _nk - The low-side control vector applies the negative rail of the linkage to the midpoint, defining the voltage across S 1 at v _0Ssi = V _IM and across S2 at v _0Ss 2 = ~ 0V. The dead-band control vector releases control of the half-bridge midpoint to the control of the greater topology, which moves V _hb according to the controlled current source, such that v _0Ssi + v _0Ss 2 = The shoot-through control vector shorts both positive and negative rails of the linkage to Vhb, such that Vossl— Voss2— ~ ov.

[0006] In a CFHB, the current source is in series with both switches S 1 and S2, and thereby regulates their current, even when both switches are conducting. Therefore, all four control vectors can be used during normal operation for a CFHB. In a VFHB, however, a voltage source is in series with switches Sl and S2, and consequently a shoot-through control vector will short the voltage source, resulting in an unregulated current. This condition is potentially catastrophic. To protect against this condition, the beginning of a turn-on signal for one switch is delayed with respect to the end of a turn-off signal for the other switch by a dead band (or dead time). This technique ensures one switch is fully off (non-conducting) before the complementary switch is driven on, thereby avoiding unintended shoot-through events. However, the duration of the dead band is typically predetermined and fixed in length. As a result, existing half-bridge converters experience thermal losses attributed to a dead band longer than strictly required in order to ensure the absence of shoot-through events.

SUMMARY OF THE INVENTION

[0007] An improved method for zero-voltage switching (ZVS) of a voltage-fed half-bridge

(VFHB) using a variable dead band is provided. The duration of the dead band is determined dynamically by a processor according to a real-time, open loop circuit model and is precisely long enough to ensure the absence of shoot-through events while also minimizing or eliminating switching losses and reverse conduction losses. Eliminating reverse conduction losses according to the present method improves the efficiency of the SMPC, reduces thermal stresses on the semiconductor devices, and allows for more easily designed cooling solutions. Any SMPC using a VFHB can be modeled as a controlled current source feeding the half-bridge midpoint and controlled voltage source feeding the half-bridge rails.

[0008] In accordance with one embodiment, the method generally includes: (a) calculating the equivalent capacitance as seen by the current source charging the midpoint of the half-bridge; (b) calculating the ZVS charge requirement based on the link voltage and the equivalent capacitance; (c) calculating the charge delivered by the current source over time during a dead band vector, equating the result to the ZVS charge requirement, and solving for the ZVS time requirement at each commutation point over the switching cycle; and (d) updating the dead bands for each commutation of each half-bridge in the SMPC. The above method is performed primarily in software in conjunction with a micro-processor and control circuit hardware adapted to accommodate real-time updating of the dead band for the SMPC.

[0009] These and other features of the invention will be more fully understood and appreciated by reference to the description of the embodiments and the drawings.

[0010] Before the embodiments of the invention are explained in detail, it is to be understood that the invention is not limited to the details of operation or to the details of construction and the arrangement of the components set forth in the following description or illustrated in the drawings. The invention may be implemented in various other embodiments and of being practiced or being carried out in alternative ways not expressly disclosed herein. In addition, it is to be understood that the phraseology and terminology used herein are for the purpose of description and should not be regarded as limiting. The use of“including” and“comprising” and variations thereof is meant to encompass the items listed thereafter and equivalents thereof as well as additional items and equivalents thereof. Further, enumeration may be used in the description of various embodiments. Unless otherwise expressly stated, the use of enumeration should not be construed as limiting the invention to any specific order or number of components. Nor should the use of enumeration be construed as excluding from the scope of the invention any additional steps or components that might be combined with or into the enumerated steps or components.

BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1 is a circuit diagram of a current- fed half-bridge converter.

[0012] Figure 2 is a circuit diagram of a voltage-fed half-bridge converter.

[0013] Figures 3(a) to 3(f) illustrate low-to-high commutation states of a VFHB under standard ZVS conditions with a fixed dead band.

[0014] Figure 4 illustrates a long dead band for control of a VFHB, in which the dead band is longer than the time required to charge the half-bridge capacitance.

[0015] Figure 5 illustrates a short dead band for control of a VFHB, in which the dead band is shorter than the time required to charge the half-bridge capacitance.

[0016] Figure 6 illustrates a long dead band for control of a VFHB, in which the dead band is precisely long enough to charge the half-bridge capacitance of a half-bridge for ZVS.

[0017] Figure 7 is a flow-diagram of a method for ZVS of a voltage-fed half-bridge in accordance with one embodiment.

[0018] Figures 8(a) to 8(e) illustrate the change in Charge Equivalent Capacitance with corresponding charge accumulated and equivalent static capacitance.

[0019] Figure 9 is a circuit diagram of a single -phase voltage-fed inverter with an ideal dead band determined by a controller in accordance with a current embodiment. DETAILED DESCRIPTION OF THE CURRENT EMBODIMENT

[0020] An improved method for zero-voltage switching of a voltage-fed half-bridge using a variable dead band is provided. As discussed herein, the duration of the dead band is determined dynamically to ensure the absence of shoot-through events while also minimizing or eliminating switching losses and reverse conduction losses. As background, Part I below includes known techniques for zero-voltage switching of a voltage-fed half-bridge. Part II below includes a discussion of the method of the present invention, namely the zero-voltage switching of a voltage- fed half-bridge using a dynamically calculated variable dead band control vector.

I. Zero- Voltage Switching of Voltage-Fed Half-Bridge

[0021] Zero Voltage Switching (ZVS) is the commutation of a semiconductor switch from an off state to an on state while there is zero voltage across its primary conduction path. The process by which a ZVS commutation takes place can be illustrated with the low-to-high commutation of an arbitrary VFHB, as shown in Figures 3(a)— 3(f). With respect to ZVS, a semiconductor switch may be considered as an ideal MOSFET in parallel with an output capacitance and an anti-parallel diode with forward voltage V _rth . Using two such switches in a half-bridge configuration, such that S 1 is the high-side switch and S2 is the low-side switch, across an arbitrary DC linkage, V _iink , referenced to the low-side rail, here referred to as GND, with the midpoint voltage, V _HB , supplied by some controlled current source (a resonant inductor for example) such that current flowing into the midpoint, i _hb , is always positive, the entire commutation process, under ZVS conditions, can be broken down as shown for the low-to-high half-bridge commutation shown in Figures 3(a)— 3(f). Here, the output capacitance of each switch is abbreviated as Cl = C _ossl and C 2 = C _oss2 . The antiparallel diode is considered to be ideal and the voltage source V _rth (D) is a function of the diode’s state - conducting = 1 and non-conducting = 0 - such that V _rth (0) = OF and V _rLfl (l) = V _rth . In this way, the behavior of any arbitrary self activating semiconductor switch can be modeled as the anti-parallel device.

[0022] Prior to commutation in Figure 3(a), i _hb flows freely through S2 and V _HB is clamped to GND. When S2 turns off, i _hb begins to migrate from the conduction channel to charging the C _oss of both S2 and Sl, as shown in Figure 3(b). This marks the beginning of the dead band, as the controller is applying a dead band control vector, driving both S 1 and S2 off simultaneously. The rate at which V _HB rises while the channel is collapsing is controlled by i _hb , the channel resistance of Sl, and the C _oss of both S l and S2 - as the channel collapses, the resistance increases, and more of i _hb migrates to charging C 1 and C2. Once the switch is fully off, the channel resistance is high enough that the current through it becomes negligible and the change in V _HB is then be dominated by the following relationship (equation 1):

[0023] If the switch has a fast turn-off edge rate, as in a MOSFET, the channel stops conducting well before the C _oss charges to the opposite linkage rail, so the dead band must be long enough to charge the midpoint from GND to V _iink , and thereby discharge Cl from V _iink to 0V. Ideally, Sl would turn on at exactly V _HB = V _link , but in reality V _HB will continue charging until Dl of Sl becomes reversed biased and clamps V _HB to V _iink + V _rth . Once the midpoint voltage is clamped, Sl is burning energy for the remainder of the dead band. Thus, it is desired to make this region as short as possible. When Sl commutates on, some switching loss will be experienced due to Cl being charged to—V _rth and discharging through the forming channel, but this loss will be small. Post S l’s commutation, i _hb should still be positive into the bridge, though, ideally, i _hb would reach 0A immediately after Sl’s commutation to minimize the resonant current.

[0024] Typically, the dead band is set in the control circuit generating the Pulse Width Modulation (PWM) for the half-bridge to a fixed value. The ZVS Time Requirement T _zvs is the minimum time required for controlling a current source (assuming a known current) to deliver the charge needed for ZVS. The ZVS Time Requirement T _zvs is inversely proportional to the current during the dead band and directly proportional to the link voltage. Therefore, a typical control circuit for a ZVS application will fix the dead band according to the lowest expected l(T _ZV s) and the highest link voltage so that ZVS can be ensured for all operating points. However, this means that when the current flowing into the half-bridge is larger, the ZVS charge requirement will be met faster, and the midpoint voltage, v _bb , of the half-bridge defined by Sl and S2 will reach the targeted linkage rail before the end of the dead band and continue charging beyond the rail.

[0025] If this charging is allowed to continue, not only will ZVS be lost, but, under high switching currents, the voltage rating of the semiconductor switch may be exceeded leading to life degradation and device failure. For this reason, ZVS topologies (and most SPMC topologies in general) provide a free-wheeling path to the midpoint current, i _hb . This is accomplished by the antiparallel body or external diodes of MOSFETs and IGBTs or the reverse conduction properties of the 2DEG in HEMTs, for example. Because this reverse conduction is uncontrolled, there must be an associated voltage drop equal to threshold of the antiparallel device, and the midpoint will therefore be clamped to the linkage rail plus the reverse threshold voltage, V _rth . In this clamped conduction state, i _hb is being forced through the device in the direction of the reverse voltage drop, generating thermal losses referred to as reverse conduction losses. This condition is depicted in Figure 3(d), in which the low-to-high commutation of the half bridge is significantly longer than the time required to deliver enough charge to the equivalent capacitance to change the midpoint voltage by the required amount. As a result, once the midpoint voltage reaches Vu _nk (60V in the present example), the current keeps charging the equivalent capacitance until the midpoint voltage exceeds the high-side rail voltage by the reverse threshold of the device. Figure 4 (depicting a short dead band) results in hard switching at about 20V in this example, while the ideal dead band of Figure 5 avoids switching losses and reverse conduction losses. Figures 4, 5, and 6 provide an example based on the general case presented in Figures 3(a)-(f) where the current source is provided with an 8 pH inductor with constant initial current of 2 A and a link voltage of 60V. The only parameter varied between the three figures is the dead time t _db .

II. Variable Dead band Control for Voltage Switching of Voltage-Fed Half-Bridge

[0026] In order to eliminate the reverse conduction losses discussed in Part I above, the dead band is calculated dynamically for each commutation point instead of using a predetermined value corresponding to a worse-case requirement. Eliminating reverse conduction losses improves the efficiency of the SMPC, reduces thermal stresses on the semiconductor devices, and allows for a more easily designed cooling solution.

[0027] Referring to the flow-chart of Figure 7, a method for zero-voltage switching of a voltage-fed half-bridge using a variable dead band control vector is illustrated. The method generally includes calculating the dead band dynamically for each commutation point instead of using a predetermined value corresponding to a worst-case requirement. More particularly, the method generally includes the following method steps: (a) calculating the equivalent capacitance C _eq as seen by the current source charging the midpoint of the half-bridge based on the link voltage, the SMPC topology, and the SMPC circuit state (step 10); (b) calculating the ZVS charge requirement based on the link voltage and the equivalent capacitance (step 12); (c) calculating the charge delivered by the current source over time during a dead band vector, Q _hb , equating the result to the ZVS charge requirement, and solving for Tzvs at each commutation point over the switching cycle (step 14); and (d) updating the dead bands for each commutation of each half- bridge in the SMPC (step 16).

[0028] Calculating the equivalent capacitance C _eq at step 10 includes determining any capacitances along the return path of the current source. This can include parasitic PCB capacitances, winding capacitances of magnetics, and intended capacitances (such as resonant tanks). The examples used in Figures 4, 5 and 6 assume only the switch capacitance, C _oeq , PCB capacitance, C _PCB , and winding capacitance, C _L , are in parallel along the return path of the inductor. Thus, C _eq = C _oeq + C _PCB + C _L . This is the case in a single half-bridge commutation in a DAB. This step is also visualized with the following example. Using a Gallium Nitride High Electron Mobility Transistor (GaN HEMT) from GaN Stystems Inc., part number GS66516T, Figure 8 demonstrates how capacitance, charge, equivalent capacitance, and energy can change over output voltage of a single GaN HEMT. Since the change in output voltage over the dead band of a half bridge commutation must equal the linkage, vds (drain-to-source voltage) in Figure 8 is interchangeable with V _iink Figure 3(a). Figure 8(a) shows the output capacitance change over voltage. Figure 8(b) shows the accumulated charge across one device over voltage. Figure 8(c) plots the equivalent static capacitance value that would hold the same amount of charge at that voltage as the C _oss , called the“charge equivalent capacitance”. Figure 8(d) shows the accumulated energy over voltage. Figure 8(e) shows the equivalent static capacitance value that would hold the same amount of energy at that voltage as the C _oss , called the“Energy Equivalent Capacitance.” [0029] Based on the equivalent capacitance C _eq , the ZVS charging requirement, Qzvs- _> i ^s then calculated according to the following (equation 2), wherein C _eq is a function of voltage because the C _oss of a semiconductor is usually not constant over output voltage, v _oss :

As noted above, Q _Z vs i ^s the charge required to move the midpoint voltage of Sl and S2, v _hb = v _oss 2, from ~0V to V _iink , which, in turn, discharges v _ossl from V _iink to ~0V. Calculating the ZVS timing requirement, Tzvs, is then performed according to the following (equation 3-4):

In the above equations 3, Qzvs is calculated as the average current during a dead band vector (the inverse ZVS period multiplied by the integral of the current). In the above equation 4, Tzvs is solved for as the required charge for ZVS divided by the average current over the time allowed for ZVS, or l(T _Z vs)· Accordingly, to properly set the initial conditions for determining Tzvs, the initial current at the start of each dead band during a switching cycle must be calculated.

[0030] As one example, the controlled current source in Figure 3(a) is realized with a load comprising a resonant inductor, L = 8 mH. Assume that at time t ₀ = 0s, L is charged to i _L = 2 A with no initial driving voltage across it’s terminals, V ₀ = 0V, meaning current is initially freewheeling through the inductor. The winding capacitance of the inductor is C _L = 50 pF and the parasitic capacitance of the PCB is C _PCB = 600 pF, which are in parallel with the C _oss of both S l and S2, as seen by the inductor. The link voltage is set to V _iink = 60F and the dead band, given by T _db o, varied to show its effect. Figure 6 shows a low-to-high commutation of the half-bridge where the dead band, t _db = 72.3 ns from t = 72.3ns to t = 144.6ns, is the precise time required for 2A to deliver enough charge to the equivalent capacitance to change the midpoint voltage by 60V. This is the ideal commutation of the half-bridge. The dead band could be larger than optimal by the time required to charge an additional 6.8V, providing a buffer region for real-world sensor errors without incurring reverse conduction losses.

[0031 ] The method further includes updating the dead bands for each commutation of each half-bridge in the SMPC at step 16. Each step of the foregoing method can be implemented in digital logic in connection with a dual-active-bridge (DAB) converter having four half-bridges, but the present method can be implemented in other SMPCs as desired. The method does not require any additional hardware, and instead relies on a mathematical model of the system topology to calculate the dead band in real time. As noted above, eliminating reverse conduction losses according to the present method improves the efficiency of the SMPC, reduces thermal stresses on the semiconductor devices, and allows for more easily designed cooling solutions.

[0032] In accordance with a further embodiment, a single-phase voltage-fed inverter for producing a square wave output is shown in Figure 9. The inverter includes a half-bridge having two series-connected switches Sl, S2 and two freewheeling anti -parallel diodes Dl, D2. The switches may not be activated simultaneously, which would otherwise short the voltage source. In accordance with one embodiment, the ideal dead band is calculated in digital logic, the dead band being the interval between the turn-off and turn-on of the series connected switches. In particular, a controller 20, for example an integrated circuit or a digital signal processor, determines the equivalent capacitance at a mid-point 22 of the half-bridge. The equivalent capacitance can be determined empirically for a given input current at the mid-point or a given rail voltage across the half-bridge, and generally includes any switch capacitance, PCB capacitance, and winding capacitance. Based on this equivalent capacitance, the controller 20 determines the ZVS charging requirement ( Qzvs ) according to equation (2) above, which represents the charge required to move the midpoint voltage from 0V to the rail voltage ( Vu _nk ). The controller 20 then determines the ideal dead band Tzvs according to equation (4) above, in which the ZVS charging requirement (Qzvs) is divided by the average current over time allowed for ZVS U(T _ZVS )), which is obtained by the controller from the initial current at the beginning of each dead band. The controller 20 then stores the ideal dead band Tzvs to computer readable memory, potentially updating the existing value for the half-bridge dead band, and creates an interval between the turn-off and turn on of the series connected switches at least equal to the ideal dead band Tzvs (e.g., with or without a buffer period). This process can be repeated periodically or in response to an event, for example in response to a change in the input current to the mid-point 22 of the half-bridge, or in response to a change to the DC rail voltage.

[0013] The above description is that of current embodiment of the invention. Various alterations and changes can be made without departing from the spirit and broader aspects of the invention. This disclosure is presented for illustrative purposes and should not be interpreted as an exhaustive description of all embodiments of the invention or to limit the scope of the claims to the specific elements illustrated or described in connection with these embodiments. Any reference to elements in the singular, for example, using the articles“a,”“an,”“the,” or“said,” is not to be construed as limiting the element to the singular.