Very High Efficiency Auxiliary Power Supply for 800V bus

Very High Efficiency Auxiliary Power Supply for 800V bus

Ionel “Dan” Jitaru  Ph.D.
CEO/President
Rompower Energy Systems Inc.
Tucson, United States
dan.jitaru@rompower.com

Andrei Savu
Rompower International SRL.
Bucharest, Romania
info@rompower.com

This document demonstrates the most suitable solution for high input voltage (300V-1000V) DC link 60 W auxiliary power supply designs using Rompower’ s technology in a single-ended flyback topology. These technologies can support designers targeting three-phase converters that include solar inverter, energy storage, EV-charger, UPS and motor driver 800V bus for AI power.

Rompower technologies are very suitable for very high input voltage applications in low power and middle range power. This bias supply safely step down the high voltage ( 300V to 1000V) to a lower, usable voltage (e.g., 12V, 15V, 24V) to power control circuits, gate drivers, or other low-voltage accessories while maintaining galvanic isolation.

This document further analyzes the key technical challenges associated with high-input-voltage auxiliary power supplies and highlights operating conditions in which the inherent limitations of conventional flyback topologies become most pronounced. This document establishes the framework for introducing Rompower’ s technology as an effective means of mitigating these limitations and enabling robust, high-efficiency operation across an exceptionally wide input-voltage range.

Considerations for High-Voltage Operation on an 800 V DC Bus

Operation from a high-voltage DC bus, such as an 800 V, introduces several critical design considerations. Among these, two dominant factors must be carefully addressed: transformer leakage inductance and switching losses. In high-efficiency flyback-based converters, where zero-voltage switching (ZVS) techniques are employed, the effectiveness of the ZVS mechanism becomes a decisive contributor to overall system efficiency.

This document analyzes and compares two ZVS approaches applicable to high-voltage flyback converters:

Primary switch ZVS obtained through the activation of the synchronous rectifier (SR), prior to the turn-on of the primary switch.

Rompower’s current injection technique, which actively enforces ZVS conditions over a wide operating range.

Both methods are examined in the following sections, with emphasis on their respective advantages and limitations, particularly in the context of 800 V input applications. Experimental results are also presented. 

The Impact of Transformer’s Leakage Inductance

In applications where the input voltage increases from a conventional 400 V DC bus to an 800 V DC bus—effectively doubling the input voltage, the transformer design must be adapted accordingly. For a given power level and magnetic core, maintaining the same peak flux density requires doubling the number of primary turns.

Since transformer leakage inductance is proportional to the square of the number of turns, a substantial increase in leakage inductance is expected when operating from an 800 V bus.

To illustrate this effect, consider an EQ25 magnetic core with an effective cross-sectional area of 100 mm², operating at 100 kHz:

400 V bus operation
Primary turns: 20
Measured leakage inductance: 2 µH

800 V bus operation (same core, same flux swing)
Primary turns: 40
Secondary turns (24 V output): 4
Measured leakage inductance: 8.2 µH

Operation of a Flyback Power Train with Passive Clamp Network

Figure 1a illustrates the power train of a conventional flyback topology, consisting of a transformer Tr with a primary winding L1 and a secondary winding L2. Due to imperfect magnetic coupling between the primary and secondary windings, a leakage inductance, Llk is inherently present.

A passive clamp network is connected across the primary winding to limit voltage stress on the main switch. This clamp is composed of a diode D1, a resistor R1, and a passive discharge network formed by resistor R2 and clamp capacitor C1. The primary role of resistor R2 is to partially discharge the clamp capacitor C1 after each switching cycle.

Figure 1(b) shows the principal operating waveforms corresponding to the circuit depicted in Fig. 1(a), including the current through the main switch M₁, the drain-to-source voltage V_DS(M₁), and the current through the clamp diode D₁.

During the interval t₀–t₁, the main switch M1 is turned on, and the current in the primary winding L1 increases linearly, reaching a peak value Ipk at time t₁. At t₁, the main switch M1 is turned off. As a result, the magnetizing current of transformer Tr is transferred to the secondary side through the synchronous rectifier SR, charging the output capacitor Co and supplying the load connected at output voltage Vo.

Simultaneously, the current flowing in the leakage inductance Llk, which initially has the same amplitude Ipk, is diverted through resistor R1 and diode D1 into the clamp network formed by R2 and C1. During the interval t₁–t₂, a charge Qa is injected into the junction of diode D1.

After t₂, the current through diode D1 reverses polarity. The charge previously injected into the D1’junction between t₁ and t₂ begins to be extracted and is fully depleted by t₃. During the interval t₂–t₃, while stored charge remains in the junction, diode D1 conducts in reverse.

At time t₄, diode D1 transitions into a high-impedance state. From this point onward, the interaction between the leakage inductance Llk, the parasitic capacitance of diode D1, and the parasitic capacitance present in the switching node (formed by the transformer connection and the drain of the main switch) results in high-frequency ringing. This behavior is illustrated in Figure 1c, where the switching-node voltage is shown together with the gate drive signal of switch M1.

It is important to emphasize that the reverse current conduction of diode D1 during the interval t₂–t₃ enables a portion of the leakage inductance energy to be transferred to the secondary side. However, in conventional implementations, the charge injected into the diode junction during t₁–t₂ is typically greater than the reverse recovery charge extracted between t₂ and t₄. As a result, residual energy remains trapped in the parasitic network, leading to voltage spikes and high-frequency ringing at the switching node, as observed in Figure 1c.

One commonly adopted solution is to increase the value of resistor R1 in order to damp the oscillations. While effective in reducing ringing, this solution dissipates the leakage energy as heat, thereby degrading overall converter efficiency.

In the ideal case where the reverse recovery charge of diode D1 is equal to or greater than the charge injected during t₁–t₂, the oscillations at the switching node would be completely suppressed, eliminating both voltage spikes and high-frequency ringing across the main switch. Achieving this condition would allow the entire injected charge to be transferred to the secondary side rather than dissipated.

Limitations of Conventional Approaches and Motivation for Alternative Solutions

For many years, power electronics designers have sought rectifier devices with extremely large reverse recovery characteristics to accomplish this objective. However, recent developments have demonstrated that the fundamental solution does not lie in increasing the reverse recovery charge of the clamp diode.

Instead, as addressed recently in [1], [2],[3],[4],[5], the problem is more effectively solved by substantially reducing the charge injected into the junction of diode D1 during the interval t₁–t₂. By minimizing this injected charge, the existing reverse recovery characteristics of the diode become sufficient to transfer nearly all of the leakage inductance energy to the secondary, thereby suppressing ringing without incurring additional dissipation.

The Novel Solution in Reducing the Charge Injected into the Clamp Capacitor

Figure 2a illustrates the power train of a conventional flyback topology comprising a transformer Tr with a primary winding L1 and a secondary winding L2. A leakage inductance Llk exists between the primary and secondary windings. A clamp capacitor Cr is connected in series with a clamp diode Dc1. The clamp capacitor Cr is further connected to circuit 100, referred to in this paper as the “energy extraction circuit.”

The energy extraction circuit 100 consists of two low-voltage rectifier diodes D1 and D2, and a voltage source VB. The voltage source VB may be implemented using a capacitor connected in parallel with a load resistor. The values of the capacitor and resistor are selected such that the voltage ripple across the capacitor is negligible; therefore, the network operates, for practical purposes, as an ideal voltage source. The voltage VB can be the bias voltage bus that can be connected to other auxiliary loads. Considering the conduction path formed by the input voltage source Vin, the primary winding of transformer Tr, the clamp diode Dc1, the clamp capacitor Cr, and the energy extraction circuit composed of D1, D2, and VB, the simplified equivalent circuit shown in Figure 2b is obtained. This equivalent circuit reduces the system to the leakage inductance Llk, energized by a current Ilk, which at the initial moment corresponds to the current flowing through the primary winding when the main switch M1 turns off. This current is denoted Ilkp, representing the primary-side current of the transformer.

Analysis of the equivalent circuit depicted in Figure 2b leads to the following observations. When VB = 0, the leakage inductance Llk, energized by the current ILlk, maintains a constant current amplitude over time, assuming ideal conditions and neglecting losses. As VB increases to a value VB1, the current through Llk decreases with a slope proportional to VB1. For a higher voltage VB2, where VB2 > VB1, the decay rate of the leakage current becomes steeper. Therefore, an increase in VB directly results in a higher current downslope for ILlk.

The shaded area under the leakage current waveform shown in Figure 2c represents the charge “Ach” injected by the leakage inductance into the clamp capacitor Cr and into the junction capacitance of the clamp diode Dc1. As VB increases, the injected charge into Cr decreases correspondingly.

Figure 2d illustrates the current through the clamp capacitor Cr, ICr where a steeper downslope and a reduced conduction interval between t₁ and t₂ are observed. This behavior indicates a reduction in the charge injected into Cr during the interval t₁–t₂. By conservation of charge, a smaller injected charge during t₁–t₂ results in a reduced charge extracted from Cr during the interval t₂–t₃. Under these conditions, a unique situation will occur wherein the extracted charge becomes smaller than the reverse-recovery charge of the clamp diode Dc1.

When this condition is satisfied, the passive clamp formed by Dc1 and Cr effectively exhibits active-clamp-like behavior. This is substantiated by the waveforms shown in Figure 2d, where current flows into the clamp capacitor during t₁–t₂, followed by reverse current flow during t₂–t₃, analogous to the operation of an active clamp.

The current flowing through Cr, as shown in Figure 2d, is reflected into the secondary current, as observed in the synchronous rectifier (SR) current waveform. During the interval t₁–t₂, the secondary current exhibits a delayed rise with a reduced slope. During t₂–t₃, the reverse current associated with the reverse recovery of diode Dc1 is reflected into the secondary winding.

The reverse recovery characteristics of diode Dc1 allow reverse conduction because the reverse recovery charge of Dc1 is larger than the charge injected into Cr during t₁–t₂. This condition is enabled by the presence of the energy extraction circuit 100, which actively removes a portion of the leakage inductance energy and stores it in the voltage source VB.

As a result, the energy extraction circuit not only limits the charge stored in the clamp capacitor but also converts the conventional passive clamp into a functional equivalent of an active clamp, while simultaneously transferring a portion of the leakage energy to VB instead of dissipating it.

Experimental Waveforms

Figure 3a shows the power train of a flyback converter supplied from an AC line through a full-bridge rectification. The converter incorporates an Energy Extraction Module, which together with the clamp capacitor Cr and clamp diode Dr constitutes the “Rompower Passive Clamp” circuit as per [2], [3], [12].

Figures 3b and 3c present the measured drain-to-source voltage VdsM1of the primary switch M1 under low-line and high-line input conditions, respectively. Figure 3b corresponds to an input voltage of 115 Vac, while Figure 3c corresponds to an input voltage of 230 Vac.

For both input conditions, the voltage waveform across M1 is free of the high-frequency spikes and ringing typically observed in conventional passive clamp flyback converters. The apparent voltage overshoot following switch turn-off corresponds to the controlled reflection of the voltage VB generated by the “Energy Extraction Module”, superimposed on the rectified input voltage Vin.

These results indicate that the “Rompower Passive Clamp” effectively controls the leakage-inductance energy and limits the voltage stress on the primary switch while maintaining stable, non-oscillatory switch voltage waveforms over a wide AC input voltage range.

Advantages of using the “Energy Extraction Circuit” in Active Clamp, converting it in “Rompower Active Clamp”

In Figure 4 is presented the switch node of a flyback converter, containing the energy extraction circuit formed by D1, D2 and CB and a current injection circuit connected to VB.

The clamp circuit in Figure 4 is formed by an active clamp Mc and a clamp capacitor Cc. The active clamp switch Mc is replacing the passive clamp diode Dc. And this type of active clamp which contains the “Energy Extraction Circuit” is contained in the reference [1],[4],[5].

The presence of the “Energy Extraction Circuit” has the same advantages presented in the passive clamp also in the active clamp. One of the key advantages is the reduction of the charge injected in the clamp Cc. That also reduces the RMS current flowing through the active clamp formed by Mc and Cc.

Figure 5 shows the variation of the RMS clamp current as a function of the transformer leakage inductance of the transformer for “Rompower Active Clamp” in comparison to a conventional active clamp 

Across the examined range of transformer leakage inductance values, the “Rompower Active Clamp” topology exhibits a consistently lower RMS clamp current when compared with a conventional active clamp configuration. For a representative leakage inductance of 2 µH, the conventional active clamp presents an RMS clamp current of approximately 1.036 A, whereas the “Rompower Active Clamp” operates with an RMS clamp current of 0.665 A.

Since power dissipation in the clamp circuit is proportional to the square of the RMS current, the observed reduction in clamp current leads directly to a substantial reduction in clamp-related losses. For the operating point considered, the Rompower clamp achieves an approximately 2.4× reduction in power dissipation relative to the conventional active clamp topology.

Additional Advantages of the Rompower Active Clamp

In addition to reducing clamp-related losses, the “Rompower Active Clamp” topology provides further performance advantages when compared with conventional active clamp implementations. In conventional active clamp flyback converters, the interaction between the transformer leakage inductance and the clamp capacitor leads to a pronounced resonant behavior, resulting in a sustained oscillatory component in the clamp current. This phenomenon is referred to in this paper as “clamp ringing”.

The “clamp ringing” is magnetically coupled through the transformer to the secondary side of the converter, where it increases the RMS secondary current. The presence of this oscillatory current component may disturb the operation of the synchronous rectifier (SR), potentially affecting timing accuracy and control stability. These effects can degrade conversion efficiency and become increasingly significant under wide input-voltage operation or during dynamic load transients.

By contrast, the “Rompower Active Clamp” topology exhibits reduced RMS clamp current and a suppressed resonant interaction between the leakage inductance and the clamp capacitance. Consequently, clamp ringing is substantially attenuated, leading to reduced RMS secondary current, improved robustness of SR operation, and enhanced overall converter efficiency.

Figure 6a presents the current flowing through the synchronous rectifier (SR) on the secondary side of the converter, while Figure 6b shows the current flowing through the clamp capacitor and clamp switch for different values of the voltage VB generated by the “Energy Extraction Circuit”.

Three operating conditions are evaluated:

1. VB = 0 V, corresponding to a conventional active clamp configuration (blue waveform).

2. VB = 5 V, corresponding to partial energy extraction (green waveform).

3. VB > 5 V (VB = 30 V in the experimental results), corresponding to increased energy extraction.

In Figure 6a, the red dotted trace represents the ideal magnetizing current IM, which is transferred to the secondary winding immediately after the primary switch M1 turns off under ideal conditions with negligible transformer leakage inductance.

For VB = 0 V, the SR current exhibits a pronounced oscillatory component caused by the resonant interaction between the leakage inductance and the clamp capacitor. When VB = 5 V, the oscillatory component in the secondary current is significantly reduced and becomes critically damped, indicating partial extraction of the oscillation energy. For VB = 30 V, the oscillatory behavior is effectively eliminated, and the SR current closely follows the ideal magnetizing current IM, corresponding to the behavior expected for minimal leakage inductance.

Figure 6b shows the corresponding clamp current waveforms. For VB = 0 V, strong oscillations are observed due to resonance between the clamp capacitor and the leakage inductance. When VB = 5 V, these oscillations are substantially damped. As VB increases further, the amplitude of the clamp current decreases, confirming that the “Energy Extraction Circuit” progressively removes the energy associated with leakage-inductance-induced oscillations.

These results demonstrate that increasing VB reduces the charge injected into the clamp capacitor, suppresses clamp ringing, and causes the secondary current waveform to converge toward the ideal magnetizing current, thereby reducing RMS secondary current and improving synchronous rectifier operation.

The Solutions for ZVS in Flyback Topology

Zero-voltage switching (ZVS) in flyback topologies has become a necessity not only to improve efficiency by reducing switching losses, but also to eliminate excessive voltage stress on the secondary rectifiers. In hard-switching operation, when the primary switch turns on, a resonant circuit is formed between the leakage inductance of the transformer and the parasitic capacitance of the secondary rectifier.

During the turn-on transition of the primary switch, the voltage across the primary device collapses toward zero as the parasitic capacitance reflected to the primary side is discharged. This rapid voltage variation is reflected to the secondary side as a sharp voltage excitation scaled by the transformer turns ratio. The leakage inductance between the primary and secondary windings, together with the parasitic capacitance across the secondary rectifier, forms a quasi-resonant network. As a result, a voltage overshoot appears across the secondary rectifier, with a peak amplitude approaching twice the reflected voltage variation, followed by pronounced ringing.

Although the ringing component can be partially attenuated by a snubber network placed across the secondary rectifier, such a snubber has limited effectiveness in reducing the amplitude of the first voltage spike, which is typically the most severe contributor to device stress and reliability concerns. Consequently, effective suppression of the initial voltage spike requires the primary switch to turn on at a substantially reduced drain-to-source voltage, ideally under true zero-voltage switching conditions.

Two principal approaches can be identified for achieving ZVS of the primary switch in flyback converters. The first approach relies on manipulating the magnetizing current of the transformer, which is then used to discharge the parasitic capacitance across the primary switch prior to turn-on. The second approach employs a current-injection technique, wherein a narrow, quasi-resonant pulse of current is injected into the primary winding. This injected current actively discharges the parasitic capacitance reflected across the primary switch, and the extracted energy is redirected back to the input supply rather than being dissipated.

The development of wide-bandgap semiconductor technologies, such as SiC and GaN, was originally embraced and widely marketed as a solution for reducing switching losses, particularly in flyback converters operating at high frequency. While these devices marginally reduce turn-on and turn-off losses due to their superior switching characteristics, their use alone does not inherently resolve the fundamental issues associated with voltage spikes and ringing across the secondary rectifiers.

In high-voltage applications, such as flyback converters operating from an 800 V DC bus, the importance of ZVS becomes even more pronounced. Under these conditions, secondary-side voltage stress, electromagnetic interference, and device reliability considerations make zero-voltage switching of the primary switch not merely a performance enhancement, but a critical requirement for robust and scalable converter operation.

ZVS_01 Zero-Voltage Switching by manipulating the magnetizing current

Figure 7a depicts a flyback converter comprising a primary winding L₁, a secondary winding L₂, and an auxiliary winding L₃. The auxiliary winding is connected to an auxiliary capacitor C_injx and a controlled auxiliary switch M_injx. Analysis of the winding polarities shows that the auxiliary winding, together with C_injx and M_injx, constitutes a functional replica of the secondary rectification stage, which consists of the secondary winding, the output capacitor, and the synchronous rectifier.

The voltages appearing across the auxiliary components are directly proportional to the turns ratio between the auxiliary and secondary windings, N₃/N₂. Accordingly, the steady-state voltage across the auxiliary capacitor is given by

During the parasitic oscillation interval, an oscillatory voltage is observed across the primary switch, as illustrated by the drain-to-source voltage waveform V_DS(M1). These oscillations are caused by the resonant interaction between the primary inductance of the transformer and the parasitic capacitance associated with the primary switching node, which is the voltage across the main switch M1. During this interval, the current flowing through the primary winding corresponds to the transformer magnetizing current I_M, as shown in Figure 7b.

At a selected instant within the oscillation interval, a control pulse is applied to the gate of the auxiliary switch M_injx, causing it to turn on. As a result, a current is injected into the auxiliary winding L₃, driven by the voltage stored on the auxiliary capacitor C_injx. This injected current is forces the magnetizing current into a controlled negative slope ΔI_M.

The injection event causes the voltage at the primary switching node to transition abruptly to a level equal to V_in +nV_o, where n=N₁/N₂ denotes the primary-to-secondary turns ratio. The resulting voltage step is denoted as ΔV₂. This voltage transition leads to switching losses referred in this paper as “initial transition losses” 

In case when the auxiliary switch is activated near the valley of parasitic oscillation, the amplitude of ΔV₂ can be significant.

This voltage step at the primary switching node is associated with switching loss, regardless of whether the node transitions from a given voltage to ground or from ground to a given voltage. Therefore, the occurrence of a voltage transition alone—independent of its polarity—results in energy dissipation during the switching event.

When the auxiliary switch M_injx is turned off, the magnetizing current—now flowing in the negative direction—discharges the parasitic capacitance associated with the primary switching node. This discharge process reduces the voltage across M1 to a lower level V_x. In the ideal case, V_x approaches zero, thereby establishing true zero-voltage switching conditions for the subsequent turn-on of the primary switch M₁.

To minimize the switching losses associated with the voltage transition ΔV₂, the auxiliary switch M_injx is preferably turned on near the peak of the parasitic oscillation present at the primary switching node, as illustrated in Figure 8. Under this operating condition, the effective voltage step ΔV₂ shown in Figure 7b is reduced, thereby lowering the switching losses associated with the “initial transition losses”  loss mechanism.

Despite this advantage, the use of hill-triggered auxiliary switch turn on, there is an increase of the magnetic flux excursion within the transformer. The manipulation of the magnetizing current produces an incremental increase in the magnetizing current ΔI_M, which directly translates into an additional magnetic flux density variation ΔB in the transformer core. This additional flux component is superimposed on the normal flux swing that develops during the subsequent conduction interval of the main switch. As a result, the total peak-to-peak flux excursion within the magnetic core is increased which leads to higher core losses.

The increased flux swing reduces the available design margin and may impose limitations on transformer size, core material selection, and maximum achievable power density and reduces efficiency.

Accordingly, although hill-triggered activation of the auxiliary switch effectively reduces the “initial transition losses”, which are the switching losses associated with ΔV₂, it introduces a fundamental trade-off by increasing magnetic losses and overall converter stress.

In summary, this ZVS implementation method ZVS_01 exhibits the following limitations:

1. The auxiliary switch M_injx must be activated within a narrow time window corresponding to the peak of the parasitic oscillation, increasing sensitivity to component tolerances and operating-condition variations.

2. The injected current increases the magnetizing current by ΔI_M, resulting in an additional magnetic flux density variation ΔB and increased core losses.

3. The elevated flux swing reduces magnetic utilization margin and constrains scalability to higher voltage and higher power flyback applications.

4. Additional switching losses associated with “initial transition losses”.

In conclusion, the zero-voltage-switching approach based on the manipulation of the magnetizing current referred also as ZVS_01 fulfills its primary objective of enabling ZVS of the main power switch by actively discharging the parasitic capacitance associated with the primary switching node. This mechanism effectively eliminates hard-switching turn-on conditions and reduces voltage stress on the secondary rectifier.

Nevertheless, the realization of ZVS through magnetizing current manipulation is accompanied by inherent trade-offs. This solution increases the peak-to-peak magnetizing current, resulting in an enlarged magnetic flux excursion within the transformer core. This increased flux swing leads to elevated core losses and reduces the available magnetic design margin. In addition, under specific operating conditions, particularly at light load, the method introduces additional switching losses referred to as “initial transition losses”. These losses arise after a prolonged parasitic oscillation interval, during which the oscillation amplitude decays due to damping. The reduced oscillation peak causes the main switch to turn on at lower voltage levels , thereby increasing the associated switching losses, referred to also as “initial transition losses”.
Therefore, while the magnetizing-current manipulation represents an effective technique for achieving zero-voltage switching in flyback converters, its application entails increased magnetic losses and potential efficiency degradation especially under light-load conditions. These limitations motivate the exploration of alternative ZVS techniques that can decouple zero-voltage switching from increased magnetizing current and associated magnetic core losses.

ZVS_02 Zero-Voltage Switching by manipulating the magnetizing current, (Alexander’s solution)

Figure 9(a) illustrates a conventional flyback converter featuring a passive clamp network on the primary side and a synchronous rectifier on the secondary side. In this section, a standard passive clamp is deliberately employed instead of the Rompower passive clamp in order to focus exclusively on the analysis of the proposed zero-voltage-switching (ZVS) solution. Figure 9b presents the principal waveforms associated with the operation of the converter, including: the gate-drive signal of the primary switch V_G,M1, the primary switch current I_M1, the synchronous rectifier current I_SR, the gate-drive signal of the synchronous rectifier V_G,SR, the drain-to-source voltage of the primary switch V_DS,M1, and the magnetic flux density Bwithin the transformer core.

The analysis focuses on the operating intervals following the instant when the synchronous rectifier current reaches zero, denoted as t₀.

Operating Sequence

Prior to t₀, the primary switch M₁ conducts for a defined interval, during which the current through the primary winding increases from zero to a peak value. During this interval, energy is stored in the transformer magnetizing inductance, and the magnetic flux density increases to a peak value B₁. When the primary switch is turned off, the magnetizing current is transferred to the secondary side, delivering the stored energy to the load through the synchronous rectifier SR.

At time t₀, the current through the synchronous rectifier decays to zero and the synchronous rectifier turns off. This event marks the beginning of a dead-time interval during which neither the primary nor the secondary switch conducts.

During the dead time following t₀, the transformer primary inductance resonates with the parasitic capacitance associated with the primary switching node, which is the drain of M1. This interaction initiates a parasitic oscillation, which is observable in the drain-to-source voltage waveform V_DS(M1).

At time t₁, the synchronous rectifier is intentionally turned on. This action forces the voltage at the primary switching node to transition abruptly to a level equal to V_in+nV_o, where n=N₁/N₂ represents the primary-to-secondary turns ratio. This voltage transition constitutes a switching event and is associated with energy dissipation, in the form of switching losses, hereafter referred to as “initial transition loss”.

Between t₁ and t₂, the magnetic flux in the transformer is driven into the negative region, reaching a magnitude denoted as B₂. During this interval, a current builds up in the secondary winding, flowing toward the secondary dot. The magnitude of this secondary current is directly proportional to the instantaneous transformer flux, which reaches its negative peak at time t₂.

At time t₂, the synchronous rectifier is turned off. The secondary current is then reflected into the primary winding and flows toward the input voltage source. If the magnitude of this reflected current is sufficient, it discharges the parasitic capacitance associated with the primary switching node to zero, thereby establishing zero-voltage-switching conditions for the subsequent turn-on of the primary switch M₁.

In Figure 10a and Figure 10b are presented the key waveforms of the ZVS-2 in a flyback topology using “shorting switch”, wherein the energy contained in the parasitic oscillations is harvested, described in more details in [6]. In Figure 10a, the synchronous rectifier is turned on during the interval between t₁ and t₂, following the same control sequence described for Figure 9b. Under these operating conditions, the current developed in the secondary winding remains insufficient to fully discharge the parasitic capacitance associated with the primary switching node. As a result, the drain-to-source voltage of the primary switch does not reach zero prior to turn-on, and zero-voltage switching is not achieved, as indicated in Figure 10a.

When the conduction interval of the synchronous rectifier between t₁ and t₂ is increased, the magnitude of the current in the secondary winding correspondingly increases. The reflected current on the primary side provides the necessary energy to completely discharge the parasitic capacitance of the primary switch before turn-on. Consequently, zero-voltage switching of the primary device is achieved, as illustrated in Figure 10b.

These results demonstrate that, in magnetizing-current-based ZVS implementations, the achievement of ZVS is highly dependent on the duration of the synchronous-rectifier conduction interval. This dependence increases sensitivity to timing accuracy, operating conditions, and parameter variations, and represents a limitation of this approach.

Relation to ZVS solution by manipulation of the magnetizing-current 

The ZVS mechanism labeled ZVS_02 described above is functionally equivalent to the mechanism previously presented and labeled ZVS_01, wherein both methods are manipulating the magnetizing current. In ZVS_01, an auxiliary winding, auxiliary switch, and auxiliary capacitor form a functional replica of the secondary rectification stage. In the implementation shown in Figure 9a, the same effect is achieved without additional components, relying solely on synchronous rectifier control to inject energy into the magnetizing inductance.

This approach was identified, based on my knowledge, by the Rompower team in 2014 and implemented in an early flyback topology, using “shorting switch”, depicted in Figure 10a and Figure 10b. Due to the limitations discussed in the previous section, this method was not pursued for patent protection and was instead publicly disclosed by Rompower through presentations at PCIM and APEC seminar. In the ongoing pursuit of an optimal ZVS solution, subsequent development efforts were redirected toward a more advanced approach based on controlled current injection.

Advantages and Limitations

Compared to the auxiliary-winding implementation, the solution shown in Figure 9a offers advantages in reduced component count, lower cost, and reduced circuit complexity, as ZVS is achieved exclusively through control of existing switches. However, the method exhibits several inherent limitations:

1. The synchronous rectifier must be activated within a narrow time window corresponding to the peak of the parasitic oscillation, increasing sensitivity to component tolerances and operating-condition variations.

2. The additional negative flux excursion B₂ increases the total magnetic flux swing beyond B₁, resulting in increased transformer core losses.

3. The elevated flux swing reduces magnetic utilization margin and constrains scalability to higher-voltage and higher-power flyback applications.

4. The voltage transition at time t₁ introduces switching losses referred to as “initial transition losses”.

5. These initial transition losses become more pronounced when the synchronous rectifier is activated after a prolonged dead-time interval, during which the parasitic oscillation amplitude has decayed.

6. As with previously discussed magnetizing-current-based ZVS methods, ZVS cannot be achieved at the first voltage valley, which represents the most efficient operating point near critical conduction mode. In high-input-voltage applications (e.g., 800 V DC buses), the first valley may occur at a relatively high voltage, further limiting efficiency.

Quantitative Core-Loss Example

To quantify the magnetic penalty associated with this ZVS method, consider a 60 W flyback converter operating from an 800 V DC input bus and delivering a regulated 24 V output at a switching frequency of 100 kHz. The transformer employs 40 primary turns and 4 secondary turns (turns ratio 10:1) and is implemented using an EQ25 ferrite core manufactured from 3C95 material.

Under nominal operation without magnetizing-current ZVS, the transformer is designed for a peak magnetic flux density swing of approximately 220 mT, resulting in a calculated core loss of approximately 0.6 W. When the magnetizing-current-based ZVS method is applied (such as ZVS_01 or ZVS_02), the additional flux excursion increases the peak flux density to approximately 310 mT. At this operating point, the estimated core loss increases to approximately 1.2 W, representing a twofold increase relative to the baseline design.

This example demonstrates that, while magnetizing-current-based ZVS enables zero-voltage switching, it imposes a substantial magnetic loss penalty that directly impacts efficiency and thermal margin.

ZVS_03 ZVS Flyback Topology using Current Injection Method 

Figure 11a illustrates a flyback converter topology that employs a current injection technique to achieve zero-voltage switching (ZVS) of the primary switch. The converter comprises a transformer with a primary winding L₁, a secondary winding L₂, and an auxiliary current-injection winding L_inj . The primary side includes the main power switch M₁ and its associated parasitic output capacitance C_oss . On the secondary side, the circuit consists of a rectifier D_o, an output capacitor C_o, and a resistive load R_o.

In addition to the main power-processing components, the auxiliary winding L_inj is connected to a diode D_inj , an auxiliary voltage source V_inj , and a controlled current-injection switch M_inj . The auxiliary winding is implemented with a reduced number of turns, typically two turns in practical designs.

The current-injection switch M_inj is activated prior to the turn-on of the main switch M₁ by a controlled advance interval Δt. Upon activation of M_inj, a quasi-resonant network is formed by the parasitic capacitance C_oss of the main switch and the leakage inductance between the primary winding L₁ and the auxiliary winding L_inj . This interaction generates a quasi-resonant current pulse with an a half-sinusoidal waveform.

The resulting current pulse flows through the leakage inductance and is reflected into the primary winding, flowing toward the input voltage source. This injected current actively discharges the parasitic capacitance C_oss to zero prior to the turn on of M1 by V_c(M1) to the main switch. As a result, the main switch M₁ is turned on under zero-voltage conditions, thereby eliminating hard-switching turn-on losses and significantly reducing voltage stress for the secondary rectifier [8],[9],[10],[11].

The energy required for the current injection process is supplied by the auxiliary voltage source V_inj through the diode D_inj . Importantly, the current injection event is decoupled from the transformer magnetizing current, enabling ZVS to be achieved without increasing the magnetic flux excursion in the transformer core. This characteristic distinguishes the current injection method from magnetizing-current-based ZVS techniques and allows zero-voltage switching to be maintained over a wide operating range with minimal impact on magnetic losses.

Analytical Description of the Self-Adjusting Current Injection Mechanism

Figure 12a illustrates a flyback converter incorporating a current injection circuit for achieving zero-voltage switching of the primary switch. The current injection network is composed of an auxiliary winding N_inj integrated on the main transformer TR₁, a current injection switch M_inj, a current injection capacitor C_inj, a diode D_inj, and an auxiliary voltage source V_inj.

Figure 12c presents the principal parameters that define the behavior of the current injection circuit, including the resonant angular frequency ω, the effective turns ratio N_i, and the characteristic impedance Z_c of the quasi-resonant network formed by the leakage inductance between the primary winding and the current injection winding and the parasitic capacitance of the main switch. The analytical expression governing the injected current is provided in Figure 12d. In this formulation, the injected current I_inj is expressed as a function of the auxiliary voltage source V_inj, the effective turns ratio N_i, and the instantaneous voltage V_i across the main switch at the moment the current injection process is initiated. The voltage V_i corresponds to the drain-to-source voltage of the primary switch M₁ at the turn-on instant of the injection switch M_inj.

Inspection of the expression in Figure 12d shows that the amplitude of the injected current I_inj is directly proportional to V_i. Consequently, a higher voltage present across the main switch at the activation of current injection results in a larger injected current amplitude, while a lower initial voltage yields a proportionally smaller injected current.

Figure 12b illustrates the voltage across the current injection capacitor C_inj (denoted as V_inj+ – V_inj- ), the drain-to-source voltage of the main switch V_DS(M1), and the injected current I_inj for two operating conditions. The solid traces correspond to the case of a high initial drain-to-source voltage Vds(M1)H whereas the dashed traces represent a lower initial voltage Vds(M1)L. When current injection is initiated at a higher V_DS(M1), the injected current begins discharging the parasitic capacitance of the main switch, and the voltage across M₁ decreases until it reaches the level Vds(M1)x. The time required for this discharge, extending from t₀ to t₂, is longer in the high-voltage case, allowing the injected current to build up to a higher peak value at t₂. The injected current reaches its maximum when the drain-to-source voltage decays to Vds(M1)x, as defined by the analytical expression shown in Figure 12e. Beyond this point, the amplitude of the injected current decays.

From the analytical relationships presented in Figures 12c, 12d, and 12e, it can be concluded that the amplitude of the injected current inherently scales with the initial voltage V_i across the main switch. This self-adjusting characteristic represents a key advantage of the current injection technique, as it automatically adapts the injected current amplitude to the voltage level that must be discharged. As a result, the discharge of parasitic capacitance from the initial voltage V_i  to zero occurs over an approximately constant time interval across a wide range of operating conditions.

This self-regulating behavior eliminates the need for computation or adaptive control of the injected current amplitude, enabling implementation with a simple control strategy, including purely analog control solutions, while maintaining robust zero-voltage switching over wide input-voltage and load ranges.

Figures 13a and 13b illustrate the drain-to-source voltage of the main switch V_DS(M1), the injected current I_inj, and the gate-drive signal controlling the main switch. In both figures, the voltage across the main switch is shown in aqua color, while the injected current is shown in green.

In Figure 13a, the current injection is initiated when the voltage across the main switch is relatively high. Under this condition, and in accordance with the analytical relationship derived in Figure 12d, the injected current exhibits a relatively large peak amplitude. The increased current amplitude provides sufficient energy to rapidly discharge the parasitic capacitance associated with the main switch, thereby establishing zero-voltage-switching conditions prior to the application of the gate-drive signal.

Figure 13b illustrates the case in which the current injection is activated when the voltage across the main switch is low, corresponding to a voltage valley of the parasitic oscillation. As predicted by the same analytical model, the injected current amplitude is significantly reduced due to the lower initial drain-to-source voltage. Nevertheless, the injected current remains adequate to fully discharge the residual parasitic capacitance and ensure zero-voltage switching of the main switch.

In flyback converter applications, this inherent self-adjusting behavior of the current injection technique results in an automatic optimization of the ZVS process across a wide range of operating conditions. The injected current amplitude scales proportionally with the instantaneous voltage across the main switch, minimizing excess circulating energy while preserving robust zero-voltage switching.

When the current injection method is combined with valley detection and the injection pulse is synchronized with the voltage valley, the converter operates with minimal injected energy and reduced switching losses. This operating mode represents the most efficient implementation of zero-voltage switching in flyback converters, providing high efficiency, reduced device stress, and consistent ZVS performance over the full input-voltage and load range.

Ideal Flyback Topology

Figure 14a presents two representative voltage waveforms of a flyback converter operating with a conventional passive clamp and hard switching. The first waveform, shown in aqua, corresponds to the drain-to-source voltage of the primary switch Vds(M1). The voltage spikes and high-frequency glitches inherent to passive-clamp operation are identified by markers (#1) in Figure 14a.

The second waveform, shown in purple, represents the voltage across the synchronous rectifier on the secondary side. When the primary switch operates under hard-switching conditions, the abrupt voltage transitions and parasitic oscillations at the primary switching node are transferred through the transformer to the secondary side. Consequently, significant voltage spikes and high-frequency glitches are observed across the synchronous rectifier, in agreement with the mechanisms discussed earlier in this work. This phenomena result in increased voltage stress on the secondary devices and contribute to elevated electromagnetic interference, underscoring the limitations of conventional passive-clamp flyback converters operating in hard-switching mode.

Figure 14b presents the same set of waveforms as those shown in Figure 14a; however, the observed behavior differs due to the implementation of the two techniques introduced earlier in this work. By employing the “Rompower Passive Clamp”—or, in higher-power applications, the “Rompower Active Clamp”—a portion of the energy stored in the transformer leakage inductance is recovered rather than dissipated. As a result, the voltage spikes and high-frequency glitches previously observed across the main switch are effectively eliminated, as indicated by marker (#1′) in Figure 14b.

Furthermore, by applying the current injection technique described in this paper, the primary switch is turned on under zero-voltage conditions. Consequently, the voltage disturbances associated with hard-switching turn-on—previously visible in Figure 14a and identified by marker (#2)—are eliminated. As shown in Figure 14b, the voltage across the synchronous rectifier on the secondary side is free of spikes and ringing, exhibiting a smooth and well-controlled waveform.

The combined application of leakage-inductance energy recovery and current-injection-based zero-voltage switching fundamentally transforms the conventional flyback topology. The resulting converter exhibits waveforms that closely approach those of an idealized flyback converter, as illustrated in Figure 14b.

“Ideal Flyback” [7] is defined as a flyback topology that harvests energy from parasitic elements and reutilizes this energy to achieve zero-voltage switching (ZVS) of the switching devices. “In Power Conversion Field we identify a superior technology by analyzing the waveforms across the switching elements and by comparing the efficiency in reference to the competing technologies”. I have to mention also that the efficiency performance though function of the technology, it is also influenced by the skill level of the designer, especially in magnetic “Know How”. 

Importantly, the advantages of this combined approach extend beyond waveform quality alone. By eliminating voltage overshoot, suppressing parasitic ringing, and ensuring zero-voltage switching without increasing magnetic flux excursion, the proposed solution achieves superior overall efficiency. This conclusion is further substantiated by the experimental results presented in this work, which demonstrate that the integration of these two techniques yields the highest efficiency among the flyback converter implementations evaluated.

65W AC-DC Adapter Efficiency Versus Line

Experimental Efficiency Results and Architectural Evolution

Figure 15 presents the measured efficiency of several 65 W AC–DC adapters based on flyback topology. The Rompower 65 W adapter employs the “Ideal Flyback”, which is the combined application of the two techniques discussed in this work—leakage-inductance energy extraction and current-injection–based zero-voltage switching—which achieves the highest efficiency among the evaluated solutions. Throughout this document, this implementation is referred to as the “Rompower 65W”. Competitor #1 and Competitor #2 which represent state-of-the-art commercial flyback-based adapters with the highest reported efficiencies currently available on the market.

As shown in Figure 15, at low input voltage (90 Vac), the Ideal Flyback achieves an  efficiency of approximately 94%. This performance represents the highest efficiency reported for AC-DC adapters using flyback topology at this power level. Compared with Competitor #1 and Competitor #2, the proposed solution exhibits efficiency improvements of approximately 1.6% and 2.75%, respectively.

At high input voltage (230 Vac), the combined use of leakage-inductance energy recovery and current-injection–based ZVS continues to provide a substantial performance advantage. Under these conditions, the “Ideal Flyback” demonstrates efficiency gains of approximately 1.6% and 2.3% relative to Competitor #1 and Competitor #2, respectively. These results confirm that the efficiency benefits of the proposed approach are maintained across the full input-voltage range.

The experimental data clearly indicate that the combined application of the two proposed techniques not only eliminates voltage spikes and parasitic ringing but also delivers the highest overall conversion efficiency. Consequently, the “Ideal Flyback” represents an optimized flyback implementation in terms of both waveform quality and energy efficiency. The gap in between the “Ideal Flyback” technology and the rest of the technologies currently used is increasing much more for very high input voltage bus such as 800V. 

The experimental results confirm that the combined application of the two proprietary techniques completely eliminates voltage spikes and parasitic ringing across all switching elements, while simultaneously delivering the highest overall conversion efficiency among other evaluated solutions. This achievement establishes the “Ideal Flyback” as a breakthrough evolution of the traditional flyback topology, offering unmatched waveform quality and energy efficiency.

The advantages of the “Ideal Flyback“ technology become even more compelling in next-generation high-voltage systems. As input bus voltages increase—particularly in emerging 800 V architectures—the performance gap between the “Ideal Flyback“ and competing power-conversion technologies widens significantly, reinforcing its position as a superior solution for high-voltage, high-efficiency applications.

Demonstration Unit for “Ideal Flyback” topology  

A demonstration unit has been developed to evaluate the performance of a 60 W DC-DC converter operating over a wide input voltage range of 200 VDC to 1000 VDC and delivering a regulated 24 V output using the “Ideal Flyback” topology with passive clamp. This document presents the technical implementation details and the corresponding experimental measurement results.

The demonstration board employs a 1700 V SiC MOSFET in a TO-263-7L surface-mount package as the primary switching device and is based on a single-ended flyback architecture incorporating Rompower’s proprietary “Ideal Flyback” technology. The measured performance of this solution is benchmarked against two comparable reference designs developed by two leading semiconductor manufacturers.

The first reference design (Reference #1) utilizes the same 1700 V SiC MOSFET and operates in quasi-resonant mode (QRM), employing valley detection to synchronize switch turn-on in order to reduce switching losses. At full load, this reference converter operates at a switching frequency of approximately 110 kHz.

The second reference design (Reference #2) is based on a 1700 V GaN device and operates at a lower switching frequency of approximately 42 kHz while relying on a ZVS-based control approach to reduce switching losses. As previously analyzed, this method introduces penalties in magnetic core loss and requires a larger magnetic core. In this case, a PQ 26/20 core is used, featuring a cross-section approximately 25% larger than the EQ25 core employed in the Rompower design.

In contrast, the Rompower “Ideal Flyback” implementation uses the same power switch as Reference #1 while achieving superior performance with a smaller magnetic. The EQ25 core used in the Rompower demonstration unit has a volume approximately 35% smaller than the magnetic core used in Reference #1 and approximately 25% smaller than the core required by Reference #2, underscoring the technology’s advantages in power density. 

Key Waveforms

As stated earlier, “In Power Conversion Field we identify a superior technology by analyzing the waveforms across the switching elements and by comparing the efficiency in reference to the competing technologies”. Figure 17 shows the drain-to-source voltage of the primary switch operating from an input voltage of 1000 V using the “Ideal Flyback” topology. As can be observed, no voltage spikes or ringing occur at turn-off, demonstrating effectiveness of “Rompower Passive Clamp”.  The apparent voltage overshoot corresponds exclusively to the bias voltage V_B, which in this operating condition has an amplitude of 40 V.

Figure 18 illustrates the same waveform, with emphasis on the turn-on transition of the primary switch. Zero-voltage switching (ZVS), enabled by controlled current injection, is initiated on the valley of parasitic oscillation, where the drain-to-source voltage of the primary switch is approximately 800 V. The voltage across the main switch is completely discharged to zero prior to the application of the gate-drive signal (purple trace). This transition is completed within approximately 40 ns, ensuring true ZVS operation at turn-on.

Efficiency Comparison

Rompower efficiency (orange color), Reference #1 (green color) and Reference #2 (blue color). 

The reference #1 and Rompower use the same SiC switch. The difference in efficiency at 800V is 7%. 

The observation of the key waveforms together with the performance chart underlines the advantages of the “Ideal Flyback” technology. 

The two major drawbacks of the flyback topology are the negative impact of the leakage inductance which leads to spikes and glices on the main switch and the hard switching operation become more significant on the 800V bus. That translates to the additional voltage stress on the switching elements and the significant increase in switching losses.

In the quest for increasing the efficiency for AC-DC adapters operating from an input voltage level less than 400V the replacement of the silicon devices to GaNs have been seen a the main improvement for fighting switching losses. For that the GaN utilization in AC-DC adapters became a very strong trends, though the improvements in this replacement were relatively modest. In most of the AC-DC adapters using flyback topology the efficiency impact was around 0.3%. The impact of the parasitic capacitance of the transformer did not change and also the spikes and ringing across the secondary switching element was not affected by the introduction of GaN primary switchers. There wer other advantages associated with GaN technology such as lower driving losses, smaller size of the die which brough a competitive advantage. 

At operation for a very high input voltage bus these two  drawbacks of the flyback became much more significant and the two solutions for ZVS such as ZVS_01 and ZVS_02 surfaces their limitation such as “ initial switching losses “ and an significant increase in core loss. 

In high-voltage applications, such as flyback converters operating from an 800 V DC bus, the importance of ZVS becomes even more pronounced. Under these conditions, secondary-side voltage stress, electromagnetic interference, and device reliability considerations make zero-voltage switching of the primary switch not merely a performance enhancement, but a critical requirement for robust and scalable converter operation.

In high input voltage applications, the advantages of the “Ideal Flyback” become more visible. That is due to the fact that ZVS is done without a penalty and as important is that most of the energy to obtain ZVS is obtained by harvesting a potion of the leakage inductance energy which usually is dissipated in conventional passive clamp.

The total elimination of the spikes and glitches in “Ideal Flyback” reduces the voltage stress. 

Reference #1 is 1450V wherein 200V is the overshot value. For 1000V input voltage the derating for the main switch is 85.3%.

Reference #2 the voltage across the main switch is 1360V wherein 120V overshot. The derating for the main switch is 80%.

Rompower implementation using “Ideal Flyback” the voltage across the main switch is 1290V wherein 40V overshot. The derating for the main switch is 75%. 

In high-voltage applications, such as flyback converters operating from an 800 V DC bus, the importance of ZVS becomes even more pronounced. Under these conditions, secondary-side voltage stress, electromagnetic interference, and device reliability considerations make zero-voltage switching of the primary switch not merely performance enhancement, but a critical requirement for robust and scalable converter operation.

Key Waveforms at Full Power versus the Input Voltage 

The bellow pictures show the measured switching waveforms of the “Ideal Flyback” converter. The upper trace corresponds to the drain-to-source voltage of the primary switch, scaled at 500 V/div, while the lower trace shows the voltage across the secondary rectifier at 50 V/div.

The experimental results demonstrate clean switching transitions on both the primary and secondary sides, with no observable voltage overshoot or parasitic ringing. These waveforms confirm the effective suppression of leakage-inductance effect  and validate the soft-switching behavior of the “Ideal Flyback” topology.

Thermal Performance 

The following pictures are presenting the temperature on the key components at full load and Vin=1000V. There is no forced air for this test. The PCB copper area on the PCB is used for cooling. 

Observation:

The unit from Reference #1 has a power dissipation 152% larger than “Ideal Flyback” board.

The unit from Reference #2 has a power dissipation 95.9% larger than “Ideal Flyback” board.

Reference #2      Transformer Core temperature           79.8C

“Ideal Flyback"   Transformer Core temperature           64.1C

Reference #2      Transformer winding temperature           90.7C

“Ideal Flyback"   Transformer winding temperature           68.3C

Reference #2      Main Switch temperature           82C

“Ideal Flyback"   Main Switch temperature           66C

Conclusion

The “Ideal Flyback” Topology eliminates the spike and glitches on the main switch through energy extraction from the leakage inductance and that energy is used to obtain ZVS without penalties unlike the  ZVS 2 solution. This gives to “Ideal Flyback” implementation a boost in efficiency of 4% comparative to Reference #2 , wherein the power dissipation in the demo unit is twice as much as the “Ideal Flyback” demo unit.

The “Ideal Flyback” in comparison with Reference #1 demo unit reduces the power dissipation in the 60W demo board by 60% due to the use of ZVS without penalties, while reducing the spike voltage across the main switch by 80%. 

The “Ideal Flyback” in comparison with Reference #2 demo unit reduces the power dissipation in the 60W demo board by 50% due to the use of ZVS without penalties, while reducing the spike voltage across the main switch by 67%. 

Notification

Some of the technologies presented in this seminar may be the subject of patent applications or patents, please contact Rompower Energy Systems Inc. for further details.

References

[1] Ionel Jitaru, “Energy Recovery From The Leakage Inductance of the Transformer ," US Patent # 10,651,748 B2 May 12, 2020.

[2] Ionel Jitaru, “High Efficiency Passive Clamp ," US Patent # 10,972,014 B2 April 6, 2021.

[3] Ionel Jitaru, “High Efficiency Passive Clamp CIP," US Patent # 11,218,078 B2 Jan. 4, 2022

[4] Ionel Jitaru, “Energy Recovery From The Leakage Inductance of the Transformer CIP," US Patent # 11,277,073 B2

[5] Ionel Jitaru, “Energy Recovery From The Leakage Inductance of the Transformer CIP," US Patent # 11,368,097 B1

[6] Ionel Jitaru, “Energy Extraction From The Parasitic Elements in Power Converters," US Patent # 11,929,665 B2

[7] I. Jitaru, “Ideal Flyback," in PCIM-Europe Conference Keynote Paper, 2021.

[8] I.Jitaru, “ Self-Adjusting Current Injection Technology”, US Patent 10,574,148.

[9] I. Jitaru “Self-Adjusting Current Injection Technology CIP ” US Patent 11,165,360.

[10] I. Jitaru “self-Adjusting Current Injection Technology CIP” US Patent 11,671,027

[11] Mark Fahlenkamp” Voltage Conversion Method and Voltage Converter”US Patent 9,673,718 B2

[12] Vijay Phadke” Systems and methods of operation for power converters having series-parallel mode active clamps” US Patent 9,774,270 B2.