Floating Step-Down Converter with a Novel Lossless Snubber

Abstract

In this research, a step-down converter with a lossless snubber is proposed, and its output is floating; therefore, it can be applied to LED driving applications. Such a structure is a modification of the conventional buck converter by adding a resonant capacitor, a resonant inductor, and two diodes to form this lossless snubber to reduce the switching loss during the switching period. Although the efficiency improvement in this circuit is not as good as the existing soft switching circuits, this circuit has the advantages of simple structure, easy control, and zero voltage switching (ZVS) cutoff.

1. Introduction

The switching power supply has the advantages of small size, high power density, and high efficiency and can be divided into hard switching and soft switching. Regarding the hard switching, when the switch is on or off, the corresponding voltage or current in it will have a cross area, so there is a switching loss. Considering the soft switching, the voltage on the switch ideally drops to zero first, and then its current rises from zero, while the current in the switch ideally drops to zero first, and then its voltage rises from zero. The first is called zero voltage switching (ZVS), occurring when the switch is turned on. The last is called zero current switching (ZCS), occurring when the switch is turned off. By means of these strategies, the switching loss of the switch can be reduced.

Consequently, various circuit topologies based on soft switching have been proposed to improve the system efficiency and performance. In the following sections, several common soft switching types are introduced.

1.1. Quasi Resonant Converter

The quasi-resonant resonant converter (QRC) is also called a semi-resonant converter or switch resonant converter [1,2]. The pulse frequency modulation (PFM) is usually used for control. Although ZCS or ZVS can be realized by the above method to improve the efficiency of the converter, this circuit will have higher switch stress when switching, which will lead to larger conduction loss.

1.2. Load Resonant Converter

References [3,4] present a type of series resonant converter (SRC). Reference [3] comments that although the symmetric PWM control can realize a wide output voltage range, it cannot ensure soft switching in some cases, such as when the range of load variation is large or when the switching frequency deviates from the resonant frequency too much. Reference [4] proposes a new strategy controlled by PWM that allows a wider voltage gain range. References [4,5] propose a type of series resonant converter (SRC). Reference [5] belongs to the type of parallel resonant converter (PRC) and presents steady-state analysis of the parallel resonant converter by phase shift modulation. Reference [6] proposes a simple technique to approximate the fundamental waveforms for the series–parallel converter (SPRC). References [7,8] present a type of LLC resonant converter. Reference [7] presents a synchronous rectifier driving scheme for the LLC converter based on secondary rectification current emulation. Reference [8] proposes an adjustable variable inductor to change the resonant frequency points so as to change the operating region of the LLC converter.

1.3. Switching Loss Reduction Based on the Snubber

Usually, a snubber is mainly constructed by a resonant inductor and a resonant capacitor. The structure of a turn-on snubber connects the resonant inductor in series with the main power switch, while the structure of a turn-off snubber connects the resonant capacitor in parallel with the main power switch. The structure can be categorized into active and passive types according to whether the structure contains an active power switch or not.

As for the active snubber, it will have an additional auxiliary switch and its auxiliary circuit, and before the main power switch is turned on or/and off, the auxiliary switch will act first, together with the auxiliary circuit, to transfer the energy from the main power switch to the auxiliary circuit so as to achieve soft switching. The concept is called zero voltage transition (ZVT) and zero current transition (ZCT) and can also reduce the stress on the main power switch. But the disadvantage is that it requires additional switching costs and complex control strategies. Reference [9] presents the ZVT technology; the auxiliary switch is not yet able to achieve soft switching, resulting in additional switching losses and limited efficiency improvement. Reference [10] proposes a dual-switch structure that can realize high boost voltage; however, its analysis is too complicated and requires an additional equivalent circuit.

As for the passive snubber, it is composed of passive components and will be the trend today because it simplifies the complexity of the circuit and the control strategy. Depending on whether this snubber uses a resistor or not, it can be categorized as lossy or lossless. The most common of the former are the RC snubber [11] and the RCD snubber [12]. Reference [11] proposes an exhaustive design method for the RC snubber by deriving the frequency-domain equations of conduction and cutoff and checking the position of the poles to reduce the oscillations. Although parasitic components are taken into account, this algorithm requires a large amount of memory and does not analyze the efficiency when the value of the parasitic components is too small. Reference [12] presents a new type of RCD snubber for the full-bridge DC-DC converter. Although the lossy snubber has a certain degree of effect on suppressing spikes and oscillations, the efficiency improvement of the circuit is limited to a certain extent due to the existence of its resistance, which causes a certain degree of loss. References [13,14,15] propose a boost circuit with a passive lossless snubber. Although the equations of the three modes in [13] are analyzed, only mode 2 can achieve ZVS cutoff, and the other two modes are only near-ZVS cutoff. The circuit mentioned in [14] has both ZVS turn-off and ZCS turn-on for the main power switch. But such a circuit requires additional snubber components in the power transfer path, so there is a higher conduction loss, and furthermore, due to the need to withstand high stresses, the component cost is high. Reference [15] proposes two snubbers to minimize the stress without giving extra stress to the primary power switch, but they use too many components and cause extra losses, so the efficiency improvement is limited, and the main improvement in the stress problem is only for light-load conditions.

In this paper, a novel lossless snubber is proposed and applied to the floating step-down converter with an LED string used as a load. This circuit is mainly derived from [13], which presents a boost converter topology. The proposed circuit structure is very suitable for being applied to low-noise optoelectronic systems such as SPADs [16,17,18], time-resolved circuits [19,20,21], or wearable EMG platforms [22,23,24]. Incidentally, the proposed circuit can be controlled by the fixed frequency with a smaller number of components used to achieve high efficiency by ZVS cutoff.

2. Main Power Stage

In the following section, the proposed circuit structure and its operating principle will be illustrated.

2.1. Circuit Structure and Its Definitions

Figure 1 shows a step-down converter with the proposed lossless snubber, whose circuit consists of a main power switch Q₁, three diodes D₁, D₂, and D₃, an output capacitor C_o, a resonant capacitor C_r, an output inductor L_o, and a resonant inductor L_r, while the load side consists of an LED string called LS1.

Prior to analyzing the circuit operation, a brief explanation of the definition of the associated symbols and the required assumptions is presented:

(1)

V_in is the input DC voltage, and the output capacitor C_o is large enough, so it is viewed as a voltage source, named V_o.

(2)

The main power switch Q₁, the main diode D₁, the output inductor L_o, and the output capacitor C_o.

(3)

The load is constructed by some LEDs connected in series, named LS₁.

(4)

For the main power switch Q₁, its parasitic output capacitance C_oss is taken into account, and the on-time of Q₁ is expressed as DT_s, where D is the duty cycle and T_s is the switching period.

(5)

The lossless snubber is built up by the resonant inductor L_r, the resonant capacitor C_r, and the diodes D₂ and D₃.

(6)

i_ds is the current in Q₁, i_Lr is the current in L_r, i_Lo is the current in L_o, and i_D1 and i_D3 are the currents in D₁ and D₃, respectively.

(7)

v_ds is the voltage on Q₁, and v_Cr is the voltage on C_r.

(8)

I_Lrk and V_Crk are the initial values of L_r and C_r, respectively, where the value of k can be 0, 1, 2, 3, or 4.

(9)

All components are regarded as ideal except the main power switch.

(10)

The circuit is operated in the continuous conduction code (CCM). Figure 2 shows the illustrated waveforms of the circuit over one switching period, and there are five operating states to be described as follows.

2.2. Operating Principle

The following will describe the operating principle of the proposed circuit:

State 1: [t₀ ≤ t < t₁]

As shown in Figure 3, at t = t₀, the switch Q₁ conducts, and the diode D₂ conducts, but D₁ and D₃ are cut off. At this time, the voltage of V_in − V_o across L_o increases, causing L_o to be magnetized. At the same time, there is a resonant loop from V_in to V_o to D₂ to L_r to C_r to Q₁ and then to V_in. During this state, the input terminal will temporarily store energy in the snubber. At time t = t₁, the resonant capacitor voltage v_Cr is charged to V_in, and the diode D₃ conducts, proceeding to the next state. Note that the snubber components L_r and C_r do not have any initial value, so the equivalent circuit is shown in Figure 4.

From the results of Figure 3 and Figure 4, the resonant capacitor voltage v_Cr(t), the resonant inductor current i_Lr(t), the characteristic impedance Z_a, and the resonant radian frequency ω_a can be expressed as follows:

$$v_{Cr}(t)=(V_{in}-V_o)[1-\cos {\omega }_a(t-t_0)]$$

(1)

$$i_{Lr}(t)={\displaystyle \frac{V_{in}-V_o}{Z_a}}\sin {\omega }_a(t-t_0)$$

(2)

$$Z_a=\sqrt{{\displaystyle \frac{L_r}{C_r}}}$$

(3)

$${\omega }_a={\displaystyle \frac{1}{\sqrt{L_r{C}_r}}}$$

(4)

Finally, by substituting the boundary condition of the resonant capacitor voltage v_Cr(t), the expression of the experienced time interval Δt₁₀ can be obtained as shown in (6):

$$V_{Cr1}=(V_{in}-V_o)[1-\cos {\omega }_a(t_1-t_0)]=V_{in}$$

(5)

$$\Delta t_{10}=t_1-t_0={\displaystyle \frac{1}{{\omega }_a}}{\cos }^{-1}({\displaystyle \frac{-V_o}{V_{in}-V_o}})$$

(6)

State 2: [t₁ ≤ t < t₂]

As shown in Figure 5, the switch Q₁ conducts, and the diodes D₂ and D₃ conduct, but D₁ cuts off. At this time, the voltages across the output inductor L_o and the resonant inductor L_r are V_in − V_o and −V_o, respectively, so the output inductor L_o and the resonant inductor L_r are in the magnetizing and demagnetizing states, respectively. At time t = t₂, due to the existence of diode D₂, the resonant inductor current i_Lr will not flow backwards after it drops to zero, proceeding to the next state.

$$I_{Lr1}={\displaystyle \frac{(V_{in}-V_o)}{Z_a}}\sin {\omega }_a(t_1-t_0)={\displaystyle \frac{\sqrt{V_{in}^2-2V_{in}{V}_o}}{Z_a}}$$

(7)

$$I_{Lr2}=I_{Lr1}-{\displaystyle \frac{V_o}{L_r}}(t_2-t_1)=0$$

(8)

$$\Delta t_{21}=t_2-t_1={\displaystyle \frac{I_{Lr1}{L}_r}{V_o}}={\displaystyle \frac{L_r\sqrt{V_{in}^2-2V_{in}{V}_o}}{V_o{Z}_a}}={\displaystyle \frac{\sqrt{V_{in}^2-2V_{in}{V}_o}}{V_o{\omega }_a}}$$

(9)

It is worth mentioning that Δt₂₀ = Δt₂₁ + Δt₁₀ ≠ π/ω_a because state 2 is not a resonant state.

State 3: [t₂ ≤ t < t₃]

As shown in Figure 6, the switch Q₁ conducts, and the diodes D₁, D₂, and D₃ are cut off. At this time, the voltage across the output inductor L_o is V_in − V_o, so the output inductor L_o is magnetized. At time t = t₃, the switch Q₁ is cut off, and due to the presence of resonant capacitor C_r, the slope of the rise in the voltage v_ds will begin to slowly go up, so the switch Q₁ can realize ZVS cutoff.

Based on the result of state 1 and state 2, the expression of the experienced time interval Δt₃₂ can be obtained as

$$\Delta t_{32}=t_3-t_2=D{T}_s-\Delta t_{20}$$

(10)

State 4: $[t_4\le t\le t_5]$

As shown in Figure 7, at the moment t = t₃, the switch Q₁ is cut off at approximately zero voltage, and the diode D₃ conducts, but the diodes D₁ and D₂ are cut off. The energy stored in the snubber will be released to the output, and the following equation can be obtained to be

$$v_{Cr}(t)=V_{in}-v_{ds}(t)$$

(11)

In this state, although C_oss, C_r, and L_o are in resonance, the output inductance is large enough that the output inductance L_o can be regarded as a current source with the value of I_{Lo_peak}, as shown in the equivalent circuit in Figure 8.

From Figure 8, it can be seen that the superposition theorem is required to solve the circuit equations. The expressions for the resonant voltage v_{Cr_Vin}(t) contributed by V_in and the switch voltage v_{ds_Vin}(t) contributed by V_in can be shown as follows:

$$v_{Cr\_Vin}(t)\approx V_{in}$$

(12)

$$v_{ds\_Vin}(t)\approx 0$$

(13)

The expressions for the resonant capacitor voltage v_{Cr_ILo_peak}(t) contributed by I_{Lo_peak} and the switch voltage v_{ds_ILo_peak}(t) contributed by I_{Lo_peak} can be directly utilized in the current divider, as described below:

$$v_{Cr\_ILo\_peak}(t)=-{\displaystyle \frac{I_{Lo\_peak}\times \frac{C_r}{C_r+C_{oss}}}{C_r}}(t-t_3)$$

(14)

$$v_{ds\_ILo\_peak}(t)={\displaystyle \frac{I_{Lo\_peak}\times \frac{C_{oss}}{C_r+C_{oss}}(t-t_3)}{C_{oss}}}$$

(15)

Afterwards, the expressions for the voltages v_Cr and v_ds can be obtained by adding the contributions of the two power sources, and the experienced time interval Δt₄₃ can be obtained by substituting the boundary condition of the resonant capacitor voltage v_Cr, and the associated expressions are as follows:

$$\begin{array}{ll}{v}_{Cr}(t)& =V_{in}-{\displaystyle \frac{I_{Lo\_peak}\times \frac{C_r}{C_r+C_{oss}}}{C_r}}(t-t_3)\\ & =V_{in}-{\displaystyle \frac{I_{Lo\_peak}}{C_r+C_{oss}}}(t-t_3)\end{array}$$

(16)

$$v_{ds}(t)={\displaystyle \frac{I_{Lo\_peak}\times \frac{C_{oss}}{C_r+C_{oss}}(t-t_3)}{C_{oss}}}={\displaystyle \frac{I_{Lo\_peak}}{C_r+C_{oss}}}(t-t_3)$$

(17)

$$V_{Cr4}=0=V_{in}-{\displaystyle \frac{I_{Lo\_peak}}{C_r+C_{oss}}}(t_4-t_3)$$

(18)

$$\Delta t_{43}=t_4-t_3={\displaystyle \frac{V_{in}(C_r+C_{oss})}{I_{Lo\_peak}}}$$

(19)

State 5: [t₄≤ t < t₀ +T_s]

As shown in Figure 9, the switch Q₁ is turned off, the diode D₁ is on, but the diodes D₂ and D₃ are turned off. At this time, the voltage across the output inductor L_o is −V_o, so the output inductor L_o is in the demagnetization state, and at the moment t = t₀ + T_s, the switch Q₁ conducts, and this state ends.

2.3. Voltage Gain

To simplify the analysis, the time interval of state 4 of the circuit is too short, so the experienced time interval in state 4 can be ignored. According to the volt–second balance of the output inductor L_o from (8), the voltage gain can be obtained to be (9):

$$(V_{in}-V_o)D{T}_s=(-V_o)(1-D)T_s$$

(20)

$${\displaystyle \frac{V_o}{V_{in}}}=D$$

(21)

It is worth mentioning that the voltage gain is identical to that of a conventional buck converter.

2.4. Boundary Current Condition for Output Inductor L_o

The operation of the boundary mode of the output inductor L_o is analyzed as in a conventional buck converter. The current ripple Δi_Lo flowing through the output inductor L_o can be expressed as follows:

$$\Delta i_{Lo}={\displaystyle \frac{v_{Lo}\Delta t}{L_o}}={\displaystyle \frac{V_o(1-D)T_s}{L_o}}$$

(22)

According to Kirchhoff’s Current Law (KCL), the average current on the capacitor is zero; it can be obtained that

$$I_o=i_{Lr}(t)+i_{Lo}(t)-i_{Co}(t)=I_{Lr}+I_{Lo}-I_{Co}=I_{Lr}+I_{Lo}$$

(23)

Therefore, the expression for calculating the DC component of the resonant inductor current I_Lr is first derived, i.e.,

$$\begin{array}{ll}{I}_{Lr}& ={\displaystyle \frac{1}{T_s}}{\displaystyle {\int }_{t_0}^{t_0+T_s}{i}_{Lr}(t)}dt={\displaystyle \frac{1}{T_s}}\left[{\displaystyle {\int }_{t_0}^{t_0+\Delta t_{10}}{i}_{Lr}(t)}dt+{\displaystyle \frac{I_{Lr1}\Delta t_{21}}{2}}\right]\\ & ={\displaystyle \frac{1}{T_s}}\left[{\displaystyle {\int }_{t_0}^{t_0+\frac{1}{{\omega }_a}{\cos }^{-1}(\frac{-V_o}{V_{in}-V_o})}\frac{V_{in}-V_o}{Z_a}\sin {\omega }_a(t-t_0)}dt+\left.L_r{\displaystyle \frac{V_{in}^2-2V_{in}{V}_o}{2V_o{Z}_a^2}}\right]\right]\end{array}$$

(24)

As I_Lo > 0.5Δi_Lo, L_o will be operated in the current conduction mode (CCM), i.e.,

$$I_{Lo}>0.5\Delta i_{Lo}={\displaystyle \frac{V_o(1-D)T_s}{2L_o}}$$

(25)

$$L_o>{\displaystyle \frac{V_o(1-D)T_s}{2I_{Lo}}}={\displaystyle \frac{V_o(1-D)T_s}{2(I_o-I_{Lr})}}$$

(26)

3. Power Component Design

Table 1. System specifications of the proposed circuit.

System Operation Mode	CCM
Nominal input voltage (Vin)	300 V
Nominal output voltage (Vo)	135 V
Nominal output current (Io,rated)	2 A
Minimum output current (Io,min)	0.4 A
System switching frequency (fs)	100 kHz

is used to describe the specifications of the proposed circuit.

3.1. Design of Output Inductor L_o

In order to allow the voltage on the resonant capacitor C_r to be charged to V_in in state 1, Equation (27) must be satisfied, as shown below:

$$\begin{array}{l}2(V_{in}-V_o)>V_{in}\\ {\displaystyle \frac{V_o}{V_{in}}}=D<0.5\end{array}$$

(27)

Therefore, the ideal duty cycle is at 45%.

$$I_{Lo,min}\ge I_{LoB}={\displaystyle \frac{\Delta i_{Lo}}{2}}$$

(28)

$$\Delta i_{Lo}={\displaystyle \frac{(V_{in}-V_o)\times D}{L_o\times f_s}}$$

(29)

Therefore, (28) and (29) are used to obtain the inequality of the output inductance as follows:

Therefore, the value of 1.26 mH is selected as the output inductor L_o.

$$L_o\ge {\displaystyle \frac{(V_{in}-V_o)\times D}{2\Delta I_{Lo,min}\times f_s}}={\displaystyle \frac{(300-135)\times 0.45}{2\times 0.4\times 100 \mathrm{k}}}=928 \mathsf{\mu }\mathrm{H}$$

(30)

3.2. Design of Output Capacitor C_o

By analyzing the operating principle of the proposed converter in Section 2.2, it can be seen that when the switch is turned off, according to the KCL law,

$$i_{Co}=i_{Lo}-I_o$$

(31)

Since the DC component of the output inductor current I_Lo is approximately equal to the output current I_o, the maximum current that can be discharged from the output capacitor is half of the output inductor current ripple. Incidentally, the maximum voltage ripple Δv_Co of the output capacitor C_o is set to be within 0.05% of the output voltage, so the following equation can be obtained:

$$\Delta v_{Co}={\displaystyle \frac{\frac{1}{2}\times \frac{\Delta i_{Lo}}{2}\times \frac{T_s}{2}}{C_o}}\le V_o\times 0.05 \%$$

(32)

By rearranging (32), (33) can be obtained as follows:

$$\begin{array}{ll}{C}_o& \ge {\displaystyle \frac{\Delta i_{Lo}}{8\times V_o\times 0.05 \%\times f_s}}\\ & ={\displaystyle \frac{(V_{in}-V_o)\times D}{8\times V_o\times 0.05 \%\times L_o\times f_s^2}}=73.33 \mathsf{\mu }\mathrm{F}\end{array}$$

(33)

Therefore, a 100 μF Rubycon electrolytic capacitor is selected as the output capacitor C_o.

3.3. Design of Resonant Tank Components C_r and L_r

As compared with the original circuit without the snubber, the design aims to extend the time Δt₄₃ so that the resonant capacitor is slowly discharged, thereby making the slope of the rise in the voltage v_ds on the main power switch Q₁ slower. However, Δt₄₃ should not be longer than the cutoff time of Q₁. As the output capacitor of Q₁, C_oss, is charged, when there is no snubber, the peak inductor current I_Lo,peak is rapidly charged to C_oss, while the proposed circuit is charged to the equivalent C_eq as shown in (48), so we hope that C_eq can be much larger than C_oss, that is,

$$C_{eq}=C_{oss}+C_r>10C_{oss}$$

(34)

By considering the worst case of (35), substituting the light-load I_Lo,peak into (35) can obtain the inequality of the value of C_eq, as shown in (36):

$$\Delta t_{43}={\displaystyle \frac{V_{in}{C}_{eq}}{I_{Lo,peak}}}< (1-D)T_s$$

(35)

$$\begin{array}{ll}{C}_{eq}& <{\displaystyle \frac{(1-D)(I_{Lo,min}+\frac{\Delta i_{Lo}}{2})}{V_{in}{f}_s}}\\ & ={\displaystyle \frac{0.55(0.4+0.2475)}{300\times 100 \mathrm{k}}}=11.87 \mathrm{nF}\end{array}$$

(36)

Eventually, the resonant capacitor C_r is selected to be two 1nF Ycap capacitors connected together.

In order to allow the current in the resonant inductor to be demagnetized before the cutoff of the switch, it is necessary to meet the requirements as shown in (37):

$${\displaystyle \frac{1}{{\omega }_a}}\left[{\cos }^{-1}({\displaystyle \frac{-V_o}{V_{in}-V_o}})+{\displaystyle \frac{\sqrt{V_{in}^2-2V_{in}{V}_o}}{V_o}}\right]< D{T}_s$$

(37)

By substituting the system specifications in Table 1 and the designed value of C_r into (37), we can obtain

$$\begin{array}{l}\sqrt{L_r{C}_r}\left[2.529+0.703\right]< 4.5\times {10}^{-6}\\ \hfill \Rightarrow L_r<969.29 \mathsf{\mu }\mathrm{H}\hfill \end{array}$$

(38)

3.4. Circuit Components Used

Table 2. Circuit components used.

Components	Specifications
MOSFET switch Q1	IPA60R380P6
Schottky diodes D1, D2, D3	C6D0605A
Resonant capacitor Cr	Y Cap: 2 nF/400 V
Resonant inductor Lr	PQ2020 core: 540 μH
Output capacitor Co	Polymer capacitor: 100 μF/450 V
Output inductor Lo	PQ3535 core: 1.257 mH
Gate driver	TLP250H

shows the components used in this circuit.

4. Experimental Results

In this section, the experimental results are displayed as follows:

From Figure 10, Figure 11, Figure 12, Figure 13 and Figure 14 it can be seen that the rising slope of the voltage v_ds on the switch Q₁ can be slowed down during the switch cutoff period, so the ZVS cutoff can be achieved. Furthermore, the lighter the load, the more obvious this phenomenon. The reason is that when the load is lighter, the current flowing through the resonant capacitor C_r, i_Cr, is also smaller, and therefore the resonant capacitor C_r discharges for a longer period of time, thereby making the rise time of the voltage v_ds on the switch Q₁ longer. In addition, some small high-frequency oscillation current can be seen during the switch cutoff period, which can be attributed to parasitic components.

From Figure 11 and Figure 15, the current spike in the resonant current i_Lr is due to the reverse recovery current of the diode at 100% and 10% loads.

In Figure 12 and Figure 16, the voltage–second balance of the output inductor V_Lo at 100% and 10% loads can be seen.

From Figure 13 and Figure 17, it can be seen that the output voltage V_o can be stabilized at a certain value under closed-loop control, while the output current I_o has a more obvious ripple at light load. Incidentally, the power switch Q₁ on the v_gs at 100% load and 20% load under the duty cycle D is about 45% and 40.2%, respectively. Therefore, the larger the load, the larger the duty cycle due to the parasitic components.

5. Efficiency Measurement

Figure 18 shows the block diagram of the proposed circuit for efficiency measurement. Firstly, a current-sensing resistor (shunt) is connected to the input and output current paths, and then a digital meter (Fluke 179, Fluke, Everett, WA, USA) is used to measure the voltage across the current-sensing resistor in order to measure the input and output currents, and meanwhile, the input and output voltages are measured using a digital meter to obtain the input power P_in and the output power P_o. Eventually, the input and output powers are used to calculate the efficiency of the actual circuit operation.

From Figure 19, it can be seen that the efficiency is above about 95% all over the load range with the proposed snubber used. From this figure, it also can be seen that when the proposed circuit is operated without the proposed snubber, it has a better efficiency of about 90% at 100% load, but for the hard switching, the efficiency is not high, especially at 20% load, where the efficiency is only about 69%.

6. Literature Comparison

The circuit in reference [13] has an output power of 340 W at full load with an input voltage of 20 V to 60 V and an output voltage of 48.5 V, with a maximum efficiency of about 96.62%. The circuit in [14] has an output power of 220 W at full load with an input voltage of 48 V and an output voltage of 96 V, with a maximum efficiency of about 96.4 %. The circuit in [15] has an output power of 200 W at full load, an input voltage of 380 V, and an output voltage of 24 V, with a maximum efficiency of about 94.5%.

References [13,14,15] all use lossless snubbers to achieve the near-ZVS or near-ZCS function.

Table 3. Literature comparison.

	Ref. No.	[13]	[14]	[15]	Proposed Circuit
Items
Rated output power (W)		340	220	200	270
Rated output voltage (V)		48.5	96	24	135
Switching frequency (kHz)		100	100	100	100
No. of snubber components (capacitor/inductor/diode)		1/1/2	2/1/2	2/2/4	1/1/2
Turn ON/turn OFF		$\times$/ZVS	ZCS/ZVS	ZCS/ZVS	$\times$/ZVS
Peak efficiency		96.6%	96.4%	94.5%	97 %
Circuit type		Boost	Boost	Buck	Buck
Snubber components on main power path		No	Yes	Yes	No

compares seven items: (1) rated output power, (2) rated output voltage, (3) switching frequency, (4) the number of snubber elements, (5) whether the ZVS/ZCS function is available, (6) the peak efficiency, (7) the type of converter circuit, and (8) whether the snubber elements are located on the main power path. From Table 3, it can be seen that compared with references [13,14,15], the proposed circuit in this paper has the highest peak efficiency of 97.0%, the least number of snubber elements, and none of the snubber elements located on the main power path.

7. Conclusions

A step-down converter with a lossless snubber is presented herein, and its output is floating; therefore, it can be applied to LED driver applications. As can be seen in Figure 18, the proposed circuit shows a significant improvement in efficiency all over the load range, about 96% at 100% load, which is due to circuit structure and operation simplicity. That is to say, there are fewer oscillations in the circuits, none of the snubber elements are in the main power stage path, and there is no energy stored in the snubber before the switch is turned on.

The following will be studied in the future:

(1)

The applications of the LSTM model to optimize the switching frequency and duty cycle in real time according to load variation [25,26].

(2)

How the proposed converter can benefit from integrating learning-based control models such as the LSTM to enhance adaptability in variable-load environments [26,27].

(3)

How energy functionals [28,29] or FEM simulations [30] could be adopted to optimize the resonant inductor design and assess magnetic saturation.

(4)

The use of soft computing techniques to dynamically optimize the snubber’s parameters in response to load variations [31].