An Active Output Filter with a Novel Control Strategy for Passive Output Filter Reduction

Abstract

This paper presents a novel control strategy for passive output filter reduction using an active output filter. The proposed method achieves the dual-function of regulating the output voltage ripple and output voltage variation during load transients. The novel control strategy allows traditional simple voltage controllers to be used, without requiring the expensive current sensors and complex controllers used in conventional approaches. The proposed method is verified with results from a 125-W forward converter.

NOMENCLATURE

 

I. INTRODUCTION

As power semiconductor technology matures, passive output filters (POF), such as LC-filters, accounts for a more significant part of the cost and volume of power converters. Therefore, reducing the POF is one of the most important issues in the design of power converters. However, this reduction is becoming increasingly difficult due to the demanding requirements of digital devices in terms of voltage regulations and efficiency [1]-[3]. Typically, voltage regulations are constraints on the output voltage ripple and on the output voltage fluctuations during load transients [1], [2]. High-frequency switching satisfies both of these regulations. However, the switching frequency is practically upper-bounded due to a proportionally increasing power loss. Thus, a number of studies have been conducted to reduce the POF without high-frequency switching.

For voltage ripple regulation, a passive ripple filter (PRF), an active ripple filter (ARF), and an active output filter (AOF) have been introduced [4]-[13]. The PRF and the ARF show good performance with regard to reducing ripples. However, these methods are overly sensitive to parameter errors [4]-[7], are ineffective owing to their dissipative schemes [8]-[10], or require complex and expensive components [11]-[13]. In addition, they are not suitable for the voltage fluctuation regulation. For voltage fluctuation regulation during a load transient, enhanced-control methods and auxiliary converter methods are typical. Enhanced control methods, such as the V2 control method [14]-[16], the sliding mode control method [17]-[18], and the linear-and-nonlinear control method [19]-[21], improve the load transient responses under identical hardware conditions. Nevertheless, the enhanced control methods have a limitation given the slew-rates of their output-inductor-currents. To overcome this slew-rate limitation, auxiliary converter methods have been suggested [22]-[25]. Several approaches change the slew-rate using a coupled-inductor for the output-inductor [22]-[23], while the AOF methods inject transient current directly into the output [24]-[25]. These methods are efficient because auxiliary converters handle only a small amount of power. As a result, the auxiliary converters can operate with high-frequency switching and small passive components. However, there are many problems with these devices, such as an expensive coupled inductor [22]-[23] and costly complex controller and current sensors [24], [25]. In addition, these approaches are only for voltage fluctuation regulation. In brief, existing studies provide a single-function for one of the two voltage regulations. Unfortunately, single function methods are generally not sufficient for POF reduction. Although post regulators [26]-[28] have a dual-function for both regulations, they show poor efficiency levels given their series structure.

The conventional AOFs [11], [24], [25] fundamentally share the same structure, indicating that this structure is able to operate for both functions. However, conventional studies have failed to achieve the dual-function because they do not include a proper voltage control strategy or a stability analysis. Moreover, previous studies were not optimized for their own single-functions in spite of their costly current sensors and the complex control methods concealed in their digital controllers.

In this paper, a novel control strategy is proposed for a dual-function AOF. The AOF has the potential for a wide bandwidth control loop given its high-frequency switching. To utilize this property, the main output voltage is controlled by the AOF rather than the main converter. As a result, the output voltage is tightly regulated naturally and a reduced output capacitor is sufficient for load transients. The main converter, on the other hand, controls the voltage of the AOF capacitor. Because there is no given regulation of the voltage, a wider voltage range is available and a small capacitor is sufficient. In addition, the output inductor of the main converter can be reduced because the inductance of the main output inductor is free from the switching voltage ripple due to the proposed method. The switching ripple current from the main converter is cancelled by the AOF and is absorbed into the AOF capacitor. To enhance the switching ripple current cancellation without a current control scheme, a duty feedforward control scheme is also proposed. Finally, it is demonstrated that the proposed dual-function AOF achieves a truly reduced POF.

In Section II, the operation principle of the proposed control strategy is introduced. A small signal analysis and the controller design characteristics are described in Section III. In Section IV, the passive filter reduction process is analyzed and compared to the conventional methods, after which a design guideline for the AOF is described. Section V shows hardware verification results obtained through a 125-W forward converter prototype. Finally, some conclusions are given in Section VI.

 

II. OPERATION PRINCIPLE

Fig. 1 shows the structure of the proposed system. The main converter is a forward converter and the AOF is a synchronous buck converter. This structure offers two benefits. First, because vA is not supplied externally, its voltage level is optimized for an efficient AOF [25]. Second, the switching voltage ripple is minimized because the current entering the output capacitor is continuous. In Fig. 2, the control strategies of the conventional AOFs and the proposed AOF are briefly described. The advantages and disadvantages are indicated in green and red, respectively. The conventional AOF for voltage ripple regulation [11]-[13] senses ripple current from the main converter and tracks the sensed current using a hysteresis controller, as shown in Fig. 2(a). The operation of the conventional AOF as a load transient assister [24], [25] is described in Fig. 2(b). Under steady state conditions, vA is controlled by the hysteresis controller. In addition, during a load transient, the duty ratio is shown to be fully on/off by the load transient detector. With the proposed method, unlike conventional studies, the output voltage is controlled by the AOF while the voltage of the AOF capacitor is controlled by the main converter, as described in Fig. 2(c). Moreover, the proposed duty feedforward control scheme is utilized to effectively cancel the ripple current without the need for current sensors (see Fig. 2(d)). Detailed descriptions of these processes are given in Section IV.

Fig. 1.Structure of the proposed system.

Fig. 2.AOF control strategy and its advantages (green) and disadvantages (red) of, (a) the conventional AOF ripple filter, (b) the conventional AOF load transient assister, (c) the proposed method (w/o feedforward), (d) the proposed method (w/feedforward).

Fig. 3 shows the key waveforms of the proposed method.

Fig. 3.Key waveforms of, (a) inductor-currents under steady states, (b) currents into the output capacitor w/ and w/o the proposed method, (c) the output voltage, the AOF capacitor voltage, and the average inductor-currents during a step-up (ΔIO) load transient at t0.

Under steady state conditions, iA is controlled, and becomes a counter current of the switching ripple current from the forward converter, as shown in Fig. 3(a). As a result, iCO follows the small switching ripple current of the AOF, as indicated in Fig. 3(b). Thus, a reduced CO satisfies the requirement of ΔVO_sw_max. In addition, LF can be reduced because ILF_pkpk does not affect iCO. During a load transient event, the dynamic response is greatly improved because the output voltage is controlled by the AOF, which has a wide bandwidth controller according to its high switching frequency [29]. Fig. 3(c) shows the key waveforms during a step-up (ΔIO) load transient at t0. The output voltage, vO, is tightly regulated, while the AOF voltage, vA follows the slower dynamic response of the forward converter. As a result, the minimum CO for load transients is also reduced. Finally, the minimum CO for both voltage regulations is reduced. Although there are additional passive components, CA and LA, a small capacitor is sufficient for CA due to the absence of tight voltage requirements on vA. Moreover, the fast switching of the AOF enables a small LA.

 

III. SMALL SIGNAL ANALYSIS OF THE AOF

A. Small Signal Model

A small signal model of the AOF is derived in this section to guarantee the stability of the proposed method. An analysis is initially conducted of a separate unterminated model, as shown in Fig. 4, after which the results are combined for the sake of convenience, as shown in Fig. 5. The main parameters, including the equivalent series resistors (ESRs) are shown in Fig. 4. The transfer functions of the separated parts are arranged in Table I. Using the results of the partial analysis, a total small signal model of the power system is derived and shown in Table II.

Fig. 4.Separate unterminated model for small signal analysis.

Fig. 5.Structure of the opened-loop small signal model.

TABLE ITRANSFER FUNCTIONS OF THE SEPARATE UNTERMINATED MODEL

TABLE IITRANSFER FUNCTIONS OF THE COMBINED SYSTEM

B. Controller Design

The corresponding closed loop gains of the proposed method are derived as follows:

The closed-loop small signal diagrams are shown in Fig. 6. HvO and HvA are the voltage feedback controllers for vO and vA, and they determine the bandwidths of TvO and TvA, respectively. Unlike the conventional cases, the proposed control strategy shows unusual phenomena in the design of the controllers. T1 shows a zero DC gain owing to the zero DC gain of GvOdA. This phenomenon occurs due to the presence of a non-external voltage source. Fortunately, it disappears after the vA control loop is closed. Therefore, only the high-frequency gain is considered in the vO controller design. Finally, TvO has an infinite DC gain and a wide bandwidth. During the HvA design, T2 shows two unnatural resonant-points. Due to these resonances, there is no phase margin of T2. However, these resonant points are removed by the other parts of TvA in Equ. (4). This procedure will be shown in detail in Section IV.

Fig. 6.Diagram of the transfer function of the closed-loop-system with the proposed control strategy.

C. Duty Feedforward

A traditional voltage feedback controller may not achieve switching ripple current cancellation due to the finite bandwidth of the feedback control loop. Thus, a duty feedforward control method is proposed. The proposed duty feedforward control achieves simultaneous ripple current cancellation with only one additional voltage sensing instance at the rectified point, as shown in Fig. 7. To cancel the ripple current of iLF, the slew rate of should be matched to the slew rate of iLF, as follows.

where refers to the average iA during a switching period, TsA. The slew rate of iLF is shown below.

The variation of dA to satisfy Equ. (5) during the forward-switch-on and switch-off times is derived as follows:

As a result, -iA is matched to the switching ripple current from the forward converter. This method is even valid when there are certain errors in the parameters, such as the transformer turn ratio, the input voltage, and the leakage inductor. In addition, the proposed feedforward scheme rarely affects the HvO design because the feedforward scheme enhances the average model in the small signal analysis of the forward converter, as presented in Section III. A. Although the feedforward affects vA, the voltage variation is attenuated as the gain of TvA at FsF. Finally, the proposed feedforward scheme does not affect the stability of the system.

Fig. 7.Proposed system including the duty feed forward.

 

IV. CONVERTER DESIGN

As mentioned earlier, an AOF with high-frequency switching is possible because the power loss of an AOF is insignificant due to its low effective power as follow:

As a result, the power loss of the AOF is determined by ILF_pkpk rather than by IO. For example, PEft is only 5% of the rated power, under the condition of ILF_pkpk=0.2IO_max.

A. Capacitor Reduction

The output voltage is determined by iCO, as follows:

Through the cancellation of the switching ripple current, the switching voltage ripple is determined by the AOF as:

and the result should satisfy ΔVO_sw≤ΔVO_sw_max. From Equ. (14), the minimum output capacitance of the switching voltage ripple regulation, CO_sw_min is determined as follows:

If the ESR of the output capacitor, RCO is negligible, a comparison of the minimum output capacitances of the proposed system, CO_sw_AOF_min and the conventional system, CO_sw_SF_min proceeds as follows.

Fig. 8 shows an example of KCO_sw versus the inductance ratio, LA/LF, and the frequency ratio, FsA/FsF, under the DA=0.5 and DF=0.25 conditions. KCO_sw decreases dramatically when FsA/FsF increases even when LA/LF is 0.2.

Fig. 8.Minimum output capacitance ratio.

According to [29], the peak output voltage variation during a load transient is derived as follows.

where fC is the bandwidth of the feedback controller and fCrt is the critical bandwidth, which is determined as follows.

Thus, the minimum output capacitance, C_Tr_AOF_min for a load transient when fC < fCrt is:

As is well known, the bandwidth of a feedback controller is proportional (1/10th approximately) to the switching frequency. Therefore, the bandwidth of the proposed method is higher by as much as FsA/FsF when compared to the forward converter, and CO_Tr_AOF_min is reduced.

Finally, the minimum output capacitance is determined with the following criterion:

The minimum AOF capacitance of the load transients, CA_Tr_min is designed using the same equations for the CO design, Eqs. (13)-(21), while the maximum variation is set to ±20% rather than ±5%, as follows.

where ΔVA_Tr_max = 0.2VA. Table III shows the passive component results with a reference to the design margin.

TABLE IIIPASSIVE COMPONENT INFORMATION

B. Inductor Reduction

First, the inductance of LF is designed. The inductance of LF for a conventional forward converter is greater than that of the other cases because the output capacitance is limited for practical reasons, such as a start-up issue.

The start-up issue is when an over current problem can occur for charging the output capacitor if its capacitance is large. Thus, this inductance is greater than the optimum value given for the voltage ripple. On the other hand, LF in the proposed method is not affected by this issue. Therefore, its inductance can be reduced. The minimum inductance of the proposed method with the given specification, LF is determined as follows:

In this paper, the inductance of LF is designed for the peak value of iLF to be less than 120% of IO_max. In general, the core of an inductor is established using the following criterion [30]:

where KAP is a constant value [30], and the area product, AP is a parameter which determines the core size. The minimum AP- value of the output inductor LF is reduced as shown below.

where:

An AP comparison between the proposed method (AP_AOF_min) and the post-regulator method (AP_PR_min), for additional inductors, is given below.

Fig. 9 shows the AP-values of the additional inductors, LA and LP, versus the switching frequency of the forward converter. The AP-value of LA is less than 2% that of LF, even in the worst case, while that of the LP is 10~30% that of the main converter. The AP result does not satisfy the ‘ILF_pkpk<1.2IO_max’ condition. The normalized total AP-value of the proposed method is shown in Fig. 10. The proposed method shows an AP-value which is reduced by half, even with the lower output inductance and lower switching frequency of the forward converter.

Fig. 9.AP-ratio of additional inductors compared to LF of, (a) the proposed method, (b) the post regulator method.

Fig 10.Total AP of the proposed method.

To verify the analysis of the proposed method, a hardware prototype of a 125-W forward converter with an AOF and a post-regulator is built and tested. First, the passive parameters are determined following the criteria in Section III. A. The results are compared with a conventional single-forward converter (SF) and a forward converter with a post-regulator (PR). Next, proper voltage controllers are designed based on the results in Section III. B. The target specifications are shown in Table IV.

TABLE IVPARAMETER VALUES OF THE CONVERTERS

C. Controller Design

Fig. 11 shows the transfer functions GvAdF and GvOdA of the hardware. As mentioned in Section III, GvOdA shows a zero DC gain due to the +20 dB/decade slope of the low-frequency region. Thus, conventional voltage controllers with an integrator, such as a type-3 compensator, cannot achieve an infinite DC gain for a zero steady state error. Fortunately, this problem will be corrected after the vA control loop is closed. Thus, only the high-frequency characteristic is an issue. After the design of HvO, T2 is designed with a proper HvA. Because HvO was previously designed, HvA is designed using the total vA voltage loop, TvA. Two-pole-two-zero with one integrator controllers are utilized for HvO and HvA. These results are shown in Fig. 12.

Fig. 11.Bode plots of transfer functions of the proposed system (opened-loop), (a) GvAdF, (b) GvOdA.

Fig. 12.Bode plot of closed-loop-gains, (a) TvO, (b) TvA.

 

V. EXPERIMENTAL RESULTS

To verify the proposed method, a forward converter hardware prototype is built and tested. The parameters follow the results shown in Section IV. A typical forward converter and a forward converter with a post-regulator are also built for comparison with the proposed method. The tests focus on three points: the voltage regulation, the volume, and the power efficiency. Fig. 13 shows the key waveforms under the steady state with the proposed method. The output voltage ripple satisfies the switching voltage ripple regulation range, 50 mV. In addition, Fig. 14 shows a 5 A step-up load transient response, which satisfies the voltage regulation, 250 mV. A comparison of the volumes between the proposed method and the conventional methods is briefly described in Fig. 15. The total volume of the inductors and capacitors of the proposed method is reduced by about 31.5% (20.9 ㎤) and by approximately 13.1% (8.7 ㎤) when compared to the conventional forward converter method and the post-regulator method, respectively. Fig. 16 shows the results of a power efficiency comparison. The efficiency of the proposed method is between those of the conventional forward converter method and the post regulator method. The proposed method shows only a 1% lower efficiency when compared to the conventional single forward converter case. The efficiency of the proposed method is 1~3% higher than that of the fast switching conventional single forward converter case and 2~8% higher than that of the post-regulator case, which are conventional approaches for passive component reduction. As a result, the proposed method achieves a passive component reduction without significantly losing efficiency.

Fig. 13.Key waveforms in a steady state.

Fig. 14.Key waveforms during a step-up load transient.

Fig. 15.Passive component volume comparison. (SF: a conventional single forward converter case, PR: a conventional post-regulator case, AOF: the proposed method case).

Fig. 16.Power efficiency comparion.

 

VI. CONCLUSION

In this paper, a novel POF reduction method is presented and analyzed. The proposed method reduces the POF while satisfying the output voltage regulations and demands for improved power efficiency. The proposed control strategy makes AOFs feasible for the voltage regulation of both the switching voltage ripple and voltage variations which arise during a load transient. A small signal analysis was done to ensure a stable and optimized system, and a design guideline was described. The volume of the POF is significantly reduced while the total efficiency is only slightly reduced. In addition, the results can be extended to other buck-type topologies such as buck converters and half/full-bridge converters, because operation algorithms and small signal models of these converters are practically the same. In addition, the proposed method can be extended to other PWM-based-converters by a similar procedure.