A Comparison of Soft and Hard-switching Losses in Three Phase Micro-inverters

Dehua Zhang, Qian Zhang, Anna Grishina, Ahmadreza Amirahmadi, Haibing Hu, John Shen and Issa Batarseh School of Electrical Engineering and Computer Science

University of Central Florida Orlando, Florida 32816, USA

Abstract—The efficiency of a PV grid-tie micro-inverter is of critical concern to its application. Three representative three-phase voltage source micro-inverter topologies have been selected to make loss comparisons: A Hard Switching Inverter (HSI), an Auxiliary Resonant Commutated Pole Inverter (ARCPI), and an Actively Clamped Resonant DC-link Inverter (ACRDCLI). Loss composition and estimation are given in detail for 350W micro-inverters. It is pointed out that the HSI has outstanding advantages in both efficiency and cost. The ARCPI losses are lower than the other two topologies. But the auxiliary circuits are costly and the controlling and driving circuits are complicated. The ACRDCLI has a simple structure but high DC-link losses. If an improved control scheme is used, the DC-link losses of the ACRDCLI can be reduced. This paper compares these advantages and limitations for assessing the selection of soft-switching topologies for micro-inverters.

I. INTRODUCTION Micro-inverters are small grid-tie inverters of 150-300W

that convert the output of a single PV panel to AC [1]. It is advantageous to extend the micro-inverter concept to large size PV installations such as MW-class solar farms where three phase AC connections are used [2]. By adopting the three-phase distributed AC micro-inverter architecture, the following advantages can be gained: (1) maximum power harvesting from each panel; (2) elimination of mismatch losses between panels; (3) easy and flexible installations; (4) less DC distribution losses. The three-phase micro-inverters have many trade offs and challenges in the areas of cost, efficiency, and life expectancy [3].

Soft-switching inverters are usually applied in the high-power region [4]. Small magnetic components are required to increase the power density of micro-inverters. A high switching frequency is required for compatibility with the small magnetic parts. Soft switching techniques help to increase the switching frequency to more than 100 kHz. The Auxiliary Resonant Commutated Pole Inverter (ARCPI) is a representative topology of three-phase high-efficiency inverters [5]. The Actively Clamped Resonant DC-link Inverter (ACRDCLI) has remarkable features of simple structure with only one auxiliary switch [6]. In addition, the Hard Switching Inverter (HSI) is able to gain both high efficiency and superior control performance if it operates near ten kilohertz. This paper focuses on exploring the

possibility of applying the above mentioned topologies to the micro-inverters.

Based on 350W micro-inverter systems, a loss comparison of the foregoing three topologies are given in this paper. The main contributors to dissipation are discussed. Assessments of the three topologies that apply to three-phase micro-inverters are also given.

II. TOPOLOGIES

A. Hard Switching Inverter Figure 1 illustrates the topology of a three-phase Hard

Switching Inverter. The CoolMOS is chosen to be the main switch because it combines extremely high on-state conductivity with ultrafast switching speed at full pulse current capability. However, in Hard Switching Inverter applications, the outstanding switching performance of the CoolMOS cannot be utilized due to the dynamic behavior of the body diode. Therefore the body diode should be blocked with a schottky diode and replaced with a SiC-schottky diode [7], as shown in Fig.1 (b).

(a) Topology (b) Combined MOSFET

Fig. 1 Three-phase micro-inverter

B. Auxiliary Resonant Commutated Pole Inverter Figure 2 illustrates the topology and key waveforms of

ARCPI. The commutation idea was addressed first by Mc. Murray and De. Donker [5,8]. Compared with Fig. 1, each phase has two auxiliary MOSFETs and a pair of resonant components to help the commutation. For phase a, the auxiliary circuits consist of Sa1, Sa1, Lra and C1, C4. According to the Fig. 2(b), suppose the load current is positive. Once S1 is going to turn on, Sa2 turns on shortly prior to the turning off of S4 and gives the resonant inductor an initial current IB1. When S4 turns off, a resonant happens between Lra and C1, C4. S1 could turn on as soon as its

1076U.S. Government work not protected by U.S. copyright

parallel capacitor C1 voltage reaches zero. The Sa1, Sb1, Sc1 help lower switches to turn on, while the Sa2, Sb2, Sc2 help upper switches to turn on. Reference [9] gives out the detail analysis of the auxiliary commutated process. The resonant inductor current iLr(t) can be expressed as:

(a) Topology

(b) Commutation waveforms of ARCPI (phase a)

Fig. 2 Three phase grid-tie ARCPI topologies and key waveforms

10 tt −

tL

Uir

dtLr ⋅=

2)( (1)

21 tt −

)sin(

2

)cos()( teI

L

RIU

teIti dt

d

Br

LrBd

dt

BLr ωω

αω αα −−

+−

+= (2)

32 tt −

tL

UtIir

dLrtLr ⋅−=

2)( 2)(

(3)

Where r

Lr

LR2

=α , RLr is Equivalent Series Resistance (ESR)

of Lr. 22

0 αωω −=d ,rr LC

10 =ω . IB is the initial current of

the resonant inductor. IB1 and IB2 in Fig. 2 (b) can set to be equal and represents uniformly with IB in equations (1) and (2). It is reasonable since the load current is small when compared to the resonant current. This assumption will not cause much extra loss on the auxiliary components.

Most researches of ARCPI were focused on the high power region. The reported efficiency is over 99% [10]. But the complexity of this circuit is significant. Furthermore, the body diode of the series auxiliary switch has a reverse recovery issue even though the auxiliary switch naturally turns off when its current reaches zero. This contributes to unexpected oscillation and much loss on the auxiliary branches. Therefore the combined MOSFETs should be used for the auxiliary switches. Otherwise, a more complicated auxiliary circuit should take the place of the current one [11].

In ARCPI the auxiliary circuits work for a short time, about one tenth of one switching period, only when a commutation is required. So ARCPI has little influence on the inverter control scheme. It is able to achieve output performance as high as hard switching inverters.

C. Actively Clamped Resonant DC-link Inverter The ACRDCLI commutation idea was first addressed by

Dr. Divan [6]. Fig. 3 illustrates its topology and key waveforms. In ACRDCLI, DC-link voltage periodically resonates down to zero. The inverter commutates at dc-link zero crossing intervals. A clamp switch is imported to clamp the dc-link voltage to a reasonable value KUd, usually K is 1.1-1.3. The outstanding feature of this topology is that only three auxiliary components, a clamp switch Sc and a pair of resonant components Lr and Cr, are needed to realize soft switching conditions for all the main switches. Because the dc-link voltage is always oscillating, there are predictable high conduction losses on the auxiliary switch and the ESR losses of the inductor. A two-amplitude resonant scheme should be used to control the DC-link for reducing the auxiliary losses [12]. As shown in Fig.3 (b), The two-amplitude control scheme is described as follows: The clamping capacitor Cc is pre-charged to (K-1)Ud. When a commutation is required, clamping switch Sc is turned on first. Thus, the clamping capacitor releases energy into the resonant inductor Lr. The Sc current ic begins to increase linearly. As soon as ic reaches the given big-amplitude turn-off current Ibc, Sc turns off under a ZVS condition. A resonant oscillation takes place between Lr and Cr. All bridge switches can be operated under ZVS when ucr, the voltage of the Cr, decreases to zero. After commutation, a new resonant oscillation between Cr and Lr occurs. When the dc-link resonates up to KUd again, the anti-paralleled diode of Sc turns on, thus Sc can be turned on under ZVS. If the inverter bridge does not require commutation, Sc turns off at a small current Isc, and the dc-link resonates at the base voltage of (K-1)Ud. In order to compensate the power losses of the resonant circuit and maintain the ZVS condition for the clamping switch, the small-amplitude resonant is kept by the periodic switching of Sc. When the next commutation signal comes, the resonant circuit starts a

1077

new big-amplitude resonance process. Reference [13] gave out a detailed analysis of the two-amplitude control scheme. Since the inverter only commutates at those zero voltage intervals, a Discrete Pulse Modulation (DPM) scheme is applied to the inverter control. According to the waveforms of ucr and iLr shown in Fig.3 (b), the dc-link inductor current can be expressed in the following equations.

(a) Topology

(b) DC-link operation waveforms

Fig. 3 Three phase grid-tie ACRDCLI topologies and key waveforms

1) Big amplitude resonant: 10 tt −

tKIti xLr 0)1()( ω−−= (4)

21 tt −

)]2

(cos[)( 0

20

απωω

−−−=−

tAeItiQ

t

xLr (5)

where22

0

])1[( bcd I

ZUKA +−=

, ]

)1([

0

1bc

d IZK

Utg−

= −α,

LrRZQ 0= ,

r

r

CLZ =0 ,

Ix represents the dc link current.

32 tt −

]2

cos[)( 00

' 20 πω

ω

−+=−

teZUIti

Qt

dxLr

(6)

where I’x represents the dc link current after commutation.

43 tt −

tZ

UKtIti dLrLr 0

04

)1()()( ω−−= (7)

2) Small amplitude resonant: 54 tt −

tZ

UKIti d

xLr 00

)1()( ω−

−= (8)

65 tt −

)]'2

(cos[')( 0

20

απωω

−−−=−

teAItiQ

t

xLr (9)

where 22

0

])1(

[' scd I

ZUK

A +−

= ， ))1(

(' 01

d

sc

UKZItg

−= −α

76 tt −

tZ

UKtIti d

LrLr 00

6)1(

)()( ω−−= (10)

III. LOSS COMPOSOTION The two main parts of power electronic devices

dissipation are power losses and control losses. Power losses include power switch losses and reactive component losses. Power switch losses are made up of conduction losses and switching losses. Reactive component losses are mainly caused by magnetic components such as inductors and transformers. The control losses are mainly driver loses.

A. Hard switching inverter Hard switching inverter losses include main switches

conduction losses and switching losses.

1) Conduction losses: The MOSFET conduction losses can be expressed by

2oonon IRL = (11)

where Io is the RMS value of the load current io(t). If a combined MOSFET is used as the main switch, conduction losses can be expressed by:

})](1[)(

)()()()({

/

02

/

01

/

0

2

dttdtiV

dttdtiVdttdtiRfL

oon

oonoonoon

∫

∫∫

−⋅+

⋅+⋅=

ωπ

ωπωπ

(12)

1078

where output frequency is fo, on resistance of MOSFETs are Ron, forward voltage drop of series diodes are Von1, and anti-parallel diodes are Von2.

2) Switching losses: The hard switching waveforms of a MOSFET are shown in Fig. 4. The switching losses of a MOSFET can be estimated by using a datasheet [14]. Turn on losses can be expressed by

Fig. 4 switching transients of the power MOSFET

dtttt

tifUfL furioodswswon ∫

+⋅= ω

π

0 2)(

)(2 (13) where tfu can be expressed by

gon

gdoondfu I

CtiRUtt )]([)( −= (14)

Igon can be expressed by

g

plateaudrgon R

UUI

−= (15)

Turn on losses include a reverse recovery process, which accounts for many of the switching losses. Reverse recovery losses can be expressed by

}]32

)([]2

)([{200

dttItidttItifUfL rrbrrmo

rrarrm

oodswswrr ∫∫ +++⋅= ωπ

ωπ

(16)

usually trra is 1.2 times trrb

Turn off losses can be expressed by

dtttt

tifUfL rufioodswswoff ∫

+⋅= ω

π

0 2)(

)(2 (17)

where tfu can be expressed by

goff

gdoondru I

CtiRUtt )]([)( −= (18)

where

g

plateaugoff R

UI −= (19)

tri,tfi, Cgd , trr, Qrr, Uplateau can read from MOSFET data sheet.

B. Soft switching inverter Soft switching inverter losses include main circuit losses

and auxiliary circuit losses. Since ARCPI and ACRDCLI do not inject resonant current to the main switches, their main conduction losses are the same as hard switching. Usually main switching losses are omitted because of the soft switching condition. Auxiliary circuit losses include auxiliary switch conduction losses, switching losses, and ESR losses of resonant inductors. Due to the non-linear feature of the magnetic materials, accurate estimation of the core loss is very difficult. An RLr proportional to Lr is defined to estimate the ESR losses.

∑∫=

+=n

iLr

t

t Lresr dtRtifL i

i0

21 )( (20)

where f is the operating frequency of the resonant inductor. The ESR losses could be estimated by inserting the deduced inductor current expression into equation (20) by using the fundamental loss estimation equation.

For ARCPI, the auxiliary switches have the reverse recovery issue. Therefore a combined MOSFET can be used as an auxiliary switch. The resonant inductor and auxiliary switches carry resonant current. However, since the auxiliary circuits only work for a short interval and the auxiliary resonant commutation pole operates at the same frequency as the inverter, the conduction losses and the ESR losses are not high. Therefore, the auxiliary losses are mainly caused by the switching losses of the auxiliary switches.

In ACRDCLI, as waveforms of iLr and iC shown in Fig.3 (b), It can be seen that a two amplitude control scheme reduces the big-amplitude frequency from a dc-link resonant frequency to an inverter commutated frequency. The RMS current of the clamping stage is thus reduced. Both the conduction losses on the auxiliary switch and the ESR losses on the inductor are reduced. However, in the micro-power region, the dc-link current is much higher than the load current. These two losses are still the main contributors to total system losses.

IV. RESULTS OF LOSS CALCULATIONS 400V input, 350W output single phase full bridge

topologies are taken as analysis objects. The main circuit parameters are shown in Table 1. The main and auxiliary switches are Infineon CoolMOS transistor SPA11N60C3 and CREE SiC diode C3D06060A [7]. The comparisons of losses are shown in Fig.5. From Fig. 5, we learn that the main losses on the hard switching circuit are switching on losses, which are mainly made up from the reverse recovery losses. If combined MOSFETs are taken as the main switches, the HSI could operate at high frequency with quite low losses. In ARCPI, the saved main switching losses are

1079

moved to the auxiliary switches. A high efficiency is available only if combined MOSFETs are taken as the auxiliary switches. But the number of switches in three-phase topology will be 12 MOSFETs and 12 diodes. It is impractical for the micro-inverters. The ACRDCLI does not have the reverse recovery issue. But the conduction losses on the auxiliary switch and the ESR losses on the resonant inductor are quite high. It should be pointed out that if the resonant frequency of the DC-link increases to 500 kHz while the switching frequency remains unchanged, a greatly reduced dc-link loss is achieved.

TABLE I. PARAMETERS OF THE MAIN CIRCUITS

Topo logies

Po (W)

Ud (V)

L (mH)

C (uF)

fsw (kHz)

fr (kHz)

control

HSI 350 400 2m 10u 20, 50

-- SPWM

ARCPI 350 400 2m 10u 20, 50

900 SPWM

ACRDCLI 350 400 5m 2.2u 15.4k av.

200500

DPM

Hard Switching

01

2

3

45

6

7

89

conductionmains

switching on switching off total

CoolMOS 20k

CoolMOS&SiC 20k

CoolMOS&SiC 50k

ARCPI

01

2

3

45

6

7

89

conductionmains

conductionaux

switchingmain

switchingaux

ESR total

CoolMOS 20k

CoolMOS&SiC 20k

CoolMOS&SiC 50k

ACRDCLI

0

1

2

3

4

5

6

7

8

9

conductionmains

conductionclamp

switchingclamping

esr inductor total

fr=200kHz

fr=500kHz

Fig.5 Loss composition and comparison

From the loss composition analysis, we learn that the soft-switching topologies save the switching losses of the main switches, but generate extra losses on the auxiliary circuit. Moreover, the control and drive for the auxiliary circuit is complex and requires many isolated DC/DC modules. The auxiliary circuit losses should not exceed the

saved main switching losses. Otherwise the benefits of the soft-switching topologies do not yield an adequate advantage in cost and power density which are essential measurements in the performance of micro-inverters.

V. RESULTS OF SIMULATION AND EXPERIMENT To facilitate the experiment, we curtailed using a three-

phase inverter by using a single 150W phase inverter for experimental verification and efficiency measurements. The efficiency measured does not include the auxiliary power. The magnetic core losses are not included in the simulation calculations, but are included in the experimental efficiency measurements. Table 2 shows the efficiency comparison results. The calculated efficiencies are from LTspice simulation, which are marked with [1]. The measured efficiencies are from experiments, which are marked with [2]. The main and auxiliary switches for LTspice are the Infineon CoolMOS transistor SPA11N60C3. The switches for the experiment are FCP20N60 for HSI and ARCPI, and STP11NM80 for ACRDCLI. The clamping switch for ACRDCLI is ST42NM65M5. The DSP control chip is dsPIC30F2023 for HSI and ARCPI, DSP28335 for ACRDCLI.

TABLE II. EFFICIENCY COMPARISON

Topologies Ud(V) Po(W) fsw(kHz) η(%)[1]

η (%)[2]

HSI 400 150 20 97.8 97.1 ARCPI 400 150 20 97.7 97.6 ACRDCLI 400/100 150 15.4 96.3 95.8

Fig.6 and Fig. 7 give out the experimental waveforms of ARCPI and ACRDCLI, respectively. In Fig.6 (a), uga1 and uga2 conduct only 1μs to assist the main switches commutation. Fig. 6 (b) shows the inductor current waveforms and the switching waveform of the main switch. The inductor current has the oscillation caused by the body diode. The main switch is soft switched. In Fig. 7, the two-amplitude resonant is realized with predictive control. The commutation time (middle waveform of Fig. 7 (b)) and the pulse width (middle waveform of Fig. 7 (a)) are predicted accurately.

The measured efficiencies show that ARCPI has the highest value among the three topologies. But as it has been mentioned, the complicated control and topology of ARCPI preclude its use in micro-power applications. The resonant dc-link creates a ZVS condition to the inverter, but the dc-link itself brings losses to the inverter, which make it impractical for the micro-inverters. The hard switching topology can compete with ARCPI at a 20kHz switching frequency. But it is hard to increase the switching frequency with PWM control and it can not match the high power density requirement of the micro-inverters. Thus developing a simpler control scheme for the hard switching topology, which can increase the switching frequency higher than 100 kHz, could be a promising solution for the micro-inverters.

1080

(a) Drive waveforms (2V/div, 4μs/div)

(b) Commutation pole waveforms (1μs/div)

Fig. 6 experimental waveforms of ARCPI

(a) DC-link resonant waveforms (10μs/div) up: voltage on clamping switch (50V/div);

middle: driver signal of clamping switch (10V/div); down: clamping switch current ic (5A/div)

(b) DC-link resonant waveforms (4μs/div) up: DC-link voltage ucr (50V/div);

middle: driver signal of main switch (10V/div); down: clamping switch current ic (5A/div)

(c) DC-link resonant waveforms (4μs/div) up: voltage on clamping switch (50V/div);

middle: inductor current iLr (10A/div); down: clamping switch current ic (5A/div)

Fig. 7 experimental waveforms of ACRDCLI

VI. CONCLUTIONS This paper makes a loss comparison of three

representative three-phase voltage source micro-inverter topologies: HSI, ARCPI and ACRDCLI. The loss analysis and calculation results show that considering both efficiency and cost, the HSI is superior to the other two topologies. The losses of the ARCPI are lower than the other two topologies. But the auxiliary resonant circuits are costly and the controlling and driving circuits are complicated. It is impractical for micro power applications. The two-amplitude ACRDCLI has a simple auxiliary circuit but brings considerable losses to the system and contributes to reduced efficiency of micro-inverter systems. Therefore, application of the soft-switching topologies to micro-inverters should be assessed with these limitations in mind.

REFERENCES [1] Quan Li; Wolfs P., ‘A Review of the Single Phase Photovoltaic

Module Integrated Converter Topologies With Three Different DC Link Configurations’, IEEE Transactions on Power Electronics, Vol. 23, Issue 3, 2008, p1320 - 1333

[2] Sahan B.; Notholt-Vergara A.; Engler. A.; Zacharias. P., ‘Development of a single-stage three-phase PV module integrated converter’, Power Electronics and Applications, 2007 European Conference, 2-5 Sept. 2007, p1-11

[3] Haibing Hu, Wisam Al-Hoor, Nasser Kutkut, Issa Batarseh, “Efficiency Improvement of Grid-tied Inverters at Low Input Power Using Pulse Skipping Control Strategy”. IEEE Trans. On Power Electronics,Vol.25,NO.12,Dec. 2010,pp.3129-3138.

[4] H.-J. Pfisterer, ‘Three Phase 160 kVA Auxiliary Resonant Commutated Pole (ARCP) Converter’, Power Electronics, Machines and Drives (PEMD), 2008. 4th IET Conference, 2-4 April 2008, p480-484

[5] R. W. De. Donker; J. P. Lyons, “The auxiliary resonant commutated Pole converter”, IEEE Proc. IAS 1990, p1228-1235

[6] Deepakraj M. Divan; Gary Skibinski, “Zero-switching-Loss Inverters for High-Power Applications”, IEEE Transactions on IA, Vol. 25, No. 4, July/August, 1989, pp. 634-643

[7] Lorenz L; Deboy G; Zverev I, “Matched Pair of CoolMOS Transistor With SiC-Schottky Diode-Advantages in Application”, IEEE Transactions On Industry Applications, Vol. 42, No. 5, 2004, p1265-1272

[8] Mc Murry. W., “Resonant Snubbers with Auxiliary Switches”, IEEE Industry Applications Society Annual Meeting, vol. 1, 1989. p829-834.

ug1 ug4

uga1 uga2

iLr

uds4 (100V/div)

(2A/div)

1081

[9] Ke Ma ; Dehong Xu ; Tao Zhang ; Igarashi, S. , “The evaluation of control strategies for Auxiliary Resonant Commutated Pole inverter”, Energy Conversion Congress and Exposition, 2009. ECCE 2009. IEEE, p810-816.

[10] Pengwei Sun; Jih-Sheng Lai; Hao Qian; Wensong Yu; Smith, C.; Bates, J.; Arnet, B.; Litvinov, A.; Leslie, S., “Efficiency evaluation of a 55kW soft-switching module based inverter for high temperature hybrid electric vehicle drives application”, APEC, 2010 Twenty-Fifth Annual IEEE, p 474 - 479

[11] W. Yu, J.S. Lai and S.Y. Park, “An improved zero-voltage-switching inverter using two coupled magnetics in one resonant pole,” in IEEE

2009 Applied Power Electronics Conference and Exposition, 2009, pp. 401-406.

[12] D.H. Zhang; J.P. Ying; T. Liu; J.T. Tan; “Three-phase two-amplitude actively clamped resonant DC-link inverter” , IEEE-PESC Jun 13-19 2003, p.1839-1844

[13] Zhang D.H., Ying J.P., Liu T., Liu Q., Wang F.B., “Analysis and design of two-amplitude actively clamped resonant DC-Link inverter”, IEEE INTELEC 2001, p. 506 – 513.

[14] Dušan Graovac, Marco Pürschel, Andreas Kiep, “MOSFET Power Losses Calculation Using the Data-Sheet Parameters”, Infineon Application Note, V 1.1, July 2006.

1082