Soft Switching Bidirectional Converter for Battery

0 downloads 0 Views 566KB Size Report
could be used as a battery discharge/charge regulator when the bus voltage is ... Figure 2. The two operating stages of the bidirectional boost converter with coupled .... classical soft switching cells (resonant capacitor, connected between drain and source of ... not stay at 0A, but reaches a relatively large value and remains.
Soft Switching Bidirectional Converter for Battery Discharging-Charging E. Sanchis-Kilders(1), A. Ferreres(1), E. Maset(1), J.B. Ejea(1), V. Esteve(1), J. Jordán(1), A. Garrigós(2), J. Calvente(3) (1)

Dpt. Ingeniería Electrónica E.T.S.E. / Universitat de València Dr. Moliner, 50, E-46100 Burjassot, SPAIN e-mail: [email protected]

(2)

División Tecnología Electrónica E.P.S.E. / Uni. Miguel Hernández Avda. del Ferrocarril s/n, E-03202 Elche, SPAIN e-mail: [email protected]

Abstract— This paper presents the results of a project that looked after a high efficiency bidirectional converter which could be used as a battery discharge/charge regulator when the bus voltage is above the battery voltage. High efficiency, high stability and simplicity are the main goals, no galvanic isolation is required. Taking into account all these parameters, our proposed solution has been a new topology based on a Boost converter with coupled inductors. The use of a bidirectional converter reduces the mass of the overall charge/discharge subsystem and lowers cost and component count. In the project, its use is intended for space applications, but telecom, automotive or similar applications can also benefit of this new concept.

I. INTRODUCTION High power buses used nowadays are normally backed up with batteries, which have to be charged and discharged depending on the bus power demand. These systems are common in telecommunications applications, space platforms and automotive electrical buses. In space applications the power source is usually a DC voltage and therefore DC-to-DC converters are needed. Telecom applications need normally an input rectifier (AC-to-DC) to feed the bus. Modularity also requires the ability of connecting modules in parallel without complex additional circuitry. The easiest way to comply with this requirement is to provide the converter with current regulation. Current regulated converters with common voltage loop can be immediately parallelized, short circuit protections are inherent to the control loop and current sharing is also guaranteed. The use of a bidirectional unit saves the need of one additional converter for the charge operation. This translates into less mass and volume and finally less cost [1]. The prototype designed and built complies with the following specifications: Input voltage (battery voltage) Output voltage (bus voltage) (no galvanic isolation required) Switching frequency Output power Efficiency

0-7803-9547-6/06/$20.00 ©2006 IEEE.

Vi = 82V..100V Vo = 120V±0.5% fs = 100kHz Po = 1kW η>95%

(3)

Dpt. d’Enginyeria Electrònica, Elèctrica i Automàtica E.T.S.E. / Uni. Rovira i Virgili, E-43007 Tarragona, SPAIN e-mail: [email protected]

To achieve a high MTBF, two basic principles have to be observed: all components must have a low temperature rise and the circuit must be as simple as possible. One way to assure a high MTBF is to apply strict derating rules [2]. II. THE COUPLED INDUCTOR BOOST CONVERTER The selected topology was already introduced in [3] and [4]. This topology is a boost and therefore a step up converter, with output filter and with its two inductors (input inductor and output filter inductor) coupled together. Bidirectionality was achieved by using bidirectional switches. The use of the coupled inductors assures that the converter behaves, from the control point of view, as a minimum phase system. This means that it does not have a right half plane zero, which can create stability problems. These systems can be easily stabilized like the buck converter. Conductance Control [5] can be easily applied and parallelizing is then straightforward thank to the current control loop.

Figure 1.

Bidirectional Coupled Inductors’ Boost.

Fig. 1 shows the scheme of the converter. It already includes the two MOSFET which permit bidirectional flow of power. Bidirectional behavior is immediate and does not need additional control circuitry. As soon as the output voltage increases, due to an additional flow of power to the load from another source (photovoltaic panels or fuel cell, for example; not shown in Fig. 1), the exceeding power will flow from output to input of the converter and charge the battery (here Vin). If M2 is substituted by a diode pointing to the output like a boost diode, the converter would be unidirectional but retain its other advantages (minimum phase system).

603

The design of the coupled inductances is explained in [3]. The coupled inductors have been substituted by two real components which are a transformer with a given magnetizing inductance and leakage inductance. As the value of the leakage inductance (54µH) is of the same order of magnitude as the value of the magnetizing inductance (108µH) an additional inductance, Lb, has to be placed in series with the transformer “to increase its leakage inductance”. We could call the transformer a “flyback-forward-transformer” because it is like a classical one but with a small magnetizing inductance which is discharged to the output. Energy transfer is done in an inductive and direct way as for example in the buck converter or the two-inductor boost [4]. This can be seen by studying its operation mode. The two switching states in continuous conduction mode are shown in Fig. 2. It can be seen that current is always flowing from the input to the output.

This behavior leads us to the only drawback of the circuit which is the mass (and volume) of its magnetic elements. The mass of the transformer is larger than the one of an inductor because of its additional secondary (it is a transformer!) and the primary needs additional copper because of the not negligible magnetizing current (large magnetizing current!). The only comparable topology would be the two-inductorboost [6] which does not couple both inductors and therefore saves the secondary of the transformer. It is also a minimum phase system and has a step up transfer function. But the transistor, in case of unidirectional operation is floating and this requires additional driving circuitry. By comparing the energy stored in the magnetic elements of both topologies designed for the same specifications and with the same current ripple, we found out that the two inductor boost stores about 8mJ and the coupled inductor boost stores 5mJ. We think that this data means that the mass of the magnetic elements will be at least similar, because the coupled inductors boost stores less energy but needs a kind of transformer. But more detailed data must be compared. The circuit itself shows also that the DC voltage at the capacitor C is equal to the output voltage. In fact Co and C are connected together by a short seen from the DC point of view. Fig. 3 shows the simulated waveforms of the converter. We can see how the current through C confirms that it is discharged during the ON state and charged during the OFF state.

Figure 2. The two operating stages of the bidirectional boost converter with coupled inductors. Please note that there is always a direct current path from input to output.

During the ON state we are charging the magnetizing inductance of the transformer and also transmitting current to the output (ILb) and during the OFF state we are discharging the magnetizing inductance to the output and again also transmitting current to the output (ILb). The DC levels of the currents are determined by the power balance of the converter and the current ripple (AC) depends on the value of the inductances. The peak currents at the input and the output are described by Eq. 1 and Eq. 2. Please note that Eq. 2 has two terms for the ripple, the one from the magnetizing inductance and the one from the output inductance, Lb, reflected to primary. I Lb _ pk = I 0max +

1⎛ N I in _ pk = I i max + ⎜ s 2 ⎜⎝ N p

1 N s Vi tON 2 N p Lb

(1)

2

⎞ Vi 1 Vi tON + tON ⎟⎟ 2 Lmag L b ⎠

(2)

Figure 3. Simulated converter waveforms at full power: Upper waveforms: Iin (=Ibat; red), ILb (blue), IC (green), ICd (magenta) Lower waveforms: Vo (red) and VC (blue). The names are after Fig. 1. The DC value of VC is Vo.

III. SMALL SIGNAL ANALYSIS The coupled inductor boost converter has the same DC transfer function as the classical boost converter and turn ratio does not affect the DC transfer ratio. Vo =

604

1 1− D

Vi

(3)

The damping (Rd-Cd) shown in Fig. 1 is needed for stability purposes and does not imply appreciable losses (PRd = 0.5W at P0 = 1kW). Without damping, the minimum phase behavior is given only under a given condition [3] that is: Ns Np

>

1 1− D

(4)

When damping the resonance properly, then the above inequality can be overridden. In fact, damping provides us with smaller inductors by avoiding the fore mentioned condition (Eq. (4)). By using the design equations given in [3] a perfect and stable behavior can be expected under all load and input voltage range. Fig. 4 shows the frequency response and how damping assures stability.

Figure 5. In red soft switching cells for both transitors based on classical soft switching cells (resonant capacitor, connected between drain and source of each MOSFET, is not shown for simplicity).

Fig. 5 shows the two soft-switching cells like in the classical design for each active switch. In order to simplify the circuit we propose a new single circuit (Fig. 6) taking into account that: a) the connection points are the same, b) the diodes correspond to the body diodes of the MOSFET and c) one single inductor can be used. Resonant inductor Lr, resonates with the parasitic drain source capacitance of the MOSFET which can be increased by adding external capacitance (to simplify not shown in Fig. 5, but shown in Fig. 7).

Figure 4.



Calculated frequency response of the converter ( i Lb ) with d

(solid) and without (dashed) damping. Figure 6.

IV. BIDIRECTIONAL SOFT SWITCHING Although theoretically no reverse recovery is present in the circuit due to the use of two MOSFET as switches, soft switching has been investigated due to the fact that the body diodes of both transistors can conduct during short periods of time. This happens because we have to provide some dead time before switching on each switch to avoid short circuit. This could force the body diode on during these transitions and can lead to reverse recovery losses. In order to apply soft switching to the converter a new bidirectional soft switching circuit has been proposed. Based on a classical soft switching circuit [7], bidirectionality has been achieved by blending the soft switching circuits of the boost topology with the one of the buck topology and of course using transistors instead of diodes. It has to be taken into account that the topology of Fig. 1 is a buck converter with input filter when working in reverse mode and a boost with output filter in direct mode. The design of the circuits is as explained in the references [7] and [8].

The new proposed soft switching bidirectional cell.

When adding the bidirectional soft switching cell to the bidirectional boost circuit we end up with the circuit shown in Fig. 7. Note that we have already added the resonant capacitors in parallel with the drain-source capacitance. A single resonant inductor is needed and two auxiliary MOSFET are used. Therefore one additional floating driver is needed for Maux2.

Figure 7.

The bidirectional BOOST with the new proposed soft switching bidirectional cell added.

This circuit can be driven with two different strategies. The one described in [7] and a much newer one proposed in [8] which simplifies the driving timing and avoids the thermal

605

run-away of active components. Both strategies provide similar (small) efficiency enhancements and both have been developed and tested with high voltage, low current circuits. The classical strategy has been chosen because the power layout does not change. This means that the auxiliary transistors are driven with a short on time corresponding to the time required by Lr to charge up to Iin plus a quarter of the resonant period. This happens just before the main switches are switched on (see [7]). After that, the inductor discharges linearly. In the following figure (Fig. 8) we show the basic waveforms of the converter with the soft switching cell. We can see that the auxiliary transistors are switched on just before the main transistors are switched on and therefore a small time gap ([t1-t2] and [t4-t5]) is left between both main transistors. The more detailed discussion on this waveform can be found in [7]. The main problem encountered in our circuit is shown by the red trace. Due to the reverse recovery of the body diodes of the auxiliary transistors and the resonance of Lr and the parasitic capacitors of the circuit, the current through Lr does not stay at 0A, but reaches a relatively large value and remains at this constant level during the time intervals of [t0-t1] and [t3t5]. This current is circulating through closed loops (M1-Maux1Lr and M2-Maux2-Lr) and is not providing energy to the output. Therefore it is only dissipating energy and contributing to the overall losses of the circuit. In our driving scheme we also drive Maux2 during direct operation of the converter for control simplicity purposes, although this action does not provide soft switching to M2. From the “ideal behavior” point of view (demonstrated in simulation where no recovery currents were present) no current change should happen through Lr by this additional on switching of Maux2. But real circuit shows how the current changes its sign and unfortunately reaches a new constant level different form 0A (see [t1-t2] in Fig. 8).

transistors. Experimental waveforms are shown in Fig. 8. A heatsink becomes necessary for the auxiliary transistors. This problem has been already mentioned in [7] and [8]. In [7] it is solved with an additional diode in series with Lr to block the inverse current, but in our case this makes the cell unidirectional. In [8] they do not add the additional diode but reduce the di/dt through Lr to reduce the reverse recovery through the diodes. However the experimental waveforms presented in [8] still show this circulating current.

Figure 9. Switching waveforms of bidirectional soft switching cell. Please observe how the current through Lr stays on an unacceptable level instead to return to zero after switching OFF. Ch1: VDS (50V/div), Ch2: ID (5A/div), Ch3 ILr (5A/div), Ch4: VIsens (5V/div). P0=1kW, Vi=82V.

To solve this problem, several solutions have been tried. First is to go back to the unidirectional circuit what means that the soft switching cell has to bee duplicated (one for each transistor). Unidirectionality is achieved with an additional diode put in series with the resonant inductor. The circulating current disappears completely.

Figure 8. Main waveforms of the soft switching boost with direct current flow (boost mode). The red trace shows the real current flowing through the resonant inductor.

These circulating currents which happen due to the reverse recovery of the body diodes of the auxiliary MOSFET appear to be relatively large and high losses are induced in these

Figure 10. Switching waveforms of unidirectional soft switching cell due to added diode. Ch1: VDS (50V/div), Ch2 ILr (10A/div), Ch3: ID (10A/div), Ch4: VIsens (5V/div). P0=1kW, Vi=82V.

Of course bidirectionality could be preserved if the diode is replaced by a switch capable of blocking current in one of the two directions depending on if we are working in buck mode

606

or boost mode. Complexity of this scheme is unacceptable for our application and this solution has been discarded. Another solution is to connect a saturable inductor in series with the resonant inductor. This solution improves the efficiency but still does not avoid completely the circulating current through the circuit (see Fig. 11).

Efficiency Interleaved Boost 99,0% 98,5%

η

98,0% 97,5% Vi=82V Vi=82V sw1D

97,0%

Vi=82 sw2D 96,5% 96,0% 500,0

1.000,0

1.500,0

2.000,0

2.500,0

3.000,0

3.500,0

4.000,0

4.500,0

5.000,0

Po [W]

Figure 13. Efficiency of an interleaved BOOST converter under hard switching conditions (Vi=82V), under soft switching conditions with a single diode (Vi=82V sw1D) and under soft switching conditions with two diodes (Vi=82V sw2D). The second diode was placed only to increase reliability.

Figure 11. Switching waveforms of bidirectional soft switching cell. Saturable inductor has been added in series with Lr. Please observe how the circulating current through Lr has been reduced. Ch1: VDS (50V/div), Ch2: ID (5A/div), Ch3 ILr (5A/div), Ch4: VIsens (5V/div). P0=1kW, Vi=82V.

No efficiency improvement is observed with any of these solutions as far as we have measured in our lab (see Fig. 12). Efficiency 99,0% 98,0% Vi=85V HSW

η

97,0%

Vi=100V HSW Vi=85 SSW w/diode

96,0%

Vi=100 SSW w/diode Vi=85 SSW w/Lsat

95,0%

Vi=100 SSW w/Lsat

94,0% 93,0% 200,0

300,0

400,0

500,0

600,0

700,0

800,0

900,0

1000,0 1100,0

Po [W]

Figure 12. Efficiency of the converter under hard switching conditions (HSW), unidirectional soft switching conditions (SSW w/diode) and bidirectional soft switching conditions (SSW w/Lsat).

From Fig. 12 we see that the hard switching circuit is the most efficient. At high power levels efficiency is increased about 0.5% by the classical soft switching cell. The saturable inductor leaves efficiency below 96% in any case. Just for comparison purposes we present also the measured efficiency of a 5kW interleaved boost converter with passive soft switching [9]. We see how the efficiency is just a little better (+0.5%) with soft switching under high power conditions, but otherwise we reach a very high efficiency under hard switching conditions. In this case we were using a passive circuit, but if an active circuit is required, the complexity is not always worth the efficiency increase due to the loss of reliability.

These experimental results have made us think of the real use soft switching circuits, which add complexity with a small efficiency benefit. Our experience confirms that with high current applications, diodes always behave “badly” in real live and it is very difficult to overcome the losses introduced by this behavior. In our application, due to the fact that we had a low voltage (200V) we could use Schottky diodes which have a much better reverse recovery behavior, but for higher voltages the reverse recovery losses become important. As all soft switching circuits need somewhere a diode which usually has to switch in a very fast manner, the problem is normally shifted from the main switch to the auxiliary switch. Efficiency improvements are therefore quite small. In some very specific cases soft switching can be used but normally it is only useful under determined load and line conditions. In our application we have concluded that the bidirectional soft switching cell does not provide the necessary practical benefits in order to be implemented in our power circuit. Only one cell for each transistor provides a very small improvement but increases complexity of the circuit a lot. Following this criteria we could also place one passive soft switching cell for each transistor like the one used in the interleaved boost [9]. But only 0.5% efficiency improvement can be expected. If we control the di/dt of the transistor (here slow turn on and fast turn off) we can optimize the losses due to the reverse recovery of the diodes in the circuit and no additional element must be added to the circuit. This provides us with high efficiency, a simple circuit and fewer components. Further measurements are going to be conduct in order to find the optimum switching times. V. EXPERIMENTAL RESULTS The built prototype is shown in Fig. 14. The system includes the main transistors on the left side attached to a heatsink and the auxiliary transistors of the soft switching circuit and the resonant inductor in lower left area of the PCB. An additional heatsink for these two transistors to avoid overheating was required in the final version. On the upper side we find the transformer with the designed magnetizing inductance and the additional inductor at its right side.

607

to see that the on transition is not too hard in the hard switching mode except for the current spike. Of course under soft switching operation the on transition is loss less. The off transition is hard in hard switching mode and a little bit smoother in soft switching mode.

Figure 14. Power section of the prototype. Additional heatsink had to be added to the auxiliary switches if no diode was placed in series with Lr. Main power switches (TO-247) are at the left, auxiliary switches are at the bottomleft (TO-220AB). At the top-left is the transformer and at its right the additional inductor, Lb. The rectangular blue shape between the two yellow MKT capacitors at the lower right is a LEM sensor.

Experimental results show that the converter is bidirectional as expected and that switching occurs in a soft manner in both directions. We tested the unit with the soft switching circuit in direct and reverse mode of operation observing that both ways of operation performed correctly. But unfortunately efficiency was not improved. The oscillograms show (Fig. 15) that the soft switching circuit actuates during the turn on of the devices achieving a zero-current-transition. The turn off is not as smooth but still reduces switching noise.

Figure 16. Safe Operating Area (SOA) of the main transistor under direct operation and handling P0=1kW. On the left we see the hard switching and on the right the soft switching transition (y-axis 50V/div; x-axis 5A/div).

VI. CONCLUSION A new bidirectional step up converter with coupled inductors has been built and tested. When coupling its two inductors and damping the resonance, a minimum phase system results. Conductance control can be applied and bidirectional behavior is naturally achieved with this converter. At system level, this topology and thank to its bidirectionality saves volume, mass and cost. A new soft switching cell have been designed and tested but no efficiency improvements have been gained. But the new bidirectional soft switching cell works as expected and provides soft switching to the converter in either direction of power flow softening the on transitions. Only when using two separate unidirectional soft switching cells, one for each transistor, we can expect a slight efficiency increase at the higher power end. We have finally chosen to not use any soft switching cell and hard switch the circuit. Reverse recovery problems can be minimized by controlling the di/dt of the transistors (here slow on transition and fast off transition). ACKNOWLEDGMENT The authors would like to thank:

Figure 15. Switching waveforms of M1 in direct mode and reverse mode with bidirectional soft switching cell. VDS 50V/div; ID 5A/div. In direct mode P0=1kW and in reverse mode P0=410W

Safe operating area of the main transistor under hard and soft switching conditions is shown in Fig. 16. It is interesting

608



Spanish Ministry of Science and Technology (SubD.G.P.I.) which has supported this research with the project ref: ESP 2003-08905-C03-03.



LEM and Mr. Stefan Lüscher for providing us with samples for the prototype and good advices for the right selection of the current transducer.

REFERENCES [1] S. H. Weinberg, A. López, “A Bidirectional BDR/BCR for Satellite Applications”, 5th European Space Power Conference 1998, ESA SP416, pp.2732, September 1998. [2] “Derating and end-of-life parameter drifts. Electrical, electronic and electromechanical components”, ESA ECSS-Q-60-11A, http://www.ecss.nl [3] J. Calvente, L. Martinez-Salamero, H. Valderrama, E. Vidal-Idiarte, “Using Magnetic Coupling to Eliminate Right Half-Plane Zeros in Boost Converters”, IEEE Power Electronics Letters, Vol. 2 (Nº2), pp.58-62, June 2004 [4] P. Rueda, S. Ghani, P. Perol, "A New Energy Transfer Principle to achieve a Minimum Phase & Continuous Current Boost Converter”, IEEE PESC Conf., 2004, pp.2232-2236. [5] D. O’Sullivan, H. Spruyt, A. Crausaz, “PWM Conductance Control”, IEEE PESC Conf. 1988, pp. 351-359, 1988. [6] J.L. White, W.J. Muldoon, “Two inductor Boost and Buck Converter”, IEEE PESC Conf., 1987, pp.387-392. [7] G. Hua, C. Leu, Y. Jiang. F.C. Lee, “Novel Zero-Voltage-Transition PWM Converters”, IEEE Transactions on Power Electronics, Vol.9, (Nº2), pp.213-219, March 1994. [8] M. Jovanovic, Y. Jang, “A Novel Active Snubber for High-Power Boost Converters”, IEEE Transactions on Power Electronics, Vol.15, (Nº2), pp.278-284, March 2000. [9] E. Sanchis-Kilders, A. Ferreres, E. Maset, J.B. Ejea, V. Esteve, J. Jordán, R. García, A. Garrigós, “High Power Passive Soft Switched Interleaved Boost Converters”, IEEE PESC Conf., 2004, pp.426-432.

609