Paper Title (Use Style: Paper Title) s103

Improved Phase Shedding Techniques in Interleaved Converters

Anagha Rayachoti

Department of Electrical and Computer Engineering

Missouri University of Science and Technology, Rolla, MO 65409

1 Copyright 2103 Missouri S&T

Abstract—Light load efficiency in interleaved converters is very poor when compared to single phase converters. Phase shedding is one of the approaches used to improve this light load efficiency. This paper analyzes the transient behavior of the system during phase shedding. The transient voltage response of the system during the process of phase shedding shows large deviations. A new method of phase shedding, called the ramp control method is used to improve this behavior. Later, this new method is combined with the current sensing control to further improve the results.

1.  Introduction

Multiphase converters or interleaved converters are a parallel combination of DC-DC converters with an interleaving difference of 3600/N between the adjacent phases. The main advantages of the interleaved converters are:

·  Smaller inductors are used which makes the converters more compact.

·  Interleaving reduces the ripple current at both input and the output.

·  Efficiency of the interleaved converters is more in comparison to normal converters.

But one main drawback of these converters is the decrease in efficiency at light loads. Light loads imply losses in switches gain prominence, thereby decreasing the efficiency at light loads. In interleaved converters, these switching losses cause more problem because of the presence of larger number of switches. Light load efficiency gains prominence because, be it in laptops or electric vehicles where this converter finds its application, the system runs in the light load condition most of the time.

Phase shedding is one way to improve the light load efficiency. Phase shedding is the procedure where the number of active phases is decided based on the load current. Load current increase causes more phases to become active while decrease in load current causes these phases to become inactive and lesser number of phases are in operation.

Major challenge encountered in this process of phase shedding is the poor transient behavior of the system immediately after the phase shedding. In this paper, a novel approach to the process of phase shedding is introduced and the improvements in the voltage and current profiles of the system are analyzed.

2.  Related work

In this paper, an interleaved, two phase buck converter with phase shedding control is modeled using the Simulink software. The converter is simulated under different phase shedding control approaches and the resultant voltage and current profiles are analyzed.

The controls for phase shedding, in general, is obtained by measuring the total load current of the interleaved converter. If the load current goes below a certain threshold, the number of phases which are in operation decreases and vice versa. But for this paper, we assume that the phase shedding decision is made at a certain point of time by an external signal. This is done to simplify the analyses of the transient behavior.

The converter with ratings, given below, is modeled in simulink and is as given in Fig.1

VIN = 42 V

VO = 36 V

L = 10 uH

C = 100 uF

IO = 150 A

C Documents and Settings any Desktop Results h1 jpg

Fig.1: Two phase buck converter model in simulink

Current mode control technique is used for this converter. The output voltage is compared to the reference voltage and the error is passed through a compensator to obtain the required reference current IREF. The phase currents are compared with this reference current IREF and the required switching states are generated. This control model is illustrated in the Fig.2

C Documents and Settings any Desktop Results h2 jpg

Fig.2: Control scheme for one phase of the two phase buck converter in simulink

2.1  ‘No ramp’ control:

The converter, discussed in the previous section, is implemented here with the phase shedding control. Phase shedding decision C2 is taken by external source which switches OFF phase#2 at 0.02 s and switches it ON again at 0.04 s. Hence the C2 control is essentially a pulse signal. This control is implemented in the converter in simulink by the use of a simple multiplier block which multiplies the signal C2 and the switching state SS2. Thus as C2 becomes OFF i.e. zero, the SS2 goes into the OFF state, thereby shutting down the phase.

This converter is simulated and the results are as given below:

C Documents and Settings any Desktop Results 1 jpg

Fig.3: Phase currents for a ‘no ramp’ condition

C Documents and Settings any Desktop Results 2 jpg

Fig.4: Output voltage for a ‘no ramp’ system

C Documents and Settings any Desktop Results 3 jpg

Fig.5: Zoomed in view of the phase currents at 0.02s for a ‘no ramp’ system

C Documents and Settings any Desktop Results 4 jpg

Fig.6: Zoomed in view of the output voltage at 0.02s for a ‘no ramp’ system

Fig.7: Zoomed in view of the phase current at 0.04s for a ‘no ramp’ system

Fig.8: Zoomed in view of the output voltage at 0.04s for a ‘no ramp’ system

For the output voltage simulations, given above in Fig.4, it is observed that a voltage dip is observed when phase 2 is switched OFF i.e. a phase shedding decision is made at 0.02 s. The falling slope for a 36 V system is given as -36/L while the rising slope is (VIN - VO)/L i.e. 6/L in our case. Hence we see that the falling slope is much larger than the rising slope. Therefore the phase that is shut down goes to zero much more faster than the rate at which the ON phase can rise to the required current level. This causes a dip in the total current which is reflected to the output voltage as well.

The dip is the voltage, as observed from Fig.6, is close to 27 V which is almost a 9 V dip which is unacceptable in any system. It is also observed from the voltage graph in Fig.8 that the voltage rise in the system, when the phase 2 is switched ON again at 0.04 s is about 40 V which is a 4 V rise but this is not really significant when compared to the dip and hence the focus, for now, is on eliminating the dip.

One way to resolve this issue would be by decreasing the falling slope to a suitable value. One way to decrease this slope would be by not shutting down the phase abruptly and employing a more gradual shut down in phase.

2.2  Ramp control

2.2.1  ‘No current sensing’ control

To improve the behavior of transient voltages in the above system, the phase to be shut is gradually brought into an OFF state. This is done by making the IREF curve of the second phase follow a falling ramp when it is shut down. Basic solution can be summed up as given below:

·  Phase 2 follows IREF under steady state conditions.

·  Phase 2 is shut down implies it follows IREF –IRAMP until its current becomes 0.

·  Phase 2 is OFF implies it follows 0 current.

·  Phase 2 is switched ON implies it follows IRAMP till its current reaches the required value.

·  Phase 2 is ON implies it follows IREF as in steady state conditions.

Implementing this control for the converter is very complicated. Definite controls for detecting the point where current crosses 0 and the point where current reaches the average value need to be defined. Doing so may prove to be cumbersome in analog domain or rather the simulink domain. Hence state machine diagrams are used for implementing this control.

For the state machine diagram, we need to first define the states and the controls required for the states. Here, two states- one for the reference current IREF and another for the ramp current IRAMP are defined. There are four possible combinations of these two states- IREF IRAMP (Q2Q1)- 00,01,10,11. At any given point of time, a system is in one of these four states. The state of the system is decided by certain controls. In this case, the controls are:

1.  C2- This control shows if the phase 2 is in ON or OFF mode. This control has already been described in the previous section. Its value is 1 when the phase is ON and 0 otherwise.

2.  Y- This control detects the point when the phase 2 current reaches 0. It is 1 if the phase 2 current is non zero else it is 0.

3.  Z- This control detects the point when the phase 2 current reaches the average value. It is 1 when the current is equal to the average value else it is 0.

Fig.9: State flow chart for the ramp control in simulink

Fig.9, shown above, is a state flow chat created in simulink and it describes the different states and the transition between the states. During the steady state operation, the system is in the state 10 (IREF is 1 and IRAMP is 0). Once the phase shuts down, then irrespective of the value of Y and Z, the state changes to 11 (IRAMP is negative here). Once Y becomes 0, the state changes to 00. Once again as the phase becomes ON, irrespective of the value of Y and Z, the system is in the state 01. And once Z becomes 1, the state of the system changes to 10.

With the controls C2, Y and Z defined, the state diagram is converted into a flip flop design and integrated into the simulation. The clock pulse used for this flip flop implementation has a frequency of 1000 KHz. The flip flop implementation is as given below in Fig.10.

Fig.10: Flip flop implementation of the ramp control

The value of IRAMP needs to be chosen such that the rising slope of phase 1 is equal to the falling slope of phase 2. The rising slope of phase 1 is approximately equal to 6 x 105 i.e. (VIN - VO)/L and hence a value of 6 x 105 or less for IRAMP will solve the problem of voltage dips. Hence, in this simulation, the value of IRAMP is taken as 6 x 105.

The value of IRAMP is positive when the phase is ON and is negative when the phase is OFF. For C2=0, IRAMP has a negative sign and for C2=1, IRAMP has a positive sign. This is implemented as shown in the figure given below:

This converter is simulated and the results are as given below:

Fig.11: Phase currents for ramp control with a ramp of 6×105

Fig.12: Output voltage for ramp control with a ramp of 6×105

Fig.13: Zoomed in view of phase currents at 0.02s for ramp control with a ramp of 6×105

Fig 14: Zoomed in view of output voltage at 0.02s for ramp control with a ramp of 6×105

The voltage dip is observed to be about 6 V, which is definitely better than the no ramp condition but it is seen that there still is a scope for improvement.

Reducing the falling ramp by ten times will give us better results which are as given below. But it is observed that there is a wide dip of 1 V in the system.

Fig.15: Phase currents for ramp control with a ramp of 6×104

Fig.16: Output voltage for ramp control with a ramp of 6×104

Fig.17: Zoomed in view of phase currents at 0.02s for ramp control with a ramp of 6×104

Fig.18: Zoomed in view of output voltage at 0.02s for ramp control with a ramp of 6×104

2.2.2  ‘Current sensing’ control

The wide dips during the transient phase can be decreased by programming the controller such that the total current passing through the load remains constant at all times i.e. if the controller senses that IO is decreasing, it should increase the duty cycle should increase thereby preventing the dip in VO. So now both VO and IO are sampled and compared with a certain reference. This type of control is also known as the current sensing control technique which is illustrated in Fig.19.

Fig.19: Current sensing control technique

The converter is simulated with the current sensing control technique and the results are as given below:

Fig.20: Phase currents for a ramp of 6×104 with current sensing control

Fig.21: Output voltage for a ramp of 6×104 with current sensing control

Fig.22: Zoomed in view of the phase current at 0.02s for a ramp of 6×104 with current sensing control

Fig.23: Zoomed in view of the output voltage at 0.02s for a ramp of 6×104 with current sensing control

For a ramp of 6×104, it is seen that the wide voltage dip is only about 0.2 V which is a very huge improvement compared to the previous case.

Droop control improves the voltage profile of the system, thereby eliminating the transient disturbances in the system. The only disadvantage with this control is that current needs to be measured to implement this control and current sensing is sometimes a challenge in itself.

3.  Conclusion

In this paper, phase shedding in interleaved converters has been introduced and the challenges in the transient behavior have been analyzed through the simulation results. Through the ‘ramp control’ approach introduced in this paper, it is seen that the transient behavior of the output voltage of the system improves. It is also seen that the ‘ramp control’ approach when combined with the ‘current sensing’ control further improves the voltage profile. Hence this approach can be used for increasing the light load efficiency of the system without disturbing the system behavior during transients.

4.  Acknowledgement

This research is currently sponsored by the Intelligent Systems Center(Missouri University of Science and Technology)

References

[1]  D. Baba, “Benefits of a multiphase buck converter,” Texas Instruments Incorporated.

[2]  B. T. Irving and M. M. Jovanovic, “Analysis, design and performance evaluation of droop current-sharing method,” in Proc. IEEE in Applied Power Electronics Conf. , 2000, pp. 235-241.

[3]  J. Sun, “Dynamic performance analyses of current sharing control for dc/dc converters,” PhD, Virginia Polytechnic Institute and State University, Blacksburg, Virginia, 2007.

[4]  M. S. Elmore, “Input current ripple cancellation in synchronised, parallel connected critically continuous boost converters,” in Proc. IEEE in Applied Power Electronics Conf., 1996, pp. 152-158.