With recent developments in power electronics and renewable energy, DC power supply systems have received renewed attention from researchers. Compared to existing power supply systems, DC power requires fewer power conversion stages and can be integrated and/or extended for use as an uninterruptable power supply (UPS) or a regenerative energy source of greater efficiency and reliability (e.g., for electrical vehicles). In IT data centers in particular, DC power systems have taken hold as a more cost effective solution, operating at around 20% higher efficiency than previous AC power systems. Despite recent research efforts, however, the greater efficiency of DC power systems has yet to be applied in more personal (i.e., household) settings, and infrastructural support for such systems remains inadequate. Even so, the overarching project to change home appliances from AC to DC power, and thereby eliminate AC-DC conversion, remains a central priority [1-10].
If DC home appliances can be designed to use the same rated power and function as existing appliances, their steady-state characteristics would be the same as those of existing AC home appliances. The problem is that, in the transient state of plugging in / out, or of heavy DC load control, difficulties arise from the different types and sizes of supplied power. These difficulties must be addressed before DC power can be brought to the home .
In this paper, problems triggered by the DC transient state of DC home appliances are classified in detail, through a comparative analysis between AC and DC home appliances. Multi-circuit countermeasures are also analyzed and experimentally verified using a standard testbed (5-kW DC power supply system and 2.5-kW DC home appliances). From these comprehensive evaluations, the most suitable techniques for solving the transient state problems inherent to DC home appliances are proposed.
2. Comparative Analysis of AC and DC Home Appliances
As shown in Fig. 1, a DC home power supply system is much simpler than an AC home power supply system, owing to its minimization of the power conversion stage. The internal structure of a representative AC home appliance, (e.g., refrigerator, washing machine, etc.) consists of multiple loads operated via relays and/or power devices. By simply removing the AC-DC power conversion stage from an existing home appliance, a primitive DC appliance can be realized, as shown in Fig. 2.
Fig. 1.Schematic diagrams of home appliances with AC and DC power supplies
Fig. 2.Block diagram of a primitive DC home appliance
Note that, after the AC-DC conversion stage, the steadystate characteristics of this primitive model of a DC home appliance are the same as those of an existing home appliance. There are, however, significant differences in the transient state at start-up (i.e., plug-in or power-up) or under heavy DC load control, owing to differences in the type and size of supplied power. If a DC home appliance is designed with no consideration of these differences, its safety and reliability cannot be guaranteed.
Fig. 2 also indicates the most problematic aspects of the primitive DC home appliance - namely, structural change and DC arc. These problematic aspects are classified and their corresponding countermeasures summarized in Table 1. For purposes of discussion and experimentation, we will assume a fixed DC supply voltage of 380-VDC, the likely standard for future DC appliances .
Table 1.New Problems of DC home appliance and its countermeasures
3. Inrush Current Reduction Technique for DC Home Appliance
When a DC home appliance is initially connected to a DC grid, the input capacitor voltage rapidly increases from zero to the rated supply voltage. This sharp increase in voltage can lead to an inrush current that exceeds the maximum rated current for many component devices, resulting in immediate breakage, repetitive damage to low current immunity devices, and malfunction of sensitive circuits such as are found in overcurrent devices [11-13]. The start-up inrush current can be approximately expressed by the DC distribution voltage, VD, and the equivalent resistance, Requivalent, as follows:
The path of a start-up inrush current and the associated parameters are shown in Fig. 3(a). The nominalized simulation waveforms of Fig. 3(b) imply that the inrush current risk, which is proportional to the inrush current value, ranges from a minimum of 1.77 to a maximum of 9 times greater for DC home appliances. Furthermore, devices and techniques that can be used to reduce the inrush current are limited owing to their inherent weakness to the DC arc. This makes the inrush current problem particularly difficult to solve for DC appliances, as well as increases the associated risk.
Fig. 3.Characteristics of start-up inrush current
In this section, various inrush current reduction techniques are compared with respect to their suitability to DC appliances, using a detailed analysis of their simulated waveforms. A series connection of two electrolytic capacitors (specification: 560 μF, Vmax = 250 V, ESR = 0.7R) is selected as the input capacitor and the settling time is defined as the point at which the input voltage of a home appliance exceeds 95% of the rated supply voltage.
3.1 Inrush current reduction method using inductor
Fig. 4(a) illustrates the circuit used in this technique. A series inductor connecting the input capacitor and the input terminal restricts the slope of the start-up inrush current. The circuit follows the RLC step response with underdamping, which is accompanied by an overshoot and oscillation, as shown in the representative simulation waveform in Fig. 4(b). The parasitic inductance and resistance of the PCB, which are much smaller than the ESR of the input capacitor and series inductor, can be neglected. The resulting inrush current and input capacitor voltage are expressed by
Fig. 4.Inrush current limiter using inductor
Eqs. (2) and (3) show how inrush current reduction can be improved by adding a large series inductor. Note, however, that the settling time (=3/α s) also increases in proportion to the inductance, resulting in a long period of unstable oscillation.
3.2 Inrush current reduction method using relay
Fig. 5(a) illustrates the circuit used in this technique. The peak value of the inrush is restricted by the added series resistor upon start-up. After start-up, the power loss from this resistor can be removed by turning on the relay connected in parallel. The inrush current and the input capacitor voltage are expressed by
Fig. 5.Inrush current limiter using relay
The peak inrush current is directly proportional to the ratio of the supply voltage to the resistance. In addition, this circuit achieves over-damping without overshoot and oscillation of input voltage. Thus, even if the settling time is increased, the stability of the system does not deteriorate. It should be noted, however, that excessive current can occur not only during start-up but also during relay turn-on, as indicated in Fig. 5(b). To keep this secondary inrush current at an acceptable level, an additional control circuit should be used to determine the relay turn-on time.
3.3 Inrush current reduction method using relay
This method controls the start-up inrush current by connecting an NTC (negative temperature coefficient) thermistor in series on the path . In the start-up transient state, the inrush current is limited by the initial resistance of the NTC thermistor. When the current flows into the NTC thermistor, the resistance of the NTC thermistor declines automatically, as it warms. The advantage of this method is that it does not require an additional circuit to prevent resistance loss, and is therefore simpler and less expensive.
3.4 Inrush current reduction method using MOSFET
This method reduces the time derivatives of the input capacitor voltage by manipulating the turn-on transient curve of a MOSFET connected in series with the input capacitor. This technique is not suitable to an AC system because its anti-parallel diode can create an unwanted path if the polarity of the applied voltage is frequently changed. Even if the polarity is fixed by an additional circuit, the method does not offer any advantages over techniques used in AC home appliances. Consequently, research into inrush current reduction using the MOSFET method is difficult to find [11-13]. In contrast, for a DC home appliance, the MOSFET method has clear advantages over existing relay methods, owing to the robustness to a DC arc with low power loss .
To control the turn-on transient state of the MOSFET, the non-linear transient curve of Fig. 6(a) should be transformed to a linear curve, like the one shown in Fig. 6(b). This linear curve can be obtained when changes in the parasitic capacitance of the MOSFET (shown in Fig. 7(b)) are not affecting the transient curve. For this to occur, a relatively large capacitor needs to be added in parallel with the gate-drain capacitance, as shown by Cedit of Fig. 7(a). With this additional large capacitor, Coss and Ciss can be approximated by Cedit for the entire transient region, regardless of changes in the drain-source voltage. By connecting this modified MOSFET in series with the input capacitor, the start-up inrush of the circuit shown in Fig. 8 is limited as Fig. 9(a). The circuit consists of Rinitial and Cinitial, which determine the initial value of the gate-source voltage; Rdamping, which reduces the oscillation in the transient state; and Cedit, which manipulates the turn-on transient curve.
Fig. 6.Turn-on transient curves for N-MOSFET: (a) Typical turn-on transient curve and (b) Modified turn-on transient curve
Fig. 7.MOSFET parasitic capacitance: (a) MOSFET parasitic capacitance circuit; (b) Parasitic capacitance versus VDS
Fig. 8.Configuration of MOSFET inrush current reduction circuit for DC home appliance
Fig. 9.Simulation waveforms of MOSET inrush current reduction circuit for DC home appliance
Because the start-up inrush current with MOSFET inrush current limiter is equal to the drain-source current of MOSFET connected to the input capacitor in series and clamped at constant level corresponding to the miller-flat level of the gate-source voltage, the inrush current is almost constant value, Ipeak,inrush. Therefore, once the desired peak value of the inrush current, Ipeak,inrush, is selected, the settling time, tsettling, corresponding to this peak value, can be determined by
Because the drain current is constant at Ipeak,inrush, when the drain-source voltage decreases, the gate-source voltage of the flat region (miller-flat), Vgs,flat, is determined by
where gfs is the transconductance indicating the proportions of the gate-source voltage to drain current. In this flat region, the slope of the input capacitor voltage is equal to the slope of the gate-drain voltage of the MOSFET connected in series. Therefore, the gate resistance related with these slopes, Rgate, is given as
To obtain the intended turn-on curve, the MOSFET should be in the off-state immediately after the home appliance is connected with the distribution board. However, because the gate-source capacitance is by far the lowest among the capacitances of all series capacitors, its voltage can suddenly exceed the threshold. In this case, an unwanted turn-on of the MOSFET can occur, resulting in failure to reduce the start-up inrush current. Therefore, a sufficiently large capacitance in parallel with the gatesource capacitance, Cinitial, (in Fig. 8) is needed to reduce the initial gate-source voltage to a level below the threshold voltage, as shown in the simulation waveforms of Figs. 9 (b) and 9 (c). This condition can be satisfied if
In addition to this condition, to stabilize the gate-source voltage at the designed flat voltage, Vgs,flat, (as in the simulation waveform of Fig. 9 (b)), the gate-source voltage should be maintained below Vgs,flat until the voltage of Cedit is fully built up and the current flowing through Rdamping drops to zero. This condition can be met if
where tdealy,initial (=6Rdamping*Cedit) is 99.9% of the time for Cedit to fully build up.
4. Input Polarity Correction Circuit for DC Supplied Home Appliance
Unlike existing AC home appliances, DC home appliances can only be connected to a DC grid if the polarity of each terminal is known. Otherwise, a malfunction or even an explosion may result from the polarity error of the electrolytic capacitor. Current workarounds to this unipolar problem are purely mechanical - i.e., modifying the shapes of the plug and an outlet  - and do not represent a fundamental solution. In this study, we devise a solution in circuitry: an input polarity correction circuit (IPCC) that automatically corrects the polarity at the moment of plugin (start-up). Figs. 10 and 11 show schematics for the two types of IPCCs, one using a MOSFET and the other a relay.
Fig. 10.Input polarity correction circuit using MOSFET
Fig. 11.Input polarity correction circuit using relay
At the initial operation of the MOSFET IPCC, the polarity is corrected by the MOSFET’s anti-parallel diodes. Simultaneously, the polarity of the input power is automatically detected by the sensing resistors connected in parallel with the low-side-switches. Through the opamp, the sensing signal that includes detected polarity information triggers a pair of switches corresponding to the input polarity turn-on (no additional controller is needed). Similarly, the relay IPCC corrects initial polarity through a diode-bridge, and, using the detected signal from the sensing resistor, operates a 4-contact relay to open a predetermined path. Using a relay in this way, the IPCC can further reduce start-up inrush, assuming that the current resistor in parallel connection with the relay is designed according to the method described in section 3. 2.
5. Experimental Result with Power Simulator
For experimental verification, we constructed the representative test-bed shown in Fig. 12(a). The 5-kW ACDC PWM converter, which supplies 380-VDC, functions as the virtual DC home supply system, while the power simulator for a DC home appliance consists of separately designed functional boards, as shown in Fig 12(c). Using this test bed, we conducted both a functional unit test and a comprehensive appliance test. Fig. 12(b) shows the specific configurations of the “universal power board” and the “new front-end board.” The universal power board was designed to represent DC-DC conversion on a wide range of home appliances. The new front-end board is designed to protect the user and / or the home appliance itself from the risk of start-up transient state problems.
Fig. 12.Configuration of experimental test bed for DC home supply system
5.1 Performance of start-up inrush current limiter
All of the inrush current reduction techniques described in section 3 were implemented using the new front-end board (cf. Fig. 12) of our DC home appliance power simulator. Because the realistic design range for the peak value of the start-up inrush current varies depending on the type of technique, it is impractical to set the same numerical standards for all techniques. Thus, each technique was evaluated experimentally according to its own specific goals.
The experimental waveform shown in Fig. 13(a) indicates that without any inrush current reduction circuit, an excessive start-up inrush current can be generated over the range (556-A) measured by our instruments.
Fig. 13.Experimental waveforms for start-up inrush current reduction circuits in a simulated DC home appliance
When using the inductor method of reduction, if the desired peak limit of the inrush current is set to 60-A, then for the given inductor size, the theoretical inductance and settling time are 8.5 mH and 37 ms by (2-3). Note, however, that the experimental waveform shown in Fig. 13(b) shows a peak current slightly different from the one desired. The reason for this discrepancy is that the inductance of the series inductor can temporarily decrease under roll-off owing to the DC bias caused by the start-up inrush current, resulting in a higher peak than the one anticipated by theoretical design. Thus, to achieve the desired peak value for the inrush current, we would need a larger series inductor to compensate for roll-off. Note that this would also make the system unreasonably bulky and increase the settling time
When using a relay with a series resistor, if the desired peak limit of the inrush current is set to 20-A, the theoretical initial charging resistor and settling time are 20 Ω and 16.8 ms, based on Eqs. (4, 5). The experimental waveform in Fig. 13(c) supports this theoretical result and verifies the inrush current reduction. Unfortunately, the relay is vulnerable to the DC arc that occurs during a turnoff transient, so this technique is not suitable for DC home appliance, in spite of its design convenience.
Fig. 13(d) shows the experimental result of the NTC reduction method, here implemented with two commercial NTCs connected in series. Although simple and inexpensive, this method is inherently unreliable. Once the temperature of the NTC thermistor reaches a level where the NTC thermistor loses its resistance, a certain amount of cool-off time is required before the NTC thermistor can recover its resistance. If the home appliance is re-powered faster before this cool-off period is complete, the circuit will not be able to reduce the inrush current. Thus, this technique is also unsuitable for a DC home appliance.
Finally, using the MOSFET method, we were able to reduce the inrush current as indicated in Table 2, on the basis of Eqs. (6-12). The experimental result in Fig. 13(e) confirms the high design accuracy of this method. Considering all of the above and theoretical analysis of section 3, we found the MOSFET inrush current limiter to be the best for reducing the inrush current while maintaining a short settling time. Note that this technique is also space and weight efficient.
Table 2.Design Example of Inrush Current Limiter using MOSFET
5.2 Performance of input polarity correction circuits
To assess the IPCCs mentioned in section 4, each circuit was applied to the front-end board for comparative testing. To simulate real-world conditions, loads corresponding to a washing machine appliance were used in all experiments. By using the entire power simulator of a DC home appliance, the experiments were performed according to the actual operation modes of a washing machine (laundry, hot water, dry). The power loss curve for each IPPC in Fig. 14(a) indicates that the power loss is less than half the loss of a simple diode-bridge under most load conditions. With respect to only loss, the relay correction circuit may be the most effective among IPCCs. Given the aforementioned risk of a DC arc, however, the MOSFET correction circuit remains as the most reasonable approach. It should be mentioned that our primitive model of a DC home appliance has an open circuit fault problem in which the electrically charged energy in the input capacitor can flow in reverse to the plug-end just after plug-out. As shown in Fig. 14(b), right after the plug of the test-bed is completely separated from the distribution board, a voltage of approximately 120-V remains, enough to pose a serious safety risk (e.g., an electric shock to a user). However, if the MOSFET polarity correction circuit is on board, it will turn off to block the reverse current path the moment the supply voltage is cut off, thus providing additional fault protection.
Fig. 14.Experimental results of various IPCCs
5.3 Performance of heavy load control using MOSFET
For safety during a heavy DC load control transient, the relay was changed for the MOSFET robust to a DC arc. Taking further advantage of the MOSFET, the pulse width modulation (PWM) control method was applied and evaluated to enhance the performance and stability of the system.
Figs. 15(a) and (b) and Figs. 16(a) and (b) show that using a low duty-ratio PWM pulse for heavy load control is more stable than an on-off control using a high duty-ratio PWM pulse. Note that in Figs. 15(a) and 16(a), the high duty-ratio control produces sag-surge and de-rating of the DC distribution voltage, reducing the stability of the entire system. Similarly, Figs. 15(b) and 16(b) show that using a low duty-ratio PWM pulse suppresses the overshoot voltage and that this detrimental overshoot voltage increases as the duty-ratio rises. Since the existing relay control method corresponds to the on-off control of the MOSFET with full duty-ratio, we conclude that heavy load control using the MOSFET is more stable than the existing relay method in all cases.
Fig. 15.Representative waveforms of heavy DC load control (Duty ratio: 0.5, Fsw: 20kHz, 2kW-220AC resistive heater)
Fig. 16.Representative waveforms of heavy DC load control (Duty ratio: 0.1, Fsw: 20 kHz, 2kW-220 AC resistive heater)
Furthermore, with the MOSFET, the existing on-off control method for a heavy load can be modified into a sophisticated PWM control for enhanced performance.
Using the resistive heater of a drum washing machine (rated power: 2 kW) as the test load, the both control methods are experimentally compared with help of the experimental test-bed and the results are summarized in Fig. 17. In the case of the existing method, temperature control is achieved through a hysteresis band and only two rapid slopes of on and off states, resulting in unavoidable temperature errors as shown in the temperature tracking curves of Fig. 17(a). In contrast, under PWM control using a MOSFET, various slopes of temperature corresponding to various duty ratios can be used, resulting in more sophisticated control as also shown in Fig. 17(a).
Fig. 17.Experimental results of heavy DC load control according to control methods: On-off control with hysteresis band vs. sophisticated variable PWM control with MOSFET (Load: 2kW resistive heater, Recorder: Yokogawa MV2000)
This difference about the control ability is also closely related to the efficient use of the energy. With the sophisticated control ability, PWM control enables to consume the power just exactly as needed while existing on-off control has limits to reduce unnecessary consumption energy. It can be confirmed by the consumption energy curves of Figs. 17(b) and (c) that the consumption energy of PWM control is less than the existing one about 50% in an equal control goal (180℃). Therefore it can be concluded that the performance of the PWM control is superior over the existing controls.
In this paper, for the development of a safe and reliable DC home appliance, inherent problems were classified and robust safety circuits against transient state were presented. Especially, the power simulator of a DC home appliance was constructed and the techniques for solving start-up transient problems were applied to the front-end board. A heavy DC load control board using a MOSFET was newly designed to complete the entire test-bed, leading to more sophisticated control performance and energy saving. As a result, as countermeasures of the aggregated start-up in-rush current problem in a DC home appliance, the MOSFET in-rush current limiter was strongly suggested over the other types of techniques. And for solving the unipolarity problem of a DC home appliance, two types of the IPCC were newly devised. The MOSFET IPCC was more suitable for a DC home appliance. The whole evaluation process was verified theoretically and experimentally. Therefore, it is expected that all of the proposed techniques will be applied to DC home appliances in the near future.