Switching power supplies in the tens of kilowatt power range have been slowly replacing traditional silicon controlled rectifier (SCR) based topologies over the past several decades. The advantages and disadvantages are well known. High frequency operation of switching power supplies enables magnetic components to be reduced in size and weight and allows faster response times to line and load perturbations. The principle disadvantage is that the demands placed on switching devices tend to make high power switching power supplies less reliable than their SCR based counterpart.
Numerous power circuit topologies are currently being deployed for high-power switchmode applications. The most common configurations consist of three power conversion stages:
The two AC to DC converters are very similar in function except for the operating frequencies; the converters consist primarily of rectifiers, low pass filters, and snubbers. The snubbers limit switching transient voltages and absorb energy stored from parasitic components. The second stage, the DC to AC converter, generates a high-frequency voltage which drives a transformer at a frequency generally at 20 kHz or above. The transformer is required for ohmic isolation and production of an output voltage as determined by the transformer turns ratio. The DC to AC converter is the most complex stage and there are numerous power processing topologies presently in production.
Most high-power DC to AC converters utilize a H-bridge configuration, four power devices, for exciting the high-frequency transformer. The H-bridge is controlled with pulse width modulated (PWM) or with other modulation strategies to produce a voltage of limited pulse width or amplitude. Modulation of the H-bridge produces a controllable output voltage.
DC to AC converter topologies fall into three groups: hard-switched converters, soft-switched converters, and resonant converters. The primary difference between the topologies is the switching device’s load line during the commutation period (switching transition). It is during the commutation period where power devices dissipate the most power.
Hard-switched converters allow the power devices and/or snubbers to absorb commutation energy. Soft-switched converters have additional passive circuitry to shape power waveforms to reduce losses during the commutation period. The advantage of reduced commutation losses is offset with increased circuitry complexity, additional on-state losses (due to waveform modification), and sensitivity to loading conditions. Resonant power converters have highly tuned tank circuits which cause either device voltage or current to appear sinusoidal. The advantages and disadvantages are similar to soft-switched converters. Resonant power converters are second-order and timing is more critical than soft-switched converters.
Hard-switched, soft-switched, and resonant converters are usually designed to operate from a DC voltage source and are commonly referred to as voltage-fed converters. Characteristically, voltage-fed converters are prone to shoot through problems which can occur when one device fails to turn off before the other series connected device turns on. While protective circuitry can be designed to minimize catastrophic problems, generally, such protective circuitry must be effective to detect shoot though problems in one to two microseconds. Variation of device parameters and abnormal modulation of voltage-fed converters can cause half-cycle voltage imbalance which can result in transformer core saturation. Protective circuitry must also have a response to detect these conditions before damage can occur in the power semiconductors.
Current-fed power converters -, the electrical dual of voltage-fed converters, is still another, but less known and used, power circuit alternative for power conversion. The advantage of these power converters over their voltage-fed counterpart is that shoot through and half cycle symmetry cannot cause device failure or core saturation. This is characteristic of SCR based converters and one of the main reasons why current-fed converters tend to be more robust. The main disadvantage of current-fed converters is that a fourth power conversion stage is required to convert the DC bus voltage to a DC current. While the added stage results in additional complexity and losses, the power conversion stages can be made to work more efficiently. Current-fed power converter topologies are implemented less than voltage-fed converters primarily because of cost.
This article describes the differences between voltage-fed and current-fed converters and the sensitivities to conditions causing power semiconductor stress. Issues for implementing the fourth power conversion stage, the voltage to current converter, are also discussed.
A simplified schematic of a voltage-fed converter is illustrated in Figure 1. The converter consists of a H-bridge, insulated gate bipolar transistors (IGBT) Q1 through Q4, power transformer T1, and output rectifier diodes D5 through D8. The input voltage source can be a battery, DC power supply, or rectified AC bus. For practical reasons, capacitor C1 is required to insure a low impedance bus at higher frequencies. Inductor L1 and capacitor C2 form a low pass filter that removes the AC components on the output.
Figure 1. Voltage-Fed Converter
With conventional, hard-switched, PWM modulation schemes, Q1, Q4 conduct for a fraction of a half-cycle and Q2, Q3 conduct for a fraction of the other half-cycle. This excites transformer T1 equally on alternate half-cycles. Averaging the rectified voltage on the secondary side of the transformer produces a DC output voltage that is proportional to the conduction period of the IGBT’s.
Timing on voltage-fed converters is critical. If IGBT’s Q1, Q2 or Q3, Q4 conduct simultaneously, current rapidly rises in the conducting devices leading to device failure in microseconds. To prevent this critical operating condition, designers introduce turn-on delays in the modulation schemes, monitor DC bus currents, and sense on-state conditions of the power devices. The challenge of successfully implementing these protective schemes is that circuitry must be both responsive to high speed faults and insensitive to electrical noise. This is a formidable challenge especially when power levels are on the order of tens of kilowatts.
A secondary issue with voltage-fed converters is the production of DC voltages with variations of on-state voltages, variations with rise and fall times, and erroneous switching states. Exciting a transformer with DC voltage causes core saturation and power device failure as previously described. Typical methods to circumvent catastrophic events are the placement of air gaps in the transformer, placement of DC blocking capacitors in series with the transformer primary winding, and deploying current mode modulation for cycle by cycle current balancing.
Current-fed converters are the electrical dual of voltage-fed converters. As illustrated in Figure 2, current-fed converters consist of a H-bridge, IGBT’s Q1 through Q4, power transformer T1, and output rectifier diodes D5 through D8. The input current source has to be created with additional power electronic circuitry. For practical reasons, inductor L1 is required to insure a high impedance bus at higher frequencies. Unlike a voltage-fed converter, the output filter consists of a single component, capacitor C1.
Figure 2. Current-Fed Converter
Current-fed converters operate in a mode where voltage and current waveforms are transposed from that of voltage-fed converters. Operation requires IGBT’s Q1 through Q4 to be PWM modulated, but in this case, with the constraint that Q1, Q3 or Q2, Q4 are never allowed to be simultaneously placed in a non-conducting state. The constraint insures the input impedance of the H-bridge is always finite; otherwise, a current source feeding into an open current would produce a destructive high voltage. (As should be noted, constraints placed on switching states and conditions of abnormal operation are the electrical dual of the voltage-fed topology.) Averaging the rectified current on the secondary side of the transformer produces a DC output current that is proportional to the conduction period of the IGBT’s.
Transformer T1, being excited with a current PWM waveform, is basically insensitive to variations of on-state voltages, variations with rise and fall times, and erroneous switching states. With current-fed converters, core saturation can be prevented as long as the ampere-turn excitation is within the bounds of normal operation even if the current is DC.
The disadvantage of current-fed converters is the fact that current sources are not commonly available and such sources must be created from a voltage source. Deployment of buck converters or choppers is an obvious choice because of their very efficient use of power semiconductors. With this extra power conversion stage, control can be placed in the current-fed converter, chopper, or both. Figure 3 shows a high-power converter with a 3-phase input rectifier, chopper, current-fed converter, and output rectifier.
Figure 3. Rectifier, Chopper, and Current-fed Converter
The novel feature of the current-fed converter combined with an input chopper is its performance under abnormal operating conditions. Transformer T1, IGBT’s Q1 through Q5, and diodes D1 through D8 can all operate in a shorted state with system level protection. Under such conditions, the rate of rise of current is a function of the applied voltage across inductor L1 divided by its inductance. Inductor L1 is typically sized to maintain a peak to peak ripple current within a fraction of it maximum value. As long as system shutdown occurs within the switching period of the chopper, peak currents are well controlled. Permitting an extended fault detection period allows fault protection circuitry to be well filtered enabling robust, nuisance free tripping operation in high electrical noise environments.
Another key feature of the chopper and current-fed converter combination is that each circuit can protect each other from abnormally high currents with a single detecting scheme. A fault in the converter stage can be protected with the chopper shutdown and a fault in the chopper stage can be protected with the current-fed converter shutdown.
The previous constraints placed on the switching states of the current-fed converter can be circumvented with the introduction of catch diode D16. This component provides a current return path for IGBT’s Q1, Q3 or Q2, Q4 when the devices are turned off. Diode D16 clamps the maximum off state voltage of the H-bridge to the voltage across capacitor C1.
This article describes the general characteristics of high power voltage-fed and current-fed converters and their sensitivity to device parameter variations and erroneous switching states. Voltage-fed converters generally have series connected power devices across an input capacitor. Abnormal switching states can permit simultaneous device conduction causing currents to increase very rapidly. In addition, voltage-fed converters can also produce DC offsets which can cause the magnetic core of the main transformer to saturate. To protect power semiconductors under these conditions, high speed fault detection is required. The protection of power semiconductors in high, electrical noise environments is difficult.
Current-fed converters are the electrical dual of voltage-fed converters and prefer a shorted state to an open state of operation. These topologies cannot create fast rising current spikes and cannot cause magnetic core saturation under erroneous conditions. Current-fed converters operate with the robustness of SCR based power supplies, but at high-frequency. Current-fed converters require an additional power processing stage which can be used for control and enhanced system protection.