Two of the top considerations in data center and server power-supply unit (PSU) designs are power density and efficiency. Meeting the strictest efficiency standard—80 Plus Titanium—is now a minimum requirement for next-generation data center and server PSUs.
The 80 Plus Titanium requires over 96% end-to-end peak efficiency, which means that the power factor correction (PFC) stage efficiency must be higher than 98.6%. The challenge is that the traditional PFC topology cannot meet this efficiency requirement because of bridge rectifier power loss. Assuming a 1-V diode voltage drop and a 230-VAC input, a 4-kW bridge PFC topology will have 33 W of power dissipation on the bridge rectifier and a subsequent 0.825% efficiency loss. Thus, data center designers must adopt topologies such as semi-bridgeless PFC or totem-pole PFC to achieve the >98.6% efficiency target for the PFC stage.
A semi-bridgeless PFC power stage requires two PFC inductors, and each inductor only performs boost operation in half of an AC cycle. A bridgeless totem-pole PFC stage requires only one PFC inductor, which performs boost operation in a full AC cycle. Semi-bridgeless PFC requires a larger footprint, while bridgeless totem-pole PFC enables high power density and high-efficiency PSU designs. That’s why bridgeless totem-pole PFC is popular in power supplies that comply with 80 Plus Titanium requirements.
Figure 1 The semi-bridgeless PFC topology is shown on top and bridgeless totem-pole PFC on bottom. Source: Texas Instruments
Average current-mode control is the most common control method for continuous conduction mode (CCM) PFC, regardless of topology. Average current-mode control uses inductor current instead of instantaneous current (such as peak current). In traditional bridge PFC, the PFC controller’s reference uses ground as the source of a MOSFET and the negative end of an output bulk capacitor, with a current-sensing resistor between the anode of a rectified bridge and ground, as shown in Figure 2. The controller detects voltage on the sensing resistor; an internal operational amplifier converts the signal to a positive value, and then sends the converted signal to the current loop.
Figure 2 This is how current sensing is performed in a traditional PFC circuit. Source: Texas Instruments
Implementing average current-mode control in traditional PFC is simple, as the current flow is unidirectional over the current-sense resistor as well as the PFC inductor. But because the current flow on the PFC inductor in bridgeless totem-pole PFC is bidirectional, you must implement different current-sensing methods for average current-mode control. This article will discuss three current-sensing methods for average current-mode control bridgeless totem-pole PFC, along with their trade-offs.
- Current sensing with a shunt resistor
Figure 3 is a simplified schematic of current sensing with a shunt resistor. In this circuit, the operational amplifier converts the voltage on the shunt resistor to a signal, with its peak-to-peak voltage less than its bias supply voltage (VCC). Adding a DC bias of one-half VCC enables the analog-to-digital converter (ADC) to read bidirectional current signals. The ground reference of the control circuit is generally placed at neutral.
Figure 3 The simplified schematic shows current sensing using non-isolated shunt resistor method. Source: Texas Instruments
Current sensing is typically accurate, without a significant propagation delay. You can increase the bandwidth of the current loop and enable the power stage to respond quickly to an overcurrent fault. A shunt resistor-based current-sensing method provides the highest sensing accuracy, so you can use the sensing results for control and protection, as well as for accurate input power monitoring.
However, shunt resistor-based bridgeless totem-pole PFC requires a complicated circuit for output voltage sensing because the output ground reference and controller (MCU shown in Figure 3) ground reference are not at the same node. Shunt resistor-based current sensing also requires more isolated rails from the auxiliary supply and more isolated drivers.
Another way to enable current sensing and convert it to a control circuit is with an isolated amplifier, as shown in Figure 4. Using an isolated amplifier enables the controller’s ground reference to output to ground, which will simplify the sensing and driver circuitry in the design.
Figure 4 Current sensing is performed with a shunt resistor and isolated amplifier. Source: Texas Instruments
Some of the most important considerations for isolated amplifier-based current sensing include:
- Requires an isolated VCC supply rail for the input-side sampling circuit.
- Requires a narrow input voltage range for low power consumption on the shunt resistor.
- Needs low nonlinearity of the output signal gain and low error drift with temperature for sensing accuracy.
- Needs a fast response to input signal transients.
For example, the AMC3302 isolated amplifier has an input voltage range of ±50 mV, enabling the selection of a shunt resistor with smaller resistance to help reduce power dissipation of the shunt resistor and improve system efficiency. The AMC3302 amplifier’s output bandwidth of 340 kHz ensures fast response to the input transient, while the integration of an isolated DC/DC converter eliminates the need for an external isolated power rail.
- Current sensing with a current transformer
A current transformer is a ferrite component that requires a magnetic reset in every switching circuit. Placing it in series with the main field-effect transistor (FET) enables sampling of the current pulse, as shown in Figure 5. Theoretically, if the controller sampling point is set at one-half the on-time (Ton), the sampled value equals the average current of the PFC choke in a switching cycle.
Figure 5 The schematic shows current sensing performed with a current transformer. Source: Texas Instruments
In bridgeless totem-pole PFC, during the positive half of an AC cycle, Q2 works as the main FET and Q1 is the synchronous FET. During the negative half of an AC cycle, Q1 works as the main FET and Q2 is the synchronous FET. Sensing current at Ton on a full AC cycle thus requires the placement of two current transformers in series with both Q1 and Q2. The controller selects a sample value of CT1 and CT2 from the polarity of the input voltage.
A current transformer provides an isolated current signal for the control circuit. If the controller ground reference is at the negative end of the output bulk capacitor, low-frequency FETs (Q3 and Q4) require only a non-isolated driver, simplifying the sensing circuit compared to the shunt resistor-based sensing method. Unfortunately, the nonlinear B-H curve and hysteresis loop of ferrite material make the sample value inaccurate. The current distortion might be worse, especially during light loads or zero crossing. And in addition to sensing accuracy concerns, having a current transformer in series with the main FETs will increase the power-loop inductance, therefore increasing voltage stresses during switching events. You may need an additional snubber to clamp the FETs’ voltage stress.
- Current sensing with a Hall-effect current sensor
Figure 6 shows a simplified current-sensing circuit that uses a Hall-effect current sensor. The Hall-effect current sensor outputs a Hall potential that is proportional to the input current from isolated electromagnetic signal conversion. The polarity of this potential is the same as the current’s direction, so the Hall-effect current sensor is suitable for bidirectional current sensing.
Figure 6 The schematic shows current sensing performed with a Hall-effect current sensor. Source: Texas Instruments
Compared to the isolated amplifier method, a Hall-effect current sensor converts signals through the magnetic field inside the chip itself, eliminating the need for an isolated power rail for the application. You can design the input conductor resistance below 1 mΩ for high current sensing, achieving lower power loss.
A Hall-effect current sensor’s bandwidth is in the range of 10 kHz to 1 MHz, so for PFC with average current control, choosing a sensor with a bandwidth higher than 10 times that of the current loop is sufficient for the system. In other words, you need more than 50 kHz of bandwidth in the Hall-effect current sensor for 2- to 5-kHz current-loop PFC.
The advantages of the Hall-effect current sensor make it a good option for current sensing in bridgeless totem-pole PFC. One notable characteristic of common Hall-effect current sensors today is that its sensible current range is related to its VCC voltage level. An example is the TMCS1100A1 Hall-effect current sensor that allows 46 A linear measurement current range with VCC = 5V while it only allows 29 A with VCC = 3.3V. Therefore, it’s tricky to use the Hall-effect current sensor’s full sensible range with higher VCC while sending the sensor output signal to a controller with lower VCC.
To understand this trick, let’s start with Hall-effect current sensor fundamentals. Equation 1 expresses the output of a Hall-effect current sensor:
Vo = S × IIN + VOffset (1)
Where S is the sensitivity of the Hall-effect current sensor in millivolts per amperes, IIN is the input current and VOffset is the offset voltage at a zero current input.
In most bidirectional sensors, VOffset is usually set to one-half VCC, so the maximum allowed range of input current is VCC/2×S. A higher VCC will allow a wider sensible current range. For example, a Hall-effect current sensor with 3.3 V of VCC would allow a 1.65-V/S sensible current range, while a Hall-effect current sensor with 5 V of VCC would allow a 2.5-V/S sensible current range.
Figure 7 The circuit configuration accommodates different values of VCC between the Hall-effect current sensor and the controller. Source: Texas Instruments
In order to have a wider sensible current range while outputting to a 3.3-V maximum ADC, an operational amplifier can set the gain of the sensing circuit and adjust the maximum input voltage to the ADC. Figure 7 shows the application circuit. First, the Hall-effect current sensor is sensing a wider input current range, with 5 V of VCC and 2.5 V representing 0 A of current. Equation 2 illustrates how to lower the Hall-effect current sensor’s output signal range based on VCC levels (or voltage reference levels):
1.65 V/2.5 V = R3/R1+R3 (2)
You can use Equation 3 to set the amplifier gain so that the maximum sensed current level is still within the allowable input voltage range of the controller:
Imax = 1.65 V/S × R1/R2 (3)
As long as you determine the maximum current level that you want to sense, you will only need to decide one resistor value (for instance, R1) and can determine the rest (R2 and R3) using Equations 2 and 3.
Pros and cons
Table 1 summarizes the pros and cons of each current-sensing method. Each system’s specifications and requirements will help you determine which sensing method to choose.
Table 1 provides a comparison of different current-sensing methods. Source: Texas Instruments
Below are a couple of shunt-resistor-based and Hall-effect current sensor-based bridgeless totem-pole PFC design examples.
Sheng-Yang Yu is application manager at Texas Instruments.