Synchronous buck switching power supply driving LED array

background

Automotive lighting assembly suppliers are considering using LED devices to compete with high-intensity discharge (HID) lighting. First, the driving circuits of LED devices are less complex than HID lamps. HID lamps require high-voltage ballast circuits to start an arc in the HID lamp, and after the arc is started, its voltage output needs to be adjusted to maintain a constant power supply to the HID lamp. From an electromagnetic compatibility (EMC) point of view, these high-voltage circuits are prone to noise, further hindering the use of these technologies in the automotive field. Finally, the cost of LED devices continues to decline, making this technology increasingly attractive to the cost-sensitive automotive market.

A typical LED headlight application requires more than 25 watts of power to the LED array. Because one of the advantages of LED components is high efficiency, the driving electronics should also improve efficiency to fully utilize the advantages of LED technology. Therefore, it is possible to consider using some kind of switching power supply (SMPS) to achieve this goal (see Figure 1). However, the goal of most SMPS designs is to regulate voltage rather than current.

Selecting Circuit Topology

For this application, a buck topology was chosen. Given the input voltage limitations (VBATT = 9 V min.) and the array's forward voltage drop (2xVF = 8.0 V, VFMAX = 4 V@IF = 350 mA), it was reasonable to expect a buck topology to meet these requirements. Other methods of driving LEDs are to generate/regulate the voltage using a switch and then regulate the current through the LED using pulse width modulation. A current limiting resistor is required in series between the LED and the switch to prevent excessive current from flowing through, potentially causing damage. This series resistor dissipates power and also results in reduced efficiency.

However, the SMPS itself has components that help stabilize the current. A simplified circuit of a buck regulator is shown in Figure 2.

A closer look at the energy storage components in this design reveals some interesting points. The current through the inductor can be viewed as both an AC and a DC component. Consider the case of an SMPS inductor operating in continuous mode (see Figure 3 for the current waveform through the inductor). The DC component is of particular interest in this application. Since current is the critical parameter, regulating current and delivering it to the load is the primary goal of this circuit. The goal of minimizing the AC components should also be kept in mind.
 

Additionally, because the output voltage is not a consideration and will vary with the LED device, there is no need to consider the regulation task at this node as with traditional voltage regulator circuits. The output capacitor provides current during this time while the inductor charges and helps provide energy to the LED array. This component of the traditional regulator will remain.

Select Controller

The ON Semiconductor CS5165A was chosen here because it has 5-bit programmability as the error amplifier reference voltage changes from 3.54V to 1.25V. With a variable reference voltage, an adjustable regulator can be designed without changing the feedback components.
 

Another favorable feature of the CS5165A is that it is a controller rather than a regulator. This allows the output switch to be selected based on the specific power handling requirements of the entire circuit. Finally, the CS5165A is a synchronous regulator, further improving the efficiency of higher power designs in this particular application.

Final Design

See Figure 5 for the following discussion. Given the above advantages, the CS5165A can be used in a design with a rated output current of 3.5A over the typical automotive input voltage range (additional protection for load disconnection and reverse battery is required externally). It is assumed that the reader is familiar with the basic concepts of a buck SMPS, so only the more unique features of this design are emphasized here.
 

The first step is to convert the desired parameters and current values ​​into voltage values ​​to be regulated by the CS5165A. This can be done by RSENSE1. To further improve efficiency, an operational amplifier is used to amplify the voltage signal on RSENSE1 and keep the losses in the resistor to a minimum. The equation for determining the Vref set point for the desired load current is as follows (where A is the gain of the amplifier circuit):

VREF=A ILOAD RSENSE1

A pair of NTB45N06 N-channel power MOSFETs were selected from the perspective of reducing conduction losses and heat. In addition, the logic-level version of the device was selected for the upper MOSFET M1. This helps drive the upper MOSFET with a higher input voltage when the charge pump peak reserve is insufficient.

To drive the upper MOSFET, a charge pump is implemented using C1 as the charge pump element. C1 pumps charge into the shunt regulator circuit formed by Q1, D1, D2 & D4, R2 & R3 and C3 & C4. When M2 turns on and drives the switch node (source of the upper MOSFET M1) to ground, C1 charges to the battery voltage through D1. Then, when M1 drives the switch node up from the battery voltage, the charge on C1 is transferred to C3 through D2. This voltage is used to drive M1 above the battery voltage and provide sufficient VGS for the device.

M1 includes D3 & R1, which form an asymmetric drive circuit. In early versions of this design, it was found that shoot-through current was a problem. Shoot-through is defined as current flowing directly from VBATT to GND due to M1 and M2 being on at the same time. Controlling the timing of driving M1 and M2 is very important, so R1 is added to delay the turn-on time of M1. This allows M2 enough time to turn off when M1 turns on. The CS5165A provides some non-overlap time, but adding this circuit has more to gain. Diode D3 reduces the turn-off time of M1 when the drive cycle is reversed. This reduces the shoot-through phenomenon when M2 must be on and M1 must turn off quickly.

Another circuit that reduces punch-through and improves efficiency is the network of D5, R5 & C6. In the presence of high dV/dT at the switch node, the lower MOSFET M2 can turn on through its own drain-gate capacitance. Adding D5, R5 & C6 reduces this effect: When the IC's lower MOSFET drive signal (VgateL) goes high, current flows through the diode and resistor to the source of the FET. This current builds up a voltage on the capacitor equal to the voltage drop across the diode. Diode D5 is a dual diode, so the voltage is about 1.2V. Then, when the lower MOSFET drive signal (VgateL) is driven to ground, the gate of M2 is actually driven below ground due to the voltage on C6. This voltage is enough to turn off M2 when the upper MOSFET M1 is turned on.

Finally, the sense voltage generated on RSENSE1 is amplified with an amplifier. The implemented circuit is a differential configuration with a voltage gain of 10. Therefore, the voltage generated on RSENSE1 varies between 125 mV and 354 mV over the entire range of current regulation. The result is 1/10 the power consumption compared to the direct forward sense resistor method. If RSENSE1 were 0.7 ohms instead of 0.07 ohms, approximately 18 watts would be wasted on the sense resistor! 
 

performance

After building the circuit according to the schematic, the following performance data is obtained. First, the actual output current, IOUT, is plotted as a function of the programmable reference voltage, VREF. VREF can be selected from 1.340 V to 2.090 V in 50 mV steps, and from 2.140 V to 3.540 V in 100 mV steps. The performance is clearly seen in Figure 6.

The values ​​plotted in Figure 6 represent operation for test input voltages from 9 V to 15 V. Note the clear reflection point where the setting changes from 50 mV steps to 100 mV steps. The overall operating range of this circuit can be changed simply by changing the value of RSENSE1. Also note that IOUT changes very little for various input voltages.

The following set of operating waveforms are shown in Figure 7. Note that the operating frequency changes because the CS5165A is a constant off-time controller. Component C12 sets the off-time value. The off-time remains fixed, while the on-time changes based on the load requirements. In this case, the load current changes, which will increase the voltage drop across the LED array. In a classic regulator, the duty cycle changes based on the step-down voltage ratio. Because the voltage ratio effectively changes with different load currents, the duty cycle also changes. Note that these results can be obtained by observing the measured values ​​of the waveforms in Figure 7. Also note the ripple current through L1.
 

Let's briefly discuss the efficiency issue. Refer to Figure 8 for the following discussion. As can be seen, the circuit is generally most efficient at lower input voltages and heavier loads. Under all operating conditions, the overall efficiency will not be less than 75%.

in conclusion

In general, efficiency is lower at higher input voltages because of the startup circuitry and the maximum input voltage allowed by the CS5165A. Zener diode D4 was chosen as an 18V device. This still allows 17.3V to be applied to the CS5165A, taking into account the ~0.7V junction drop across Q1's base-emitter. Although this is slightly above the maximum VCC value on the datasheet, it still allows the upper MOSFET to be driven slightly at higher VBATT values. If the upper MOSFET is driven too lightly, it will operate in the ohmic region, causing more conduction losses in the MOSFET than expected.

The circuit described in this article meets the goal of driving a high power parallel LED array. Some limitations of this approach are in the configuration of the LED array itself. The various parallel and branch circuits will carry different currents depending on how the LED devices are matched. Trying to monitor and control the individual branch circuits requires more effort than rearranging the array. A better array that handles this limitation is to connect all the LED devices in series and boost the voltage from the car battery to meet the requirements. This approach also has its disadvantages. However, once you have a parallel LED array, this circuit can provide many useful functions to drive such a configuration.