# How to use current source drivers to boost series-connected LED output

For optimum performance, high brightness LEDs should be driven by acurrent source rather than by a voltage source.

This article examines the design of a constant- current LED driverthat can be used for driving a string of series-connected LEDs. (Figure 1, below )

The number of series-connected LEDs can vary from one to several. Todrive the LEDs, the power-stage topology is a modified buckboostconverter.

Figure1. The LED driver is designed to operate from inputs ranging from 7V to28V, and the LED current varies from 0.4A to 1.2A by adjusting thepotentiometer K2. |

(Clickhere to view an enlarged view ofFigure 1 )

There are many integrated driver solutions in the industry; onepossible integrated driver solution to implement the controllersubsystem is using the MAX16818, an average current mode controllerwith a transconductanceamplifier for the current error amplifier.

The current-sense voltage across the current-sense resistor isamplified internally by a factor of 34.5. The transconductance of thecurrent error amplifier is 550 microseconds and the peak-to-peaksawtooth is 2V. In this circuit, the input current is sensed by aresistor, R _{s} , in the return leg.

The value of the current-sense resistor is set by the averagecurrent limit required. The maximum voltage across the LEDs is where nn is the number of LEDs and is the maximum voltage drop in the LEDs atthe full-load current, I _{f} .

The maximum output power is P _{max } =V _{LEDmax} * I _{f} and the efficiency is n(Eta).Thus, the maximum input current is:

The minimum value of the average current limit threshold is 24mV.The current-sense resistor value can be calculated as:

To prevent oscillations in the PWM comparator section of thecontroller, the slope of the signal on the negative input of thecomparator should be less than the slope of the sawtooth on thepositive input. The slope of the sawtooth is given by Vsfs. while thegain of the current error amplifier is given by

G_{ca} = 34.5 * gm * R _{c} .

In this equation, gm is the transconductance of the current erroramplifier (CEA). The output of this amplifier goes to the negativeinput of the pulse-width modulation (PWM) comparator.

Positive input of the PWM comparator is the sawtooth that has apeak-to-peak amplitude of V _{s} with a switching frequency, f _{s} .

This is the AC gain of the current-error amplifier from thecurrent-sense voltage across R _{s} to the output of the amplifier atthe high frequencies below the pole from the compensation capacitor, C _{p} . This is the gain of interest atthe PWM comparator section.

The maximum gain of the current error amplifier G _{ca} is set by the equation

where V _{LEDmax} /L is the input current down-slope.

From this, we can then see that the maximum value of Rc is given by G _{CA} = 34.5 x gm x R _{c} . The value of C _{c} should be chosen such that thezero due to R _{c} C _{c} is at frequency below the currentloop crossover frequency, fc, for adequate phase margin at crossover.

The equivalent small-signal model for the power section of the LEDdriver circuit is described by the following equations:

The small signal control-to-output gain of the boost regulatorcurrent-loop power section, from V _{ca} at the CEA output to V _{rs} ,the voltage across R _{s} , is:

where R _{s} is the current sense resistor and L is the inputinductance, I _{l} is the DC current in the inductor, V _{in} is the DC input voltage, and V _{LED} is the total DC voltage acrossthe LED string.

The overall open loop gain of the input current section is given bymultiplying the G _{CA} and V_{rs} /V_{ca } equations.The result is set equal to 1 to solve for the loop gain crossoverfrequency:

The maximum value of the crossover frequency, fcmax, is given bysubstituting for the maximum value of R _{c} intothe previous equation:

After the current loop has been designed, we can complete the designof the outer voltage loop.

Case in point

Let's examine a typical design example where the number of LEDs in thestring is 3, the input voltage can range from 7V to 28V, the switchingfrequency is 600 kHz and the chosen inductor is 5.1 microH.

The maximum current required in this example is 1.2A and the numberof LEDs in the string can vary from 1 to 4 and the maximum voltage onthe LED string will be 18V. The total output power is P _{max} = 21.6W.

Assuming an efficiency of 90 percent, we can calculate the maximuminput current as 3.428A. If we set the current sense resistor to 0.007ohms, R _{cmax} can then be calculated as 2.55k-ohms.

Choosing a value of 2 k-ohms for R _{c} less than R _{cmax} .Thus, f _{cmax} = 132.6 kHz for an 18V output.We need to set the zero, f _{z} , at a frequency lower than f_{cmax} . With C _{c} = 2200pf, f _{z} =36.17kHz. The pole, f_{p} , should be set higher than twice theswitching frequency. Cp at 47pf sets f_{p} at 1.693MHz .

The LED is modeled as a voltage source in series with a resistor. Inthe present model, the voltage source is 3.15V with a dynamicresistance of 0.6 ohms for each LED.

If we have three LEDs in series, then the total voltage sourcewould be 9.45V. The total resistance would be 1.8 ohms. The calculatedcrossover frequency at 9V input for a case with three LEDs in series is

uses a current-sense amplifier to sense the LED current and thiscurrent signal is converted to a voltage with respect to ground.

One implementation of this approach uses the MAX4073, which includesthe CSA and a current mirror.

In the circuit, the overall open loop gain of the inner current loopis measured as shown in the circuit by injecting an AC sweep source.The current loop crossover frequency from the simulation is 85.5kHz.This agrees very closely to the calculated crossover frequency of82.445kHz.

The actual voltage gain for the voltage for V _{out} /(R _{sense} * I _{load} ) in the MAX4073T is 20V without anyresistor connected from Vout pin to GND. This gain can be changed byadding an external resistor from Vout pin to GND. The bandwidth of thehigh side sense amplifier is 1.8MHz. A type 2 compensation adequatelycompensates the voltage loop and keeps the LED driver stable over therange of operation.

A network analyzer can be used to optimize the values of the type 2compensation values in the voltage loop. The crossover frequency shouldbe set much lower than the crossover frequency, fc, of the averagecurrent mode control loop.

Dimming control

LED brightness can be controlled using PWM to dim the LEDs. Thisapproach fixes the output current of the LED driver and modulates thetime that the driver is on.

Analog dimming, which varies the analog output current of the LEDdriver, results in color distortion. For this reason PWM dimming is thepreferred solution to vary the LED brightness.

When PWM dimming is used, the time that the LED driver is on ismodulated and the applied duty cycle roughly equates to displaybrightness, with 100 percent being the maximum.

PWM dimming can also be done by changing the programmed LED currentfrom zero to full load current, but this will not result in a widedimming range because the control loop is too slow. The circuit wouldhave to be modified to get a much faster response. One circuitimplementation for the control loop provides PWM dimming on theMAX16818 driver.

When the PWM dimming signal goes low, the LED current is immediatelyreduced to zero by opening a switch in series with the LEDs. At thesame time, the gate drive for the power switching MOSFET is turned offby shorting the CLP pin on the MAX16818.

This is accomplished by turning on Q1. At the same time, a switch(Q3) in series with the compensation components on the outer voltageloop opens, thus maintaining the voltages on the compensationcapacitors in the outer voltage loop voltages. Shorting the CLP pin toground immediately programs the input current to zero. Since theaverage current mode control loop has a very high crossover frequency:

It is not necessary to put a switch in series with the inner currentloop compensating capacitors.

Once the PWM dimming signal goes high, the switch across the CLP pinis turned off and the switch in series with the outer voltagecompensation components is turned on.

The switch in series with the LEDs is turned on, allowing the LEDcurrent to flow. In this manner, the control loop recovers to the samestate as it was prior to opening the LED current path and the LEDcurrent recovers to its operating current extremely fast with minimumovershoot.

The LED driver is designed to operate from inputs ranging from7-28V, and the LED current varies from 0.4 to 1.2A by adjusting thepotentiometer R _{2} . The number of LEDs in series inthe LEDstring can be varied from 1 to 4.

Suresh Hariharan is SeniorScientist at Maxim Integrated Products. To view a PDF version of thisarticle, go toBoost LED performance via current source.