LED Constant Current Driver Circuit Design

An LED converts current into light, and the voltage across it is a consequence rather than a control. A constant current driver is therefore the normal way to power one, and the design questions are about where the current is measured, how much voltage headroom the driver needs and what happens when the junction gets hot.

Why Current Control Is Necessary

The forward voltage of an LED falls by about two millivolts per degree as it warms, and its light output for a given current falls at the same time. Driving it from a fixed voltage therefore produces a current that rises sharply with temperature, which increases the heating further.

That positive feedback is the reason a simple resistor from a supply works only for a small indicator where the current is low and the thermal rise is modest. Above a few tens of milliamps, the current has to be regulated rather than set by a voltage.

The current also has to match between the LEDs of a string. LEDs in series share the same current by definition, while LEDs in parallel share it according to their individual forward voltages, which is why parallel strings need their own ballast or their own driver.

Linear and Switching Topologies

A linear driver drops the difference between the supply and the LED string across a pass transistor. It is quiet, simple and free of magnetics, and it wastes the difference as heat, which limits it to a low current or a supply that is close to the string voltage.

A switching regulator converts the supply efficiently and tolerates a wide difference between input and output. A buck driver is used when the supply is above the string voltage, a boost when it is below, and a buck boost when it can be either.

The choice follows from the power level and the thermal budget. A linear driver that dissipates a watt is a packaging problem, while a switching driver at the same power is an electromagnetic compatibility problem. Both are solvable, and the decision should be made deliberately.

Where the Current Is Measured

current sensing is usually done with a resistor in series with the string, and the voltage across it is compared with a reference. The resistor value sets the current for a given reference, and its tolerance and temperature coefficient appear directly in the accuracy of the drive.

A sense resistor in the high side of the string requires a differential amplifier that tolerates the string voltage, while one in the low side is simple but moves the LED cathode away from ground. The low side arrangement is easier and is used whenever the mechanical design allows it.

Kelvin connections to the sense resistor remove the effect of the solder joint and the trace resistance. At a sense voltage of a hundred millivolts, a few milliohms of track resistance is a measurable error, so the sense pair is taken from the inside of the pads. The layout treatment is similar to any low resistance measurement and follows the practice in our guide to mixed signal board design.

Headroom and Dropout

Every driver needs a minimum voltage across it to regulate, made up of the sense voltage, the comparator offset and the saturation voltage of the pass element. If the supply falls below that sum the current falls with it, and the light output dims.

In a battery powered product the headroom sets the point at which the light begins to dim, which is a user visible behaviour. Designing the string voltage to leave a margin at the lowest battery voltage keeps the output constant until the battery is nearly empty.

A switching driver does not have a dropout in the same sense, but it has a maximum duty cycle and a minimum on time. Those limits set the extremes of the conversion ratio it can achieve, and outside them the current regulation is lost.

thermal derating and Junction Temperature

thermal derating in this context means reducing the drive current when the LED junction or the driver gets hot. A light engine that runs at full current in a warm room will exceed the maximum junction temperature of the LED unless the current is reduced or the heat sink is enlarged.

The driver itself also has a thermal limit, and many integrated drivers include a thermal foldback that reduces the current as the die temperature rises. The behaviour is a smooth reduction rather than a shutdown, which is preferable for lighting but has to be characterised.

The thermal path is part of the electrical design. Measuring the case temperature of the LED and its forward voltage at a known current gives the junction temperature, and that measurement is what turns a derating curve from a guess into a design input.

Dimming and Control

Analog dimming reduces the current and shifts the colour of the light slightly, because the emission spectrum changes with drive current. Pulse width modulation switches the current on and off at a frequency above the visible flicker limit and preserves the colour at the cost of a switching waveform.

A pulse width modulated driver has to be fast enough that the edges do not dominate the waveform, and the supply has to tolerate the pulsed load without a large ripple. The decoupling capacitor at the driver is chosen from the pulse current and the acceptable ripple.

Where several channels are dimmed independently, the phase of the pulse trains can be staggered so that the total current drawn from the supply is smoother. That reduces the size of the input capacitor and the emissions at the dimming frequency.

Layout and Thermal Design

Keep the current loop of a switching driver small. The inductor, the switch and the sense resistor form a loop that carries a fast current, and the area of that loop sets the radiated field. A compact layout with a ground plane underneath is the standard answer.

The switching node is the noisiest point on the board and should be kept as small as possible in copper area. A large copper pad on that node acts as a capacitor to the plane and couples the switching waveform into everything nearby.

Separate the analogue sense traces from the power path and bring them to the driver as a pair. The sense voltage is small at the switching frequency, and a shared return with the inductor current puts the ripple directly into the measurement. The general method for containing a switching converter is described in our guide to EMI suppression design principles.

Verification and Common Faults

Verify the driver by measuring the current with a current probe or a shunt and an oscilloscope, not with a multimeter. The average value can be correct while the ripple is large enough to stress the LEDs, and only the waveform shows it.

Test the whole assembly hot and cold. A current that changes by more than a few percent between the temperature extremes points to a reference or a sense resistor with a poor temperature coefficient, and the fix is a component change rather than a layout change.

A driver that flickers at low dimming levels usually has a control loop with too little compensation at low current, or a minimum on time that is longer than the pulse. The release checks that keep these boards consistent are collected in our PCB design release checklist and the assembly points we inspect are listed in judging PCB quality.

FAQ

Can I drive an LED from a fixed voltage with a resistor? Only for a low current indicator. The forward voltage falls as the LED warms, so the current rises and the heating increases.

What sense voltage should I use? A hundred millivolts is a common compromise. Lower values waste less power, and higher values are easier to measure accurately.

Why does my LED light dim when the battery is nearly flat? The driver has run out of headroom. Reduce the string voltage or use a boost driver so the current stays regulated.