Flicker Free LED Driver Design

Flicker in a light source is measured by how much the light output changes and how often, and the eye is far more sensitive to it than most designers expect. A driver that passes an electrical specification can still produce a light that causes headaches, and the cause is usually the output ripple.

What Makes Flicker Visible

The modulation depth is the ratio of the variation to the average output, and the frequency is how often it repeats. A small variation at a low frequency is more visible than a large one at a high frequency, because the eye integrates over a short interval.

Frequencies below about a hundred hertz are the most objectionable, and the mains frequency and its harmonics sit exactly there. A ripple at a hundred or a hundred and twenty hertz from a single phase supply is the classic source.

Visible flicker is not the only effect. Even a modulation that cannot be seen consciously causes eye strain and headaches in some people, and stroboscopic effects on rotating machinery are a safety issue in industrial lighting.

Sources of output ripple

A switching driver with a poor output capacitor produces a ripple at the switching frequency and, if the loop is slow, at the mains frequency as well. The second is the one that matters, because it is far more visible.

An off line driver without power factor correction draws its current in short pulses and the output has a component at twice the mains frequency. That ripple is inherent in the topology and has to be filtered out afterwards.

A linear driver fed from a rectified supply reproduces the supply ripple directly. A small capacitor on the output is not enough, because the driver passes the variation through rather than smoothing it.

Choosing the output capacitor

The output capacitor is the main tool against ripple, and its value is set by the current, the frequency and the acceptable variation. The calculation is straightforward and the result is often surprisingly large.

An electrolytic capacitor has the capacitance per unit cost and volume, and its equivalent series resistance limits the ripple current it can handle. Exceeding that rating dries the electrolyte and the capacitor fails early.

A film or ceramic capacitor in parallel with the electrolytic handles the high frequency component, where the electrolytic has become inductive. Both are needed, and the ceramic should be placed closest to the load.

electrolytic capacitor lifetime is the parameter that decides the life of the whole driver. It halves for every ten degrees of temperature rise, so the thermal design around it is part of the flicker design rather than a separate topic.

dimming and Modulation Depth

dimming by pulse width modulation switches the current on and off, and the light follows the current almost exactly. That produces a modulation depth of nearly a hundred percent, and the only thing that saves it is the frequency.

The frequency has to be high enough that the eye integrates the pulses, usually several kilohertz, and high enough that any stroboscopic effect is negligible. A dimming frequency of a few hundred hertz is visible to many people, particularly in peripheral vision.

Analog dimming reduces the current without switching, so the modulation depth depends only on the residual ripple. It is the better choice for comfort where the colour shift does not matter, and the colour shift is the reason it is not always used.

Measuring Flicker

Flicker is measured with a photodiode and an oscilloscope or a dedicated flicker meter. The photodiode must be fast enough to follow the modulation, and a slow one filters out exactly what is being measured.

The two numbers to record are the percent modulation and the frequency. A meter that reports a flicker index combines them into a single figure that can be compared with a specification, and the index is easier to use as an acceptance criterion.

Measure with the light at several dimming levels, because the behaviour at ten percent is often different from the behaviour at full output. The worst case is usually at a low level where the loop gain is low.

The Interaction with the Mains

An off line driver has a ripple at twice the mains frequency that is inherent in the conversion. The output filter reduces it, and a larger capacitor is the direct remedy with the cost of size and life.

A two stage driver with power factor correction and a regulated output removes most of it. The extra stage costs efficiency and money, and it is the standard solution above a certain power because it also meets the harmonic current limits.

For a driver powered from a direct current supply, the ripple comes from the switching only and is much easier to filter. That is one reason a low voltage lighting system with a remote supply can be made flicker free more cheaply.

Thermal and Lifetime Effects

Flicker can appear over time rather than at the start. A capacitor that dries out loses capacitance and the ripple rises, which is a common failure mode in a driver that has been in service for years.

Measuring the ripple at commissioning and again after a period of service is a useful diagnostic. A ripple that has doubled indicates the capacitor, and replacing it is cheaper than replacing the driver.

The temperature of the capacitor should be measured in the enclosure rather than in free air. A capacitor rated for a thousand hours at a hundred and five degrees can last for many years at sixty, and the difference is entirely a matter of where it is mounted.

Driver Topology and Flicker

A buck driver with a large output capacitor gives a low ripple, and its output current is continuous. A boost driver has a discontinuous output current and produces more ripple for the same capacitance.

A flyback driver stores energy in the transformer and delivers it in pulses, so the output ripple is larger and the filter has to work harder. The topology is chosen for isolation or for a wide input range rather than for its ripple performance.

Where the light output has to be smooth, the choice of topology is part of the flicker design and not only an efficiency question. Comparing the ripple of two topologies into the same load is a quick way to decide.

Verification and Faults

Verify the design with a photodiode and a scope at full output and at the lowest dimming level. Both the frequency and the modulation depth should be recorded, and the modulation at the low level is usually the failing one.

Check the temperature of the output capacitor after an hour of operation in the enclosure. Its life is set by that temperature more than by any other single factor in the driver.

A light that flickers only occasionally is usually being dimmed by a controller that is sending a low frequency signal, rather than having a hardware fault. The release checks that keep such a driver consistent are collected in our PCB design release checklist, the assembly points in judging PCB quality, and the thermal measures in our guide to PCB thermal management design.

FAQ

What modulation depth is acceptable? Below about five percent at low frequencies and below twenty percent above a kilohertz for most applications. Lower is always better.

Why is my dimmed light more flickery than the full output? The loop gain is lower at low current, so the ripple is a larger fraction of the output. Increase the output capacitance or the dimming frequency.

Can I use analog dimming instead? Yes, and it removes the switching modulation entirely. The trade is a small shift in the colour of the light as the current changes.