Driving LEDs With NXP Sensorless Sensing

There’s no doubt that LED technology has improved a lot over the last few years. However, LED lamps are still perceived by consumers as inferior to incandescent bulbs, largely because despite the variety of colours of LEDs that are available, they could not match the colours produced by a traditional light bulb. So-called ‘Cold LEDs’, which produce a blue-ish colour light, have gradually been superseded by ‘Warm LEDs’, which can mimic the colour temperature of an incandescent almost flawlessly.

However, the problem is still very pronounced when the lights are dimmed. An incandescent bulb follows what’s called the black body radiation curve, producing a warmer colour temperature when dimmed. By contrast, an LED’s colour temperature does not change when it’s dimmed, making a dimmed LED’s light output much colder when compared to an incandescent. This difference in light quality is one of the issues holding back adoption of LED lamps in the consumer marketplace. This problem is the subject of a new development from NXP which promises to make dimmable LED lamps that mimic the warm light of incandescents in an affordable and reliable way.

Dimming Problems

So, why has it taken until now to come up with a viable solution to the LED dimming problem? Radu Surdeanu, Principal Scientist at NXP, explains that the problem is actually three-fold.


Figure 1: Using RGB LEDs can produce an approximation at the black body radiation curve (the black line in this figure), but it’s a complex and expensive solution

Firstly, the LED system has to follow the black body radiation curve as closely as possible, ideally to within the resolution of the human eye. An approximation of the black body curve can be achieved using a combination of red, green and blue LEDs (Figure 1) but this misses the warmest part of the curve, and as Surdeanu points out, it makes for a complex and rather costly system. A simpler option is to use a white LED in combination with a red one, which is cheaper, but the colour output is worse (Figure 2). NXP’s proposed solution uses a warm white LED alongside an amber LED (Figure 3).

“In this way we can cost effectively mimic the black body radiation curve with the required accuracy — human eye resolution,” Surdeanu says.


Figure 2: White and red LEDs are cheaper but the colour profile is worse

The second requirement for the system is a logarithmic correction for the eye’s sensitivity to variations in light that has to be done when dimming. The eye is very sensitive to changes in light at low light levels, but in bright light, the eye is less sensitive and this must be taken into account. The third and biggest challenge is to accurately compensate for the effects of temperature on the LEDs’ performance.

“When the LEDs are dimmed, the LED temperature reduces, which has an effect on the light output,” he says. “For accurate colour point control, the driving circuitry has to correct accurately for this effect at every moment of dimming.”


Figure 3: NXP’s proposed solution uses a warm white LED alongside an amber LED, which can mimic the black body radiation curve closely

There are several ways in which light fixture manufacturers tackle this challenge. Some simply do nothing and hope for the best! Obviously this gives a poor result. Some manufacturers try to avoid getting the LED hot by over-engineering the heat sink and de-rating the nominal current through the LED. Unfortunately, says Surdeanu, this over-engineered heat sink can cost up to a third of the total cost of the lamp. A better solution is to also add a temperature sensor near the LEDs, then make the assumption that the LED temperature is close to the one given by the sensor and have a thermal model incorporated in a microcontroller to adjust LEDs’ output as necessary. This is the most costly approach since it requires an extra temperature sensor, microcontroller and connections between them.

“It also requires complex thermal modelling which is dependent on the design of the LED module and the heat sink architecture,” Surdeanu says. “Because the LED temperature is not measured directly, and due to the large thermal mass of the heat sink, the method is slow and rather inaccurate.”

Manufacturers using this method have compensated for speed and accuracy by calibrating the system at certain dimming points, then using a lookup table, but this makes the system really rather complex. NXP calls its solution to this problem ‘sensorless sensing’.

Sensorless Sensing

“The sensorless sensing method makes use of the physics of the LED die, which says that, in theory, the forward voltage across the LED should be linearly dependent on (inversely proportional to) the temperature of the LED,” Surdeanu explains. “This is true if the voltage is measured at the right level of current through the LED, which should be low enough not to have self-heating effects while measuring and not to have series resistance effects.”

“In our method, we use the correct range for measurement current levels, which gives us accurate measurement of the voltage across the LED, in real time. We use only the two wires already needed to drive the LED, and since we measure directly at the LED, we have a direct temperature measurement at the source, therefore, independent of the LED system design or heat sink architecture,” he says.

If there is already a microcontroller in the system, this can be used to perform the measurements and the feedback loop, but it’s not required. These functions can be implemented in the LED driver, which then outputs the appropriate current level needed to drive the two LEDs to obtain the correct colour point in any conditions. NXP has demonstrated this feedback loop clocking in at 20µs, fast enough to dynamically adjust the light colour. Their demo (Figure 4) also self-calibrates to compensate for ageing.


Figure 4: The NXP sensorless sensing demonstrator at CES 2013 showing a warm, dim light

Cost-Saving

Surdeanu points out that without a microcontroller, and since an external temperature sensor and its connections are not required (‘sensorless’), system cost is reduced. There are also a couple of more subtle ways that sensorless sensing can make LED lamps cheaper. Typically, LEDs are driven at about 80% of their nominal current to avoid overheating, so more LEDs need to be used to get the same light output. Using the NXP method, the LEDs can be driven closer to their rated current, so fewer LEDs are required. Also, a smaller heat sink can be used since temperature is measured directly at the heat source.

In fact, sensorless sensing can be used as a form of overheat protection since it’s accurate to around 1°C — the drivers receive the temperature measurements anyway and this information can be used to make sure the LEDs don’t exceed a set temperature. This, combined with removing as many external components as possible, can help improve the reliability of the system.

It’s hoped that sensorless sensing will help make LED lamps a more attractive choice for consumers since they’ll be able to mimic the incandescent bulb very closely. Outside of lighting, the technology can be used in applications such as mobile phone camera flashes — to stop the LED overheating while filming in flash mode, to increase the light when used as a torch, or to get the correct colour point for a double LED flash. Automotive applications also provide some uses for the overheat protection capability, as well as failure detection and LED lifetime indicators. Given LEDs’ overall attractive energy consumption profile, and the potential market for these devices globally, these developments should help make a difference to energy saving worldwide.