LED Color Shift Under PWM Dimming

By Steven Keeping

Contributed By Electronic Products


Light-emitting diodes (LED) demand different dimming techniques than the analog methods used for traditional lighting because the correlated color temperature (CCT) of solid-state lighting changes with voltage. Such color change can be detectable to the consumer and undermine claims made about the light quality of LEDs.

The established solution is to eschew analog dimming in favor of a digital technique such as pulse width modulation (PWM). Manufacturers claim the use of PWM ensures that LEDs will be immune to color changes during dimming.

However, recent research suggests this is not the case because different PWM duty cycles change the junction temperature of the LED; and junction temperature itself is known to alter the color emitted by the chip (the “chromaticity”).

This article reviews PWM dimming, takes a closer look at its effect on junction temperature, and considers whether any changes in chromaticity are perceptible to the consumer.

Driving and dimming LEDs

The efficient operation of an LED relies on a carefully managed power supply because the devices feature a non-linear forward voltage/forward current relationship.

An LED is, as the name states, a form of diode. In normal operation, a constant forward voltage of sufficient magnitude (typically with the LED in series with a resistor) such that the device is operating in its (narrow) conduction region, is applied. Forward voltages for commercial high-brightness devices vary, depending on whether the chip is a low-, medium-, or high-voltage device, but a typical operational range for a low-voltage device would be 2.7 to 3 V.

The forward current determines the relative luminous flux (essentially the brightness) of the LED. Figures 1a and 1b illustrate the forward voltage versus forward current and forward current versus relative light output for a Philips Lumileds Luxeon TX LED. This chip produces 289 lm at 1 A and 101 lm/W (at 700 mA/2.8 V).

Forward voltage vs. forward current for a Philips Lumileds Luxeon TX LED

Figure 1a: Forward voltage vs. forward current for a Philips Lumileds Luxeon TX LED.

Forward current vs. relative light output for a Philips Lumileds Luxeon TX LED

Figure 1b: Forward current vs. relative light output for a Philips Lumileds Luxeon TX LED.

The graphs illustrate how a relatively minor variation in the forward voltage significantly affects the forward current and hence the luminous flux.

Lighting manufacturers would prefer that consumers could dim LED lighting in the same way that they dim incandescent lighting. Such analog dimming techniques are simple to implement and many inexpensive solutions are on the market.

Unfortunately, things are not that simple. First, the forward voltage/forward current relationship is not linear. Second, a small change in voltage dramatically alters forward current, making precise control difficult. In addition, a change in voltage not only alters the luminosity, but also the CCT of the LED.

Most white LEDs use a blue LED “photon pump” allied to an yttrium aluminum garnet (YAG) phosphor. Some of the photons from the blue LED pass through the phosphor unaffected, while others are absorbed by the material and re-emitted in the yellow part of the spectrum. The eye perceives the combination of blue and yellow light as white. At low current, the light looks ‘warmer’ (more yellow). However, at high current, the phosphor becomes less efficient and the blue emission becomes more dominant, making the light ‘cooler’ or bluish. LED manufacturers specify that their devices will produce the nominal CCT at a particular forward voltage.

Digital dimming using PWM has become a popular solution for dimming because the ‘on’ part of the pulse train cycle drives the LED at the manufacturer’s specified forward voltage (thus maintaining the CCT), but by altering the duty cycle of the pulse train, the (average) current can be varied to increase or decrease the LED’s luminosity. Because LEDs are semiconductor devices, they can be switched at well above the frequency where the eye detects flicker.

Silicon vendors offer a range of dedicated LED drivers that look after both the drive voltage requirements and PWM control. A key characteristic of these drivers is the speed at which they can switch (‘slew’) the current between ‘on’ and ‘off’. A rapid slew rate improves the precision of the dimming control. The most popular solution is a switching voltage regulator in a step-down (“buck”) configuration because this topology features the fastest slew rates.

Figure 2 shows a Texas Instruments LED driver using PWM dimming and the resultant output waveforms. The TI LM3404 device is a monolithic switching voltage regulator designed to deliver constant currents to high-power LEDs and targets automotive, industrial, and general lighting applications. Hysteretic controlled on-time and an external resistor allow the converter output voltage to adjust as needed to deliver a constant current to series and series-parallel connected LED arrays of varying number and type. LED dimming via PWM is part of the standard feature set.

TI’s LM3404 LED driver

Figure 2: TI’s LM3404 LED driver and PWM dimming waveforms produced by the chip.

The effect of PWM duty cycle on junction temperature

LED manufacturers produce white LEDs with outputs classed as “warm white” (2,600 to 3,700 K CCT), “neutral white” (3,700 to 5,000 K CCT) and “cool white” (5,000 to 8,300 K CCT) grouped into “bins” comprising devices of very similar color output. Manufacturers spend a lot of time and money on the blue photon pump and YAG phosphor to ensure that the emitted light matches the specification and consumer’s expectations. Many manufacturers suggest that PWM is the best solution for dimming LEDs while maintaining nominal CCT.

However, it seems that this viewpoint may not be quite correct. Recent studies have highlighted that PWM dimming may not be as immune to chromaticity instability as first thought. In one study,¹ the researchers checked the performance of yellow, orange, and red LEDs against PWM dimming of varying duty cycle.

The researchers subjected the LEDs to PWM dimming schemes at a current of 20 mA and modulation frequency of 1 kHz. The duty cycle was varied from 3 to 100 percent. The experiments revealed that the peak wavelengths of each LED moved toward shorter (“bluer”) wavelengths as the duty cycle decreased. Figure 3 illustrates this effect.

Change in peak wavelength

Figure 3: Change in peak wavelength of yellow, orange, and red LEDs vs. PWM dimming duty cycle.

It turns out that the change in peak wavelength (and hence chromaticity) is due to the fact that lower duty cycles heat the LED p-n junction less than higher cycles. The physics is complex, but in essence, junction temperature alters the chromaticity because the LED’s band gap (which determines the wavelength of emitted photons) narrows as the temperature rises.

Figure 4 illustrates how the junction temperature changes with duty cycle for the three LEDs.

Junction temperature increase

Figure 4: Junction temperature increase of yellow, orange, and red LEDs vs. PWM dimming duty cycle.

How much of a change in peak wavelength is required before the chromaticity shift becomes perceptible to the consumer? To find out, the researchers plotted how the three LEDs’ chromaticity coordinates varied on the CIE color space between the lowest and highest duty cycles. The difference in the coordinate positions was 0.012 for the yellow LED, 0.007 for the orange, and 0.002 for the red. The human eye is able to perceive a difference of 0.003 in the yellow and orange region, and 0.004 in the red. Therefore, the chromaticity shift of the yellow and orange LEDs was easily detectable.

The conclusions that the researchers reported were that it was the junction temperature change of the LED that caused the chromaticity shift, and the temperature was influenced by the PWM duty cycle.

In this study, the experimenters worked with yellow, orange, and red LEDs rather than the white devices used in mainstream lighting applications. Nonetheless, the physical principles that govern the operation of the colored LEDs and the material used in their construction are virtually identical to those of the blue LED photon pump at the heart of a white LED.

Another study² conducted experiments with PWM dimming schemes with duty cycles from 3 to 100 percent on white LEDs. In this study, the researchers used both white LEDs employing a blue die and phosphor and devices that combine red, green, and blue (RGB) LEDs to produce white light.

The effect of PWM duty cycle on chromaticity was again observed. Again a shift was noted for both types of device, with the chromaticity shift for the RGB system being more marked than the white LED, primarily due to the significant change in wavelength suffered by the red device.

Commercial LED vendors such as Seoul Semiconductor, Cree, and OSRAM do not include information about how PWM duty cycle temperature alters the chromaticity of their white LEDs in datasheets. However, they do provide data on how chromaticity is affected by junction temperature. (See the TechZone article “Thermal Effects on White LED Chromaticity.”)

Figure 5 shows the relative variation in the CIE color space coordinates of Seoul Semiconductor’s Z-Power chip-on-board (COB) white LED with temperature. (The CIE color space defines the chromaticity of a light source; see the TechZone article “Defining the Color Characteristics of White LEDs.”)

The Z-Power COB comes from the company’s ZC12 Series high-power LED modules and can generate 2,000 lm at 350 mA at a nominal CCT of 5,600 K.

From Figure 5, it can be seen that the coordinates shift by around 0.8 percent as the junction temperature increases from 25° to 125°C.

Seoul Semiconductor’s Z-Power COB

Figure 5: Junction temperature vs. CIE X, Y shift for Seoul Semiconductor’s Z-Power COB.

The Z-Power COB is a high-voltage device (37 V), but the effect of junction temperature on chromaticity can be significant in lower-voltage devices as well. Consider Cree’s X-Lamp MK-R. This is a 12 V LED that produces 1728 lm at 1.25 A (with an efficacy of 129 lm/W). The LED’s nominal CCT is 6,200 K; Figure 6 shows how the chip’s CIE color space coordinates vary with junction temperature.

Cree’s XLamp MK-R

Figure 6: Shift in CIE x, y coordinates with junction temperature for Cree’s XLamp MK-R.

Test the effect

PWM dimming is a proven technique for LED lighting control. However, the designer should be aware that its effect on chromaticity, particularly for white light produced from RGB devices, could be more significant than expected.

Recent research suggests that the influence of the PWM dimming on LED color is so large that it must be taken into account when designing a satisfactory light source. The scientists even go as far as to suggest that information about spectral and color variations under PWM control should be included in the datasheets by the LED manufacturers.

Comprehensive testing, particularly at the extremes of the PWM range, is encouraged to determine its influence on junction temperature and the consequent change to chromaticity. Often the variation is insignificant, but if a lighting fixture’s thermal management is less than perfect, junction heating from varying PWM duty cycles could cause the color change to become noticeable. Worse still, increasing junction temperature lowers the LED’s luminosity.

Such variation in color risks potentially disappoint consumers who have grown up with the consistent glow of incandescent bulbs or fluorescent tubes, slowing the uptake of LEDs for mainstream lighting at a time when their environmental advantages are growing in significance.

For more information on the parts mentioned in this article, use the links provided to access product pages on the DigiKey website.

References:
  1. On spectral and thermal behaviors of AlGaInP light-emitting diodes under pulse-width modulation,” P. Manninen and P. Orreveteläinen, Appl. Phys. Lett. Vol. 91, No. 18, 181121, 2007.
  2. Impact of Dimming White LEDs: Chromaticity Shifts Due to Different Dimming Methods,” Marc Dyble, Nadarajah Narendran, Andrew Bierman, and Terence Klein, Lighting Research Center Rensselaer Polytechnic Institute, 2005.
 

Disclaimer: The opinions, beliefs, and viewpoints expressed by the various authors and/or forum participants on this website do not necessarily reflect the opinions, beliefs, and viewpoints of DigiKey or official policies of DigiKey.

About this author

Image of Steven Keeping

Steven Keeping

Steven Keeping is a contributing author at DigiKey. He obtained an HNC in Applied Physics from Bournemouth University, U.K., and a BEng (Hons.) from Brighton University, U.K., before embarking on a seven-year career as an electronics manufacturing engineer with Eurotherm and BOC. For the last two decades, Steven has worked as a technology journalist, editor and publisher. He moved to Sydney in 2001 so he could road- and mountain-bike all year round, and work as editor of Australian Electronics Engineering. Steven became a freelance journalist in 2006 and his specialities include RF, LEDs and power management.

About this publisher

Electronic Products

Electronic Products magazine and ElectronicProducts.com serves engineers and engineering managers responsible for designing electronic equipment and systems.