Merelle Dr., Thomas

2015-2017 Pi Lighting
2012-2015 Member of Technical Staff at GlobalFoundries, Fab1
2006-2011 Applied Research at Philips-NXP, IMEC
2002-2005 PhD in Physics, Montpellier – Research Collaboration with STMicroelectronics

Colorimetry and Variability of LEDs

Variability is part of the DNA of LEDs. From “tolerances” to binning, LEDs are devices which characteristics of color, voltage, wavelength, reliability vary from one device to another. In this presentation we will share our findings on the impact of parameter variations on the final performance of an LED Light engine. In a first part, we will review the variability causes from a typical LED manufacturing process on the final LED device, and the different parameters we will look at. A rapid review of the “new” colorimetry parameters, Circadian Index, EML, TM-30 parameters will be presented. In a second part, the other causes of variability will be studied. LEDs are devices which are produced, measured and binned. This binning process itself has a significant impact on the “parameter tolerances” we find in LED datasheets. This measurement process during the binning also has significant consequences. The differences between inter-lot variability and intra-lot variability will be explained. During its lifetime, the characteristics of an LED changes, adding more variation, which we will consider.

In a third part, we will combine those effects and show the overall performance a manufacturer can guarantee, and what a final LED user can expect, on all typical parameters, with a special focus on spectra related deduced parameters such as CRI, Rf, Rg, circadian indexes, but also on variability related to flux, efficacy, voltage, etc. With this part, at LED level we will simulate a batch of LEDs and describe the statistical distributions which are obtained in reality.

Finally we will consider those variations, and implement them in a system with several LEDs, and review the final system performance variability a user can expect, and a manufacturer can reach. As a consequence, we will propose some simple and less simple design strategies to reduce this variation, and ensure the end user gets the performance he/she expects.