This article draws inspiration from less common practices that some light source and luminaire manufacturers use to improve the general colour rendering index (_Ra according to [1]). Although these methods have pitfalls and some can be considered tricks, it is worth exploring them in more detail using algorithms for calculating colour rendering indices.

Red refill

The stimulus for this experiment was a sample of an LED replacement bulb with_Ra = 90, which contained white LEDs with lower colour rendering and a smaller amount of red LEDs. The spectrometric measurement confirmed the high_Ra value, but the light appeared somewhat pinkish.

In the experiment, we use two types of light emitting diodes: a white one with parameters _Tcp = 6030 K and_Ra = 71 and a red one with wavelength _λp = 631 nm. Let us denote the blue peak intensity of the white LED by_IB and the peak intensity of the red LED by _I1.

_Ra increases with increasing proportion of the red component (_I1/IB), up to 16 points at the maximum. At the same time, the alternate chromaticity temperature decreases, by approximately 2000 K at its maximum. However, a precise algorithm for calculating the color rendering [2] shows an increase in _Rf of only 7 points. As the proportion of the red component increases further above the maximum shown in Fig. 1,_Ra no longer decreases. The progression of the parameters is shown in Table 1. A similar procedure was also described in [3].

In addition to colour rendering, however, the "whiteness" of the mixed light, i.e. its deviation from the reference source, must also be monitored. This can be expressed as the distance of the two lights in the coordinates u', v' (denoted Δu_',v'). Using the approximation of MacAdam ellipses by circles of radius 0.0011 in the coordinates u', v' according to [4], the deviation can be given as N times the base MacAdam ellipse. For N=1 the difference is not visible to the human eye, for N=2 it is slightly visible, for N=3 it is visible. For example, for double-pole fluorescent lamps according to EN 60081:1999, the tolerance is N=5 (SDCM).

From Table 1 it is clear that the limitation of the deviation from the reference source N ≤ 5 is satisfied only for the contributions of the red _I1/IB component in the order of one percent. Here, however, the improvement in_Ra is only a few points. Further increasing_Ra by adding red leads to exceeding the white light limit and it is already coloured light. The experiment therefore shows that high_Ra does not automatically imply high quality white light.

Single channel replenishment

Looking at Figure 1, the question arises whether the wavelength _λp = 631 nm is indeed the most appropriate choice. To give an idea, we can shift _λp of the complementary light across the visible spectrum, mix it with white light and observe the response in_Ra and _Rf. In this experiment, we use a model of the LED spectrum in the form of a Gaussian curve with a width at half maximum (FWHM) of 16 nm, which corresponds to the original red LED. The amplitude _I1 is chosen to be 0.25_IB. The result of this experiment is shown in Fig. 2. The local maxima at 623 and 627 nm indicate that the wavelength is correctly chosen for the sample.

Two-channel replenishment

The results of the previous experiment directly encourage us to use the local maxima around 493 nm, i.e. in the blue (Fig. 2), to improve the colour rendering, which in turn leads to an increase in the chromaticity temperature (Fig. 3). Is it possible to further improve the colour rendering in this way and to bring the coordinates of the mixed light closer to the line of the reference emitters? LEDs with _λp ≈ 490 nm are also available ( Ice Blue).

For the complementary red (_λp1 = 625 nm) and blue (_λp2 = 493 nm) channels,_Ra, _Rf and _Tc were calculated for all combinations of 100 different intensities _I1 and _I2. The ideal combination with the smallest distance from the daylight line is _I1 = 0.47_IB, _I2 = 1.24_IB (see above). The colour rendering results are surprising:_Ra = 96, _Rf = 94 at chromaticity temperature _Tcp = 6300 K. The rendering index of the deep red _R9 is 81, the indices of the other deep colours are greater than 95. The spectrum with multiple peaks resembles a fluorescent lamp rather than a light-emitting diode (see Figure 4).

By controlling the ratio of the complementary components, a chromaticity temperature of _Tcp 5750-6900 K ,_Ra 93-97 and _Rf 91-94 can be achieved while maintaining a deviation of N ≤ 5. Recall the parameters of the white LED: _Tcp = 6030 K and_Ra = 71.

This colour mixing method can achieve impressive results - a 25-point increase in the colour rendering index of the white LED.

In the practical application of this procedure, it must be taken into account that the mixed light can easily deviate from the white light tolerance band, for example due to variations of individual diodes within a series, due to variations of the supply current, due to spectral changes with temperature, or due to different aging rates of different types of diodes. These effects can cause an undesirable visible change in the chromaticity of the light source during its lifetime. They can be compensated for by calibration during manufacture, the use of precision power supplies, temperature dependent control, integration of different LEDs into a single housing, correction of the control according to the aging pattern, manual calibration or automatic calibration of the luminaire using a built-in colour sensor.

In the next part of the miniseries, we will look at whether high colour rendering index light can be obtained by mixing lasers and how many wavelengths need to be combined. We will also explore a filter that improves colour rendering.

Literature used

[1] Method of measuring and specifying colour rendering properties of light sources. Austria: CIE Central Bureau, 1995. CIE Publication, no. 13.3. ISBN 9783900734572.

[2] CIE 2017 Colour Fidelity Index for Accurate Scientific Use. Austria: CIE Central Bureau, 2017. CIE technical report 224. ISBN 9783902842619.

[3] MÁCHA, Marek. Mixing of light sources as a tool for increasing_Ra. Světlo. vol. 2011, no. 5, pp. 62-63. ISSN 1212-0812.

[4] Chromaticity Difference Specification for Light Sources. Austria: CIE Central Bureau, 2014. CIE TN 001:2014.

Author. Antonín Fuksa NASLI & Blue step
Published in Světlo 1/2020