Complexities of Bicolor LED Lights:
An Extensive Color Analysis
By Tim Park
LEDs come in all types of shapes and sizes. There are panels, bulbs, ribbons, and COB (“chip on board”) fixtures often found in fresnel type point-source lights. For a long time these LEDs were single-color, balanced to either tungsten or daylight. However this year as we tested over 165 LED lights at the annual NAB trade show to update our comprehensive LED database, we found that over half of the LEDs at the show were bicolor. This was an incredible change from last year when there were only a handful.
What is a Bicolor LED?
The idea behind bicolor LED lights is that there are tungsten balance LEDs and daylight balanced LEDs mixed within the same fixture. By adjusting the intensity of each type of emitter, the entire range of color temperature spectrum between the two LED classes is now available to the filmmaker. This way a filmmaker just needs one type of light that they can dial in to match incandescent practicals, or the sun outside, or brands of single-color lights that have color temps that are a bit off, etc.
Initially this sounds like a wonderful idea, which is probably why so many LED manufacturers are doing it. They already have each type of LED, all they have to do now is repackage them into the same fixture and allow the user to dim them independently. Some manufacturers even provide a readout so you know where on the color temperature range the you are. (Word of caution: most of these readouts are off by 200 K and often even more.)
The more I thought about the intricacies of creating a bicolor LED, the more I wondered if this was even good for film and video makers. Mixing two dissimilar lights does not necessarily create a useful blended light.
Tungsten LED? Daylight LED? Please Explain
Tungsten light is based on blackbody principles, where a blackbody emits visible light if you heat it up enough. As it gets hotter and hotter, the color balance of the light changes, going from a more orange color to a more blue color. This is where the idea of correlated color temperature (CCT) come from, and why the units are kelvin (K), which is a measure of temperature. Looking at the spectrum from a true tungsten light source shows how full and even the color spectrum is.
However, daylight is VERY different. While the sun is a great light source, that light is significantly changed as it goes through our atmosphere. It’s why the sky is blue, the clouds are white, and the sunsets are so full of reds and oranges. This explains why the color temperature of daylight depends on time of day and if you are in the sunlight or in the shade, and if it is cloudy or completely sunny. In fact tungsten light and “daylight” are so different that the color quality is measured using very different standards. These standards — called “standard illuminants” — abruptly change at 5000K, which is why all spectrometers give very odd graphics and data as you cross over this temperature point; below 5000K CIE Standard Illuminant A is used, at and above 5000K CIE Standard Illuminant D65 is used. (More on the abrupt change below, in the section titled “Specific R-Values.”)
So if a LED manufacturer plans to mix these two types of light together, what happens when they both are on at full intensity to give you the color temperatures in the middle? In our video “Are LEDs Ruining Your Projects” we show the problems you can have if you mix different brands of LEDs together without doing any camera tests first. And the LED lights in that experiment were all supposedly daylight-balanced. Throw in some tungsten LEDs and very new problems could potentially crop up.
Let’s say the LED manufacturer decides to modify their phosphor formulations so that the resulting mixed light is still useable. What does this do to the tungsten and daylight extremes of the color range? The tungsten and daylight ends might now be lower quality as the manufacturer compromises the LEDs.
What is a filmmaker to do…
With the potential for so many problems with bicolor LEDs, and with so many brands building them, it was imperative to figure out what is going on along the entire bicolor range, as well as what happens when those lights are dimmed.
So I took a bicolor LED from one of the top LED manufacturers on the market and ran some tests. Yuji International is a huge innovator in LED phospors and are not shy about putting all the data from their LEDs right there on their website. So if they can pull off bicolor LED lights, there is hope. If they can’t, well…
(Note: We are not affiliated with Yuji International in any way, and gain nothing by mentioning their lights or linking to them. They have twice sent us small samples of LEDs free of charge for our tests, but that is it.)
The Light We Tested: Yuji BC Series Hybrid
Yuji International sent me a half-strip of their BC Series 12x Multirow LEDs (24V). The BC Series is a line of high CRI LEDs that Yuji sells in different forms, from bulbs to COB emitters to different sized ribbons. However, it isn’t their highest CRI line. Their VTC Series of LEDs has even better CRI values since they use a violet die instead of a blue die to excite the phosphors, and then add blue phosphors to the mix to still get the blue values. However Yuji does not have bicolor lights in their VTC Series, so the BC Series Hybrid was going to have to do. Still, it is a great line of LED so this was not a problem.
We have previously tested the VTC line, which is currently ranked at the top of our database of 165 LED lights. The BC Series being reviewed in this article ranked 2.0 points lower, which is still excellent.
Price (USD): $120 (Yuji Int’l)
|Yuji BC Series Multirow Hybrid|
Product Page (Measured 5/2016)
|Rank||Color (CCT)||CRI (Ra)||CRI (Re)||TLCI||CQS||Rf||Rg|
Method for Testing the Bicolor LED
In a blacked out studio the Yuji LED sample was positioned vertically with black duvetyne placed directly behind it to block any reflections from the C-stand it was mounted on. We mounted an Asensetek Lighting Passport SMART Spectrometer to another C-stand so the spectrometer sensor was exactly 1 meter in front of the LED ribbon light pointing directly at the light. The spectrometer was synced to the Spectrum Genius app to collect and process the data from the meter. We attached two Yuji Flicker-Free Power Supplies (24V) to the bicolor strip since each color — tungsten and daylight — needs to be powered separately.
For the first test the tungsten emitters were set to full power with the daylight emitters off. The daylight emitter were then barely turned on and their brightness gradually increased so that the resulting mixed light increased in approximately 200K increments. Once the daylight emitters were at maximum brightness, the tungsten emitters were gradually turned down so the mixed light continued in 200K increments until the tungsten emitters were completely off.
The second test involved increasing only the tungsten emitters in ten evenly spaced increments on the effective range of the power supply dial. (Note: the knob had to be turned about 33% before any emitters were illuminated. Furthermore, maximum brightness was reached at about 66% on the dial, with no change in light quality seen for the remainder of the range of the knob. This was observed for both power supplies and both the tungsten and daylight emitters.)
The third test was like the second except was only with the daylight emitters.
The fourth test was to increase both the tungsten and daylight emitters at the same rate so that the color temperature remained at 4100K at all brightness levels. This is the effective midpoint of the emitters as seen when both types are at maximum power.
Color Temperature (CCT) vs. Brightness
For both extremes of the bicolor range — 2700K and 6500K — when only the tungsten-diodes or only the daylight emitters are illuminated, the correlated color temperature (CCT) was very stable as the LEDs were dimmed.
When both types of emitter (tungsten and daylight) are combined, as color temperature increases, brightness increases and decreases in a predictable fashion.
Therefore, color measurements are highly accurate and repeatable in a near-linear fashion. This makes sense since the color spectra are additive in both color and brightness. (It is not exactly linear because the black body locus is a curve. So while the bicolor LED is truly linear, when referenced to the curve the result is reported as a slight deviation near the midpoint.) Since the color temperature within the range is stable with dimming, and the spectra are additive, and the midrange increases and decreases predictably, it is safe to assume dimming won’t affect overall color temperature. (Some individual colors do drift a bit with dimming, which we discuss in the section “Specific R-Values.”)
To demonstrate how each LED diode is additive, I’ve overlaid the 2700K and the 6500K spectra with the spectrum at 4100K, which is when both emitters are at their maximum. While this diagram shouldn’t be taken analytically — since the maximum from each spectrum is relative and not quantitative — you can see how the middle of the 2700K and 6500K spectra could add together to create the 4100K spectrum. (Since each spectrum has a relative y-axis that is adjusted to the dominant wavelength, I distorted the vertical axis of the 2700K and 6500K spectra to have a more correct relationship with the 4100K spectrum, which is much brighter than either of these two spectra.)
Adding daylight emitters to tungsten increases the color temperature almost 1400K, from 2730K to 4100K. In contrast, adding tungsten to daylight emitters decreases the color temp 2400K; daylight-only has a color temp of 6478K. Furthermore, the daylight emitters are 17% brighter than tungsten at 1 meter, 709 lux and 612 lux respectively. Therefore the tungsten emitters seem to have more influence over color temp, despite not being as bright. Possible the manufacturer made the daylight emitters brighter to increase their influence.
Conclusion this far: it appears that the tungsten and daylight LEDs in this bicolor LED are behaving very well.
Color Quality vs. Color Temperature
Spectra & TM-30-15
The spectra across the complete bicolor range remains very broad with no major dips. Additionally at the tungsten end — 2700K and 3200K — there are no major peaks other than the red peak which should be there. The mid-range 4100K spectrum as well as those at the daylight end (5600K and 6500K) show a strong blue peak, which ideally wouldn’t be there but is normal for LEDs. The blue peak is because the phosphors in most LEDs are triggered with a blue LED. However this blue peak isn’t extreme and so doesn’t significantly affect the color quality so it is acceptable. So everything is looking good.
As we mention in the “How To Read The Results” section of our article explaining our LED database, the TM-30-15 is based on 99 colors samples. When reading the graphic, take note to see how well the red line traces the reference’s black circle. The TM-30-15 graphics suggest that the tungsten end has higher fidelity of color (Rf) and that it slightly decreases as the bicolor LED moves toward the daylight end of the range. Additionally the middle of the range has a little higher gamut (Rg), meaning the color is a slightly more saturated. There are barely any hue-shifts (shown with green arrows) on the tungsten end and in the middle of the temperature range. However in the 5600K and 6500K graphics you can see some larger hue-shifts in the reds. Nothing earth shattering, but it explains the decrease in color fidelity. But I’m still very impressed with the results.
Both CRI (Ra) and CRI (Re) dip when both emitters are illuminated equally (2.8 and 4.6, respectively) resulting in 4100K light– which is on the warmer side of the CCT midpoint – while both types of emitters have excellent color alone. As mentioned earlier, the dip is due to the black body locus being a curve, so the linear behavior of the bicolor as it is tuned results in some deviation around 4000K. In both CRI (Ra) and CRI (Re) this decrease is from the R1, R2, R5, R6, and R8 values, as is discussed below in the section “Specific R-Values.” This dip is increased in CRI (Re) heavily by the behavior of the saturated red (R9) value, as well as saturated yellow (R10) and the medium skin tone (R15) values, while only slightly from the light skin tone (R13). Obviously saturated blue (R12) also is involved, but in different ways. Therefore, emitters might have been optimized for working alone and not when mixed.
Across the entire color temperature range, TCLI tends to give the highest values of all color quality scales. Despite this both TLCI and CQS drop as color temp increases above 3000K (1.7 and 5, respectively). If these color quality scales are important for your production, such as if you are in television broadcast, you will want to favor the tungsten side of the range.
Color Quality at Set Color Temperatures vs. Brightness
When the bicolor are mixed evenly (4100K), CRI (Ra), CRI (Re), TLCI, and CQS all increase as brightness increases (3.3, 3.8, 1.1, and 2, respectively). Therefore, for mixed light, brighter light might result in higher color quality.
With tungsten emitters, all color scales increase as brightness increases (CRI (Ra): 2.8; CRI (Re) 3.1; TLCI: 2.8; and CQS: 5). Daylight emitters behave similarly, with Ra, Re, and TLCI all increasing with brightness (2.8, 2.9, and 2.1), although CQS is unchanged. Therefore all mixtures of light have better overall color quality as brightness increases.
All of the individual values that make up CRI (Ra) — R1 through R8 — are above 91 for the entire color temperature range. Therefore this scale might not fully reveal all that is going on with the color spectrum.
Within the additional R-values that make up CRI (Re) — R9 through R15 — much more variation is seen: R9 (saturated yellow), R10 (saturated red), R12 (saturated blue), and R15 (medium skin color) all change dramatically over the color temperature range. The R9 value is the most significant as it drops 17 points at the midpoint of the color temperature range, before returning nearly to where it started. This means that if you are filming skin tones or objects containing saturated red, you should stick with either full tungsten or full daylight. R12 also drops 17 points as the color temperature increases, but it never recovers. This means that, ironically, the saturated blue is much better with the tungsten end than the daylight end of the range. The other two big movers – R10 and R15 – both dip and recover about 8 points.
Something to note is that right at 5000K some R-values show a jagged move either up or down. Sharp examples of this are with R8, R11, R12, although a few others show it slightly. This is because the correlated color temperature standard illuminant changes right at 5000K, and so spectrometers switch to a different calculation at this point. It isn’t that the color suddenly gets better or worse at this point, it is just that the reference point changes.
Dimming When Both LED Diodes Are Evenly Mixed (4100K)
Since much of this change happens at the color temperature midpoint, we looked to see how dimming the LED affected the color quality. Saturated red (R9) and saturated yellow (R10) both continue to decrease as the LED is dimmed at 4100K, dropping an additional 9.5 and 10 points, respectively, from full brightness to just above zero brightness. This is also seen with saturated blue (R12) at 4100K, which drops 7.5 points as it is dimmed to zero. Other R-values also drop as the LED is dimmed at 4100K, so if you are working in the middle of the color temperature range you will probably want to work at full brightness and decrease the brightness using other methods, just as diffusion or neutral density (ND) filters.
This decrease in R10, and R12 is also seen in the tungsten and daylight ends of the color temperature range. At the tungsten end, R10 and R12 both decrease 6.5 points upon dimming down to zero, while at the daylight end R10 drops 3 points and R12 drops 5 points.
Dimming At The Tungsten Extreme (2700K)
Dimming At The Daylight Extreme (6500K)
Conclusion on Bicolor LEDs
As I suspected, the tungsten and daylight LEDs produced much higher quality color individually than anything produced when when mixing them. Melding dissimilar lights brings in too many variables to control. That being said, this specific bicolor LED still did a fantastic job at producing high quality light.
Looking at the big picture, the light from the Yuji BC Series Hybrid LED is better than many other LED lights on the market. In fact, take the worst light from this LED — let’s say at 4100K when CRI (Ra), CRI (Re) and the saturated red values (R9) are all at their lowest — and say that is the only light this LED produces. It is still on par with single-color LEDs that are ranked only a couple points below where this LED is ranked as a bicolor. So from this perspective this bicolor LED does very well across its entire color temperature range, it’s just that the front and end of the range are the best. Additionally, as mentioned earlier, at all color temperatures this specific bicolor LED does better at full brightness.
What Does This Say About Bicolor LEDs As A Whole?
This gives me hope. Yes, this light is from one of the best on the market. Lower quality brands still have plenty of opportunity to screw up their bicolors. But these results show that bicolor LEDs can be a very useful tool for those who need to constantly adjust the color temperature of their lights. And since LED technology is still in its relatively infancy, the future is very bright. When I look at all the improvements in phosphors and other LED technology just in the past 12 months, I’m excited to see what shows up at NAB next year. Perhaps many of lingering problems will be fixed!
Also, not all bicolor LEDs will behave as this one did. Some might produce better color quality when dimmed. Or different specific colors might struggle or do better. Remember that these results are only for this bicolor LED.
What Do You Think?
Do you agree? Perhaps you have a very different perspective. Let us know your thoughts by leaving a comment in the comment section below!