Emitter Measures

Something new that I’ve decided to add to my reviews are direct measures of colour temperature, colour tint, and colour rendition.

I will explain these terms down below, but this is something that just wasn’t feasible a decade ago when I was first reviewing lights. As you can see on the other Methodology pages here, I’ve always looked for simple, reliable, low-tech solutions for all my testing needs. Back in the day, there was no such thing as an affordable spectrophotometer or spectrometer, as those were intended for heavy-duty industrial applications (where calibration and absolute accuracy are critical). They therefore typically cost several thousands dollars. Not as bad as the tens of thousands of dollars for calibrated integrating spheres, but still out of my league.

Upon my return to flashlight testing, I’ve noticed a lot of reviewers are providing these colour numbers and charts. Some are using monitor calibration tools (a cool, relatively lower-budget idea), but the most popular devices by far are the bluetooth-based Opple Light Masters that link up to an app on your phone. You can pick up the basic model II for under $25 CDN (for simple colour temperature measures). For The models III/IV offer a lot more, including CRI, Duv, as well as circuit flicker/PWM for ~$50/55 CDN. Even if absolute calibration accuracy is not high, that’s a pretty impressive package for the price – and still perfectly fine for relative comparison results here.

However, the first thing you need to keep in mind is that any results posted here are simply from a single given sample (i.e., n=1). It is very likely that other flashlight specimens from the same manufacturer would behave differently, given the natural variation in emitters (especially for those who don’t specify temperature or tint bins). But it is still useful to know how close a given specimen is to its rated CCT and CRI specs, and how the tint in particular might shift over different drive levels (a common phenomenon for current-controlled circuits). Plus, there are lot more different makes and models of emitters out there now, and I thought it would be interesting to see how their CRIs compared. But as always, caveat emptor – please don’t try to over-generalize from a single given sample to all lights from that maker.

What is “White” Light?

To start, let’s ask a deeply philosophical set of physical questions: what is colour, and what is “white” light?

To begin to answer that, physical scientists tend to start with an explanation of black body radiation. This is the thermal electromagnetic radiation emitted by a idealized black body (i.e., an opaque, non-reflecting body). It has a very specific continuous spectrum of wavelengths that are inversely related to intensity, and that depend only on the body’s temperature (which, again, is idealized as uniform and constant for the sake of calculation – this is known as Planck’s law). Although our sun (or any star) is not a perfect black body, this principle is still a good general approximation for the energy it emits. The sun’s radiation, after being filtered by the earth’s atmosphere, thus characterizes “daylight”, which we evolved to use for vision.

As the temperature of a black body decreases, the emitted thermal radiation decreases in intensity and its maximum moves to longer wavelengths. In other words, the colour shifts with black body temperature (which is why we measure colour in temperature degrees, as I’ll describe in a moment).

Another key colour concept is chromaticity. Chromaticity is an objective specification of the quality of a colour regardless of its luminance (aka intensity). Chromaticity consists of two independent parameters, often specified as hue and saturation (aka colorfulness). These can be plotted on a two-dimensional x,y coordinate system, as shown below for the classic CIE 1931 (an old standard from the French commission internationale de l’éclairage).

Again, the colour (chromaticity) of black body radiation scales inversely with the temperature of the black body. The line of all such colours is shown above in CIE 1931 space as that main black curvilinear line (the black body curve is also known as the Planckian locus).

In essence, the “whiteness” of a light source is judged by reference to these idealized black body radiators. A black body can thus be described by its temperature, and produces light of a particular hue, as depicted above.

Correlated Colour Temperature (CCT)

For light sources that approach a black body (so-called Planckian light sources), we typically refer to their correlated colour temperature (CCT). This is the temperature of a Planckian radiator whose colour best approximates it (i.e., correlates to it).

Correlated colour temperature (measured in degrees Kelvin) is the first key characteristic of a white LED that we tend to concern ourselves with – and which most makers will specifically identify in their lights. A warm white light (e.g. an incandescent bulb) is typically ~2700-3000K or so, a neutral white light is typically ~4000-5000K or so, and a cool white light is typically ~5500-6500K or so (aka “daylight”). 6500K is pretty much the standard cool white emitter you will find on most lights who don’t specify a colour temperature.

Tint (Duv)

Tint is measured by a change (delta) in the hue-saturation chromaticity coordinate system referred to above. Since we have taken luminance out of the equation, we have a two-dimensional hue-saturation space referred to as x,y in the CIE 1931, but now typically referred to as u,v in the CIE 1960 Uniform Colour Space (UCS), as shown below.

Delta u,v (Duv) refers to the point in this chromaticity coordinate system measured in relative distance from the standard black body curve (i.e., perpendicular to the Planckian locus). Basically, it gives you a measure of the relative hue and saturation as you move away from the black body Planckian locus. A negative value indicates that the colour point is below the black body locus (i.e., magenta or pink) and a positive value indicates a point above the black body locus (i.e., cyan, green or yellow).

Obviously, you want to be as close as possible to the black body Planckian locus, to avoid distracting colour tints. You can expect to pay more for this privilege, through the use of premium tint bins.

In an ideal world, manufacturers would love to make only the agreed-upon “whitest” version of each of these colour temperatures (i.e., closest to the black body Planckian locus). But that isn’t possible in any manufacturing process, where there is bound to be some variability. Instead, manufacturers can “bin” the production output of their LEDs across defined areas of the chromaticity spectrum (just like they can bin for output, and for forward voltage, etc.). Here is an example of an old chromaticity chart for Cree emitters, with the black body Planckian locus (“BBL”) shown as a dotted line through the centre of the bins (with each bin identified by a 2- or 3-digit code):

Some makers – especially the more bespoke or custom makers – will buy specific defined “premium white” tint bins that closely straddle that locus (and advertise as such in their products). But for the most part, it is hard to predict what you will get in exact tint for any give temperature class, as it varies depending on the production run unless you purchase a specific defined tint bin.

Colour Rendering Index (CRI, or Ra)

A colour rendering index (CRI) is a quantitative measure of the ability of a light source to reveal the colours of various objects faithfully – in comparison with a natural or reference light source. Basically, it is a measure of how well the colour of objects appear to the human eye for a given light source (or put another way, how well subtle variations in colour shades are revealed under that source). Our friends in the CIE defined CRI along a scale of 0 to 100, with 100 being given to a source that is identical to the spectrum of daylight (i.e., very close to that of a black body). The lower the number, the less well you can differentiate colours.

The way this is done is through a set of 8 or 15 predefined standardized colours, collectively known as test colour samples (TCS). Each TCS is scored individually, and given a value known as a Rendering (R) score (on a scale of 0 to 100), numbered for each TCS. So for example, referring to chart below, R1 (for TCS1, aka “light grayish red”) would be measured and given a certain score, along with all the the Rs, from R2 through R8. These would then all be averaged, producing the final score up 100, known as Ra (or simply CRI).

At least, that’s the case in North America. In other parts of the world, an extended set of TCS up to R15 is used (this is sometimes referred to as Re, or Extended CRI, to differentiate it from the general Ra/CRI). The reason for this is that strongly saturated colours around the periphery of the core general set (like “strong red” R9, “strong yellow” R10, etc.) are missing from the general CRI measures. R9 in particular is very significant for LED flashlights, as modern LEDs are relatively deficient in the red part of the visible spectrum. So at a minimum, LED flashlight enthusiasts tend to like to know Ra + R9, if possible.

Again though, the CRI of a light does not indicate the apparent colour of the light source – that is given by the CCT and the tint (Duv). Rather, it tells you how good that light source is in helping you differentiating different colours, either on average (for Ra/CRI), or for a specific colour (e.g. R9 for red).

Taken together, these three measures (CCT, Duv, and CRI) give you the key quantitative estimates you need to know how a light compares to an idealized light source – that is, what we all know as “daylight”.

My testing setup

For my preliminary testing, I purchased an Opple Light Master II (LM2), for fairly basic measures (CCT and computed Duv). This is not a calibrated spectrophotometer – but based on my testing, I’m reasonably confident in using it for relative comparison emitter testing.

I’ve since purchased the new Light Master IV (LM4), but there are issues with the accuracy and reliability of this initial batch (especially on high CCT and low CRI lights). However, it does appear to be reliable for CRI measures across the range of CCTs. I’m waiting to see if they can resolve some of the issues through software updates, but it may require a replacement device once they have it sorted. In the meantime, I will stick with my LM2 for more consistent and reliable CCT and Duv measures, but will report CRI measures from my LM4.

Let me make one point clear right up front – there is no perfect way to measure light on an uncalibrated device. The best I can do is come up with a consistent way of measuring flashlights that appears to be valid (i.e., matches the relative experience of my very personal eyeball/brain combo).

My LM2 unit reports values that match what I personally perceive, in relative terms. For example, for cool white emitters, CCTs typically gets warmer and Duvs more positive when driven to lower outputs (with current-controlled drivers). For neutral-warm tint high CRI emitters, some produce a more rose-tinted (i.e., negative Duv) relative to others. The relative magnitude of this effect to my eye matches what my LM2 unit is giving me. Of course, that doesn’t mean its values are accurate in an absolute sense – only that they match what I perceive in a relative sense (e.g., more or less rose-tint, etc.).

Other reviewers are likely to go about this differently, but here is how I’m setting up my tests for summary measures:

  • All measurements are taken in complete darkness, with the LM2/LM4 on the floor facing upwards.
  • Flashlights are mounted (or held) at a fixed point somewhere between 2-4 feet above the sensor, pointed straight down. This is to ensure only the absolute center (hotspot) of the beam is measured.
  • Reported measures are taken at the highest fully-regulated output level of the light (i.e., if a light has a step-down from a heavily-driven Turbo, I will take the first Hi level where it is able to maintain flat regulation for an extended portion of the run).
  • I’m trying to keep lux measures within a band of ~10K-25K for the most part, to ensure consistent operation of the LM2/LM4 (I do this by varying the height of the flashlight). Although the meter is meant to be used with lower intensity light sources, I find this to be the most reasonable range to work with (given the brightness of modern lights).
  • I’m using the LM2 for CCT and Duv results (as they match my visual experience and are internally consistent), and the LM4 for CRI (Ra) measures.

Note that tint in particular can fluctuate across a beam, so ideally you would want to use an integrating sphere to get the best overall average measure. But I don’t have an IS, and this hotspot measuring method is simpler.  Note as well that both CCT and tint can fluctuate across output/drive levels, which is why I’m focusing on the highest fully-regulated level for my summary results. I do plan to do additional measures across output levels, and will report on anything interesting that I find in the individual reviews.