The brightness you see versus the brightness you read about: on light levels, LEDs, and the shortcomings of the lumen

You are replacing a 60-watt incandescent light bulb with an LED - how do you know what LED will produce equivalent light? Most likely, you select a product for which the manufacturer has printed "60-watt-equivalent" on the box. But how is "equivalent" determined? The industry uses a measure called the "lumen" to measure light volume across light sources. Based on our installations, we believe the better quality the light, the "brighter" a lumen appears. Here we'll explore why we think the lumen fails to indicate brightness properly where LEDs are concerned (full spectrum lighting appears brighter to the human eye) and what we do about it (standardize our own lumen requirements based on the light quality of the source being replaced).  

How bright is a lumen?

Lumens are designed to be the apples-to-apples of the light world. If you’re replacing one light source with another, you need a medium of exchange, and lumens exist to serve as that medium. As examples, a 40-watt A-bulb emits 450 lumens, a 32-watt 4-foot long fluorescent tube emits 2,600 lumens, an industrial high-bay light drawing 1000 watts emits upwards of 50,000 lumens.

In practice, however, the lumen let's us down in translation: while you want 350-400 lumens to replace that 450-lumen A-bulb, 1,500-1,700 good LED lumens will more-than-adequately replace the 2,600 coming out of the fluorescent tube. Why the vast discrepancy? Let's take a step back, and look at how the lumen's measured.

How is light converted to lumens?

A given beam of light (coming out of your computer, from the light bulb over your head, from the sun) is composed of photons oscillating at a blend of wavelengths within the visible light portion of the electromagnetic spectrum. Think of white light as a mix of the rainbow’s colors, emitted at various strengths. The makeup of wavelengths in a beam is unique by light source.

Sunlight, for example, has a "spectral distribution" that looks like this:

This is a view of natural light's spectral distribution within the visible portion of light. As you might guess, sunlight also emits "light" that is ultraviolet (UV, will give you sunburn) and infrared (also known as heat). Light bulbs all emit some volume of infrared/heat energy, and some lights (fluorescents, metal halide) also emit significant UV light.

This is a view of natural light's spectral distribution within the visible portion of light. As you might guess, sunlight also emits "light" that is ultraviolet (UV, will give you sunburn) and infrared (also known as heat). Light bulbs all emit some volume of infrared/heat energy, and some lights (fluorescents, metal halide) also emit significant UV light.

The human eye, meanwhile, is more sensitive to some colors and less to others. Our eyes find it easier to see, for example, amber and green than to see deep red or blue. Our range of color sensitivities can be shown graphically in what’s called the luminosity curve. 

The human eye's sensitivity to colored light during daylight hours (aka photopic vision).

The human eye's sensitivity to colored light during daylight hours (aka photopic vision).

Unsurprisingly, the color-sensitivity of the human eye parallels the wavelengths emitted by sunlight.

In order to convert a light source to lumens then, the industry multiplies each wavelength reading coming out of the light source by the human eye’s sensitivity to those wavelengths. This, the theory goes, allows us to answer the question “how bright will this particular beam seem to the human eye” no matter what the color composition of the light. Apples to apples.

Where the lumen begins to struggle

Let's have a look at some spectral power distributions for common artificial light sources (from right to left: incandescent, fluorescent, metal halide):

Our experience, replacing countless incandescent, fluorescent, and metal halide bulbs, has revealed that the further the spectral distribution is from natural, the larger the lumen discount you need to apply in order to get light levels correct. Incandescent light, redder than sunlight but escalating consistently through the spectrum, only requires a slight lumen discount. Fluorescent and MH lighting, however, which emit light by running electrical current through gas and then filtering UV light through phosphors, ends up requiring significant lumen drop. 

An LED spectral distrbution, however, can combine the blues associated with natural light with consistency across the visible spectrum.

 

A spectral power distribution for a good, single-chip LED. We tend not to specify red/green/blue multi-chip LEDs because they have a lumpy spectral distribution. 

A spectral power distribution for a good, single-chip LED. We tend not to specify red/green/blue multi-chip LEDs because they have a lumpy spectral distribution. 

We implement LEDs according to observed and expected brightness. The lighting industry does not provide data through which specifiers should make adjustments, but we constantly monitor brightness needs as LED spectrums improve and reduce lumens as necessary given the lighting application. For incandescent applications, we find 85-90% of original lumens to be the correct adjustment, in T8 fluorescents the discount ranges from 30-55% depending on ballast factor and what phosphors are used in the original tube, and in industrial high-intensity-discharge applications we are discovering discounts of 50-75% are necessary.

Why (we believe) a good lumen looks brighter than a bad one.

Theoretically, the lumen accounts for the eye's sensitivities, so a bad lumen should look bad but equally bright. What's causing the sense of dimness with bad lighting?

The parts of the eye that sense colored light, cones, are grouped by wavelength to which they are attuned. There are three types of cone, and they generally fit into three groups with sensitivities focused on the red, green, or blue parts of the visible spectrum (where do you think we get our "primary" colors?). When we see "reddish orange" per the below chart, it means our brain is receiving a strong signal from the red cones, a moderate signal from the green ones, and none from blue. Each color is seen per a specific averaging of these cone outputs.

15 million cones per eye, grouped by their three categories and showing to what colors they are most sensitive.

15 million cones per eye, grouped by their three categories and showing to what colors they are most sensitive.

However, cones fatigue. When a specific wavelength, or color, of light is emitted in high intensity, the relevant cones become fatigued and less sensitive to that wavelength. For a do-it-yourself example thereof, here’s an “after-image test.” If you look a bright image for 20 seconds, the cones in your eyes that see the colors shown (red and bright green in the below example) get tired. When you look at a white space thereafter, your eye will not "see" those shades of red or green until the cones have rested - this means that when you look at a white space, the fatigued parts of your eye will be missing red and green, making those areas look blue and pink, respectively.

An after-image test. If you want to give it a try, look at a point near the center of the image for 20 seconds, then scroll down to the white space at the end of the post. You might see a blue box with pink birds - the after-image of the red and green.

An after-image test. If you want to give it a try, look at a point near the center of the image for 20 seconds, then scroll down to the white space at the end of the post. You might see a blue box with pink birds - the after-image of the red and green.

When a light like fluorescent or metal halide lighting generates white light by averaging three or four wavelengths across the spectrum, everything in the room becomes an after-image. Your eyes quickly (20 seconds) fatigue to the predominant wavelengths, and the light seems dimmer as a result.

A second factor contributing to what we believe to be the bad light lumen discount is the effect of the eye's rods. Rods perceive light and dark (no color) and are used primarily at night (which is why you don't see colors at night). The luminosity curve, shown at top and used for lumen measurement, is cone-only. Rods, though less active during the day, are particularly sensitive to blue light.

The blue line here shows rod brightness sensitivity by wavelength. Rods do not see color, but like cones (the green, red, and dotted black lines) have wavelength preferences.

The blue line here shows rod brightness sensitivity by wavelength. Rods do not see color, but like cones (the green, red, and dotted black lines) have wavelength preferences.

Though they do not see color, rods control the dilation of the pupil even during the day. Thus, a very strong blue spectrum spike like that often found in fluorescent and HID lighting tricks the eye's rods into thinking that the ambient light is extremely bright. They in turn close the pupil aperture, causing blue-heavy lighting to seem dimmer. 

We conclude.

The lighting industry has grown up around the lumen, and related measures like the footcandle and candela, in an effort to standardize light output and regulate the minimum brightness required for a workspace, hospital, school, and so on. It is not our experience, however, that the lumen translates across light sources. We believe, based on available literature and common sense, that cone fatigue and the pupil-closing reflexes of the rods contribute to the sense that a bad lumen is not as bright as a good one.

The LEDs we work with tend toward the spectral distribution of natural light, allowing us (or forcing us, depending on perspective) to reduce lumen count for a given interior while maintaining appropriate brightness. As our clients look to LED-light ever more diverse spaces, we will continue to hone our understanding of how we perceive good light, and how best to move forward from the light buildings currently use.

Further reading on the topic includes some interesting notes on the relevance of rods AND cones in this discussion from Philips, good solid explanation of the eye, and a pretty thorough review of color science and LEDs from the perspective of a chip engineer. Meanwhile, a GE engineer has a look at "spectrally enhanced" light measurement programs ("GE has no expressed opinion on the validity" of these programs). Finally, there's a whole pile of research on lighting measurements that incorporate rods and cones (one summary here), in an effort to properly equalize lumens - but it's mostly focused on low-lighting applications like outdoor areas, streets by night, and so on.