Ensuring safety in LED lighting

Electronics Weekly’s LED colour rendering guide
Electronics Weekly’s history of the LED
Electronics Weekly’s guide to LED optics
Electronics Weekly’s guide to cost-sensitive design

Lighting LEDs are a route to energy efficiency that also allow creative effects that were not possible with incandescent bulbs and fluorescent tubes.

But the output spectrum, like that of fluorescent tubes, is nothing like the sunlight human eyes evolved to deal with.

Is this important? What do you need to know about the health implications of artificial light?

The short answer is: There are safe levels of every kind of light and dangerous levels of every kind of light. Knowledge and standards help manufacturers and lighting designers to navigate the vast area in between.

A little bit of biology

Eyes have to cope with a huge variation in light levels and spectra.
Two types of adaptive visual sensor in the retina, plus a variable diameter pupil, and eyelids that can squint, allow vision over at least nine decades of intensity and perception of colours from purple (400nm) to red (700nm), with some sensitivity out to 380 and 750nm.

In the dimmest light, ‘rod’ sensors can detect moving objects or images.

There is only one type or rod, so no colour vision is available in dim light.
Left in the dark, rods and connecting nerves become more sensitive, taking around 30 minutes to reach maximum, although much of the gain occurs in the first few minutes. Conversely, all additional sensitivity can be lost in seconds – so night vision which took minutes to gain is ruined if someone flashes a torch in your eyes.

Peak rod sensitivity is at around 498nm, extending between 400 and 600nm. By 640nm there is no response, which is why instrument panels on some aircraft and cars have deep red backlighting – red ‘cones’ (see below) allow instruments to be seen clearly, at a wavelength that cannot ruin night vision for forward viewing. Blue or green backlighting misses this important point.

As light gets brighter, rods saturate and a second kind of retinal sensor called cones take over.

Cones are 100x less sensitive than rods, and colour vision is possible because there are three types of cone.
Called S, M and L, they are sensitive to blue (peak at 419-420nm), green (531-534nm) and yellow (558-564nm) respectively through three different ‘photopsin’ chemicals.
L (long wavelength) cone sensitivity extends to wavelengths longer than 700nm, which is why they are frequently called ‘red’ cones.

Variable mixtures of S, M and L cone response allow many colours to be distinguished in bright light.

Cones are concentrated at the centre of the back of the eye, with rods around the outside, which is why astronomers seeking dim stars have to through a telescope ‘out of the corner’ of their eye.

Incoming light is controlled over a 12:1 range by the iris which varies pupil diameter from 7-2mm diameter automatically according to signals from both rods and cones.

Cells that sense, but don’t see

Discovered only a few years ago, there is a third type of light sensor in the eye which does not form images. Called ‘intrinsically photosensitive retinal ganglion cells’, these also have control over the pupil, and appear also to be the cells responsible for synchronising the brain’s natural circadian rhythms with daylight.

Photo-ganglion cells are the subject of intensive research. Sensing is through a substance called melanopsin, which reacts across a wide range of colours peaking at 480nm (around the dip between blue and green cone sensitivity), and may also have other sensitive peaks – the jury is still out on this difficult-to-study subject.

Importantly for eye protection, ganglion cells can keep the pupil small, while rods and cones can only shrink the pupil briefly, from which state it grows large again over about 10 seconds (see 480nm below).

“The inner retinal photoreceptive system [photo-ganglions] generates sustained and enhanced responses under continuous high levels of illumination, thus conferring continual pupillary constriction under bright light conditions,” said Missouri researchers Yanli Zhu and Daniel Tu et al in their 2007 paper, ‘Melanopsin-dependent persistence and photopotentiation of murine pupillary light responses’.

In addition to short-term adaptations of seconds and minutes, there are longer-term adaptations.

For example, the retina of young children is vulnerable to UVA from sunlight, while adults are protected by filtering chemicals that accumulate in the eye-front over years. Other multi-week protective adaptations have been reported – it has been suggested that people dwelling at high latitudes take extra care on short sunny holidays.

The adult eye, combined with physiological reactions like eyelid closure and flinching away, is able to cope with sunlight at its home latitude – with a few exceptions, see below.

If damage is to occur, light has to be bright, and its spectrum becomes important.

Shorter wavelength light has more energy than longer wavelengths – blue photons can do more damage than red photons.

Ultra-violet photons would be the most damaging. However, lighting LEDs do not produce UV.

If the eye is exposed to ultra-violet, from sunlight for example, it is absorbed at the front of the eye. At high levels this will cause ‘arc eye’ in the cornea – painful but normally reversible UV damage. Long-term UV absorption in the lens from sunlight appears to contribute to cataract formation.

Blue light hazard

Without UV, blue is le ft as the most destructive part of the visible spectrum, causing damage to the retina through photo-chemical action rather than heating.

And occasionally damaging UV reaches the retina as well – this happens in young children and those whose eye lens has been removed for medical reasons, for example.

Our faces owe something to the power of blue light.

“We have evolved to be protected from sun exposure when sun is high in the sky – it is why we have eye brows and brow ridges. When the sun is low, it tends to be redder,” Dr John O’Hagan, head of the UK Health Protection Agency’s laser and optical radiation dosimetry group told Electronics Weekly.

The science of blue light damage at high intensity is well studied, and it is known to be cumulative over durations up to hours.

The International Commission on Non-Ionizing Radiation Protection (ICNIRP) is a body which gathers scientific evidence as it is revealed and, through expert review, produces best-practice guidelines. As such, its guidelines are well respected and are the basis for global standards in light safety.

ICNIRP suggests the “use of hats, eye protectors, clothing, and sun-shading structures” to protect the eyes and skin from the harmful effects of sunlight UV.

Long-term eye heath

There is a possibility that levels of blue light exposure which do not cause damage in hours, may cause cumulative damage over years, and a link has been proposed to age-related macular degeneration (AMD).

This said, there is yet no strong evidence either way on blue light and AMD.

“Recent studies suggest that the blue end of the light spectrum may also contribute to retinal damage and possibly lead to AMD,” said the American Macular Degeneration Foundation. “The retina can be harmed by high-energy visible radiation of blue/violet light that penetrates the macular pigment found in the eye. According to a study by The Schepens Eye Institute, a low density of macular pigment may represent a risk factor for AMD by permitting greater blue light damage.”

What does O’Hagan think?

“I would be very surprised if blue LEDs increase AMD. Blue light exposure from the sun is far more than this,” he said. “The science isn’t there yet. From the physics, it is high-energy blue or UV that could cause the damage. It would have to be exposure to sunlight in childhood when these can still reach the retina. This suggests a long latency period before AMD is identified”

Another thing to think about with accumulated exposure is that threshold effects may also come into play, which would mean only blue light above a certain level is accumulated. Again, potential effects are small and this is at the boundaries of known science.

ICNIRP publishes a blue hazard curve (fig 3 here) which quantifies the risk wavelengths, with a background assumption that exposure will be for a working shift once every working day. itemid-55380-getasset.jpg

It shows that, if it is strong enough, light across and beyond 400-500nm causes eye damage, with eye lens removal extending this range to 310nm.

The eye is particularly vulnerable between 420 and 470nm, which is exactly the range the InGaN die of ‘white’ lighting LEDs emit: blue indigo and violet.

This is true of lighting LEDs from all manufacturers, and it is also true of many fluorescent lights and HID (high-intensity discharge) lamps.

Emission spectra from different fluorescent tubes vary widely. Fluorescents starting from the 1950s have a tall narrow blue spike somewhere between 430 and 450nm.

Investigated at least since the 1980s, these spikes have not been linked to eye damage.

Incandescent bulbs have similar continuous spectrum to sunlight, with less blue, so they are inherently safe so long as they are not too bright.

So where are the limits? How strong does this light have to be?

Between staring at the sun and viewing the TV – no one is yet suggesting looking at a 300cd/m2 TV screen will cause retina damage – there are boundaries beyond which exposure is harmful.

These limits are a function of intensity, spectrum and time – and scientific knowledge.

Damage guaranteed

Beyond certain intensities, the eye will suffer permanently.

Look directly at the sun (1.6×109cd/m2 intensity), and irreversible damage begins in a fraction of a second.

Electric arc welding without eye protection can do the same thing. As little as half a second of blue light from welding, accumulated over a day, can damage the eye. “Arcs in CO2 arc welding of mild steel should not be viewed for 0.47-4.36s in total per day.” said the authors of this paper.

On the other hand, despite its blue-dominated spectrum, the eye appears to cope with the 5,000cd/m2 intensity of a clear blue sky – although it remains to be seen if staring at the blue sky for years accumulates low-level damage, or indeed if living on a sunlit planet eventually accumulates damage.

Fluorescent tubes can hit 10,000cd/m2, and are probably the upper limit of safe continuous viewing, while the filament of a light bulb can be 10 million cd/m2.

These figures come from a detailed review of LED lighting safety lead by Professor Francine Behar, and ophthalmologist of the Universite Paris Descartes, called ’Light-emitting diodes (LED) for domestic lighting: Any risks for the eye?’. Also available here and here.

This report points out that on a bright sunny day a white south-facing wall can exceed 50,000cd/m2, requiring appropriate sunglasses to bring this to a safe level. The same is true of snow or sand in similar conditions.

Studies by the research team put the die of a 212 lm LED at an average luminance of 6.2×10^7 cd/m2 – so no illumination-class LED should be viewed directly at close range.

Good LED makers do it right

Responsible lighting LED makers issue information on the safe use of their devices.

For example: LED maker Cree has safety tested its LEDs, including the type that has most potential to do damage the retina – blue power LEDs.

According to Cree’s LED eye safety document: “The results of this testing show significant health risks from some of Cree’s visible light LED lamps when viewed without diffusers or secondary optical devices. These risks warrant an advisory notice to indicate the potential for eye injury caused by prolonged viewing of blue light from these devices. To date, the testing shows that Cree’s blue and royal blue LEDs (450-485 nm dominant wavelengths) pose a higher potential eye safety hazard than its white LEDs. Other colours of LED lamps, such as green and red LED lamps, do not pose as significant of an eye safety risk.”

As a rule of thumb for people in general, the ICNIRP suggests that the spectrum of a source that appears to be white is unlikely to be important for safety calculations as long as its intensity is below 10,000cd/m2.

Always safe

Through years of research, international standards have been written that describe the boundary between safe and unsafe lights.

IEC 62471 ”Photobiological safety of lamps and lamp systems’, currently includes the international best estimate of safe limits for LED lighting. This is the parent of Europe’s EN62471, which is the master specification for national standards including BS EN 62471:2008 BS EN 62471:2008 Photobiological safety of lamps and lamp systems

And there is help available to those wishing to apply it.

“IEC/TR 62778:2012 brings clarification and guidance concerning the assessment of blue light hazard of all lighting products which have the main emission in the visible spectrum,” said LED maker Vishay. “By optical and spectral calculations, it is shown what the photobiological safety measurements as described in IEC 62471 tell us about the product and, if this product is intended to be a component in a higher level lighting product, how this information can be transferred from the component product [LED or light engine] to the higher level lighting product [luminaire].”

At one time, LED products were included in the laser safety standard IEC 60825. This is no longer the case.

IEC 62471 originated from the US ANSI/IESNA RP-27 specification, which the Commission Internationale de l´Eclairage (CIE) adopted as S 009.

This document from European luminaire makers lists on-going standardisation activity.

Directly or indirectly, most of these look to the ICNIRP.

According to O’Hagan, the British Standard is due for maintenance. “The intention is that the BS EN will adopt the new ICNIRP limits, the main changes are retinal thermal, not blue light limits,” he said.

Risk groups

Through testing to the standards, LEDs, luminaires and other light sources are divided into four groups: exempt (completely safe), and risk groups (RGs) 1, 2 and 3. All are tested with 20cm between unprotected eye and light source, which is assumed to be the worst case as closer observation defocuses the image on the retina and lowers power density – except in short-sighted people who are not wearing their glasses as they can focus closer and are at maximum risk somewhere below 20cm.

Exempt means that no matter what you do, you can’t hurt yourself. Indicator LEDs for example, said O’Hagan.

RG1 is similar to exempt because, through normal everyday behaviour, no one would stare at the light because it would be uncomfortable. “The assumption is you would look for 10sec maximum,” said O’Hagan.
RG1 products can be put almost anywhere, except for example in operating theatres where anesthetised patients might have uncontrollably exposed and dilated pupils.

RG2 is where protection is from aversion response. It is assumed people will blink or flinch away in 0.25s.
“Children will try to overcome their aversion response. It is not a good idea to sell to these to consumers. You might sell RG2 security lighting that you would not put at ground-level,” said O’Hagan. “If you have very cool white LEDs with a big blue peak in, you start to shift to Risk Group 2. LED makers are managing this by decreasing the blue peak and using a slightly different phosphor.” itemid-55381-getasset.jpgTypical white lighting LED spectra for cool (> 5,000K, in blue), neutral (4,100K, green), and warm-white (3,100K, red). A blue power LED will only have the left hand peak.

RG3 light have to be handled with care by trained people. “High-bay lighting is probably RG3, but where they are mounted you cannot get exposed,” said O’Hagan.

Single die lighting LEDs are generally Exempt, RG1 or a few are RG2 – this data can be downloaded as part of a safety declaration from responsible LED makers.

Although RG1 LEDs are nominally safe, advice from many bodies including the French government is that the LED die within a luminaire should not be directly visible from the outside of the luminaire if it can be viewed closely.

Don’t play with…

And some common household objects need care, said O’Hagan: “Even some incandescent torches can cause damage if they are stuck in front of the eye.
Some LED torches could damage the eye rather more quickly.”

As an aside, LEDs in toys are currently restricted by another IEC standard that treats them as lasers.

Safety declarations can also specify a maximum viewing time.

This is the maximum time the eye should be exposed to the source at 20cm range in a working shift.

Viewing is not expected at this distance. The number allows a permissible duration to be calculated for the actual distance, taking into account additional attenuation or focussing.

“An LED might be viewable for 2,000s at 20cm. You can take it from there with inverse square for point source, or 1/d for arrays,” said O’Hagan. “The science isn’t mature on blue light long-term. ICNIRP says repeated exposure for eight hours a day is not advised.”


All lighting LEDs have a dip in their spectrum at 480nm (see above) – the wavelength at which photo-ganglion cells are most sensitive.

Peak LED emission is usually at 460nm or shorter wavelengths, where the ganglion cells are less sensitive. Published by PCM, a paper by Zaidi, Hull, Peirson et al suggest that light at 460nm elicits only half the pupil response of light at 480nm.

So in intense light conditions, the pupil will be more dilated in LED light than it will be in apparently similar sunlight – the retina will receive more high-energy blue. The effect is greater with cool white LEDs compared with warm white LEDs,  and worst in pure blue LEDs.

Whether this is important for long-term eye heath at normal in-door intensities has yet to be determined.


It has long been known that the body’s daily ‘circadian’ rhythm, and its level of the sleep hormone melatonin, are affected by light.

Recent research suggests that photo-ganglions more responsible for this physiological link with light than rods or cones are.

The circadian rhythm is a biological frequency-locked oscillator with a natural period of just over 24 hours, which gets synchronised by being reset once a day by changes in light. It affects a number of health-related aspects of the body.

“We evolved to wake with relatively red spectrum and somewhere mid-morning blue triggers the circadian rhythm,” said O’Hagan.

While retina damage needs intense exposure, sleep disruption can occur at normal household light intensities.

In people kept in the dark for a number of hours, 100 lux is enough have a strong effect on melatonin levels, and a lesser effect if the people are not dark adapted.

This result came from research by Jeanne Duffy and Charles Czeisler of Brigham & Women’s Hospital in Massachusetts, also published by PCM.

Night lights work

Overall, it suggest melatonin levels in dark-adapted people are not affected by light below 10 lux, and maximum effect is around 1,000 lux or above. Half sensitivity is close to 200 lux.

Duffy and Czeisler also investigated spectrum, using light at 460nm and 555nm, and report significantly more melatonin and circadian sensitivity through ganglion action than through rods and cones.

Standards groups are well aware of the safety and rhythm questions raised by 480nm sensitivity.

“The CIE is working on guidance for the appropriate light spectrum to trigger the photo-ganglions, and [European standards body] CEN has a working group,” said O’Hagan. “Something like 15 research groups around the world are working on this: the University of Surrey, Manchester University, and the department of ophthalmology at Oxford University are looking at it.”

Whatever detail 480nm research projects add, O’Hagan has some simple advice for people seeking to minimise sleep disruption from all forms of artificial light. “Don’t have a bright light in your face in the evening. Light should drift down to dark with the sun. In the middle of the night, use red light,” he said.

Artificial light in this case includes incandescent sources: At 500nm, halogen light is still at 50% of its spectral peak (580nm) and even standard light bulbs are around 25% of their 650nm peak.

“Halogen lamps, as well as LED types, are blue-rich and depress melatonin production,” O’Hagan said.

Even if they never replace every light bulb, LEDs are almost certainly going to be a feature of every home, office, factory, and street.

At normal in-door intensities any heath effects they might have are similar or the same as fluorescent tubes or light bulbs.

What to do?

If you are designing luminaires, let standards related to IEC 62471, and therefore ICNIRP guidelines, be your guide.
Avoid designs where consumers, and particularly children, can stare straight at the LEDs themselves from close up.

In manufacturing where illuminated die are exposed or, very bright sources are being made (for example, high bay or street lamps), provide blocking screens, filters, dark glasses, filtered glasses, diminished test current. Similar precautions are applicable to installers.

If you are putting a Risk Group 2 light source into the hands of the public, think about the warnings you will include with it.

In the home and workplace, avoid high intensity cool-while LEDs, particularly when surface brightness approaches 10,000 cd/m2.

As a user, don’t stare at high-powered light sources of any kind, or locate them where children can stare up close – high power torches, LED or otherwise, are not toys, so keep them away from unsupervised children.

Electronics Weekly intends to maintain this article.
If you understand the subject, please suggest amendments or additions to tech@electronicsweekly.com

Tags: eye, led, lighting, safety

Related Tech News


  1. Rob Martin
    April 09, 2015 01:07

    This article was very informative. Thank you.

  2. steve bush
    April 08, 2015 10:39

    Hi folks.
    Thanks to Carlos for answering PD’s question.

    the article is about lighting LEDs, which do not produce UV. Even Sorra’s violet-pumped GaN-on-GaN (which have yet to penetrate the market significantly (in April 2015)) only produce minute amounts of UV.
    And the sentence is clear: “However, lighting LEDs do not produce UV”.
    So I think “bunch of BS” is incorrect on your part, unless I am misunderstanding the ‘BS’ bit.
    If you are aware of lighting LEDs that do produce UV, please get in touch.

  3. April 07, 2015 05:11

    There is a sentence in this article stating that LEDs do not produce UV. That is a bunch of BS. Is that why one can buy a UV LED?

    Come on, who has ever told you this and you believe this? UV is a wavelength, not a heat. It is very damaging and they make LEDs that create the UV wavelength specially for medical equipment, as I have purchased these leds and used for a machine I built myself to market for killing fungi and skin/air borne viruses.

    It is bad to tell the public there isn’t anything to worry about when these wavelengths do exist, however, they do not exist unless you particularly purchase the correct colors as the led light wavelengths are very small bandwidths per color level.

  4. Carlos
    April 12, 2014 22:25

    No light has no blue, but most household LED/Energy saving bulbs are warm white in the 2700-3100K Range, which have a low blue peak emission. You should be safe with those.

  5. PD Howells
    September 06, 2013 10:26

    What brand of light can I buy in the UK which has no or less blue light for a macular degeneration sufferer?

Share your knowledge - Leave a comment