IR Illumination and Eye Safety
At TIS, we work with infrared light (IR) in order to capture irises for biometric matching, so we have had to get smart on IR light and eye safety. Most people are familiar with IR light because it allows your home security camera to "see" in the dark. You are probably familiar with LED light bulbs; they are becoming common for home use because they are so efficient. Well, LEDs come in all kinds of different wavelengths, including IR. Home security cameras will typically use IR LEDs to illuminate your porch/garage/etc. so you will be able to see things that are happening when it is dark.
To help the cameras, vendors have started providing IR floodlights which provide a LOT of IR illumination. But no one has ever really examined these to check whether they are safe, so that's what we are going to do today. Fair warning, there is a fair bit of math ahead, but I will try to make it as simple as possible. Feel free to skip to the conclusion if you just want the answer.
You may not have considered whether IR could be a hazard, and I admit I never did either before I started working with it. However, it can very easily potentially be a hazard. Normally, bright visible light isn't a problem because we have evolved to deal with that - our pupils contract, we feel pain, we shut our eyes, etc. However, even visible light can be a hazard if it is concentrated in the form of a laser; we never had to evolve to deal with high intensity laser light. Although IR is around us all the time in the form of heat, we haven't been faced with highly concentrated IR light until the last few years.
But let's step back a bit and look at what IR light is, and why cameras see it at all. You may remember the colors of the rainbow/visible light spectrum from high school physics - ROY G BIV. Humans can see visible light as shown below:
We can see in the fairly narrow range of 400 nm - 700 nm. Near infrared (NIR) is just slightly above red, in the 700-900 range. However, it turns out cameras weren't specifically designed to see in NIR, it was just a fortunate side effect of how they were made. You would think a camera would just stop at 700 nm because that is all humans can see, so why make the sensor see past that? Well, it turns out to be expensive to make a sensor that cuts off sharply at one point like that - it's simpler and easier to make it respond across the spectrum so that you cover the region you are interested in. To get the color red at 700 nm, you naturally "bleed" over into NIR for the sensor. Here is what that looks like:
You can see that humans respond primarily to mid-range/green colors. This is primarily an evolutionary advantage - primates with more green discrimination had an easier time finding fruit to eat. You can see that the camera responds a lot differently than the human eye, and can see more of the color spectrum. This works out great for security cameras. Without NIR, you would have to use visible floodlights to be able to monitor things at night and that would likely annoy your neighbors. We can use NIR floodlights instead, which are mostly invisible to humans.
NIR floodlights come in two flavors - 850 nm and 940 nm. As you can see in the camera graph above, cameras respond almost 2X better to 850 nm than they do for 940 nm. However, 850 nm lights have a problem - they are visible to humans at night, although they are dim. That is because of the same problem we had with making sensors - we can't make an LED that just outputs at 850 nm, it "bleeds" over into the red visible spectrum a bit. 940 nm are pretty much invisible to the human eye because they are a bit farther from the visible spectrum.
The difference between these two types of LEDs in terms of eye safety will become apparent later. When we consider NIR eye damage, there are a few key things to remember:
- Our pupils will not contract to protect us from intense NIR light.
- If our eyes are damaged, we won't be aware it is happening.
- Damage is actually a biological/chemical process that will take a long time to manifest.
That last bullet is the most scary to me. If you want to see real-life examples of this check out this link (not for the squeamish) that shows examples of "Glassblower's Cataract." Glassblowers in the past were subject to massive amounts of IR light which resulted in damage that accumulated over time. We certainly want to avoid anything like that.
I obviously can't analyze every IR floodlight on the market, but I thought I would pick one of the strongest available on Amazon, namely this beast. It has 196 LEDs, more than any other IR floodlight I found. I bought one of these lights and tore it apart so I could do some measurements on it. Before we can go into detail, I have to cover a few more background items so the rest of this analysis makes sense.
Beam Angle and Steradians
LEDs come in almost every configuration you can imagine, and one of the main variations is beam angle. An LED has a tiny emitter inside (typically 1 square mm or less), but manufacturers will vary the lens on the LED to make it output a narrow cone of light or a more spread light pattern. If you have ever used a flashlight that lets you vary the intensity of the beam, you are familiar with this. Here is an illustration:
It is important to understand that the same LED emitter can output a small concentrated area of light, or a broad diffuse area of light, but the overall light energy output is the same.
So now onto steradians. You are probably familiar with radians from high school trigonometry - they are like degrees, just more annoying. But radians/degrees are in a 2-D plane. A steradian is like that, but in 3 dimensions. You can think of it like a cone cut out from a sphere.
It turns out this is a very useful way to talk about light sources and people looking at them. When you hear the word "steradian" imagine yourself at the edge of the sphere, looking in towards a light at the center.
LEDs are usually specified in units of radiant intensity, measured in W/sr (watts per steradian). You can think of that as light energy spread out over some area. If you picture the sphere above growing or shrinking, the same source light energy will be spread out over a smaller or bigger area, but the radiant intensity is independent of that distance. That turns out to be useful.
An LED basically takes electricity and converts it to light, and it does it pretty efficiently - about 90%. As an interesting aside, the old style incandescent bulbs were the opposite - most of their energy was IR, with only a small amount being visible. That's why we got rid of them.
So let's say we have two LEDs that are identical, except for their lens, like LED A and LED B. Both will typically use 100 mA of electricity and both will output the same amount of light energy. However, A has a beam angle of 40 degrees vs B's beam angle of 120 degrees. That means A puts its light into a smaller area than B, but remember they both output the same amount of energy. A has a radiant intensity of 130 mW/sr and B has a radiant intensity of 12 mW/sr.
Which would you think is safer for the eye? If you are not sure, then think about it this way. Imagine a tiny flashlight using a AAA battery and a laser pointer also using a AAA battery. They will have roughly the same amount of light energy, but the laser pointer is much more hazardous than the flashlight. The more concentrated the light (greater W/sr), the more potentially damaging it is to the eye.
I looked around at a wide variety of IR floodlights and no vendor published the specifications about the LEDs they use or even a model number of the LED. However, CMVision helpfully sells replacement LEDs for their lights as shown here. That doesn't have the detail I need either, but it is a start. We know that the LEDs use about 60 mA when they are on all the time, and they have a beam angle of around 45 degrees.
When I opened up my unit, I measured the current draw for the LED board itself at about 1.5 A. With the specified LEDs, it should only be able to drive a string of about 6 LEDs, so this 1.5 A gets split into (196 LEDs/6 per string) = ~33 strings. That would give us around 45 mA per string flowing through each LED, so we are in the right ballpark.
I had to hunt around to find an equivalent LED. The LEDs on this floodlight are huge - 10 mm, which I assume is partially because they are cheap to buy in China. The closest thing I could find was this LED, which is a 5 mm LED. It turns out that smaller LEDs are more hazardous, so this is an OK substitution to make. This LED has a beam angle of 44 degrees and otherwise looks very similar to our unknown LED. My guess LED also can handle 100 mA vs the floodlight being able to only handle 60 mA. We will err on the side of caution and use the higher number. At that current, we would output 70 mW/sr. That's the magic number we are looking for. It is worth noting that the floodlight's LEDs are guaranteed to output less radiant intensity than my assumptions, so we are being conservative.
Eye Safety Calculations
The eye safety calculations are really complex because they have to cover a lot of territory, so I am going to be simplifying a lot of things from here. A really good resource is this paper from the vendor Osram. If you want to do your own eye safety calculations, you will need to spend a significant amount of time learning or hire someone.
One complexity on eye safety standards is the LED lights come under the "bulb" standard, which means that eye safety is measured at 200 mm (~ 7 inches). This is an "over abundance of caution" kind of thing. Even with a floodlight that will likely be on a garage, we can't always know what kinds of crazy things people will do with it.
There are three kinds of exposure hazards the standard addresses:
- Exposure damage to the cornea. This is caused by too much emitted light.
- Exposure damage to the skin. We will ignore this one because we are never getting anywhere near that level in any conceivable light.
- Retina damage. This is caused by the most intense element in the light.
Cornea damage is basically heating up the surface of the cornea because there is too much light coming into the eye - think of looking at a Hollywood searchlight directly. Retina damage is caused by an intense single source of light being too strong on the retina. Think of looking at a laser. Retina damage is relevant for arrays of LEDs because each LED translates to a point on your retina, even if you can't see the individual LEDs. If that point is too intense, then retina damage could occur.
Cornea Damage Calculation
The standard of safety for an LED light for long term exposure times, exposures greater than 1000 seconds, is an intensity less than 100 W / m² at 200 mm from the light. To know whether we are safe we need to convert our radiant intensity in W/sr to W/m^2 at 200 mm.
The output of the entire floodlight is simply 70 mW/sr * 196 (# LEDs) = 13.7 W/sr. This assumes an ideal radiation which we don't have, but that's OK because we are assuming worst case. For an LED light to be 100% safe for exposures > 1000s it has out less than 100 W / m² at 200 mm from the light. We can calculate our emission as 13.7 W/sr / (.2)². Basically - the strength of our light drops off as the square of the distance from it. That calculation give us 343 W/ m². Oops. We aren't eye safe for unlimited exposure at 200 mm. Even if we over-estimated our LED output by a factor of 2, we still aren't safe for unlimited exposure at 200 mm.
But of course, that doesn't really match our use case because we are farther away than 200 mm and we wouldn't be looking at this light for > 1000 seconds. But the standard doesn't care about our use case. This light should be in "Risk Group 1" which allows for exposures of less than 570 W/m^2 for 100 seconds. That would simply require a warning on the product. To be fair, the vendor has never made a claim about eye safety.
It is worth looking at a more realistic measurement for a floodlight, though. We can probably assume that anyone looking at the floodlight is at least a meter away. That would make our emission 13.7 W/sr / (1)² = 13.7 W/m², perfectly fine for looking at for > 1000 seconds. Other factors such as the angle of the lights (people will not likely look directly at them) and the time also come into play to make cornea damage essentially a non-issue for this light.
Retina Damage Calculation
We can make the math on this one a little easier. This calculation involves the size of the emitter on the LED, and all we care about is one LED. Retina damage is independent of the number of LEDs, it only depends on the brightness of a single LED. To make the math easier, we will assume the emitter is 1 mm². It is certainly significantly bigger than that, as will the emitters be for any floodlight on the market today.
For a 1 mm emitter, the limit for radiance is 545.5 mW/mm²/sr for unlimited exposure, or about 0.5 watts/mm^2/sr. That lets us divide by the inverse to easily calculate our actual radiance by dividing the radial intensity of one LED (70 mW/sr) by 2, the result being 35 mW/mm²/sr. That is well below the limit. In fact, our LED would have to output a whopping 1 W/sr to get to the limit. Those exist, but they get insanely hot and you will never find one in a floodlight.
So retina damage is a non-issue. That's good to know. But it is important to note that retina damage is independent of distance from the source. Shorter exposure times do mitigate it, though.
Summary and Conclusions
If you buy the largest IR floodlight available on Amazon and you stare at it 200 mm (~7") from your face for more than 1000 seconds (~ 16 minutes) there is a possibility you might damage your cornea. So don't do that. And I doubt anyone would do that because these things get pretty hot at that distance.
In general, you should avoid staring at very bright light sources, IR or visible, whether they are designated "safe" or not. However, under any conceivable use case for an IR floodlight in a security application, you should be completely safe.