Bright LED lamp for seasonal affective disorder (SAD)

Commercial lights for treating seasonal affective disorder (SAD, winter blues) are not nearly bright enough. They max out at 10,000 lumens, which—in my experience—has only a small effect. Recently I built a 30,000 lumen light, which seems much more effective. I suspect even more would would be even better.

Clinical studies have found that bright light treats SAD effectively. To work, it has to be really bright. Most commercially available SAD lamps are essentially fraudulent; they are much less bright than those used in the research. The brightest one I could find, when shopping online several years ago, was 10,000 lumens. That was enough that I noticed a clear improvement, but not nearly enough to restore full brain function. So over several years I’ve experimented with various homemade alternatives that are significantly brighter—and more effective.

Update: In a more recent post, I explain using more powerful LED lights to light up a whole room. They are less expensive than the technology I describe in this post, and easier to install. I think they will be a better solution for most people. They’re not suitable for all purposes, though, so I’m leaving this post up!

Lumens vs. lux

Lumens are a measure of how much light a lamp produces. Lux are a measure of how much light you get. These are not the same; the further you are from a lamp, the less light you get. Converting one measure to the other is not straightforward; it depends on how much of the light is directed toward you (vs. away in other directions) as well as how close you are.

The clinical studies were done with 10,000 lux provided by a bank of fluorescent tubes in a box that directed most of the light forward. Very roughly, if you are a couple feet from a box like that, 1 lumen produces 1 lux. So a 10,000 lumen light box (the most powerful I could find commercially) roughly replicates the clinical protocol—if you work two feet from it.

Many commercial products claim to provide 10,000 lux, which is meaningless—it depends on distance. Even a flashlight can provide 10,000 lux if you put it right next to your eyeball. Most commercial SAD lamps produce much less than 10,000 lumens; 2500, or even less, are common. (They almost never say how many lumens they produce, but you can calculate it from their wattage and the luminous efficacy of their bulb type.)

Anyway, there’s nothing magic about 10,000 lux. In fact, based on my experience, it’s inadequate. Full daylight is about 100,000 lux, and I think the closer you get to that, the better.

This year’s model

My most recent construction is shown above. It uses three LED “corn bulbs,” 10,000 lumens each. These ones, available from Amazon, have the highest lumen-to-dollar ratio of any LED bulbs I could find. They cost $120 apiece, and you need non-standard sockets. Total parts cost was more than $400, which seems like a lot, but is totally worth it if you get your winters back. (Below, I also suggest some lower-cost alternatives that I haven’t tried out yet.)

Update, September 2016: The bulb I used in 2015 is no longer available. This one is similar, and has excellent reviews on Amazon, but I have not tried it myself. Last year’s model was 100 watts, the most available; this one is 150 watts, and a little more efficient, so you could get more light with two of them than I got from three last year. They are $170 currently. It also comes in lower-watt versions that fit standard sockets.

If you have access to wood-working tools, you can build a lamp like this in a few hours.

I sized the stand so it fits behind my 30” computer monitor, with the bulbs showing over the top. It also can sit on the floor by the beanbag chair I sometimes work in, putting the bulbs a little above eye height.

The light is turned off in the picture above. When it is turned on, it blinds the camera, and mostly all you can see in a photo is the bulbs.

30,000 lumens is equivalent to about 1800 watts of regular incandescent bulbs, or 1500 watts of halogens. How much light you actually get depends on how close you are to it. I work with the light one to two feet from my face—so it is quite bright! I’ve compared it visually with midday sunlight in October in Northern California. One foot away from the lamp, it’s about equivalent. (Halving the distance quadruples the brightness—the inverse square law—so if it’s roughly 30,000 lux at two feet, then it’s roughly 120,000 lux at one foot, or the equivalent of daylight.)

I’ve found that this new LED lamp has almost completely reversed SAD. I built it mid-October, at which time I could already feel significant SAD symptoms. Within a week, they were mostly gone. I was cheerful, energetic, and capable of complex thought in December, which was unusual for me before I started bright light treatment.

I do think MORE would be better, however! It would be more comfortable to have the light further away. Also, illuminating a broader area would be a better daylight simulation. I’m thinking of adding another couple bulbs, for 50,000 lumens total.

Using a SAD lamp

If you aren’t familiar with bright-light therapy, read up on it before proceeding. The Wikipedia article could be a good starting point.

There is compelling evidence for efficacy in treating SAD, including mild “winter blues” as well as major clinical depression. It also appears to be effective in treating sleep cycle disorders. (Many SAD sufferers have longer-than–24-hour cycles, resulting in “permanent jetlag.”) There’s also moderate evidence of efficacy for treating non-seasonal depression. Bright light therapy can trigger mania or hypomania, so you may want to consult a doctor.

What is not clearly known is how light therapy works, and what is the best way of applying it. Unfortunately, this uncertainty has left room for dubious claims and outright pseudoscience, promulgated mainly by makers of devices that probably don’t work at all. The rest of this section treats uncertain issues, based on a combination of my personal experience and reading; so it too is not very well-grounded scientifically.

SAD is basically hibernation. For me, it is not primarily affective. My brain slows down after the equinox, and—without treatment—by December it’s impossible to do serious thinking. I do tend to get depressed in winter, but I suspect that’s mostly because I can’t think properly.

SAD is probably triggered by visual cues: the length and brightness of the day. Some purported SAD treatments shine light on the skin, not in the eyes; I doubt those work.

From my experience, the more sun-like the light is, the better. It’s better if the light seems to come from the same angle above your eyes as on a summer day; it’s better if it is extremely bright; it’s better if it is truly white. (More about whiteness below.)

To simulate a long summer day, light therapy is most effective in the couple of hours before the natural daybreak, and starting at the natural twilight. I turn mine on as soon as I wake up, and as soon as it starts to get dark, and also run it in the middle of the day if it’s gloomy outside. There’s clinical evidence that ramping up the brightness gradually before you wake up (“dawn simulation”) is highly effective. I haven’t tried this. Informally, I do “sunset simulation” by dimming the light gradually in the evening, and also shifting its color toward yellow from white.

Color matters

Some of my previous experiments used halogen bulbs, which did work, but they produce a huge amount of heat, and it’s hard to get close enough and be comfortable and safe. They also produce much yellower light than the sun. The commercial fluorescent SAD light I have used also produces poor quality light. I believe that truly white light is significant in treating SAD. I don’t know of good clinical evidence for this—in fact, some says the opposite—but it fits my experience, and many anecdotal reports I’ve read.

The LEDs I used “taste” much more like sunlight than halogens or fluorescents do. Sunlight has a “color temperature” of 5600k—pure white—where halogens are 3200k, or somewhat yellow. Most commercial SAD lamps use fluorescent bulbs, which are generally around 4500k, or rather greenish. Most fluorescents also have a poor “color rendering index” (CRI), so their light looks sickly and alien. CRI is a measure of how well-balanced the different wavelengths of light are. 100 CRI is ideal; 90+ is very good. Some fluorescent bulbs, and some commercial SAD lights, are marketed as “full spectrum,” which has no specific definition, and often these are not good matches for sunlight.

The LEDs I used have a color temperature of 6000k. Their CRI is not specified; subjectively, it seems good. There are LEDs available that provide CRI as high as 96.

Some LEDs allow you to vary the color temperature; mine do not. If I add another couple of bulbs, I will make at least one of them “warm white” (3000k color temperature), separately switchable, for sunset simulation.

Relatedly, I use a nifty, free program called f.lux, which simulates sunset by gradually adjusting the color temperature of your computer’s display. I find it significantly helpful. By default, it simulates sunset based on where you are, which is exactly the wrong thing in winter if you have SAD. During winter, you can set it to think you are in the opposite hemisphere, so you’ll have a summer’s day length before sunset, rather than a winter’s day.

Two alternative lighting technologies

The LED corn bulbs I used are not ideal. It would be better if all the LEDs were pointed in the same direction; the corn bulbs are meant to radiate in all directions, which is not what’s wanted in a SAD lamp. The metal reflector helps, but some is lost to the sides and upward.

I have experimented with LED strip lights. Mounting a large number of them on a flat surface would be more efficient (and better looking) than the corn bulbs. They’re also cheaper per lumen. Each of these strips produces around 14,000 lumens, for $15! (You’ll need a power supply and other components that add to the cost, but they’re still a steal.) [Update: Commenter Ivan measured these; it appears they are probably mis-advertised, and produce at most half that much light. He is experimenting with this alternative strip, which is supposed to produce 9,000 lumens for $14. Strips seem like a cost-effective solution, even if their light output is significantly exaggerated by the vendors.].

However, the strips are something of a pain to work with. After some futzing around, I decided to go with LED “bulbs” to save engineering and construction time, but you may want to go the strip route. (I’d love to hear and see that how works out, if you do!)

My friend Druktsal Pawo pointed me at metal halide bulbs. I hadn’t known of their existence. These come in several types, of which halogen metal iodide (HMI) bulbs are probably best-suited for SAD use. These have natural color temperature and high CRI, and are nearly as energy-efficient as LEDs (so heat is less of an issue than with regular halogens or even fluorescents). They are used for daylight simulation on indoor film sets.

Excitingly, single HMI bulbs can be much brighter than single LED sources, and are startlingly inexpensive. For example, this 575 watt bulb costs $14.95 and produces 47,000 lumens! Other HMI bulbs can handle several kilowatts, at around 100 lumens per watt, so several hundred thousand lumens are feasible. (Sounds of salivation from me.) Taking it to eleven, you can get a 24 kilowatt HMI bulb, rated to produce 2,300,000 lumens. That may be more than you really need.

HMI bulbs have some drawbacks, however.

So this would take some engineering.

The advantage is that a single bulb could be set near the ceiling to illuminate a large area at daylight brightness. That’s probably a better sunlight simulation than something two feet from your face.

Also, one reason many people give up on using SAD lights is that it’s a nuisance having to work so close to one. Broad illumination would solve that problem.

The main reason

I think the main reason people give up on using SAD lights, though, is that commercial ones just aren’t bright enough to do the job!