Abstract: Seasonal Affective Disorder (SAD, e.g. you get sad in winter) can be resolved by applying enough 460 nm photons per day to cells at the bottom of the eye. This often caches out to advice about how long you need to sit by a lightbox or how bright it should be, but the underlying mechanism is that you need to get a certain number of 460 nm photons into cells at the bottom of your eye every day. Practically, you can build a device that does this efficiently for $60-100. A number of considerations on how to do this not available anywhere else are given below.
Section 1: Abstract
Section 2: Make Your Own
Section 3: Floor Lamps/Industrial Engineering and Manufacturing
Section 4: Why it works
Section 5: Patent and disclosures
Section 2: Make Your Own
The cells you want to get light to (ipRGCs) are in the bottom of your eyes, so ideally you want to deliver any light therapy from a device that’s positioned above you. Also, glare is generally worse the more lux get to your eyes, and it’s also worse the more that that light comes from eye level (and better the more that it comes from above eye level), so that’s another reason to make the light therapy device be positioned above you. And you want a lot of lux in your eyes since you need 460 nm photons to get in your eyes, so putting the light above you is the only way to do this without glare. Like the sun, on all counts, which is what ipRGCs evolved in response to.
There, that’s the most efficient way to do it. You can sit under it and do whatever you were going to do anyways without it being awkward. You don’t need a floor lamp (expensive), you don’t need a visor (expensive plus you can’t do much or read while wearing it), you don’t need a light box (literally doesn’t work), you don’t need to string lightbulbs across your room (most will be too far from you to matter, because inverse square law), you don’t need a thing that only works if you have a wooden ceiling or tools plus a couple hours to install ceiling joist mounting brackets. You just need to order some stuff off amazon and spend 5-10 minutes putting it together.
You get more photons if you’re closer to the light, so you want to have the light close to your eyes without it being awkward. Experimentally, I get the most out of a 50 degree angle.
You can stop reading now and just go buy this stuff and build one. Excel sheet with links. I really do recommend buying high-CRI bulbs if you can afford them, it’s much more pleasant to sit under them. I bought a bunch of 2700 K (evening) and 5000 K (day) Hyperikon bulbs and put them in everything in my house, the light quality difference is really noticeable. Currently the brightest bulbs you can buy are around 83 CRI, so there’s some tradeoff between CRI and brightness if you’re limiting yourself to 7 bulbs.
On hanging: 85% of people have drywall ceilings, you just follow the instructions from here. Drywall ceilings are kinda flimsy so you need to use your brain if you’re planning on hanging more weight than what I’ve hung, if there’s a tub in the story above you that might leak and weaken the ceiling, etc. Credit to David Chapman for doing an earlier prototype of this stuff. His design only works with drywall ceilings (what I have) if you put up ceiling mounting brackets to hold the weight of the light bars, so trying to engineer around that constraint is how I ended up with the above design.
People complain to no end that this involves drilling a very small hole in your ceiling to install, and claim they would rather spend $500 or more on a floor lamp with the same functionality, rather than $60-100 on a ceiling lamp like this, because they’re worried about their landlord. You can buy a $7 tube of drywall repair putty and just use it on the hole after you take the lamp out.
Section 3: Floor Lamps/Industrial Engineering and Manufacturing
I talked to a bunch of people with SAD and they told me they’d rather pay $500 on a floor lamp than $60 on a ceiling lamp because of fear of landlords, so I spent more time than I should have looking into manufacturing SAD floor lamps at scale. Here’s a prototype:
This includes, a heat sink and some quiet fans on top (it overheats without them and they aren’t noticeable, you can optimize for finding quiet fans), a power supply and a printed circuit board (PCB) with all the lights automatically soldered on, a diffuser (sheet of plastic to spread out light), a ball swivel joint, and an off the shelf (bought off Amazon) music stand.
Going from memory, the total parts cost is like, $40 for the stand, $30 for the LEDs, $40 for the heatsink (?), either $35 or $90 for the power supply depending on if it does pulse width modulation (PWM, allows for dimming the lights rather than just on/off), around $50 for everything else together. That’s $185, but there are lots of one time costs: you need to pay to have a factory set up every time you do a production run, you need to pay for each unit made after that, and you need to pay a tooling cost to have plastic injection molds made if you want the lamp hood and diffuser to be injection molded (high one-time setup cost, e.g. 10-30k, low per item cost) rather than say 3d printed (high per item cost, very high time cost) or e.g. made with deep drawing. So the real per-item cost depends on how many lamps you want to manufacture, e.g. on how much startup capital you have. A production cost of $400-500 per lamp would be a decent estimate if you’re spending around 50k on a first manufacturing run. As much as there are people willing to spend twice that to have their SAD dealt with (all the software engineers I asked were), and as much as you could make a profitable business out of this (at massive opportunity cost), that’s still totally ridiculous from the perspective of solving problems efficiently. Just hang a lamp from your ceiling for $60-100, or do a startup that isn’t limited to, at the best, only being able to double your capital a single time a year.
(If you ever do manufacturing, you really want to get a bunch of quotes from different factories, as the quotes they give vary widely. In addition, you really want a contract-hired, e.g. not full time, product engineer, who has worked with a bunch of factories and can tell you which ones are worth reaching out to e.g. are not going to try to “cheat” very much/make a lower quality product than you expected and only start giving you signs that that’s whats happening after a contract has been signed).
Section 4: Why it Works
Most lights have totally whacky spectra, and basically all of them have dips in their spectra around 460 nm, which is about the wavelength needed to stimulate the cells that matter for SAD, ipRGCs. Quite a few lights, including basically all fluorescent bulbs, do not have any intensity at all around 460 nm, so they’re totally worthless (the cheap, useless SAD lightboxes you can buy off Amazon mostly use fluorescent bulbs). A general trend (there are exceptions) is that higher-CRI lights tend to have higher intensities at 460 nm, and higher color temperature lights tend to have higher intensities at 460 nm as well. A typical trend is for lights that aren’t advertised in terms of technical specs like CRI to have 0-5% or maybe 5-10% relative intensity around 460 nm, for a typical “5000 K 90 CRI LED” to have 10-20% relative intensity around 460 nm, and for a light optimized for relative intensity around 460 nm to have 40% relative intensity there.
That’s the spectra from p.8 of this, it was the best light I was able to find in terms of “what has the highest intensity around 460-480 nm”, and it is what I put in the prototype. Someday, I’d like to have a graph of “ipRGCs respond to [460, 465, 470, …600] nm light with [x1, x2, x3, etc] firing rates”, so I’d have a cleaner understanding of exactly what wavelengths to consider by how much.
There was a point where I considered using blue LEDs instead of white ones, but from a practical perspective it’s more comfortable to use daylight-colored ones, and many blue ones don’t even have peaks around 460 nm. The prototype pictured in Section 3 is 40,000 lumens, so when you weight that against the intensities of light in something like 100,000 lumens that’s made with lower quality lights, you get more ipRGC stimulation from the 40,000 lumen device.
Heatsink design is annoying; you have to use around 3x more area of heatsink if you want to avoid using a fan, but ultimately it all comes down to what works experimentally. You typically want to do light therapy in the mornings or sometimes afternoons so it doesn’t keep you up/helps you wake up early, but you can do it at other times to adjust your sleep schedule on purpose.
If you want to take an exact spectrum of a light, you can go buy a tiny integrating sphere for 2k, or pay a facility around $400 to take one spectrum for you.
I want to talk about lux vs lumens, so I’ll copypaste the section from the patent on that:
Lightboxes are characterized by their luminous flux and their illuminance. The SI units for luminous flux and illuminance are the lumen (lm) and the lux (lx), respectively. The mathematical relationship between luminous flux and illuminance is 1 lx = 1 lm/d^2, where d is the distance from the light source to the individual, in meters. In other words, the illuminance (in lx) is defined as the luminous flux of the light source (in lm) divided by the square of the distance between the light source and an observer (in m). Accordingly, for a light source with a constant luminous flux, illuminance decreases with greater distances and increases with lesser distances.
One effect of this with ordinary lightboxes is the requirement that the individual be very close to the light source to receive the industry standard illuminance for light therapy, which is 10,000 lux of white light. Because many ordinary lightboxes operate with a single 1600 lm bulb, for the individual to receive 10,000 lux of white light, the individual must be 0.4 meters, or 15.7 inches from the lightbulb inside the device to receive effective therapy. However, requiring the individual to closely view the ordinary lightbox for a prolonged period limits the individual’s ability to perform other tasks while receiving therapy and often results in the individual viewing the ordinary lightbox from a greater distance, which decreases the dose received and decreases therapeutic effectiveness.
Of course, the “industry standard” of “you need 10,000 lux of white light” just means someone made that up, and of course what matters is lux as weighted by wavelength weighted by how responsive ipRGCs are to that wavelength, or like we said in the abstract, “how many 460 nm photons get in your eye”. The reason to talk about lux and lumens is, to give you a handle on the fact that the inverse square law applies to how much light you get, just as “how long do you need to sit under it” does (twice the time means twice the photons). So in practice, you are first making sure that the light isn’t awkwardly close to you or uncomfortable, but you’re also trying to get it as close to you as you can before it starts getting awkward at all. Really you want to titrate to individual needs; some people more ipRGC stimulation than others, and you just do enough light therapy to feel as good as you would in summer.
Having high quality/enough lighting can also have a “my inner animal is pleased and knows it’s safe and comfy” quality of life effect that’s mostly separate from ipRGC activity, and it’s hard to separate this cleanly from SAD (and the conditions can be comorbid/present at the same time). Running a SAD light for longer than necessary can be nice for that reason, though if that’s all that you need, you can just get a 2700 K set and a 5000 K set of 90+ CRI bulbs. There’s one way to “cheat” on CRI value tests, by having a low R9 value and high values on the rest of the CRI test, so if you’re just doing this for general comfort, you may want to check the R9 value. Or just try different bulbs and see what feels best. Experimentally, I’ve tried about 8 different high-CRI bulbs, and Hyperikon br-40s were by far the best, with Cree a19s in a distant second place.
Section 5: Patent and Disclosures
I originally planned to patent the floor lamp, but never filed the patent after it was written. It’s pretty easy to file and write your own patents, and if you have any STEM BS/BA degree, you can earn six figures if you go pass the patent bar exam (I passed an online retired/sample exam without studying), and advertise yourself as a patent writer on one of those freelance sites. Also you can hire patent writers for way cheaper on those sites than patent lawyers, and technical writers who haven’t passed the exam or don’t have a STEM degree (and thus can’t prosecute patents, but can write them) might be willing to draft a patent for even less. But patenting things is playing at the wrong level, and you can come up with a better world saving plan (or a better selfish plan) than that.
Publicly releasing this patent and the above document makes it all count as prior art, even without releasing it under a specific license, meaning nobody will be able to patent it now. I consider it obvious that there are ways other than floor lamps/visors/ceiling lamps to treat SAD, because it is obvious that light should come from above eye level, and there are many ways of doing that. My intention is that it’s fine if anyone copies or improves on this work publicly, makes one for themselves or a friend, makes floor or ceiling lamps to sell including for profit, etc. I should probably find a copyleft license for physical devices to use for future things like this. I’ll add any edits to this post in the comments so that it’s clear what date things are being disclosed on (post published Oct 1 2020).