The Call of the Open Sidewalk

The small area of LED light emitters makes it somewhat challenging to use them effectively in low level lighting applications. The light is hard to shade and diffuse efficiently. In the past I have used light coloured ceilings and walls as diffuse reflectors. This works fairly well but it can be a problem avoiding a bright area on the wall/ceiling where the light is mounted. Here is a description of a light I built that physically moves the light emitters away from the wall to deal with this issue. This works but the resulting fixture is quite large.

The ideal angle of the wedge of light from a wall mounted fixture is something a bit under 90 degrees. We want to avoid illuminating the wall behind the fixture while preventing people from seeing the small bright emitters directly. We also do not want a lot of light on the ceiling closest to the fixture as this tends to make a bright spot, particularly if the fixture is close to the ceiling. The raw LED chip itself emits light from a flat surface. This means that the light distribution tends to be proportional to the cosine of the angle from the normal. Cutting this distribution in half would produce a fairly good light distribution for the application. That could be done by adding a horizontal reflector. Since we have no need for optics other than the reflector we might as well use surface mount LEDs and make the fixture small.

I wanted to experiment with monochrome light as we are targeting night vision here. This plot:

from here suggests that the best wavelength for night (scotopic) vision would be 505 nm. That by coincidence is the wavelength that is perceived as the North American traffic signal "green" colour. As a result LEDs that emit this odd wavelength are relatively easy to find. The colour ends up being sort of a cyanish green. Light with a wavelength of 505 nm is fairly active as circadian light but the light levels achieved here are much too low to be disruptive.

My thinking up to this point inspired me to go out and buy these from here. I ended up with the larger 1206 packages because the supplier I ordered from would only provide the wavelength I wanted in that size. The extra width (1.6mm) ended up not being a problem and they were easier to handle than the 0805 packages I normally prefer for hand made surface mount based prototypes.

I spent some time planning to use a thin sheet of stainless steel as a reflector until I stumbled on this at a local home improvement store:

It was originally intended as an edging for a counter top. It was made out of a shiny alloy and was close to the desired shape. It comes with a prefabricated wall shade in the form of the overhang.

The use of the surface mount devices pretty much mandated the use of the PCB. Rather than measure anything I made a drawing at 5X scale:

and then printed it at 20% to use as a guide. Much to my surprise the Xfig program on Linux got it exactly right with no fiddling. That's not all that expected for a mere figure creation program. Here is the drawing in various formats: fig, postscript, pdf, svg, png.

A sharp tool was used to transfer the dimensions from the 1X scale drawing to the copper:

followed by a layout:

followed by some use of a cutting disk:

Note that the edges of the copper have been ground off to prevent shorts through the metal reflector. I just dragged the cutting disk along the sides. A more appropriate tool would make for a straighter edge.

I first applied liquid flux to the PCB. This is probably optional for such a simple assembly but I had it available and it did make things easier. A popular method of hand soldering surface mount devices starts by adding solder to just one pad. In my case this meant I would first add solder to each alternate area of copper like this:

Then the device is moved into position and held in place with something like a round toothpick. Heat from a soldering iron is applied to the solder (but not directly to the device) to sweat solder the device in place. Then it is fairly easy to solder the other pins. Once the device is well secured the original pin can have solder added if needed. This was the scene after soldering was complete:

I have ended up with 3 LEDs and 2 resistors. When I made the drawing I was enough of an engineer to want to cover the case where 4 LEDs were required to make up the voltage. The idea was to just skip one cut for the 3 LED case. When I made the board I was apparently enough of a technician to be compelled to make things as shown on the drawing. Thus I managed to make a classic error involving an ambiguous drawing in a way that normally requires two people. Using 2 resistors solved the problem in the assembly phase which meant I had also achieved pointless symmetry which is another sort of classic error. There is probably some insightful observation on the human condition available to me here but it escapes me.

I picked a LED current of half the normal specification based on the hope that it might increase the lifetime of the fixture. I have no idea if that makes sense with LEDs. Heat might not be an issue. LED manufacturers like to quote a lifetime of 100,000 hours (11 years) which really just means they think their product might last a long time. With a target current of 10 mA we end up with (12V-(3.2V*3))/0.01A = 240 ohms of resistance. I ended up using two 110 ohm resistors.

After soldering the wire connections and cleaning off the rosin based flux with pure isopropyl alcohol I was done with the electrical assembly:

I used a small amount of RTV silicone sealant to glue the board in place in the reflector. The board would not stay where it was supposed to go. I ended up using the weight of the wires to keep it in place:

The preceding image shows that the reflector is quite a lot wider than would be required just to reflect the light from the LEDs. The idea was to prevent light from escaping from the ends of the channel without any extra fabrication. This worked but the channel itself was still a bit bright where it could be seen on the ends.

Testing the light distribution against a white surface showed some problems:

To prevent light from projecting below the horizontal plane the assembly had to be tilted back towards the LED side. This is inconvenient for the eventual installation of the fixture and caused two other problems. The largest amount of light was going more or less directly up to the ceiling above the fixture creating a bright spot. The other problem was that a significant amount of light was finding its way to the wall directly behind the fixture which created another bright area. Blackening the top of the channel helped but I eventually took a different approach and bent the end of the reflector up like this:

The edge of the reflector is bent up enough to be in the same plane as the top of the channel. This means that there is no light projected below the horizontal plane when the fixture is sitting flat against a horizontal surface. The result was:

This is much better. I consider the design complete.

Next I'll go into the various thoughts and observations that came out of this activity...

When using monochromatic light even illumination can create a situation that destroys visual information. The light from the designed fixture is very uniform. Evenly illuminating large areas of the ceiling results in a situation related to an outdoors whiteout condition. Objects are still visible but with little detail. I suspect that this would only be an issue in rooms small enough to allow a single fixture to evenly illuminate the entire ceiling. A fix is to move the fixture closer to the ceiling to produce sharper shadows at the cost of a more intense hot spot.

The Homo sapiens sapiens set depend on visual transitions in their peripheral vision to help with things like walking around. The sharp vertical light transition produced by the reflector means that a strong cue should exist for orientation and balance. This artificial horizon effect might end up being the more important visual information. This system is after all intended to improve navigation. I am not sure how well this actually works at low light levels. There is probably a research opportunity here.

An important performance feature of a low level lighting system is the length of time it takes for vision to adapt enough to navigate after the high level lighting is turned off. Testing this produced a result that initially seemed odd. It seemed that the nighttime vision system (rod based) was adapting significantly faster than the daytime vision system (cone based). I would see a monochrome image before I could perceive the colour of the light. This lag was something like 5 seconds. The oddness here came from the impression that nighttime vision takes a much longer time to adapt to darkness than daytime vision. Some research revealed that this isn't really true in general. It turns out that the traditional way to check the time it takes people to adapt to the dark is to expose them to really bright light for a long time and then plunge them into complete darkness. This presumably is to simulate the common situation where one is walking along in a sunlit meadow before falling into a fast moving underground river. In more reasonable transitions from light to dark it is the nighttime vision that kicks in first (Ref, See Figure 2). At the light levels I am using the effect would be close to the maximum. The daytime vision system has to adapt to a level close to the daytime minimum while the nighttime vision system only has to adapt to a light level 100 times higher than the nighttime minimum. The transition is driven harder for the nighttime system. This I think strengthens the contention that 505 nm is the magic wavelength for low level lighting.

Since the edge of the reflector is to be bent up it might make sense to cut the reflector in such a way as to allow the bent up part to shade the bright ends of the channel. In other words, the reflector could be cut off at an angle to make the LED side shorter than the bent side. This could also hide associated wiring.

Light fixtures that point upward accumulate dirt. This creates a maintenance issue. For some applications it might be better to put the light sources below eye level pointing downward. Hallways and walkways are sometimes lit this way.

Blue LEDs at 470 nm are fairly effective for nighttime vision but are a lot less visible to daytime vision than 505 nm LEDs. This might make blue LEDs preferable where it is important for the night lighting to be invisible during the day.

The next step is to make some of these and deploy them around the house to try out the low level lighting lifestyle for myself. I should be able to hide the fixtures on the top of door and window frames. This will make it possible to hook the wires up into the attic for an easy installation. I will report on the results when and if there are any.

EDIT: proceeding -> preceding

posted at: 15:49 | path: /ledlight | permanent link to this entry | Comments (0)

Did you ever watch a moth express a unnatural and disturbing desire to become one with the filament of a light bulb? I tend to feel sorry for the moth and, I must confess, a little bit superior. After spending way too much time reading about circadian light I think that in the future I will feel a little less superior.

In a previous post I mentioned something about blue light and circadian rhythm. It turns out that mammals have an entire separate visual system devoted to synchronizing circadian rhythm to the local light/dark cycle. The existence of this system was entirely unknown until recently. I would like to state at this point that I think that the people who figured all this out are awesome. All the important stuff was worked over a period of less than 10 years. It was definitely a triumph. I suspect that even now it is still pretty difficult to get funding to study something no one learned about in school. Imagine what it was like to propose that there was this huge thing that everyone else had missed in the beginning.

Basically there are slow ambient light sensors in the eye that are connected directly to a biological oscillator with a period of 24 hours. This oscillator is phase locked to the light/dark cycle and in turn locks various systems to the light/dark cycle. The system is most sensitive to light in the range from 450 to 480 nm. Light of this wavelength is perceived by the daytime visual system as blue. Here is a nice summary of what was known in 2003. Check out the figures.

Following this I will freely mix the ideas of myself and others in infotainment style.

Much of this is still not understood

... least of all by me. Even the reasons for having a circadian rhythm at all are not clear. The explanations I have seen state something to the effect that the circadian rhythm is important because it regulates important stuff. Some arctic animals get rid of of it entirely in the dark months but start it up when the light returns ((Ref)Thanks Gord). This implies that maintaining a circadian rhythm is expensive. At the same time it implies that the cost is worth it somehow.

The thing with these systems created by evolution is that evolution can't actually design anything. It doesn't matter that no one can understand it or even that it makes sense. It just has to work. Stuff like this is why I avoided taking anything like biology in school. Every path leads down the rabbit hole. I strongly prefer things that make sense in a simple way. Evolution sucks.

What is understood is that people are directly controlled by light levels in a very direct way. We just don't know what all the implications of this are.

"Wait, how many marks do I get for this?"

While reading through related papers I was struck by the apparent level of dedication shown by the human experimental subjects. These sort of things are mostly inflicted on students. Students tend to think that the most outrageous form of physical humour involves drawing stuff on drunk people. Thus it must of seemed that they were part of someone else's hilarious practical joke. When pushing back the frontiers of this particular bit of science you get to:

• Put up with people shining bright lights directly in your face.
• Sit alone in the dark for hours extending to days ... often with a thermometer in your butt.
• Stare at an illuminated surface ... without moving your eyes ... for up to 6 hours. At least one experiment had a video system to allow someone to yell at you if you attempted to move your eyes.
• Wear incredibly dorky glasses in public. Funniest when applied to teenagers (Ref).

Melatonin

Circadian light significantly reduces the level of a hormone called melatonin. The hormone is reasonably easy to measure and is dramatically effected by light exposure. Thus people like to use it as an indicator that circadian light is being detected. Since it is the thing everyone is measuring people tend to fixate on it when coming up with theories. Light makes it go away fairly fast (15min) but it can take 3 hours to return in darkness.

Age

Older people are significantly less sensitive to circadian light than younger people. In an analogous way to the way people loose the ability to hear short wavelength sounds as they grow older they also lose the ability to sense shorter wavelengths of light. Circadian light is at the short wavelength end of the spectrum and gets attenuated. The first graph from this shows the spectral characteristic of the lens of the eye with respect to age. Fluorescent lighting has a large spike of energy at 435nm. This would be effective as circadian light. Past the age of 20 this wavelength is strongly attenuated. It might be reasonable to assume that children and teenagers are much more effected than adults by fluorescent lighting. I suspect that is part of the reasoning that led to studies like this.

Circadian light is good

People living close to the poles tend to live without much natural light for much of the year. This can lead to various problems. The most famous of these is seasonal affective disorder (SAD). A common treatment involves sitting in front of a box full of fluorescent lights for a while in the early morning. Circadian light is generally assumed to be involved somehow. One idea is that melatonin prevents the production of serotonin which in turn prevents the person involved from feeling good. So... circadian light can make you happy...

In general people like brightly lighted places. I bemoaned that fact in this post when I was discussing low level lighting. Those people back in the day were not just wasting energy with their extravagant light levels. They were possibly improving the mood and health of the people who worked there. The present standard for office lighting calls for 500 lux. You need 1000-2000 lux to get a reasonable circadian effect. Typical factories and warehouses tend to be dimmer than offices.

Circadian light is bad

Bright light in the evening can retard the circadian cycle in people. The effect can be increased where there is no exposure to light in the morning. This could mean that the affected person would stay up later and have trouble getting up for work/school in the morning. The net effect would be a general lack of sleep.

The question of "How much light is too much?" is a bit complicated. Fluorescent lighting at 2000 lux would certainly have an effect after sundown. The traditional well shaded 60W incandescent sitting on an end table in a living room is not likely to make much difference. TVs and computer monitors might have an effect. For example, my monitor is rated at 300 nit. That is equivalent to 940 lux. That sounds a bit bright but the monitor only occupies maybe 1/4 of my visual field. That is important because the circadian light sensors are well scattered in the retina. We can't stop here because when people say that 2000 lux is significant they mean illumination. Only the light that is reflected from the surroundings ends up on retinas. The rule of thumb seems to be a factor of 5. So we are down by 4 and up by 5 which may mean an equivalent of 1200 lux of fluorescent lighting. That still sounds a bit high.

There are of course programs to play with for people that want to experiment with reducing evening light exposure caused by their monitors. A program like Redshift that adjusts to a warmer colour temperature at night might help. The idea is that warmer means less blue light from the screen. For those trapped in the MS Windows environment F.Lux does the same sort of thing. The less technical solution for those concerned is just to turn down the brightness of the monitor/TV at night.

The "C" word

This section could be subtitled "Circadian light is real bad". I get the impression that there is a relatively large amount of money available for cancer related studies. When you look harder you find more stuff. It is then no surprise that there is a theory linking circadian light to cancer.

A study showed that women that work night shifts tend to have more breast cancer. Also, people in developed countries tend to have more cancer than people in less developed countries (citation needed). What do night shift workers and people in more developed countries have more of? One thing is artificial light. The growth of at least some cancers has been shown to be reduced by the hormone melatonin. More light means less melatonin so we end up with more cancer.

I am sceptical myself. As mentioned earlier, most workplace artificial light environments do not count that highly as circadian light. Since shift workers tend to create a dark environment to sleep in (and close their eyes) the extra melatonin suppression could only be what counts as the evening hours for the particular person. That would be only something like 3 hours down from the normal 8 or 9 hours of high melatonin levels. Significant, but most of it is still there. If artificial light does in fact cause cancer this would be one of the most significant unintended effects related to a technology in human history.

Nightlights make no difference

This all started with a night night that had bluish/white LED's. This well done experiment shows that a lower limit of 0.01 W/m2 for circadian effect would be reasonable. My calculations were quite rough. I assumed that someone managed to stimulate their entire visual field for 15 min (the time used in the study) with the light from my night light. I also assumed that all the light energy was moved to the circadian wavelengths. Since that light energy ended up being less than 0.01 W/m2 I think it is safe to say that my night light was not going to disrupt anyone's circadian stuff. It would be hard to think of a reasonable night light that would. From time to time the news media does a popular science article on the circadian light thing. Sometimes this gets livened up to "Is your childs night light harmful? Find out at 9!". For extra excitement they throw in the "C" word. So to that I say, no, it isn't. If it bothers anyone they can just stick the thing behind a dresser. "For extra safety!".

There is a lot of bogosity here

One of the nice things about writing articles about things wildly outside my expertise is that I can say things that might be a lot more insulting coming from someone more knowledgeable. Anyone with a light meter, access to a freezer, and a test lab can do a study involving circadian light. The catch is that a lot of this stuff is wildly unknown and it is hard to know what needs to be controlled. That does not mean that I think that the simple studies are worthless. It is just that they should not be used to back up things like news articles, wikipedia entries, and products.

Let's play along at home...

It might be fun do some some wild speculation of our own. For extra fun let's be slightly political.

Daylight savings time causes unhappy/tired people! This applies to the situation where just as the day becomes long enough to see light in the morning daylight savings time kicks in plunging morning commuters back into darkness. Since the lack of morning light is bad then daylight savings time is bad for a few months in the spring. In the height of summer daylight savings time results in a lot more light late in the evening. Also bad.

The ban on incandescent bulbs will cause tired/unhappy people! Canada is planing to ban light sources that use hot filaments. Incandescent bulbs create relatively little blue (circadian) light. Compact fluoresents could be expected to have more of a disruptive effect in the evening. Children/teens would be more affected by this. The "Find out at 9!" article writes itself at this point. If you dim an incandescent it produces virtually no circadian light. This option does not exist with compact fluorescents.

With great knowledge comes great responsibility

People are already trying to use this new found knowledge to abuse the working class. Yes, 3% extra productivity justifies any means. Employers have been fiddling with lighting for years. The problem is that now lighting at work is known to have an effect on life outside of work. I would not be surprised to see circadian light issues show up in future employment contracts. Even if an employee accepts a particular situation is it then ethical to fiddle with stuff that might conceivably have an effect on things like reproduction (Ref1, Ref2)?

What should we do?

I don't know about other people. Poorly understood? Potentially powerful and/or dangerous? Great! Let's try to hack it! More later...

posted at: 16:40 | path: /ledlight | permanent link to this entry | Comments (0)

I seem to be stuck on the low level lighting thing. An advantage to not working is not having anyone to tell me to get back to work...

Here is an approachable introduction to modern colour vision theory. It helped me a lot. The rest of the site (watercolours) is a bit too technical for me but I think I now know what to expect when I mix single colour LEDs.

For my bathroom night light I now think that just green and red LEDs would of worked fairly well. I could of balanced them to give some sort of yellow. A problem would occur if the colour was exactly the same as the perceived light colour. It would not be visible against white. The 3rd colour in between removes that ambiguity. That is assuming that the green and red LEDs do not balance out to the in between colour. The apparently excessively bright green LED is thus a feature as it moves the balance point well away from the yellow LED wavelength.

Another possible problem with just using green and red (or blue and red) is that you would end up with more green light. Green (and blue) can blind night vision when not used in moderation. I now think of my bathroom night light as primarily yellow with a bit of some other wavelengths thown in to allow some colour discrimination.

posted at: 17:13 | path: /ledlight | permanent link to this entry | Comments (0)

Some wisdom came out of my night light projects. I'll share that here.

Are light emitting diodes ready for prime time residential lighting? That question is surprisingly controversial. I am not really sure myself about the answer to the general question. I am sure that for low level lighting LEDs entirely rule.

Up to this point in the history of the world there has not been a practical and efficient way to make a small amount of light for a long time. Incandescents are particularly bad for efficiency and life. Electroluminescent were the closest thing pre-LED era but they are an area emitter and are thus hard to shade effectively. They also need high voltage wiring.

The result of this is that the system of vision intended to allow people to function at night is not really used by most people. People can see at levels down to 0.01 lux. Typical residential living spaces are lit to levels between 50 to 100 lux. That is as much as 4 orders of magnitude more than the absolute minimum required.

It may be that we have hit a technological threshold. If low level lighting is more practical then perhaps it will become more common. It is interesting to consider how a hypothetical low level lighting enthusiast would arrange the lighting in a house.

Let us first invent a specification. All living areas should be lit to the level of moonlight on a clear night. One source claims this is .3-1 lux. Brighter would interfere with things like sleep. Darker would mean a longish time to adapt after turning off a light. Tropical moonlight would be a bit bright I think so let's aim for the lower end of .3 lux.

For a 3m X 3m room we would have around 10m2 to illuminate. Lux is lumens per square metre so to hit .3 lux we would need 3 lumens. LEDs with an efficacy of 30 lumens/watt are available so our power budget could assume around 0.1W per lamp. Since this is mostly best case I'll double the assumption to 0.2W. It would likely take 7 of these lamps to light my house. So the total power draw is 1.4W giving us a yearly power consumption of 12 kW/h. That would cost me $0.72CAN a year.

The fixtures could be built in at time of construction. Retrofit would be easy as you would only have to snake in something like alarm wire to power the lights. The lights could be designed to run off the door bell transformer. If instead you ran them off a small sealed lead acid battery charged from a wall wart you could claim that you had established a sort of residential emergency egress lighting system.

The proposed level of lighting would of course mostly serve for navigating through the house. This would still simplify things. Most rooms would not need any ceiling lights and associated switches/wiring. Hallways and stairways would not need three way switched lights.

In general safety would be greater at night. I used to fall over the stool in my kitchen on a semi-regular basis before I installed my always-on counter light. I did get quite good at it but it would be unrealistic to think I would not eventually do it badly. The compact fluorescent that lights my basement stairs sometimes likes to wait just long enough for me to miss the top step before producing light after I turn it on. A little bit of light would help there too.

Our house of the future would probably need some built in higher level lighting. A ceiling mounted light for the dining room table is nice. Vanity lighting for the bathroom mirror is essential to allow the residents to maximize their apparent reproductive fitness. Kitchen counters and the stove would need bright lighting. Pretty much the rest of the lighting inside the house would be provided with lamps.

When one deals with lighting one is dealing with some fairly primal stuff. I always need to turn on the main light in the bathroom before having a bath. I think I am afraid to go in the dark water. Perhaps at some level I am afraid of alligators. When fluorescent lighting become popular people immediately started illuminating work spaces to some almost insane levels. Legend has it that photographers of that era used to have trouble believing the result of their light meter readings. They were seeing outside daylight levels of lighting inside buildings. A decade or so later "de-lamping" became popular when it became fashionable to care about power consumption. My point is that people like it when it is bright indoors and the brighter the better. Perhaps low level lighting as a way of life will never become popular simply because of that.

Here are some interesting/related links;

• Don Klipstein's Lighting Info Site
• The LED Museum
• Color Vision
• Scotopic Vision

posted at: 22:37 | path: /ledlight | permanent link to this entry | Comments (0)

Greg Moeller said:

I run normal incandescent in the main and second floor lights, motion sensors turn them on at about 5-10% for just a few seconds so you can walk up/down stairs without tripping. The kitchen has a compact florescent for dim lighting if you're grabbing water, also motion sensed. (5 minute timer on this one)

... which is a good point. You can use your regular lighting at a really low level at night especially if you can tell there is someone there. I guess incandescent is the only common type of lighting that this can work with.

I suppose I can generalize this a bit more. This is really all about low level lighting.

posted at: 16:08 | path: /ledlight | permanent link to this entry | Comments (0)