LED lighting, part 3 by Tim Thorstenson

(This is the third and final installment of this series. The first part, LED technology essentials, appeared in issue #127, Jan/Feb 2011. The second part, LED-based products, appeared in issue #128, Mar/Apr 2011.)

In this LED series, we've touched on the idea of building your own lights. The merits of building vs. buying are always very dependent on individual circumstances. With LEDs, building allows one to use the most efficient LEDs in the design of lights that are suited to specific needs. But to help you decide if building is right for you, Part 3 on LEDs will consider the details. We will also examine a related and critical topic: delivering light efficiently.

Basic requirements

We'll discuss projects that require only some fine soldering and a little basic electrical and mechanical technique. The final light or fixture is limited mainly by your needs and imagination, but the basic light-making assembly will be similar regardless of what you want to build: you need an LED or LEDs that will meet your requirements and a driver (electronic controller) that is matched to both your power source and LEDs. Finally, a heat sink is required to keep the LEDs from getting cooked. Once you have these basics, they can be built into an almost unlimited array of lights. This could include portable lanterns for camping and emergencies, trouble lights for use on car power, portable spotlights or floodlights with handy mounting attachments that could be run from a small rechargeable battery, and various fixtures for use in a camper, cabin, or home.

A battle plan

In Part 1 and 2, we've tried to focus on general principles. With the changing nature of the technology, pages filled with details about specific LEDs and products could quickly become useless. But it is also necessary to use real-life examples in order to talk sensibly. We'll use the best currently-available LEDs to illustrate the principles, but the principles themselves will remain useful even if (or when) these specific LEDs become outdated.

Sources of components

The most straightforward source of components is a retail supplier, such as LEDSupply (www.ledsupply.com, 802-728-6031). This company supplied components for the example projects pictured here and they receive very positive reviews for service. Of course, there are other good vendors, too. There are also bargain components available from discount outlets and websites like eBay. You can get some good deals, but there is a huge range of quality and customer service may be difficult or impossible. Remember that, especially in a cabin or home, even low voltage electronic components could pose a fire hazard if they fail. I'd suggest considerable caution and some appropriate research on forums like www.candlepowerforums.com prior to using "bargain" components.

LED selection

There are several different categories of LEDs (see Part One) and any of these could be considered, depending on the amount of light you need and the availability of suitable drivers. Here, we will stick with single-die LEDs because they are relatively easy to configure and because drivers are readily available. Brighter "multi-die" LEDs can also be useful in some applications, but suitable drivers are more limited at the moment. In addition, single-die LEDs have at least a couple of general advantages. First, for area lighting, several single-die LEDs can often be the better choice because they allow both the light and the heat to be spread out over a larger area. Second, continued improvements in efficiency are likely to turn up first in the single-die category.

Previously, we used the Cree "XR-E" LED to illustrate many concepts, but we also learned about a newer and more efficient Cree LED called the "XP-G." When I started working on this series of articles, the XP-G was only available in cool white and its availability was limited. But neutral and warm white color tints have now been released. Cool white is readily available and, by the time this article gets to print, it is likely that the same will be true of the more pleasant color tints. Because this LED represents the best efficiency available at this moment, we will use it for our examples. But, as I mentioned earlier, the principles of building will almost surely be similar when the "next new thing" rolls out.

Regardless of what type of LED you select, there are a couple essential numbers that are needed. One is the maximum current that it will handle, which is usually provided in milliamps (mA). The second is the voltage (usually called the "forward voltage") required to run the LED at maximum or at lower current levels. Such information, along with lumen outputs and lumen per watt efficiencies at various current inputs, can be had from a vendor or from the manufacturer's data sheet. As we go, we'll see examples of these numbers and how to use them. For home brewing, purchasing LEDs pre-mounted on what is called a "star board" (see the accompanying pictures) is often a sensible idea. Good mounting and connections can be a real challenge when trying to use tiny, bare LEDs. In contrast, a star board is easy to solder and it allows for simple mechanical mounting. In most cases, you can simply secure it to the heat sink with a heat-conducting epoxy or with a non-hardening heat transfer paste and some small self-tapping screws. Vendors that sell LEDs can also supply these materials.

It should be noted that there are also pre-made assemblies that can be useful for some projects. For instance, there are pre-made strips available that have LEDs mounted every so many inches on a long, narrow piece of "printed circuit board" (PCB) material. Some of these strips even include on-board drivers, so firing them up can be as simple as mounting the strip to an appropriate heat sink, providing a power feed, and adding a suitable cover or diffuser. Note that some of these options require the unit to be electrically isolated from the metal mounting surface. Make sure to consult the technical information.

Heat management

As we saw in Part One, heat management is essential. For many applications, there are pre-made heat sinks which are readily available and fairly inexpensive. The back of the heat sink (the cooling fins) should be left exposed to the open air and should not be enclosed in any sort of housing. In addition, the heat sink will work best if it is mounted with the fins pointed up or on its side with the fins running vertically. Within these limits, the heat sink and mounted LED can be incorporated into virtually any sort of lantern, lamp, fixture, or housing that you can brew up for your needs. A single small heat sink (about 2.5 inches square and one inch tall) will be adequate for a single small LED like the XP-G and it would be sufficient for a couple of them if the intended use will generally involve operation at lower power or for fairly short periods. If you are experimenting with your own materials, note that surface area is generally more important than mass and that a flat black surface will actually emit heat much more efficiently than a shiny one.

Powering the LEDs

Most LEDs can be operated over a wide range of electrical currents and there are some huge tradeoffs. As the current is increased, more light will be produced. But more heat will be produced, too. In addition, there is a diminishing return: total lumens will increase, but the lumen per watt efficiency will decline as the current input is ramped up. On top of this, published outputs and efficiencies (the ones I will cite) assume a relatively low operating temperature for the LEDs. As the LEDs are pushed to higher current levels, it becomes progressively harder to maintain a low operating temperature, even with good heat sinking. As a result, the gap between "ideal performance" and "practical field performance" will likely widen as we go to higher current inputs.

In any case where maximum efficiency is a must, lower operating currents are a necessity. To prevent thermal deterioration of the LED, they are also a good idea when a very long service life is desired. The LED itself has some inherent "thermal resistance" and more is added at the interface between the LED and the star board and at the junction between the star board and the heat sink. This all means that the LED will unavoidably run at a temperature that is at least somewhat higher than the surroundings, even if your heat sink is a fifty pound chunk of copper with huge fins. As the current input is increased, this temperature elevation will also increase. If we throw in the additional fact that home-brew heat management is likely to be less than ideal, the conclusion is that modest current inputs are the best way to insure a long service life. Of course, the drawback is that more LEDs will be needed to produce a given amount of light.

But when we get to things like trouble lights, camp lanterns, and cabin lights, the situation can be somewhat different. Even at fairly high operating temperatures (within reason), LED service life can still be many thousand hours before any significant reduction in brightness will occur. In these uses, it can be perfectly acceptable to run the LEDs at full power (assuming that adequate heat sinking is in place) because such lights may not get a thousand hours of use even in several years of service. Efficiency will drop off, of course. But, as a practical matter, it may be both easier and cheaper to use a little bigger battery for such projects than it would be to build lights that use larger numbers of more gently-powered LEDs.

It is important to understand that this does not need to be an either/or choice because an adjustable driver with a top current output equal to the maximum rating of the LEDs can be selected. For cabin or home, the lights could be designed so that a moderate brightness setting would prove adequate for most needs. But when you need extra light for intermittent tasks like cooking or reading, it would be available by turning a knob. Portable lanterns and work lights may get run at maximum brightness for routine use, but they could be turned down if an emergency requires you to conserve battery power. Of course, generous heat sinking should be used to allow good heat removal during highest-power operation. Electrical current measurements with a good meter can give you a handle on how much juice is being used by adjustable lights at various brightness settings.

Diffusers & lenses

A broad, uniform field of light is usually desirable for general area lighting and the natural pattern of the bare LED may be fine in some settings. But if the LED is subject to possible damage, it would be prudent to cover it with a clear plastic or glass cover. If glare is a problem or an even more diffuse light is desired, a textured or frosted diffuser may be preferable. However, extreme care should be used in selecting diffusers because heavy frosting or texturing can greatly reduce the emitted light.

In some settings, a tighter focus than the LED's natural output may be desired and a particularly useful accessory here is the "focusing lens." These small, inexpensive plastic lenses mount over the LED and they function by bending the light into a tighter pattern. Several different patterns are available and the right one can greatly improve the useful efficiency of your project because it will squeeze the light into the area where it is needed. For example, an unfocused LED might make an excellent area light. But when it comes to a trouble light, you may find that a lot of the lumens will wind up landing outside of where you really need them (be it a fuse box or an engine compartment). In contrast, a "wide" focused lens will provide a pattern that is still broad enough for such tasks but which is quite a bit tighter (and therefore brighter) than a bare LED. Lenses with a "medium" or "tight" focus could prove very useful for vehicle-mounted work lights or when you want to light up something like a walkway or outbuilding via a fixture located at some distance. One caution with these lenses: a given model of lens will be designed to work best with a certain type of LED. When ordering, make sure that you are getting the correct ones.

It's helpful to mention the difference between a focusing lens and a reflector. A focusing lens sits in front of the LED and it can be designed to gather the vast majority of the light output into a specified pattern with a very sharp cutoff between bright light and virtually no light. In contrast, a reflector sits behind the source (and to the sides), so it will only catch part of the LED's light and put it into a beam. The rest (which can be a major portion of the total output) will escape directly out of the front of the LED without being captured and focused. The result is a bright central beam and a broader, dimmer surrounding area of "spill" light. In many flashlight applications, this spill light is by no means wasted. The central beam allows you to spot things at a distance, but the spill light simultaneously allows you see closer hazards over a much wider area. In a defensive situation, this could be a menacing intruder. In more pristine environments, it could also be the tree branch hanging at forehead level or the gopher hole at your feet. But when it comes to an area lighting job (i.e. lighting up the path to the outhouse), the spill light from a reflector may indeed be wasted. Here, the right focusing lens can provide much more efficient use of the light.

Drivers

Good commercial drivers are available for single-die LEDs. Because the current inputs are fairly standard, drivers designed for older LEDs like the Cree XR-E will work just fine with the new XP-G. Because most readers are probably interested in using power sources such as 12-volt batteries, we will focus on the "buck" drivers. These take voltages which would otherwise be too high and reduce them so as to push a constant current flow to the LEDs. "Boost" drivers are also available for battery sources that would otherwise provide too low of a voltage.

A good example of a buck driver is the BuckPuck which is made in the USA by a company called LEDdynamics. It is no larger than a piece of bubble gum and can have standard outputs of 350, 500, 700, or 1,000 milliamps (mA). Drivers like these are available with pre-attached wiring harnesses and, if desired, a "trimpot" (adjustable resistor) that allows the brightness of the LED to be varied (turning the knob on the trimpot will progressively reduce the driver's current output from its specified maximum). The driver can be mounted in a small box and the trimpot is designed to be mounted through a hole in the enclosure with the adjustment stem sticking out. Attaching a knob to the stem will then provide easy adjustment. The standard DC version of the BuckPuck will handle up to 32 volts, so it is very suited to 12- or 24-volt DC batteries. Some drivers may draw a very small amount of power even with the trimpot turned all the way down, so an on/off switch or other positive means of disconnect is a good idea to prevent a small drain when the lights are not in use. One very important thing to note is that drivers like the BuckPuck can often power more than one LED. The maximum number will depend on things like the voltage supplied by the source and that required by the LEDs.

Design considerations for drivers

There are some extra precautions that should be taken when the drivers will be connected to a battery that is simultaneously attached to a charging system. Most importantly, a charger can provide up to several volts more than the "nominal" rating of the battery system, so it is important to insure that the voltage will never exceed the driver's rating. Some charging systems can introduce an "AC ripple" and this can be a problem with electronic devices like drivers. Better quality off-grid systems may provide filtering to eliminate this "ripple" along with regulation to prevent harmful voltage spikes and other problems, but this is always a good thing to check out specifically. Care may be required if the driver will be mounted remotely from the LEDs or if there is a long wire run from the driver to the power source. Long connecting wires can act like an antenna and pick up electrical "noise" that could interfere with the driver's electronics. Manufacturer's data sheets can provide information on how to deal with problems like this. It should be noted that the efficiency of a given driver can vary depending on factors like the input voltage. For applications like a camp lantern or work light, this is probably a fine detail to lose sleep over. But if you are planning a more extensive project, driver efficiency can become significant. The manufacturer's data sheet will once again provide helpful details and a good vendor can often assist with finding the best driver for a given project.

A caution about homemade drivers: designs for simple current-regulating circuits can be found in many basic electronics manuals (they are sometimes called "constant current power supplies"). These can provide a stable current output over a wide range of voltage inputs, but the pitfall with many simple circuits is their efficiency. They often do their job by essentially dissipating some of the input power and turning it into wasted heat. Depending on factors like the input voltage of the source and the voltage required by the LED, it is possible for a simple current regulating circuit to use as much power as the LED itself, or even more!

Configuring the LEDs and drivers

NOTE: Our use of the Cree XP-G LED as an example can lead to some confusion. Cree originally specified a maximum current of 1,000 mA, but they raised it to 1,500 mA after further testing. However, I am going to treat 1,000 mA as "maximum" for a few practical reasons: first, the higher limit only applies to cool white XP-Gs. Second, simple drivers are not as readily available at higher outputs, at least as of right now. Third, for reasons mentioned earlier, higher inputs could result in a high operating temperature under typical homebrew conditions. Finally, even at ideal temperatures, operation at 1,500 mA will result in an efficiency drop of around 40% relative to performance at 350 mA and this is somewhat self-defeating.

When it comes to powering LEDs, the first thing to address is the number of LEDs that can be connected to a single driver. To determine what is possible, we need the voltage required by the LEDs, the voltage required by the driver, and the minimum voltage that will be supplied by the power source when it is at a state of maximum discharge. Here's an example: At 1,000 mA, an XP-G LED requires a "forward voltage" of around 3.3 volts and a BuckPuck driver specifies an "overhead" of about 2.0 volts. This means that two such LEDs connected in "series" with a single driver would only require a total of 8.6 volts (2 LEDs at 3.3 volts + 2.0 volts for the driver). This could be supplied very comfortably from a 12-volt lead acid battery (which should never be discharged to a voltage anywhere near that low). If your battery won't be discharged too heavily between charges, it is possible to power three XP-G LEDs at 1,000 mA (required voltage would be 11.9 volts). It is important to note that a battery run down below this voltage (such as during an emergency) will still continue to power the LEDs, but their brightness will decline progressively as the voltage drops further.

When connecting multiple LEDs in this kind of setup, they are wired in series: the (+) output lead from the driver is connected to the (+) terminal of the first LED, the (-) output lead from the driver is connected to the (-) terminal of the last LED and all intervening LED-to-LED connections are made in an alternating (-) to (+) to (-) fashion. Of course, a proportionately larger heat sink (or multiple small ones) is essential when setting up multiple LEDs. A higher voltage system can allow even more LEDs to be powered from a single driver, but there are a few cautions here. First, it is important to make sure that the maximum voltage of the power system never exceeds the voltage limit of the driver. Second, even though the driver is still delivering the same amount of current through the LED string, the total power it is handling will increase as the LED string is made longer. Thus, it is important to check with a good vendor or with the manufacturer's data sheet for the acceptable maximum number of LEDs.

Output vs. efficiency

As we know, higher current will give us more lumens, but with a reduced efficiency. To see the impact of this tradeoff, let's consider the performance of a string of three XP-G LEDs. As discussed earlier, it becomes progressively harder to maintain low operating temperatures for the LEDs as current input is increased. Thus, the spread between these ideal values and typical field performance will likely become bigger as the current input is increased. The numbers cited here are ideal LED outputs for the "R5" bin (best available at this writing) in cool white. We'll consider the performance of neutral and warm white tints in a bit. If you are visualizing what you might need for a project, it is handy to recall that a standard 60 watt incandescent light bulb will produce about 840 lumens.

At a 1,000 mA input, three XP-G LEDs could provide an ideal LED output of 1,040 lumens and efficiency of around 103 lumens per watt. In contrast, operation at 350 mA would produce only 420 lumens, but efficiency will climb to about 132 lumens per watt (a 25% increase). In many practical settings, operation at a compromise current somewhere between these levels can provide a good balance between output and efficiency while keeping the heat output reasonably modest. For instance, operation of the three LED string at 500 mA would generate about 585 lumens while retaining an efficiency of around 120 lumens per watt. Increasing the current to 700 mA would result in about 730 lumens with efficiency in the neighborhood of 108 lumens per watt. Remember that operation at higher current inputs will result in more heat, so it is important to plan your heat-sinking accordingly. BuckPucks are available in a choice of the current outputs just mentioned.

Advanced wiring strategies

If you are certain that you only want to run the LEDs at lower levels (350 or 500 mA), there is a slightly more complicated wiring arrangement that may be worth considering. In a series-wired string of LEDs, each LED will receive the full current supplied by the driver. But if two identical, series-wired strings are connected in electrical "parallel" to a single driver, the driver's output will split between the strings and each LED will receive about one-half of the driver's current. Using the example of a 12-volt system again, this means that you could power six XP-Gs at 350 mA by purchasing only one 700 mA driver instead of having to buy two 350 mA drivers. Alternately, a single 1,000 mA driver could be used to power six LEDs at 500 mA. Drivers with adjustable outputs can still be used and data sheets will provide schematics for wiring strategies like this.

It is easy to miss important points in the midst of all the numbers and there is one that definitely needs to be emphasized here. This technique allows for the use of few drivers when the goal is to run the LEDs at lower current anyways. But we lose the option of cranking the LEDs up to "maximum" for occasions where the extra light would be handy, so it is good to be clear about what we are giving up. Powering six LEDs from one 1,000 mA driver (as just described), will allow for a maximum output of about 1,170 lumens (six LEDs at 500 mA). But using two adjustable drivers to power the same LEDs as two separate strings of three would allow for up to 2,080 lumens (six LEDs at 1,000 mA). The question here is whether having access to the extra light is worth the cost of the second driver. The answer, of course, depends on your intended uses.

Color tints and relative efficiency

In Part 1, we saw that neutral and warm white LEDs can provide light that many will find more pleasant, but that they also suffer from lower lumen outputs and efficiencies relative to cool white. This remains true with the new XP-G LEDs: best bin efficiencies for the XP-G LEDs at 350 mA are 132 lumens per watt in cool white, 116 lumens per watt in neutral white, and 102 lumens per watt in warm white. Cree is also offering an "outdoor white" which sits between cool and neutral white and provides an efficiency of 124 lumens per watt at 350 mA. There's still a tradeoff between aesthetics and efficiency, but it's definitely worth noting that even the warm white XP-G (102 lumens per watt) is more efficient than the cool white tint of the older XR-E LED (92 lumens per watt). As with cool white, the neutral and warm tints are available in a number of bins that vary in efficiency.

One more power strategy—but use with caution!

It can be tempting—and it is possible—to connect an appropriate string of LEDs in series and feed it directly from a given power source without the use of a driver. This approach has some serious limitations and it is very easy to burn out some expensive LEDs if it is done improperly. But there are a couple of reasons for mentioning it. First, those who pursue this topic further are likely to run across it, so a discussion of the pros and cons is in order. Second, as we will see, the operating specs for the XP-G work out nicely for this technique, so it could actually prove useful in a few cases.

Most of the problems with "direct feed" are a result of the touchy relationship between current and voltage in an LED and the XP-G can be used to illustrate the problem: about 3.0 volts is needed to drive the LED at 350 mA, but 3.3 volts will increase the current to 1,000 mA. In other words, increasing voltage by 10% will almost triple the current flow. The numbers vary with different LEDs, but the same general touchiness holds true and this makes it very easy to accidentally overpower the LEDs. Of course, this can lead to a shortened service life and, in extreme cases, it could even burn out the LEDs in pretty short order. Making matters worse is the fact that the published forward voltages for a given LED are "nominal" values. As with brightness and efficiency, the exact required voltages will vary somewhat from one LED to the next. So even if you are supplying the correct voltage in theory, it is still possible that the LEDs could be getting an excess dose of current. It is for these very reasons that the LED makers strongly suggest powering their products by controlling current rather than voltage.

That said, the XP-G is about as good as it gets for this approach. First, the voltages are convenient: a string of four such LEDs wired in series would require about 13.2 volts in order to produce a 1,000 mA operating current (4 LEDs at 3.3 volts each). This, as many readers will know, is in the neighborhood of the maximum voltage supplied by most 12-volt lead acid batteries when in a state of full charge (various types of "12-volt" lead acid batteries will specify somewhat different full-charge voltages, but most of them are at or below 13.2 volts). In addition, the XP-G has a maximum current rating of 1,500 mA (in cool white). To get the current this high, a nominal voltage of around 3.5 volts per LED is required and this works out to about 14.0 volts for a four-LED string. This is above what we are likely to get from most "12-volt" lead acid batteries and it provides a cushion with respect to accidentally overpowering the LEDs. If considering this approach, please make sure to confirm the actual maximum voltage output of the power source you will be using!

Here are some precautions for those thinking about this strategy. First, it would be very prudent to conduct actual current measurements with a good meter to insure that the LEDs are not being overpowered. Second, I personally would avoid a setup that involved the simultaneous connection of a charger. In a setting like a cabin where the batteries will be charged by sun or wind while you are away, the charger and lights could be connected to the battery with a "double throw" switch so that one or the other (but not both) would be engaged to the battery at any given time. Third, a battery in a full state of charge will run the LED strings at fairly high current inputs, so generous heat-sinking should be used for such setups.

It is also important to understand that this strategy has some other problems that are less dramatic but are still very significant from a practical perspective. First, there is no way to efficiently dim the LEDs. An adjustable resistor could be placed in series with the LED string, but this illustrates the problem with using resistors to directly control the current to LEDs: as the resistor is turned up and the LEDs become dimmer, the power that would otherwise go into the LEDs is simply being dissipated by the resistor in the form of heat. So much for efficiency! The other big snag is that brightness will fade dramatically as the battery voltage drops during discharge. The output of four XP-Gs at 13.2 volts will be over 1,300 lumens. But at 12.8 volts, it will be a little under 1,000 lumens. At 12.4 volts, it is only around 800 lumens and at 12.0 volts it is a humble 550 lumens (these numbers are calculated from "nominal" LED voltages, so actual results could vary by a fair bit). In contrast, a three-LED string with driver control will be able to maintain full brightness (or a reduced output of your choice) down to at least 11.9 volts. With a two-LED string and a driver, constant full brightness would be available down to 8.6 volts.

The problem of fading brightness can be minimized by using a battery that has a capacity well in excess of the minimum needed for a typical usage cycle; this will minimize the voltage drop that will occur between charges. Whether your intentions are routine or emergency uses, good drivers can allow for much more flexible power conservation along with convenient (and efficient) brightness adjustment. In addition, the cost of some decent drivers can pale pretty quickly in comparison to the expense of a few cooked LEDs!

That having been said, this strategy could be useful in some settings. If you can tolerate the brightness swings (and/or minimize them with system design), then this approach could save some money in places where the lights will either be "on" or "off" and where adjustable brightness (at least of the intentional kind) is not needed.

Please watch the numbers!

It is fortunate that the specifications in terms of electrical current and forward voltage are generally consistent within a given LED product line and it is only the lumen output and efficiency that vary as we go from tint to tint and from bin to bin. However, check the numbers carefully for any specific type of LED that you will be using. For instance, the "older" Cree XR-Es have a maximum current limit of 1,000 mA in cool white but only 700 mA in neutral and warm white. At the moment, only cool white XP-G LEDs have received approval from Cree for operation at 1,500 mA. The suggested limit is still 1,000 mA for neutral and warm white.

Cost

At retail from LEDSupply (as of this writing), an adjustable BuckPuck driver will run about $20 and one with a fixed output will be a little less. Cree XP-G LEDs are around $9 apiece (best bin versions pre-mounted on a star board). Small heat sinks are $4 apiece or less. In other words, figure about $60 to brew up a basic skeleton that uses three XP-G LEDs and a BuckPuck driver (I'm figuring three heat sinks). Of course, this does not include components like lenses, housings, power cords, and so forth. As with other home-built projects, value often lies in the eye of the beholder. The other powering alternatives described above can reduce the number of required drivers in cases where the ability to run the LEDs to maximum output is not needed or eliminate them entirely when the consequences of "direct drive" are acceptable.

Delivering light efficiently

Efficiency has clearly been an important topic in this series of articles, but the truth is that we've actually covered only half of that story. Lumen per watt ratings can tell us how efficient a source is at making light. But we also need to consider the critical issue of delivering that light efficiently. We can get a grip on this topic by imagining a bare incandescent light bulb hanging in a rustic cabin with open rafters. Simply replacing this bulb with a good quality compact fluorescent (CFL) bulb could easily increase the lumen per watt efficiency by a factor of three or four or more. But the point we are considering now is that the improvement, as described so far, is not nearly as good as it could be. The CFL, like the incandescent bulb, will emit light in essentially all directions. Much of that light will shine up into the rafters and be wasted.

An easy and effective solution is to mount a suitable reflector over the bulb. This will capture the light that would otherwise be lost and redirect it down to where it will be useful. A reflector is a very good thing in a situation like this, but note: even the shiniest reflector will not be perfect. Even good quality, mirror-smooth, silver reflectors will often reflect only 85 to 90% of the light that hits them. Polished aluminum, stainless steel, or white-painted reflectors can be even less efficient. A related concern is diffusers mounted in front of the light source. These may sometimes be a necessity in order to prevent objectionable glare, but they can also absorb and scatter a significant amount of light and reduce overall efficiency. In fact, even a plain sheet of glass or plastic can absorb around 10% of the light that strikes it and an opaque white diffuser can soak up 20% or more.

It is important to appreciate the significance of these "optical losses," so let's consider an example. I'll note up front that this is oversimplified, but it is also very instructive. Let's say that we have a fixture with a 90% efficient reflector and a 90% efficient diffuser. We are going to put a 1,000 lumen CFL bulb in the fixture and, for simplicity, we will assume that half of the light (500 lumens) will shine directly down to the diffuser while half will shine up to the reflector before getting redirected down to the diffuser. Of the 500 lumens that hits the diffuser directly, 90% will make it out of the fixture (450 lumens). But the 500 lumens which hit the reflector first will fare somewhat worse. The 90% efficient reflector will reflect only 450 of those 500 lumens. On top of this, only 90% of the 450 lumens which survived the reflector will make it through the diffuser, so our net here will be 405 lumens. Of the 1,000 lumens we started with, only 855 will make it out of the fixture (450 "direct" lumens plus 405 "reflected" lumens). In other words, even this optimistic example will eat up almost 15% of the available lumens! A poorer reflector with 85% efficiency and an opaque white diffuser with 75% efficiency would increase the total optical loss to around 30%. If we were starting with a bulb that had an efficiency of 100 lumens per watt, our final fixture efficiencies would be 85 and 70 lumens per watt, respectively.

An immediate lesson to be drawn from this is to be careful about making apples to oranges comparisons when selecting products. The efficiency rating for a bare bulb will not include the optical losses that result when said bulb is put into a separately purchased fixture. But an honest and accurate rating for a pre-made fixture (that includes its own light source) should factor in these additional losses. In the same vein, things like extremely opaque diffusers, tinted lenses, and colored shades should be avoided if your goal is maximum efficiency, and this applies whether you are buying or home-brewing.

This discussion also leads us to seeing a definite advantage of LEDs, at least in some settings. Conventional bulbs and tubes—whether they are incandescent, fluorescent, or metal halide—will emit light in all directions. Using a reflector is far better than nothing in situations where some portion of the light would otherwise be wasted. But, as we have just seen, even the best reflector will not redirect all of that light and some of the lumens will still be lost. In contrast, an LED emits almost all of its light in one direction—out of the face of the chip. In a setting like our rustic cabin example, an LED mounted in the rafters and pointed downward would naturally send the majority of its light into the area where it was needed. In addition, it may be possible in some cases to avoid the use of heavily frosted lenses or diffusers because the LED can be situated to produce minimal glare when viewed from the side.

Now the "aimed" or "directed" lighting provided by an LED (or by a CFL with a very good reflector and no diffuser) will provide the most efficient delivery, but it is also important to consider the other side of this coin before designing a lighting system: a potential problem with such directed lighting is that it can produce very sharp and very dark shadows. In some cases, such as with yard lights or trouble lights, we live with this because the benefit of efficient light delivery is worth the drawback. In other cases, the shadow problem can be reduced or eliminated by using several dimmer sources placed at different locations instead of relying on a single brighter one. LEDs are perfect for this approach, but there are cases where this may not be practical.

To see the benefits of "un-directed" lighting, let's move from our rustic cabin to an off-grid home with finished walls and ceiling. Let's say that we want to light up a pretty decent-sized room from a single location, so we can't use the trick of sprinkling several LEDs around in the ceiling (perhaps we don't have decent access to the attic or maybe we just don't want to fool around with multiple fixtures). If we hang a bare CFL bulb in the middle of the ceiling, it will emit light in all directions. The light that gets thrown up to the ceiling and out to the walls will not get used with the greatest efficiency, but at least some of it will get reflected and bounced back into the living area. This means that any given spot in the room will actually receive light from multiple directions—both directly from the bulb itself and indirectly as a result of light that has been reflected from the walls and ceiling. The net result will be much fainter and softer shadows.

If you've never thought about the significance of this effect before, find a room that is decently lit by only a central ceiling fixture. Sit down with your back to the fixture, put your magazine in your lap, and see if you can still read it! The reason that you are not looking into a pitch dark shadow is because indirect light from the ceiling fixture is reaching the magazine after being reflected from the walls and ceiling. In an area where good quality light is important, the benefits of this type of lighting may be worth the sacrifice. To salvage as much efficiency as possible, there is one fairly obvious thing to keep in mind: white walls and ceilings may not be 100% efficient reflectors, but they are a far cry better than darker alternatives.

Such shadow-free lighting could also be accomplished with a single LED fixture by either buying or brewing something that uses reflectors and/or diffusers to spread the directional light produced by the LEDs. Alternately, one could bounce the LED light off the ceiling instead of aiming it directly down into the room. But in doing so, we are forcing the naturally directional LED to behave like an omni-directional bulb. This can create the same kind of optical losses that result when we force a CFL into behaving as a directional source. The conclusion here is that an LED has a natural advantage with respect to delivery efficiency when directed lighting is desirable, but an "omni-directional" bulb or tube (CFL or otherwise) can enjoy a similar advantage where diffuse, shadow-free lighting is needed. If you are choosing between options that have a roughly similar starting efficiency in terms of raw lumen per watt output, this is something to consider.

Of course, this ignores other potentially significant pros and cons (such as the efficient dim-ability of LEDs or the cheaper current purchase price of CFLs). When you start looking at the actual products, it is important to consider a large number of factors to find the best options for your needs. To recap what we've covered, these include total lumen output, lumen per watt efficiency, service life, purchase price, the "natural directionality" of the source, and the availability of efficiently adjustable brightness. All of these need to be evaluated and dovetailed to fit with your needs. There are no perfect or easy answers here, but I hope that this background information will be of some help.