LED Flashlight White Paper

By Henry Schneiker
www.HdsSystems.com
September 17, 2004, updated August 2011
Copyright 2004 - 2011 - all rights reserved

A flashlight (headlamp, torch) designed around an LED can offer significant advantages
over the same flashlight designed around an incandescent light bulb or other light
emitting device. In this paper, I will cover the luminous, electrical, thermal, optical ,
mechanical and interface aspects of flashlight design as they pertain to LEDs as well as
how the human visual system adapts to different lighting conditions. Although certain
aspects of this paper are somewhat technical, even the non-technical person will be able
to gain a better understanding of the issues and trade-offs involved in designing an LED
flashlight.

By necessity, this paper talks about specific devices. However, I do not mention any
manufacturers or part numbers because that information is not important to the points
being discussed. As with any project, you should fit your design to the actual parts
being used - taking advantage of their capabilities and being careful of their limitations.

The LED

LED stands for Light Emitting Diode. A diode is a device that passes current in the
forward direction while blocking it in the reverse direction. When you pass current
through an LED in the forward direction, the LED emits light. The color of the emitted
light depends on the material and construction of the diode. The amount of light varies
roughly linearly with the amount of current passing through the LED, which provides
us a convenient way to efficiently change the light output.

LEDs are solid state devices and therefore quite reliable. Mean time to failure is often
listed at over 100,000 hours of operation under specified conditions. LEDs are far more
reliable than the ubiquitous incandescent light that you normally find in a flashlight.
LEDs can also tolerate much higher shock and vibration loads than other common light
sources including incandescent lights, florescent lights and other arc lights.

But as good as LEDs are, they not perfect.

The most significant issue from a production point of view is the variability of LEDs.
When LEDs are manufactured, considerable effort is made to get all the LEDs to come
out the same. However, even on the same wafer, there is tremendous variation from
LED to LED. The variations are so significant that the manufacture must test all the
LEDs and sort them into bins. Common bin categories include color tint, light output
(flux) and forward voltage.

So to get consistent production results all you have to do is specify the bin you want
when you order the LEDs, right? Wrong. To illustrate why this is so, let’s take a closer
look at what the bin ranks actually mean.

If you take a color bin for a white LED and convert the CIE xy coordinates of the bin
corners to CIE L*a*b* values - where each whole number is approximately one
distinguishable shade of color or brightness - even a small bin can have 13
distinguishable shades of color diagonally across the bin. A larger bin can easily have
30 or more shades of distinguishable colors diagonally across the bin. This means that if
you take two LEDs from opposite diagonal corners of the same bin and view them side
by side, you will see that the white light being emitted from each LED is of a different
shade of white. Further, people often find some shades of white more pleasing than
other shades, with shades closest to the black-body temperature curve being the most
pleasing. And people also have preferences as to color temperature - some preferring
cooler shades (more blue) and some people preferring warmer shades (more red).

A similar situation results when you look at the light output at a specific current (the
flux) and the forward voltage. A typical bin can cover a flux range of n to 1.3n and a
forward voltage range of v to 1.1v. When combined in the worst case scenario, this can
produce a 40% difference in overall efficiency within the same bin. This difference can
be seen easily if you compare the two lights side by side. Now add to this the
measurement error tolerance of 10% on both parameters and you get a 51% difference
in possible overall efficiency within the same bin. Talk about a quality control
nightmare.

If a customer compares two lights and sees an easily distinguishable difference between
them, the customer will generally assume that one of the two lights is defective.
Educating the customer to the realities of production variations only goes so far.
Significant variations will often result in customer dissatisfaction and returns.

There is, of course, a statistical side of this. The actual distribution within the bin is
statistical in nature and the probability of any two lights exhibiting a sufficient
difference that the customer will notice or be concerned about the difference is also
statistical.
The unit-to-unit variation is much higher in LEDs than in most other light sources.
Some of these differences can be reduced through a calibration process. However, it
might be easier if you have multiple product lines and carefully sort the LEDs for each
model - thus reducing the intra-model variations.

LEDs have another problem that is worth noting. LEDs become less efficient as you
increase current and temperature. And since LEDs have a negative temperature
coefficient, the hotter they get, the more current they want to draw. This makes them
unstable and prone to thermal run-away unless careful attention is paid to how they are
driven. Further, high temperatures will increase the rate of permanent lumen (light
output) loss beyond the typical 8% per 1000 hours.

When I talk about high power LEDs, I am referring to LEDs that are designed to handle
one or more watts of continuous power per LED. However, the same principles can
often be applied to the low power LEDs or arrays of low power LEDs. The high power
LEDs typically provide a fairly low thermal impedance which can be coupled to an
effective heat sink to minimize the LED’s junction temperature. Every effort should be
made to keep the LED cool for maximum efficiency. If the LEDs will be driven at high
power levels under conditions where effective heat dissipation is not possible, the LEDs
must be thermally protected to ensure reliable operation.

And finally, when you dim an LED by reducing the current, the LED can and often does
undergo a significant color shift. The severity of the color shift is dependent on the
individual LED. This effect is most noticeable to a human observer with white LEDs.
What happens is that as the current drops, the dominant frequency also drops. The
amount of frequency shift is quite variable from one LED to the next and a
manufacturer seldom tests for it. With white LEDs, the dominant frequency (blue) is
used to excite a phosphor that re-emits at the lower end of the spectrum to provide
yellow, orange and red. As the frequency shifts down, the blue dominant frequency
shifts toward green making less of the required blue frequency available for down-
conversion to yellow, orange and red and so the color shifts toward green. The only
way to reliably prevent this is to always drive the LED at recommended power levels
while dimming.

The Power Supply

The purpose of the power supply is to supply and regulate power to the LED and
ensure a stable operating environment. The power supply can vary in sophistication
from a voltage source in series with a resistor - such as a battery with or without an
additional resistor - to a fully regulated constant power system with thermal regulation.
The characteristics of the power supply directly affects cost and the stability of the light
output.

LEDs have a steep Vf/If curve. This means that a slight change in forward voltage (Vf)
will have a relatively large affect on the forward current (If). Further, this curve is
different for each LED. To drive this point home, if you assume you have LEDs that all
come from the same bin and the bin provides a Vf tolerance of 10%, the power
consumed by LEDs in that bin can vary from w to 2w if powered with a constant
voltage.

As I mentioned earlier, LEDs have a negative thermal coefficient. This means that Vf
decreases with increasing temperature. With a constant voltage power supply,
increasing the temperature slightly decreases the Vf and increases the current by a
relatively large amount - causing a net increase in power and a further increase in
temperature. For any given thermal path, there is a power level at which the heat
transfer can no long keep up the the increasing power and falling Vf. Thermal runaway
is the result and it will destroy the LED if nothing limits the power to safe levels.

Regulating the current or power through an LED effectively prevents thermal

runaway. Regulating power through an LED has the added benefit that it tends to
provide better brightness regulation over a wide temperature range. A good
regulating circuit will keep the flashlight brightness the same throughout the life of the
battery. Both systems can include thermal regulation to prevent dangerous heat build-
up if the thermal path is inadequate at a high power setting.

A direct drive flashlight - i.e., batteries in series with an LED, with or without an
additional resistor - cannot provide a constant output brightness. The brightness/time
graph of this class of flashlight produces a smoothly descending curve. This is because
the battery voltage sags quickly and thus the power through the LED drops rapidly.
Suffice it to say that in the first quarter of the battery life, the brightness drops
significantly. Battery life in this case is often defined as the amount of time it takes for
the brightness to drop to 25% of the original value with fresh batteries. Another
problem with direct drive flashlights is that power through the LED can be at
dangerously high levels until the internal resistance of the battery increases sufficiently
to limit the power to safer levels. As a result, direct drive flashlights are usually less
reliable due to a higher LED failure rate.

The efficiency of a system is controlled by the efficiency of all the parts. In a typical LED
flashlight, you have a battery, a power supply (or regulator) and the LED. The
efficiency of the battery drops as the power increases. Most power supplies and LEDs
have an optimum point where they are most efficient and are less efficient at higher or
lower power settings. This optimum point is often around 10 to 20% of rated power.
In general, the increase in battery efficiency at low power settings more than offsets
any decrease in power supply and LED efficiency at low power settings. The net result
is that there is little or no net decrease in total efficiency at typical low power settings.

At the higher power settings, a compounding affect can be observed that decreases
total efficiency in a non-linear fashion - often faster than N squared. The primary cause
of this rapid drop in efficiency is the I squared R (current squared times the resistance)
losses found throughout the circuit - especially where high currents are found. The next
biggest loss in a switching power supply is usually coil saturation. And finally, as the
LED junction temperature and current density goes up, the LED efficiency drops.

Batteries have a large affect on a flashlight’s design because they supply the power.
Alkaline batteries develop a high internal resistance as they are used up. Therefore,
alkaline batteries are not well suited to high current applications because large amounts
of their power will be lost due to the internal resistance - dissipated as heat. Further,
alkaline batteries stop producing useful amounts of power by the time the temperature
drops below freezing, making them inappropriate for cold environments. Lithium
batteries (in many different chemistries) have a fairly low internal resistance
throughout their life and work at temperatures well below freezing. The economics of
lithium batteries has changed to the point that they are now similar in cost to alkaline
batteries, making them the first choice in battery power for most flashlight applications.

You must be careful when using rechargeable batteries not to over-discharge them.
Over-discharging rechargeable batteries can damage or destroy them. When a
regulating switching power supply is used, the battery can be over-discharged before
the light output drops by a noticeable amount. A flashlight should detect rechargeable
batteries and prevent them from being over-discharged. Further, lithium rechargeable
batteries often contain a special protection circuit that turns off the battery to prevent
accidental over-discharge - causing instant darkness. Once the protection circuit turns
off the battery, only recharging the battery will turn the battery back on. A flashlight
should detect a lithium rechargeable battery and avoid the conditions what will activate
the protection circuit and suddenly leave you in the dark.

As your batteries nears the end of their life, a flashlight should signal you and reduce
the brightness gracefully. You should be notified that the batteries are nearing the end
of their life so you can find a safe place to change your batteries. If you cannot change
the batteries, the light should continue to provide light for as long as possible - long
enough for you to return to safety - and fresh batteries. Even if that takes you a day or
two. Once you know the batteries are at the end of their life, you should be able to
manually reduce the brightness further to make the remaining battery power last even
longer.

The Optics

The LED chip emits light in all directions. An optical system is used to gather the light
and shape it into a beam. The LED itself has a built-in lens. This lens is limited in its
ability to gather and shape the light into an acceptable beam and the LED manufactures
only offer a limited number of lens options. An optical system consisting of just the
built-in lens in the LED may produce acceptable results for inexpensive flashlights, but
for higher quality results, a more advanced optical system is required.

A parabolic reflector can theoretically generate a beam of perfectly parallel light rays
the same size as the exit aperture (the large open end of the parabola) given a point
source of light at the focal point. Some percentage of the light from the point source
will not be incorporated into the beam because it never contacts the reflector surface.
The solid angle that intersects the parabola relative to total spherical solid angle is used
to define the reflector capture efficiency - the smaller the exit aperture solid angle, the
higher the capture efficiency and the brighter the beam - all other things being equal.

In a real reflector you have imperfections that tend to scatter the light beam. First, your
light source is never a point source - it is always a complex 3-dimensional surface.
Second, the reflector’s surface is never perfectly smooth or exactly the correct shape.
This leads to angular errors that tend to spread the beam. By designing in and
controlling the angular errors you can shape the light distribution (beam pattern) to suit
your purposes. Reflectors tend to provide the highest quality beams and the highest
overall efficiency.

A transmissive lens, such as a common magnifying glass, uses refraction to achieve a

similar result. Instead of light being reflected off a surface, light is bent at the air/lens
interface. Transmissive lenses have three main problems. First, they have a relatively
low capture efficiency - any light that falls outside of the aperture is lost. Second, as the
angle between the lens surface and the light source deviates from perpendicular, an
increasing amount of light is reflected off the surface and lost instead of being bent into
the lens. Third, the lens projects an image of the emitting surface - as a result, the focal
point is normally chosen to be different from the emitting surface so the image is
blurred.
There are also compound reflector-refractor such as Total Internal Reflection (TIR)
optics that combine properties of both systems. These optics often use the high
reflectivity of the air/optic interface to form the reflector surface. These optics tend to
be more efficient and have better beam quality than a pure transmissive lens but are
not as good as pure reflectors. However, these optics have the advantage that they
tend to be more compact than reflectors.

An interesting property of lenses is that they are all about angles. Since the same
angular error will produce the same result regardless of size, the only limits to small
size are the emitter size and surface tolerances. As a result, a smaller emitting surface
can use a smaller optical system and produce the same beam pattern. Or, you can
increase the size of the optical system to reduce the angular errors and improve the
beam pattern definition. The variations are endless.

The Human Visual System

If you look at typical flashlight marketing literature, you might think that brighter is
always better. But allow me show you why always brighter is not always better.

Your eyes can adapt to a wide range of light intensities. At noon on a summer day the
sun can illuminate a surface to 120,000 lux while an overhead full moon might
illuminate the same surface to only 0.1 lux. Your eyes use three main methods when
adapting to changing light levels. Under bright conditions, your eyes exclude excess
light by closing the pupils. As the light level falls below the point where the pupils are
fully open, the eyes undergo a chemical change to increase the cone (color) cell
sensitivity. Finally, as the light level falls below the capability of the cone cells, the rods
(no color) cells take over and become maximally sensitive. It takes roughly 20 minutes
after you leave a brightly illuminated area to become fully dark adapted.

So how much light do you actually need to perform a task? 1000 to 2000 lux is
recommended for fine detail work where maximum visual acuity and color recognition
is required. 100 to 200 lux is appropriate for most office work. By 10 lux, colors are less
vibrant while visual acuity is good. By 1 lux, unsaturated colors cannot be distinguished
and visual acuity is only acceptable. By 0.1 lux, no colors are visible and visual acuity is
poor. By 0.01 lux, visual acuity is lousy and you can see objects better if you don’t look
directly at them.

The forgoing illustrates that providing 50 to 200 lux of illumination is plenty for most
tasks that require good color recognition and good visual acuity. Much less light is
needed for tasks that only require object shape identification.
It takes one lumen - that’s one candela per steradian - to illuminate one square meter
(about 10 square feet) with 1 lux. Note that luminous flux - in lumens - is mapped
against the color sensitivity curve of the human eye. However lux can also be defined
in terms of radiometric flux at a single frequency and is equivalent to 1.46mW of radiant
electromagnetic power per square meter at 555nm (the color to which you are most
sensitive - a shade of green).

The amount of light illuminating an object depends on how bright the light is and the
distance the light is from the object. If you know the amount of light illuminating an
object at one distance, you can calculate what the amount of light illuminating an object
will be at another distance using the inverse square law. For example, if you move the
light twice the distance from an object, the illumination drops to one quarter (2 squared
is 4 and the inverse of 4 is 0.25 or one quarter). As another example, if you have a light
that generates 500 lux at 1 meter (just over 3 feet), it will generate 100 lux at 2.2 meters
(about 7 feet) or 10 lux at 7.1 meters (about 23 feet) or 1 lux at 22.4 meters (about 73
feet). The drop in illumination corresponds to an increase in the illuminated surface
area. Put another way, you need 4 times the light to see twice as far with the same
illumination level.

The eye responds to light in a logarithmic manner. That is, for the eye to indicate a
significant increase in brightness - i.e., an easily recognizable increase in brightness - the
amount of light must double. For dealing with the large variations of light in your
environment, this is a wonderful thing. For flashlights trying to get ever brighter, this
is a terrible thing because it takes over twice as much power to generate twice as much
light while the battery life drops to less than half - due to the escalating inefficiencies
discussed earlier - for a single incremental increase in brightness. However, if you want
to maximize battery life by using lower brightness settings this is a great thing. Why?
For every incremental decrease in brightness, you more than double the battery life
and after a few minutes, your eyes adapt to the new brightness level so the apparent
drop in brightness actually decreases. By reducing the amount of light to match your
task, you can dramatically improve the battery life.

The eye’s logarithmic response to light suggests that brightness levels should be spaced
in a logarithmic way. That is, the steps between brightness settings should be a
constant factor - such as 2x or 1.4x - instead of using a more linear approach. This
provides the appearance of the steps being equally spaced. This same method is used in
the audio industry for volume controls. Because of the eye’s adaptive characteristics, a
continuous control is inappropriate for a flashlight. The reason is that you are quick to
turn up the light but slow to turn it back down. In fact, if you are asked to turn up the
light for a minute and then turn it back down to where it started, the resulting
brightness is usually brighter than before starting the exercise. Human factors should
never be forgotten when designing control systems. Thus, s stepped approach is the
preferred method.

Taking all of the above together suggests that we can use lower levels of light for many
tasks and gain benefits in the process. Not only do you increase battery life with lower
output levels, you will be able to see further when you really need to and you reduce
your visibility footprint. Here is the scenario. Imagine you are using a low light level
to navigate - such as following a trail in the mountains at night - which you can easily
do on less than 1 lumen. Also imagine you can get maximum output with a simple
press of the button. After hiking for a while you come to a place where you need to
look far ahead. You hold your light above your head, point your light to where you
want to see and then press the button for maximum light output. Because your eyes
are dark adapted, you will be able to see much further with the maximum amount of
light than if you had been following the trail with a brighter light. Depending on
circumstances, you may be able to see twice as far on the same amount of light. That is
a huge increase in distance without buying a bigger flashlight.

For the tactical crowd, select 0.1 lumens to sneak up on your enemy. The light footprint
is so low that they will not see you coming. Upon arrival, a quick press of the button
not only blinds your enemy, but it gives you the tactical advantage because the enemy
is also brightly lit. Releasing the button allows you to disappear back into the darkness
with enough light to move undetected.

Putting all the light in a narrow beam is great for looking at an object at a great
distance. However, the contrast from the bright beam center to the dark area just
outside the beam is so great that you will have difficulty seeing objects just outside the
beam - to the extent that trying to walk using such a beam can be difficult or
dangerous. A general purpose light should provide enough light outside the main
beam so your eyes can move comfortably between the beam center and the outer
areas and take in a large viewing area. Better still is to provide a smooth transition
between the two areas. The broader the transition zone, the lower the contrast and the
safer you will be when using the light to navigate rough terrain. You will find that you
can use less total light with a smooth broad beam. If fact, you might want to carry two
lights - one with a broad beam for walking and general use and another with a narrow
beam for spotting distant objects.
User Interface

As you may have gathered from the previous discussion, there is a lot to be gained by
providing a flashlight with more than an on/off switch. Adding a brightness control
extends the utility. Having the flashlight automatically detect and compensate for
different battery systems allows operational flexibility. And there are always a few
extra features that would be nice to have, such as automatic emergency signaling, a
way to find the flashlight in the dark or a way to prevent accidental turn-on.

As soon as you start adding features to the flashlight, you increase the complexity of
the user interface - not to mention the design and cost. Let’s take adding a brightness
control as an example. Will there be a single button (click codes), two buttons
(brighter/dimmer), n buttons (one for each brightness), linear slider or a rotary switch
(volume control)? Is the brightness control integrated with the on/off switch or is it
separate. Do you have to worry about making the controls water proof or making
them work after being abused? How many brightness levels will there be? Will the
user have access to all brightness levels all the time? Will the brightness control be
continuous or a step function? How will brightness levels be spaced. Are there other
features that will need controls and how will they interact with the brightness control?

And how will the flashlight’s electronics detect the user’s input? On/off switch?
Change in resistance? Change in capacitance? Change in inductance? Mechanical
coupling? Optical coupling? Magnetic coupling? Digital or analog? The possibilities
seem endless.

While you are thinking about all this, you should never forget that the main purpose of
a flashlight is to turn on and generate light. So this function should remain as simple to
use as any other flashlight. Features that are used the most often should be the
simplest to use - a user should never have to scroll through a list of brightness settings
to get to his favorite brightness setting.

A microcomputer can be used to help implement the user interface, whether it involves
a single control or multiple controls. The microcomputer is also handy in implementing
automatic functions and performing other tasks. Having the intelligence in software
instead of in hardware provides us with the opportunity to tailor various characteristics
and provide automatic controls that are not easily done in hardware. Further, it allows
us to add new features with relative ease.
Summary

Designing a flashlight around an LED allows us to design a very rugged and

sophisticated light. A typical laundry list of desired features includes small size, high
maximum brightness, multiple brightness presets, rotary brightness control, constant
brightness regulation, thermal regulation, a smoothly tapered beam pattern, support
for multiple battery chemistries and rechargeable batteries, long battery life, battery
end-of-life warning, simple controls, automatic emergency signal, switch lock-out,
automatic turn off after some time, a find-me-in-the-dark capability and the ability to
customize these things. These features can now be found in a single compact flashlight.

HDS Systems designs advanced flashlights and incorporates the superior capabilities
discussed in this paper. The latest flashlights can be seen at www.HDSsystems.com .