What you should know about HBLEDs – and nobody will tell you – Part 2
In Part 1, "A Brief History of Light", Ed established a working definition of the lighting market and began to address some of the misconceptions that even engineers have about HBLEDs. In the process, he identified the market segments where HBLEDs were already enjoying widespread use, as well as those where HBLEDs could be competitive as the technology matured. Part 1 concluded with a quick recap of lighting technology from the Edison bulb to modern-day solid-state devices. In part 2, we’ll take a deeper dive into what makes LEDs work, how they’ve evolved, and some of the design challenges modern HBLEDs present to the unwary engineer.
Part 2- Light from Chips
The concept behind the LED was known early in the 20th century and it is ironic that the first light-emitting structures were created with silicon carbide, the same material that in the late 1990’s became one of the "new" materials making possible some of the first new blue LEDs. Work was done on light emission with silicon carbide in the late 50’s by Dr. Carl Acardo at MIT. However, materials such as Gallium Arsenide, Gallium Nitride and, very recently, Indium Gallium Nitride, took over and continued to the present for infrared and visible emitters.
In the simplest manner, we can say the LED is a signal diode which emits light at the junction of its positive and negative layers ( the "junction" being analogous to the boundary between the two metallic contacts in a switch), in proportion to the current passing through that junction. It differs from the incandescent and gaseous discharge lamps in that it has no filament or electrodes to wear out but, like the other two, it does indeed need to factor in temperature. It must be treated, in almost all respects, just like any other diode in terms of heat removal, voltage rating, the mathematics of its operating characteristics and assembly/packaging techniques.
Unlike the other light sources, the LED chip is monochromatic; i.e., it emits light at only one wavelength. Putting a colored lens in front of it will not provide another color as with an incandescent lamp. To get something that looks more like daylight, one must mix light from multiple LED’s. Alternatively, a UV- or blue-emitting chip can b coated with a special yellow phosphor. The result is what is called ‘secondary emission" as the phosphor emits a multi-wavelength light when stimulated with the LED’s single wavelength. Because it costs much less than packaging several LEDs together, virtually all so-called "white" LEDs use phosphor-based secondary-emission technology using blue or ultraviolet chips.
(It should be noted that research devices have been recently fabricated whereby multiple regions in the same chip emit different wavelengths, thereby creating white light without phosphors. While initially very compelling from a scientific standpoint, these devices have not yet proven to be commercially viable—a common outcome for 99% of R and D laboratory "breakthroughs")
It’s interesting to note that this idea of secondary emission is exactly how a fluorescent lamp has operated all these years. The arc inside the fluorescent tube generates ultraviolet light, which excites the phosphor coating on the inside of the tube to produce secondary emissions in the visible spectrum. The various fluorescent lamps called "cool white" or "warm white" etc’ simply mean a different mix of phosphors. The chemistry of achieving the proper phosphor secondary emission characteristics reflects serious science. That’s why white LED manufacturers are also obliged to become masters of the phosphors.
But there are additional complications in making a properly-functioning HBLED. Managing the light, after it has left the microscopic chip junction is one of the largest challenges of the HBLED maker. Many problems arise because light traveling from the quantum junction of that tiny chip (which is typically no larger than .040" X .040") must travel through the chip material itself and then through the phosphor coating and then through additional clear encapsulating material to the viewer.
Because different wavelengths are refracted differently as they pass through different materials, a prism effect can occur. This means that the beam of light from of a white HBLED can have "fringes" of light that may not be the same color as in the center of the beam. This is caused by slight frequency shifting of the original light as it refracts (bends) while passing from one material into another.
Without getting into the detail, let’s just say that two of the major challenges to makers of white LEDs today is in a) trying to improve what is called light extraction and b) creating phosphors which more efficiently create secondary emission and are much less subject to degradation with elevated temperature.
Intensity vs. Illumination
One of the most misunderstood, or at least misused, aspects of light source comparison is the meaning of lumens per watt and its application to HBLEDs versus other sources. The confusion arises from the different ways that light from each source is measured.
When we talk about a 60 watt incandescent lamp having a 750 lumen output, we mean all the light emitted as might be collectively measured by sensors at an infinite number of locations inside of a sphere, with the bulb suspended in the center.
On the other hand, all HBLED light output for indicators has historically been specified based on its perceived intensity as seen from a distance and within a certain "viewing angle". Indeed, the HBLED, especially if it has the right lens or plastic package dome, can "seemingly" have more lumens per watt as seen from a point in the distance, but overall, the total lumen output, if measured in that sphere, would be a fraction of the incandescent lamp’s light.
One way to bring home the point is to compare the concepts of illumination and intensity. If we stare at a one-watt white HBLED, having a focusing lens, as in an LED penlight, from 10 feet away, we will be blinded by the intensity of the "spot". ( high lumens-per-watt as measured by our eye) but the rest of the room will be dark. If we stare at a 40-watt bulb from the same distance, our eye will observe dramatically less intensity (much lower lumens per watt as observed by our eye) and we can stare at the bulb for a good while and in fact the whole room will be illuminated. .Since very few people have actually participated in such a simple experiment, the true meaning of lumens/per watt versus intensity has escaped them.
A Simple Experiment
It’s nice to talk theory, but a little practical experience goes a long way to getting a real feel for a technology of "directed light". So, if you are more of the empirical type who prefers direct experience, you can try this simple experiment:
Take some operating incandescent lamps rated at 7, 11, 25, 60 and 150 watts and put them 15 feet from daylight-sensing photo detector or regular light meter (a calibrated equivalent of your eye). Also 15 feet away, place a flashlight, a 60-degree angle white HBLED (not collimated or reflectorized), a 5 watt MR11 halogen lamp, and an HBLED with 10-degree lens.
If you don’t have the time to run the experiment yourself, here’s roughly what the photo detector will register in terms of brightness:
Light Source Relative brightness
–100 lumen per watt/5000K HBLED with no lens at (300mA/one watt),
typical 120 degree beam angle 1
–Small high intensity flashlight (reflector removed)-3 watts 1
–11 watt incandescent bulb 4
–25 watt incandescent bulb 6
— 100 lumen per watt/5000 K HBLED (at 350 mA),
with a 60 degree collimating lens 12
–5 watt MR 11 halogen lamp (integral reflector) 12
–60 watt incandescent bulb 13
–100 watt incandescent bulb 18
–Same single HBLED (at 350mA/one watt) with
10 degree collimator 100
–Regular 2 D cell flashlight 100
–Small high intensity incandescent flashlight -3 watts
(set to narrow beam) 100
Besides the relative brightness of each source, notice that the 60 and 100 watt bare bulbs are easily seen to light up the room, while the two flashlights and the collimated HBLED exhibit proportionately bright circles on the wall with the rest of the room being nearly dark. It becomes obvious that reflectors and collimators, or even just the domed lens of the LED itself, can provide a dramatically greater perception of intensity (or brightness of light on a target) over a bare bulb. Put another way, a one watt 100-lumen 5000K LED, with a 10 degree lens can put 300 times the light on a target as a 60 watt incandescent bulb !
What’s happening here is that we are seeing two effects. When light is collimated, you perceive a very bright "indicator" with very high "intensity" and, when light is emitted in a near-spherical pattern, you perceive it as a bright "illuminator". Understanding this differentiation is the first step in intelligently and creatively applying the new LED technologies. A LED reading light, such as those used in aircraft cabins, where almost 100% of the light is used only where wanted, can indeed be more energy efficient than an incandescent source lighting the same page. It is not that the LED is only more efficient but rather that the more productive manner of directing the light creates the legitimate equivalent of more energy efficiency.
This lesson illustrates how much easier LEDs are to collimate with lenses or reflectors than either fluorescent or HID technologies. This gives HBLEDs a significant edge in applications where light must be directed toward a given area, as with track lighting for supermarkets, mall retailers , museums etc…
LED Industry Evolution
In the late 60’s and early 70’s, technologists were focused on developing infrared LED emitters for optocouplers and IR sensing in general (as in motion detectors), as well as visible-red indicators for simple applications.
In the 80’s, development of green, amber, and other colors beside red gave visible emitters a little more attention, but it was not until blue LED’s became available in the mid 90’s that the press-release machinery went into high gear. White LED’s became feasible with what is called the RGB (red/blue/green) LED. With any RGB device, red, blue and a green emitting chips were combined to generate white light. The RGB LED today is a major contributor to high-end color displays of all kinds, including TV.
With those developments, the equipment industry could replace just about any incandescent panel lamp with an LED and the days of the colored lens were ending. Today, the LED dominates the panel indicator market. The LED attributes of compatibility, longevity, immunity to vibration are impossible to beat for an incandescent.
The progress of HBLEDs is not the result of some single breakthrough but rather numerous incremental improvements and parallel market forces.. Once the white RGB LED was introduced, there was immediately attention given to possibilities other than simple white light panel indicators or simple color mixing. These possibilities demand not just indication but high brightness so as to be seen from a good distance.
The result has been the refinement of substrate materials, phosphors for secondary emission and recognition that both and wafer chip sizes needed to get larger and assembly methods needed to change for the first time in over 30 years.
it’s only since 2010 that we finally begin to see decisions being made by LED makers to invest in equipment to make much larger wafers (up to 8 inches in diameter), a decision necessary for any meaningful ultimate price reductions. While efficacies have steadily improved since 1999, pricing for a given size of chip has changed very little until 2013, after which there was a 2: 1 reduction(sometimes more) in chip pricing.
It seems that a tipping point, driven by this very significant price drop, was reached. This price reduction changed the design equation for great many commercial LED end products. In other words, LED pricing had always been a driving factor for luminaire design but now the priority swung over to heat exchanger, power supply and housing costs and weight. As the industry entered 2014, it became apparent that the these developments were having a profound impact on the industry and the markets it serves.
One immediate result of this rapid shift in priorities is that a great many products are now being viewed as "first generation", are being obsoleted by those companies which are faster on their feet and capable of taking better advantage of changing cost structures. We can look to the recent past and see a similar scenario unfold in the early computer industry when there was great emphasis on computer hardware technology and cost issues. Once computer hardware experienced the the cost reductions associated with the "commoditized" standard products and high volume production, the pendulum swung to software. Some companies saw the wring on the wall, but others did not…
It is useful to make one distinction here between the IC industry and the LED industry. With computer chips getting larger and larger—more than a quarter inch on a side, large wafers provide the only way to obtain adequate yields and cost effectiveness. Presently there is virtually no benefit to make LEDS as such a large chip. Very large LED chips actually tend to have worse light extraction and efficacy (i.e lumens per watt.) and are more costly/tedious to collimate.
It can be noted that in 2014 , fewer than 2% of white LED products employed LED chips much larger than 1.0 X 1.0 mm. What this means is that large LED wafers will indeed bring down costs but not to a "cost per PN junction degree seen with computer chips unless or until LED makers can better address the "light extraction" and "spreading resistance" issues.
In 2010, two events triggered a change in industry thinking. First, Bridgelux and Philips announced 6-inch wafer capability, something said to be almost impossible for LED type wafers just a few years before,
Secondly, over a dozen firms, most notably Bridgelux, Cree, Philips, Nichia and Citizen, have begun mass producing higher power multichip LED arrays, referred to as chip on board (COB) and adoption of so-called called "remote phosphor" technology. Remote phosphors simplify manufacturing by replacing the individually-coated LEDs with an array that’s covered with a single phosphor-doped lens cover or conformal coating.
These improvements have helped eliminate the whole "binning" and CCT inconsistency present with individual LEDs. Furthermore, it means unsophisticated light fixture makers can spec and buy a fully-assembled COB array almost as easily as they would a light bulb. So, while the LED cost remains the same, the additional labor, testing, cost and headache of assembly can be sharply reduced.
The fact remains that one can buy a packaged IC chip with a 1mm X 1 mm size, having over one thousand transistors elements at less than 25 cents while an equivalent size LED, until recently cost 6-8 times that – even in high production quantities. For LED lighting to go truly mainstream, manufacturers needed to achieve least a 3:1 cost reduction. Much of this has been accomplished since 2013, as processes, materials, and larger substrates became available which allow high efficacy LEDs to be produced using the same automated fabrication equipment used by low -complexity IC makers.
For over 30 years, LEDs found applications as indicators and couplers, tasks
which used 20-30 milliamps of current and required little thought about thermal resistance or heat sinking. Likewise, the "ballast" used to regulate the current flowing through the LED could be implemented using by a simple resistor in these low-power applications. HBLEDs, on the other hand, need special circuits to maintain the proper current if all of the efficiency advantages of the LED is to be realized. IC makers have stepped up to the plate with current-controlled driver chips.
The HBLED market is also maturing as lens and reflector companies introduce new products to enhance the LED light output effectiveness. Meanwhile, heat sink manufacturers are also rolling out their specialized products for thermal management solutions that deal with the unique thermal issues that arise in high-powered LEDs. We’ll take a closer look at this in the next section.
So while LED development used to mean just a few new IR or visible emitter chips here there and a few new packages here and there, the HBLED chip itself is now just one piece of the product puzzle just as the maker of a microprocessor must become expert in the ways of software and the nuances of the end market.
Thermal Management & Other "Hot" Issues
It was earlier noted that most LED chips have been no larger than 15 mils on a side (the smaller the better in terms of manufacturing cost). At 20-30 milliamps this is of no consequence, but at 350-700 milliamps (typical HBLED operating levels for an individual chip), it is a big deal. If that heat is not kept to below certain levels, the chip will be destroyed or will certainly cease to function as desired. When that happens, the chip ceases to be a "semi" conductor and becomes simply a "conductor"
It also turns out that the LED chips themselves are quite robust and typically can function safely (disregarding light output) well up to and even 150C. This would be fine except that once the phosphor coatings currently in use are heated much above 125C, there is such a reduction in the performance that it renders the white light emitting coated chip essentially useless for commercial applications.
Helping LEDs maintain their cool begins with understanding that all semiconductors exhibit what is called "thermal resistance" considerations, (i.e. how hard it is to transport heat from its internal light emitting "junction"), usually specified in degrees C per Watt. In other words, it’s a measure of how much the chip’s internal temperature will rise above the mounting surface temperature for every watt of power in the device. And once heat from the chip gets to the case or other mounting surface, it then has to travel further to a heat sink or before it is finally transferred to the surrounding air.
There are three things that are critical to keeping an LED junction and its phosphor coating from "melting down". The ability of the heat to get out of the chip at this first barrier is determined significantly by:
a) how far the heat has to travel (even a few thousandths of an inch from the bottom of a chip to a copper or aluminum surface is not a good thing)
b) the heat conduction properties of the LED material itself and
c) how large the chip is (the larger the surface touching the metallic mounting surface of the case, the more easily heat can transfer just as a large window in a house lets more heat escape than a small window).
HBLED makers strive, by various new techniques, to have the dissipative PN junction as close to the case or substrate as possible. This has not been an easy job due to the many tradeoffs and processing limitations involved, but there have been substantial improvements over conventional, historical, lower current LEDs. Manufacturers also face the material challenge, but not much can be done here.
Materials such as Indium Gallium Nitride are substantially worse than silicon as heat conductors so that built-in penalty will exist for the foreseeable future.
The LEDs typically used as panel indicators and optocouplers are rarely more than 15 mils (.015") on a side and have very high thermal resistances, well over 150 degrees C per Watt. This prevents them from being driven with much more than 30-40 mA. To operate above 300 milliamps (and as much as an ampere), HBLED makers have had to increase the area of the chip by more than a factor of 10 and relocate the actual junction location closer to where the chip meets the heat sink—so as to drop the thermal resistance by 12:1 or more
Today’s chips, which (aside from specialty items) can be as large as .050 X .050", have essentially hit the wall in terms of thermal resistance and power capability for a single device. Thermal resistances of 4-5 degrees C per watt are about the best which can be achieved today for the raw die, plus another 2-3 degrees C per Watt between the package and its heat sink.
There is an unfortunate (and common) perception in the HBLED industry that one can simply add a finned heat sink and the sum of the areas of the fins will act to greatly magnify the cooling ability. That is, it is assumed that a finned heat sink with a total fin area of 100 square inches can cool an LED array as well as a flat piece of aluminum also having an area of 100 square inches.
Nothing is further from the truth, but the myth lives on.
The reality behind the myth is that unless the fins are exposed to moving air, the finned heat sink is only marginally better (30-35% better in the most optimistic possible situation) than a block of metal of the same volume or a simple, six-sided folded sheet metal box having the same outside dimensions as the heat sink
CFM is like voltage and LFM is like current .Every electrical engineer knows it is current which heats an incandescent lamp filament, not voltage. A fixed voltage is there but it is what we do with it to create some level of current is what determines the exhibited result.
And it’s also important to remember that a heat sink with very tightly spaced fins, even with air flow, will have little effectiveness unless the air is forced through the fins…..a task easier said than done. It is a common mistake by design engineers to think that packing more fins will automatically improve cooling. The Internet is filled with ads for high density finned heat sinks which are of little value because it is almost impossible to create adequate air speed through the fins without ducting or special plenums to channel air flow.
There is a "pivot point" beyond which decreased fin spacing actually can increase back pressure, retard air flow and degrade cooling. Because college thermodynamics courses do not get into these real-world things, engineers have to learn by trial and error the dynamics of cooling with finned structure is an extremely nuanced are. Even experienced power circuit people are often surprised when physical bench-test results are different from computer modeled heatsink designs.
For example, some LED literature suggests that customer use a "5 degree C per watt heat sink". It roughly takes a quadrupling of the surface area of a heat sink to halve the thermal resistance (and therefore handle twice the wattage). This means as power reaches levels above 10, 20, or 50 watts, the heat sink size, without air flow can get out of control.
With air flow, there is no such thing as a fixed 5 degree C per watt heat sink, since the thermal resistance is a function of the airspeed in LFM (linear feet per minute). With heat sinks, CFM, while relevant, is not what causes cooling—it is the LFM. Many are confused by the distinction.
The thermal resistance of a heat sink is a function of its physical design, coupled with the rate of air flow across its surfaces (this can range from no flow all the way up to 400 or 600 linear feet per minute). A heat sink maker will usually say that there is such and such a figure but only under such and such specific air flow condition.
Finally, it can be noted that LED fixture makers are realizing that the use of moving air, via small, high reliability DC brushless fans or other methods can dramatically simplify cooling as compared with "brute force" approaches with large heat sinks.
If you enjoyed this article, you will like the following ones: don't miss them by subscribing to :
