Can LED lighting deliver? Incandescent lamps are cheap but tend to have short lifetimes. The light emitted by incandescent sources is perceived as particularly pleasant because these hot radiators generate a continuous (or full) emission spectrum. But this very fact means that they waste much of the electrical energy supplied to them. Compact fluorescent lamps (CFLs) are actually the least compact of all the various lamps commercially available at present (Figure 1). While CFLs are certainly energy efficient, replacing an incandescent light bulb with a CFL is not just a matter of screwing out one lamp and screwing in the other. While the CFL thread fits the lamp holder socket, the lamp doesn’t always suit the luminaire (see Figure 2) – and it is not uncommon for users to reject the energy-saving CFL on aesthetic grounds. Other consumers dislike the ‘cold’ light that they claim CFLs produce, although there are now plenty of so-called warm-tone lamps available on the market. And if the CFL is located in a cold environment, the luminous flux produced by the lamp drops as a result. Frequent switching also tends to shorten the lifetime of a CFL. In fact, frequent switching shortens the lifetime of incandescent lamps, too – but this is less of a problem because replacing an incandescent filament lamp is a lot cheaper than replacing a CFL. Incandescent lamps are also used in headlamps and spotlights and they can also be dimmed without difficulty. CFLs on the other hand are only dimmable in certain cases, and with an angle of radiation of approximately 120°, CFL-based spotlights are spotlights in name only. Finally, the efficiency of both fluorescent and incandescent lamps decreases rapidly, the more they are dimmed. In the light of all these limitations, there is a need to develop a source of light • whose lifetime does not deteriorate with frequent switching, • that provides the full luminous flux immediately after being switched on, • that can be dimmed to zero with relatively little technical effort, • whose efficiency remains high at both full and partial loads, • that can provide both a point beam and a broad cone of illumination depending on lamp design, • that has a long lifetime, • and is still affordable.
Figure 1: Too hot: the mains-voltage halogen lamp (left) is power hungry and short-lived. Too bulky (see also Figure 2): the (theoretically) compatible CFL, shown in the middle, is hard to focus. Is the LED lamp pictured right the answer?
LED.doc
Figure 2: The aesthetic issues with some CFLs often means that these lamps are not used to save energy in living areas.
Page 1 of 10
Lamps that meet these requirements are now available. What is regarded as ‘affordable’ is of course a matter of opinion. However, when a lamp is available that costs ten times as much as a halogen spotlight, but has an operating life twenty times as long, is that lamp expensive or not? Light-emitting diodes (LEDs) have been used for over three decades as indicator lamps on electronic devices. However, the last few years have seen a number of significant developments, particular the manufacture of high-brightness LEDs, so that light-emitting diodes seem to be a realistic alternative for space lighting purposes. Or are they? In what follows, we shall be taking a close look at the properties of these new light sources.
Back to school: The physics of LEDs Light-emitting diodes are, as the name says, diodes. To generate light, LEDs have to be forward biased. If reverse biased, they would also light up – but only once! The reverse or ‘cut-off’ voltage of an LED is relatively low as they are not designed to be used for rectification. Like all diodes, an LED exhibits an exponential characteristic under forward bias conditions. An LED does not therefore obey Ohm’s law, in which voltage and current are linearly proportional to one another, but exhibits an exponential dependence. The dependence of the diode current ID on the voltage VD applied to the diode is given by the following expression: ⎛ V ⎞ ⎜ nV ⎟ − 1⎟ [1] I D = IS ⎜e ⎜ ⎟ ⎝ ⎠ D
T
where: n is the emission coefficient (essentially a correction factor for the individual diode); VT =
kT is the so-called thermal voltage; q
k = 1.38*10-23 J/K is the Boltzmann constant; q = 1,6*10-19 As is the elementary charge (the charge on a single electron); T is the absolute temperature of the diode junction at the moment of measurement;
60mA 50mA 40mA 30mA 20mA
280mW 240mW
ID theor (kalt) ID theor (warm) ID mess PD theor (kalt) PD theor (warm) PD mess
200mW i Î
70mA
320mW u Î
80mA
Figure 3: Measurement on a single (large) LED of the type used in LED lamps
120mW 80mW
10mA 0mA 1,5V
160mW
40mW t Î 2,0V
2,5V
3,0V
3,5V
0mW 4,0V
Figure 4: Voltage-current characteristics of a single (large) LED of the type used in LED lamps
IS is the diode’s (reverse) saturation current, which – essentially irrespective of the applied voltage – also flows in the high-resistance direction provided that the diode’s cut-off voltage is not exceeded. In the case of a germanium diode, IS is of the order of 100 nA; IS for a silicon diode is about 10 pA. Even though the absolute values of both these saturation currents are LED.doc
Page 2 of 10
very small, the relative difference between the two currents is 1:10 000. This value is of particular significance when attempting to compute the behaviour of the diode in the forward bias direction (see Figure 4). It is clear from the above equation that as the temperature T is part of the exponent, it has a strong influence on the magnitude of the diode current. But as the current changes so does the temperature. Fortunately, as T and hence VT rises, the exponent itself gets smaller, so preventing the diode current from rising catastrophically. This temperature response is evident in measurements of a large LED designed for use in a solar-powered light (Figure 3). The thin lines represent the theoretically computed curves for the ‘cold’ and ‘warm’ states, based on the assumption that diode heating is negligible at a diode current of 0.8 mA and is 53 K at 80 mA and assuming a value of n = 5.7 for the diode’s correction factor (i.e. emission coefficient). The ambient temperature was 19°C or 292 K when expressed on the absolute temperature scale required here. It is clear from Figure 4 that the measured current-voltage characteristic curves in the current range of interest (0.8 - 80 mA) lie between the theoretical curves computed for the ‘cold’ and ‘warm’ states. Assuming a correction factor of n = 5.7, the diode equation shown above is thus seen to provide a realistic model of the behaviour of this randomly chosen diode. Putting theory to one side, in what follows we shall only be interested in the thicker lines in Figure 4 that show how the measured I-V characteristic curves respond to the unavoidable heating effects.
Keeping the current under control The measurement data in Figure 4 show that very small changes in the applied voltage lead to extremely large changes in the diode current and diode power. It is therefore necessary to limit the current in some way. The simplest way of limiting the current is to connect a resistance in series with the LED. A somewhat more elegant solution is to use a transistor-based protective circuit. But in both cases, the excess voltage is simply dissipated as thermal energy. If that excess energy is to be saved and not wasted, the solution is to use a high-frequency switchedmode power supply that is more complicated, and hence more sensitive and expensive. These devices are particularly useful in isolated loads when the availability of electrical power is limited, such as in solar-powered traffic signs, garden lamps, etc. It is in these low-power, isolated devices (e.g. torches, bicycle lights) that LEDs have been most successful at replacing filament incandescent lamps. 300mA
LED-Leuchtmittel 12V 1,25W weiß
4,0VA
LED-Leuchtmittel 12V 1,7W warmweiß
I Î
2,5VA
250mA
150mA
2,0VA
3,5VA
I Î
200mA
3,0VA
200mA
50mA
0mA 7V
8V
9V
10V
11V
12V
13V
I= I˜ P= P˜ S˜ Q˜
150mA
1,0VA
100mA
0,5VA
U Î 6V
1,5VA
P; Q; S Î
100mA
P; Q; S Î
I= I˜ P= P˜ S˜ Q˜
6V
Figure 5: 12 V, 1.25 W LED lamp from Osram on DC and AC supplies
1,5VA
0,5VA
U Î
0mA
0,0VA
2,0VA
1,0VA
50mA
14V
2,5VA
7V
8V
9V
10V
11V
12V
13V
0,0VA
14V
Figure 6: 12 V, 1.7 W LED lamp from Megaman on DC and AC supplies
Which current limiting strategy has been selected can sometimes be determined relatively easily by measurement. A comparison of two LED lamps, both of which are designed to be driven by 12 V DC or AC supplies, illustrates how the distinction can be made apparent. The current in the Osram lamp initially increases exponentially but as the voltage rises the lamp clearly begins to LED.doc
Page 3 of 10
limit the further growth of the current (Figure 5). This result is in itself proof that the Osram lamp contains some form of electronic control. In the Megaman lamp, the current is limited by a simple resistor (Figure 6). Interestingly, the Osram lamp was older than the simpler Megaman model.The power of commercial LED lamps has been rising continually, and the higher power rating of the Megaman lamp is enough to indicate that it is the younger of the two models tested. The oscilloscope images offer a clearer picture of how the diode current behaves. Figure 7 and Figure 8 display the voltage, current and THD profiles that result when these two LED lamps are driven by a 50 Hz AC voltage supply of 10, 12 or 14 V (rms). It is clear that the Osram lamp does not attempt to restrict the current at voltages up to 10 V AC, at higher voltages the amplitude of the current waveform is truncated at a maximum current limit of 200 mA. When the lamp is driven by a DC voltage source, increasing the light output at voltages above 10 V can only be achieved by enlarging the conduction angle. If the lamp is connected to a DC supply, no further increase in light intensity is possible. The problem is that the power loss in the transistor rises with increasing voltage. An overvoltage therefore means that the lamp is less economical to run. This could be avoided by using a switched-mode power supply but given the modest power levels in these lamps the cost would be prohibitive. Figure 5 shows that the current in the LED lamp is negligible at sinusoidal rms voltages of 8 V or lower, i.e. when the instantaneous AC voltage (or the DC voltage in the case of a DC supply) drops below about 10 V – as is apparent from a careful examination of the upper row of voltage and current curves in Figure 7. The threshold voltage is most clearly seen in the measurements made using a DC voltage supply. An LED has a higher conducting-state voltage (or ‘cut-in’ voltage) than, say, the approximately 0.8 V exhibited by a typical silicon diode. The model tested here (Figure 4), for instance, has a cut-in voltage of around 3.8 V. To make optimum use of the supply voltage, multiple LEDs are connected in series such that the supply voltage is just greater than the sum of the ‘cut-in’ voltages of the individual diodes. LED lamps are in fact composed of one or more parallel-connected groups, with each group comprising one or more serially connected LEDs.
Figure 7: Behaviour of the 12 V, 1.25 W LED lamp from Osram
The LEDs in the Megaman lamp are obviously being fed a DC voltage irrespective of whether the lamp is connected to an AC or DC source. This can be seen in the current profiles and the harmonic spectra in Figure 8, which are identical to those found for a non-PF-corrected CFL that operates according to the same principle. The LEDs are only grouped in series (with corresponding resistors) after the voltage has been rectified and then smoothed. The advantage of this solution is that the LEDs are illuminated all the time and not just when the voltage curve passes through its peak. It is possible that the higher rated power of the Megaman lamp is simply a reflection of this fact. The optimum solution would be to combine this type of rectified voltage with the current limiting system used in the Osram lamp. LED.doc
Page 4 of 10
Figure 8: Behaviour of the 12 V, 1.7 W LED lamp from Megaman
Running LEDs (or other semiconductor diodes) in parallel is not usually an option as the rapid exponential rise in the current-voltage characteristic means that even minor differences in the cut-in voltages can lead to a very unequal distribution of current in the branches of the parallel circuit. The fact that the LEDs are connected in groups becomes apparent when the voltage is lowered and not all the LEDs darken to the same extent, but they don’t all behave independently of one another either. In Figure 9 the DC voltage has been reduced to 7 V, a mere two LEDs continue to glow – and that surprisingly brightly.
Figure 10: As this advertisement for a 12 V LED lamp makes clear, the lamp has to be run with a conventional transformer. Clearly ‘full compatibility’ is still some way off.
Figure 9: Behaviour of an LED lamp at a low supply voltage
Using LED lamps in practice A particular advantage of 12-volt LED lamps is that they can be used as a direct replacement for existing halogen spotlights. Although in some cases the LED lamps can only be run on conventional halogen lamp transformers (see Figure 10 and ref. [2]). The reason why the manuLED.doc
Page 5 of 10
facturers do not allow them to be driven by electronic halogen lamp transformers is not specified. A brief test indicates that the lamps also function on an electronic transformer, the question is of course: for how long? It has to be stated, though, that in this case only one out of four lamps being run on one electronic transformer had been replaced (Figure 11). The restriction made in Figure 10 is probably not to be blamed on the LED but on the transformer, since most electronic halogen lamp transformers prove unable to cope with part load. When one out of a number of lamps connected fails, still nothing much happens, but as the second one fails a complete failure or continuous flickering of the remaining lamps is very often the result. As the benefits of an electronic transformer are often not quite as far-reaching as the advertisements would have us believe, it is probably safer to use conventional transformers made of copper and iron when installing halogen lamps. Toroidal-core transformers are particularly well suited to these applications, because the power requirements decrease substantially when halogen spotlights are replaced by LED lamps and toroidal-core transformers exhibit significantly better efficiencies when partially loaded. In contrast, the efficiency of a conventional transformer drops significantly when run under partial load, and the performance of an electronic transformer will more often than not end up in a disaster. 100% η Ï 80%
60% η, 60VA standard transformer
40%
η, 60VA toroidal core transformer η, 400VA standard transformer
20%
η, 400VA toroidal core transformer I sec / I sec,nom Î
0% 0%
Figure 11: One out of four 20 W halogen spotlights in a system replaced with an LED lamp 1.25 W
25%
50%
75%
100% 125% 150% 175% 200%
Figure 12: Variance of efficiencies in traditional and toroidal core transformers 60 VA and 400 VA, respectively, with varying load
Improved efficiency? According to current forecasts, LED lamps have a great future ahead of them. These predictions refer to a number of features of these lamps, including their energy efficiency. There are reports of overall luminous efficiencies of 100 lumens of luminous flux for every watt of input power, making the LED lamp as efficient as the most efficient fluorescent lamps. Lab measurements of a single LED lamp alone already yield an overall luminous efficiency of up to 100 lm/W, while the system as whole, i.e. including the voltage converter, results in values of up to 50 lm/W. The job of the voltage converter is to reduce the mains voltage to the required low operating voltage. Most voltage converters are electronic devices that feed the strings of LEDs with a highfrequency voltage or in the optimum case with a stabilized DC voltage. Occasionally LED lamps are run directly at mains frequency, for instance when they are used to replace existing halogen spotlights, i.e. the step-down transformer is already in place and only the control electronics (in the simplest case: a rectifier and a series resistor) need to be built into the lamp. It should be mentioned that the loss in power that results in the drop in overall luminous efficiency from 100 lm/W to 50 lm/W is due to the fact that the sample lamps used here are of very low power. Just as higher efficiency ballasts and transformers have been designed for use with fluorescent lamps and halogen lamps respectively, more efficient LED-based lamp systems can be built for higher power applications.
LED.doc
Page 6 of 10
The LED lamps currently available commercially can be thought of as lying somewhere between incandescent lamps and fluorescent lamps. The 1.5-W lamp in Figure 10 is specified as having an illuminance of 240 lux. As it stands, this information is not particularly meaningful. Illuminance is a measure of the total luminous flux incident on a unit area of the illuminated surface. The more of a lamp’s luminous flux that can be directed onto the surface, the greater its illuminance. The total luminous flux from a light source is measured in lumens. According to the experts at Osram, it is very difficult to measure accurately the total luminous flux of a directed (focused) light source. Advertisements for these products will sometimes quote illuminance values in a way that gives the impression that customers can use these figures to make a meaningful comparison between competitor products. However, if sharper focusing enables the emitted light from a lamp to be concentrated into an area one quarter of that previously illuminated, the illuminance of the lamp is suddenly four times as great, although the overall luminous efficiency (strictly ‘overall luminous efficacy’) of the lamp has not changed at all. While the total amount of visible light falling on a certain area of surface is obviously an important quantity in describing a lamp’s properties, it says nothing whatsoever about the lamp’s energy efficiency. Nevertheless, the relatively simple replacement of a halogen spotlight by an LED spotlight represents a major step towards achieving more efficient use of energy resources. It cannot be repeated too often: halogen lamps are incandescent filament lamps and as such are only marginally more efficient than the ordinary incandescent filament bulbs found at the low end of the energy classification scheme published by the EU. Typically, the standard household incandescent lamp will be given a class E rating, with classes ranging from A (most efficient) to G (least efficient). And if the lamp has a power rating below 40 W or if the glass bulb is decorated in some way, the lamp more often than not slips down into class F. Halogen lamps with a clear quartz envelope usually have a class D energy rating, though if the envelope has a matt finish, they move down a level to class E. Once again we are dealing here with a classification scheme with an inbuilt and unnecessary limitation. By defining ‘class A’ as the most efficient category, the scale has been needlessly truncated at the wrong end. While it can be extended to include classes of ever poorer efficiency by simply adding another letter to the series, extending it to include new products exhibiting previously unattained levels of efficiency is not so simple. With the exception perhaps of wax candles and torches, light sources with class G efficiency have never been available, except for the unspeakable ‘Linestra’ lamps which, on top of all, are often confused with fluorescent lighting tubes. But as the classification scale has been around for some time, during which the lighting industry has not stood still, household lamps are now available that can only be classified by using constructions such as ‘class A+’ and ‘class A++’. In the case of T5 fluorescent tubes, both the high-efficiency (HE) lamps and the brighter (but much less efficient) ‘high-output’ (HO) models are given a class A rating. And T8 lamps, which are claimed to be technically outdated, are also found in the (obviously very broad) class A category, while the essentially dead category ‘G’ still forms part of the overall classification scheme. Seen against this background, it is all the more irresponsible when a well-known manufacturer launches a series of improved halogen lamps that consume 30 % less energy and then markets these lamps under the banner ‘energy saver’, which is of course the term already used for compact fluorescent lamps. 30 % less power consumption relative to earlier halogen lamp models is still 400 % more than the power consumed by an equivalent CFL and 200 % more than currently available LED lamps from the same manufacturer! Interestingly, the catalogue entry for these ‘energy saver’ halogen lamps does not say anything about their energy saver classification. Additionally, there are a number of very common lamps, such as reflector lamps, that are not included in the classification scheme. The same is also true of lamps with a wattage of below 4 W [3], which means that the LED lamps current available on the market do not yet have to be classified in terms of their energy efficiency. LED.doc
Page 7 of 10
The power problem Light-emitting diodes are by their very nature directed sources of light. If one wants to have a more diffuse light source, steps have to be taken to scatter the beam – something that is easier to achieve than the reverse process of constructing a directed beam of light from a widely dispersed light source. It is the well-defined and directed nature of the light beam from halogen spotlights that have made them so popular. Fortunately, the light from LED spotlights has the same characteristics. Replacing a halogen spotlight by an LED spotlight couldn’t be more straightforward. Unfortunately, LED lamps do not deliver the required power. A 1.7-watt LED lamp is just about bright enough to replace a 5-watt halogen spotlight, but certainly not one of the far more common 20 W or 35 W models. As a result, LED lamps have not so far had a major commercial impact in the domestic lighting market. However, Osram, for example, is already conducting laboratory tests on 1000-lumen LED lamps.
The colour problem Light-emitting diodes generate monochromatic light, i.e. light of a single colour or, more precisely, light of a single frequency or wavelength. To create white light using these monochromatic sources requires mixing the light from a red, a green and a blue LED. It has in fact proved possible to combine three such LEDs in a single housing. The Tridonic company has managed to create a white LED light source using a very different approach. As in a fluorescent lamp, a phosphor conversion coating is applied to the envelope of a blue LED. Light of longer wavelengths, i.e. of lower frequencies and hence of lower energy can then be created by downconversion of the higher frequency blue light. However, achieving the right mixture of colours and thus the desired type of ‘white’ is a challenging technical problem in both techniques. It is probably more accurate to call the white light generated by ‘normal’ white LED lamps as light blue (Figure 14) and the ‘warm white’ LED lamps now available have a definite yellowish appearance (Figure 13). But according to both Osram and Tridonic, rapid progress can be expected in this area.
Figure 13: Three CFLs compared to a ‘white’ LED lamp (on the right at the back) and a ‘warm white’ LED lamp (on the left at the back)
Figure 14: Corridor lighting in a hotel in Berlin’s ‘blue light district’ LED.doc
Page 8 of 10
It is not known how long the lamps in Figure 14 have been in use. It could well be that the hotel proprietor bought them more or less as prototype LED lamps a long time ago, only to discover that they seem to last forever. If the colour had been better, they would probably at that time have been so expensive that he wouldn’t have bought them in the first place. When it comes to lighting living spaces, there is a definite preference in Europe for a more yellowish light. However, the distinctiveness of the bluish white light is particularly appropriate for bicycle lamps where it is not so much a question of seeing but more of being seen. And the added safety that a cyclist or other road user gains from having distinctive illumination at night will last until these new lamp types have become the norm. As it stands, however, LED-based technology is simply not bright enough at present for use in car headlamps, but developments are clearly moving in the right direction. Audi, for instance, is already using LEDs for daytime running lights. LED lamps are in fact ideal for use in a vehicle’s rear and indicator lights, as the required colours can be generated directly without, as is currently the case, having to use an inefficient incandescent bulb to generate a continuous spectrum from which all the other colours are then filtered out. Once again, it is the bicycle, which typically has only 3 watts of electrical power available, that is currently best able to profit from LED technology. Not only do LED lamps make bicycles more visible from the rear, the conventional front headlamp is being increasingly replaced by the far more conspicuous LED flashing lamp, whose lifetime is not shortened by frequent switching and that does not generate switching peaks. Almost all other lamps produce these effects. In the case of fluorescent lamps driven by a low-loss magnetic ballast the cause is core saturation, in (smaller) fluorescent lamps with an electronic ballast and no active power-factor correction it is the smoothing capacitor, in (larger) fluorescent lamps with an electronic ballast and an active power-factor correction circuit it is the filter capacitor, and in the case of the incandescent lamp it is the low resistance of the cold filament.
Outlook One could draw the conclusion that LED lamps are expensive and weak, that their light is ‘cold’ and that their energy efficiency is only half as good as that of a CFL – making them pretty much useless as replacements for anything. However, their efficiency is far superior to that of an incandescent lamp, they have extremely long lifetimes, are able to work at all typical operating temperatures, they do not need time to stabilize or warm up, they are insensitive to frequent switching and do not interfere with the switching device. They are small, compact and produce a focused beam of light. The development from simple indicator lamps to light sources that are now beginning to find practical applications has been extremely quick. It seems therefore safe to assume that further progress will be made and that in a few years LED lamps will be available for a wide range of lighting applications. LED lamps require an electronic controller that is usually built in to base of the lamp, but as the design is simpler than that required for CFLs, it seems unlikely that this will delay or hinder development. In view of the relatively low power of the LED lamps available up until now, there has been little demand for dimmable devices, but making dimmable LED lamps should not prove to be a technical problem. The efficiency when running at partial load is as good or better than at full load, a fact that makes LED lamps particularly interesting as sources for emergency and continuous lighting. The LED lamp thus looks set for a bright future. And the availability of LED lamps that are compatible with lowvoltage halogen spotlights and mains-driven lamps with conventional lamp holders (see Figure 16) simplifies replacement considerably. Even at their current high prices, the very long lifetimes of these lamps makes them economical to use. They are therefore particularly practical in locations where changing lamps is not a straightforward procedure. The same conclusion was reached by the technical manager of a luxury hotel in which, as so often is the case, power hungry halogen spotlights were being used to light the corridors day and night. The weekly lamp inspection is now being simplified by gradually replacing the halogen spotlights by LED spotlights, starting with those in difficult-to-access locations. The noticeably bluish light that they emit will however remain a constant reminder that the lamps were installed before the ‘warm LED.doc
Page 9 of 10
tone’ LED lamps became available, because even if they run uninterrupted day and night the same LED lamps will still be operating in five years time.
Figure 15: This bathroom lighting arrangement consumes 5*50 W of power and two to three lamps per year
Figure 16: Three lamps were replaced by so-called reflector CFLs (left) and two by LED lamps – one ‘white’ and the other ‘warm white’ (right); exposure details as in Figure 15; total power consumption now about 36 W (i.e. 15 % of the previous load). The CFLs deliver the brightness, but not the required operational characteristics; the LED lamps deliver the right operational characteristics, but not the brightness – but progress is being made!
If for political reasons a ban on the use of incandescent lamps is introduced (Figure 15), LED lamps will be able to replace incandescents in those situations where a CFL would fail due to its cold start behaviour or its sensitivity to frequent switching (Figure 16). But political initiatives frequently get watered down as the legislative process progresses. It is quite likely that lowvoltage halogen lamps will be excluded because they are not classified as incandescents or because they are not regarded as replaceable, even though by that time LED lamps will probably represent a viable and more energy efficient alternative.
[1] www.krucker.ch/Skripten-Uebungen/AnSys/ELA4-D.pdf and numerous other websites that can be found by typing ‘LED’ and ‘thermal voltage’ into a search engine [2] see for example: www.conrad.de [3] www.eup4light.net
LED.doc
Page 10 of 10