Chapter 3
Driving LEDs
Abstract
This chapter builds on the previous chapter about light-emitting diodes (LEDs). We discovered that LEDs need a constant current and here methods of producing a constant current are discussed. Both linear and switching power supplies are introduced, with advantages and disadvantages being discussed. Parallel and series connection options are also discussed, with the effect of temperature included. Suggestions for methods of testing LED drivers are given.
Keywords
light-emitting diode
LED
constant current
linear regulator
switching regulator
parallel connection
series connection
current limiter
temperature dependency
power supply
3.1. Voltage Source
We have seen in Chapter 2 that a light-emitting diode (LED) behaves like a constant voltage load with low equivalent series resistance (ESR). This behavior is like a Zener diode—in fact Zener diodes make a good test load, rather than using expensive high power LEDs!
Driving a constant voltage load directly from a constant voltage supply is very difficult because it is only the difference between the supply voltage and the load voltage that is dropped across the ESR. But the ESR is a very low value, so the voltage drop will also be low. The LED forward voltage and ESR are also temperature dependent. A slight variation in the supply voltage, or the LED forward voltage, or the ESR, will cause a very large change in current; see curve A in Fig. 3.1.
Figure 3.1 LED Current Versus Supply Voltage.
If the variation in supply voltage and forward knee voltage (Vf) is known, the variation in current can be calculated. Remember that there are variations in LED voltage drop due to both manufacturing tolerances and operating temperature. Most supply voltages from a regulated supply have a 5% tolerance, but from unregulated supplies like automotive power, tolerances are far greater.
These equations assume that ESR is constant. In practice, the Vf and voltage drop across ESR are combined since manufacturers quote the voltage drop at a certain forward current. The actual Vf of the LED diode junction can be determined from graphs, or can be measured.
Consider that there is a large difference between the source and load voltage, which is dropped across a high value series resistor. In this case, changes in LED forward voltage cause very little difference between the maximum and minimum LED current. This may be perfectly adequate for low current LEDs, up to 50 mA. However, in high power LED circuits, a large voltage drop across a series resistor will be inefficient and may cause heat dissipation problems. Also, the ESR of LEDs is lower as the power rating increases. As previously described, a standard 20-mA LED may have an ESR of about 20 Ω, but a 350-mA LED will have an ESR of 1 Ω typically. Thus a 1 V difference in supply voltage could increase the LED current by 1 A in a power LED. Even in low current LEDs, the proportional change in current can be very high.
3.1.1. Passive Current Control
Looking again at Fig. 3.1, if the LED voltage drop increases, curve A of the graph shifts to the right. The slope of the graph is unchanged and is just due to the ESR. However, low current LEDs can have a relatively high value resistance added in series, to reduce the slope of the current versus voltage graph; see curve B in Fig. 3.1. Now a change in supply voltage has much less effect on the LED current, If.
With a series resistor added we are able to calculate the variation in current, provided that the variation in supply voltage and load voltage is known. In the following equations, the load voltage includes the voltage drop across ESR, at the rated current, so only the external resistor value is needed.
As an example, let us drive an LED from an automotive supply; this is a nominal 13.5 V, but for this exercise we can set the limits at 12–16 V. Let us select a red LED for tail-lights (Lumileds Superflux HPWA-DDOO), with a forward voltage drop of 2.19–3.03 V at 70 mA forward current. Choosing to connect two LEDs in series, with a series resistor, we have a typical voltage drop of 5 V. So the typical voltage drop at 70 mA needs to be 8.5 V; this means that the series resistor should be 121.43 Ω. The nearest standard value of resistor is 120 Ω, rated at 1 W since we will have a typical power dissipation of 588 mW.
At the high limit of source voltage, the LED is overdriven by 38%. But there is almost a 2:1 ratio between Imax and Imin, so if we increase R by 38% the worst case current levels are 70 mA maximum and only 35.78 mA minimum.
In the previous calculations, the voltage drop across ESR (0.672 V) was included in the minimum and maximum load voltage values, so we ignored ESR. From the manufacturer’s data sheet of the Lumileds HPWA-DDOO LED, graphs show that the ESR is about 9.6 Ω. Suppose we now want to operate at a lower current. Using the same example, but operating with a typical LED current of 50 mA, we must modify the results. Now the voltage at the current knee is Vf = 1.518–2.358 V. With a typical 13.5-V supply and 50 mA, the value for Vf is 1.828 V. The total resistance needed is 196.88 Ω, but we already have 9.6 Ω ESR. An external resistor value of 180 Ω is the nearest preferred value for a current of 50 mA.
The series resistor has a higher value, so the variation in current is reduced to 1.6:1 ratio. The maximum current is now below the LED current rating of 70 mA.
What happens when we connect LEDs in parallel? Unless the LEDs are matched (or “binned”) to ensure the same forward voltage drop, the current through one string could be considerably different from the current through another. Even if the forward voltages are matched initially, differences in power dissipation from one LED to another will cause a temperature difference in the LEDs. As the temperature rises, the forward voltage drop reduces. This causes more current through the LED, the ESR dropping a higher voltage, and so the temperature rises further. At the same time, parallel LEDs will take less current and cool down. We end up with thermal runaway and ultimately LED damage.
When multiple LEDs are used to provide lighting for an application, they are frequently connected in an array, consisting of parallel strings of series connected LEDs. Since the LED strings are in parallel, the voltage source for all strings is the same. However, due to manufacturing variations in forward voltage for each LED, the total voltage drop of each string differs from the other strings in the array. The forward voltage also depends on the LED diode junction temperature, as previously mentioned. To ensure uniform light output for all LEDs, equal current should be designed to flow through each string of LEDs.
The traditional way to drive a parallel LED array is to connect a current limiting resistor in series with each string and power all the strings using a single voltage source. A substantial voltage drop is required across the resistor to ensure that the current will stay in the desired range, in the presence of temperature and device-to-device voltage variations. This method is inexpensive and suffers from power inefficiency and heat dissipation. It also requires a stable voltage source.
A better way of powering the LED array is to regulate the total current through all the strings and devise a means to divide that total current equally among the LED strings. This is active current control and is the subject of the next section.
3.1.2. Active Current Control
Since a series resistor is not a good current control method, especially when the supply voltage has a wide tolerance, we will now look at active current control. Active current control uses transistors and feedback to regulate the current. Here we will only consider limiting LED current when the energy is supplied from a voltage source; driving LEDs using energy from current sources will be discussed in Section 3.2.
A current limiter has certain functional elements: a regulating device, such as a MOSFET or bipolar transistor; a current sensor, such as a low value resistor; and some feedback (with or without gain) from the current sensor to the regulating device. Fig. 3.2 shows these functions.
Figure 3.2 Current Limiter Functions.
The simplest current limiter is a depletion-mode MOSFET; it has three terminals called gate, drain, and source. Conduction of the drain–source channel is controlled from the gate–source voltage, like any other MOSFET. However, unlike an enhancement MOSFET, a depletion-mode MOSFET is “normally-on” so current flows when the gate–source voltage is zero. As the gate voltage becomes negative with respect to the source, the device turns off, see Fig. 3.3. A typical pinch-off voltage is −2.5 V.
Figure 3.3 Depletion MOSFET Characteristics.
A current limiting circuit with a depletion-mode MOSFET uses a resistor in series with the source to sense the current (Fig. 3.4). The gate is connected to the negative supply (0 V). As current flows through the resistor, the voltage drop across it increases. The voltage at the MOSFET source becomes in higher potential compared to the 0 V rail and the MOSFET gate. In other words, compared to the MOSFET source, the gate becomes more negative. At a certain point, when the voltage drop approaches the MOSFET pinch-off voltage, the MOSFET will tend to turn off and thus regulate the current.
Figure 3.4 Depletion MOSFET Current Limiter.
The main drawback of using depletion-mode MOSFETs is that the gate threshold voltage (Vth) has a wide tolerance. A device with a typical Vth of −2.5 V will have threshold range of −1.5 V to −3.5 V. However, the advantage is that high drain–source breakdown voltages are possible and so a limiter designed using a depletion-mode MOSFET can protect against short transients (longer periods of high voltage would tend to overheat the MOSFET).
A simple integrated current limiter is a voltage regulator in the place of the depletion-mode MOSFET, as shown in Fig. 3.5. This uses an internal voltage reference and so tends to be quite accurate. The disadvantage is that there is a minimum dropout voltage, which can be as high as 4 V. This type of current limiter can be used for current sink or current source regulation, depending on whether the load is connected to the positive or negative supply rail.
Figure 3.5 Linear Regulator as Current Limiter.
In Fig. 3.5, the well-known linear voltage regulator LM317 is used as a current limiter. The LM317 has three terminals, labeled “IN,” “OUT,” and “ADJ.” The feedback pin labeled “ADJ” (adjust) controls the regulation of the current. A current sensing resistor is placed between the “OUT” terminal and the load, which is also connected to the “ADJ” terminal. Due to the series connection, the voltage across the current sensing resistor is proportional to the output current and will cause the voltage at the “ADJ” terminal to be lower than the “OUT” terminal. The LM317 feedback limits the current when the “ADJ” terminal voltage reaches 1.25 V below the “OUT” terminal.
If accurate current limiters are used in series with each LED string, parallel strings of LEDs and their limiters can be connected to the same voltage source. In this case, each string will have approximately the same current. With the same current flowing through each LED, the light produced will be almost the same for each LED and thus no “bright spots” will be seen in the LED array.
3.1.3. Short Circuit Protection
The current limiting circuits described in the previous section will provide automatic short circuit protection. If the LED goes short circuit, a higher voltage will be placed across the current limiter. Power dissipation is the main issue that needs to be addressed.
If the power dissipation cannot be tolerated when the load goes short circuit, a voltage monitoring circuit will be needed. When a voltage higher than the expected voltage is placed across the current limiter, the current must be reduced to protect the circuit. In the previously described LM317 circuit, the regulator itself has thermal shutdown.
3.1.4. Detecting Failures
If we have a short circuit condition in the LEDs, the voltage across the current limiter will increase. We can use this change to detect a failure. In the circuit shown in Fig. 3.6, a 10-V Zener diode is used in series with the base of an NPN transistor. When the voltage at the “IN” terminal of the LM317 reaches about 11 V, the Zener diode conducts and turns on the transistor. This pulls the “FAILURE” line to 0 V and indicates a short circuit across the LEDs.
Figure 3.6 Shorted Load Protection.
3.2. Current Source
Since an LED behaves like a constant voltage load, it can be directly connected to a current source. The voltage across the LED, or string of LEDs, will be determined by the characteristics of the LEDs used. A pure current source will not limit the voltage, so care must be taken to provide some limit; this will be covered in more detail in the next section.
Parallel strings of LEDs can be driven by current-sharing circuits. The simplest of these is a current mirror, which shares the current equally between strings based on the current flowing through the primary string. Fig. 3.7 shows a simple current mirror. The basic principle relies on the fact that matched transistors will have the same collector current, provided that their base–emitter junctions have the same voltage across them. By connecting all the bases and all the emitters together, every base–emitter junction voltage must be equal and therefore every collector current must be equal.
Figure 3.7 Current Mirror.
The primary LED string is the one that controls the current through the other strings. Since the collector and base of transistor Q1 are connected, the transistor will be fully conducting until the collector voltage falls low enough for the base–emitter current to limit. Other transistors (Q2–Qn) have their base connections joined to Q1, and will conduct exactly the same collector current as Q1 since the transistors are matched. The total current through Q1 to Qn will equal the current source limit.
The voltage drop across the LEDs in the primary string must be higher than any other string in order for the current mirror to work correctly. In the slave strings, some voltage will be dropped across the collector–emitter junction of the transistors Q2–Qn. The slave circuits adjust the current by raising or lowering this surplus voltage drop across the transistor.
3.2.1. Self-Adjusting Current Sharing Circuit
As an alternative, the current sharing circuit shown in Fig. 3.8 automatically adjusts for string voltage.
Figure 3.8 Self-Adjusting Current Sharing Circuit.
Assuming that the LED array is driven from a current source, there will be equal current division among all connected branches. If any branch is open due to either a failure or no connection by design, the total current will divide evenly among the connected branches. Unlike the simple current mirror, this one automatically adjusts for the maximum expected voltage difference between strings of LEDs, which is a function of the number of LEDs in the string and the type of LED used. The components must be able to dissipate the heat generated by the sum of each string current and the headroom voltage drop across the regulator for that string.
In high reliability applications, the failure of a single LED should not significantly affect the total light output. The use of the current divider will help the situation. When an LED fails short circuit, the voltage of the string containing the shorted LED will have less voltage. The self-adjusting current sharing circuit will accommodate the change in voltage and still distribute the current equally. When an LED fails open circuit, the current divider will automatically redistribute the total current among the remaining strings, thus maintaining the light output. In this application, an extra diode string can be added for redundancy, so that any single failure will not cause the remaining LEDs to operate in an overcurrent condition.
Equality of current division among the branches is dependent on the close matching of the transistors, which are in close vicinity (ideally a single package with several matched transistors). When any of the transistors saturate due to large variation of the string voltages, equal current division will be lost.
Diodes connected to each collector detect the voltage of each branch. The highest branch voltage (corresponding to the LED string with the lowest forward voltage) is used to bias the transistors in the linear operating region. The cathode of each diode is connected to a common “bias bus.”
To accommodate variations in string voltages and keep the current divider transistors from saturation, diodes are connected between the “bias bus” and the “transistor base bus.” More than one external diode can be used to accommodate large voltage variations. If the string voltage variation is less than one diode drop, the two buses can be joined.
When a branch is not connected, there will be higher base current flowing in the associated regulating transistor. This could interfere with the current division in the connected branches, so a resistor (about 1 kΩ) is connected from the “transistor base bus” to each transistor base to ensure correct operation of the overall circuit.
3.2.2. Voltage Limiting
In theory, the output voltage of a constant current driver is not limited. The voltage will be the product of the current and load resistance in the case of a linear load. In the case of an LED load, the voltage limit will depend on the number of LEDs in a string. In practice, there will be a maximum output voltage because components in the current source will break down eventually. Limiting the LED string voltage is necessary to prevent circuit damage and the voltage level will depend on the particular circuit. In linear regulators or buck (step-down) switching regulators, the voltage is limited by the source supply voltage. But in boost regulators the output voltage can rise to very high levels and a voltage feedback scheme to provide overvoltage protection should be provided.
Safety regulations will be covered in Chapter 10, but Underwriters Laboratories (UL) Class 2 and safety electrical low voltage requirements limit any potential to 60 V DC, or 42.4 V AC; so equipment designed to meet these requirements should consider both mains supply isolation (if applicable) and output voltage limiting. The number of LEDs in a string will be restricted in this case, so that the total string voltage remains below 60 V.
3.2.3. Open Circuit Protection
Some constant current drivers, especially switching boost converters, will produce a sufficiently high voltage to destroy the driver circuit. For these types of driver a shutdown mechanism is required. Using a Zener diode to give feedback when the output voltage exceeds a certain limit is one method. Most integrated circuits (ICs) have an internal reference and comparator circuit to provide this function. A potential divider comprising two resistors is usually used to scale down the output voltage to the reference voltage level. Some overvoltage detectors within ICs have a latched output, requiring the power supply to be turned off and then turned on again before LED driver functions are enabled. Other circuits will autorestart when the open circuit condition is removed (i.e., when the LEDs are reconnected).
3.2.4. Detecting LED Failures
In a constant current circuit, a failure of a single LED within a string of series connected LEDs can mean that either a whole string is off (open circuit LED) or a single LED is off (short circuit LED).
In the case of an open circuit LED, the load is removed and so the output voltage from the current source rises. This rise in voltage can be detected and used to signal a failure. In circuits where overvoltage protection is fitted, this can be used to indicate a failure. Similarly, in the case of a short-circuited LED, the output voltage from the current source drops, so the drop in voltage can be used to indicate a failure. In some switching circuits, indirect measurement can be made by detecting a change in the MOSFET gate drive signal (the gate drive pulse width will change as the output voltage changes).
If a current mirror is used to drive an array of LEDs with a number of parallel strings, the result of an open circuit LED will depend on which string the LED is located. In a basic current mirror as shown in Fig. 3.7, a failure in the primary string will cause all the LEDs to have no current flow and not be lit. Detection of the rise in output voltage would be a solution. However, if the failure were in a secondary string, there would be higher current flowing in the other strings and the output voltage would not rise very much (only due to the extra current flowing through the ESR). The voltage at the transistor collector of the broken string would fall to zero since there is no connection to the positive supply and this could be detected.
Another technique, for low current LEDs, is to connect the LED of an optocoupler in series with the LED string. A basic optocoupler has an LED and a phototransistor in the same package. Current through the optocoupler LED causes the phototransistor to conduct. Thus when current is flowing through the LED string and the optocoupler’s internal LED, the phototransistor is conducting. If the string goes open circuit, there is no current through the optocoupler’s LED and the phototransistor does not conduct.
3.3. Testing LED Drivers
Although testing an LED driver with the actual LED load is necessary, it is wise to use a dummy load first. There are two main reasons for this: (1) cost of an LED, especially for high power devices, can be greater than the driver circuit; and (2) operating high-brightness LEDs for a long time under test conditions can cause eye strain and temporary sight impairment (snow blindness). A further reason is that some dummy loads can be set to limit the current and so enable fault-finding to be made easier.
3.3.1. Zener Diodes as a Dummy Load
Fig. 3.9 shows how Zener diode can be used as a dummy load. This is the simplest and cheapest load. The 1N5334B is a 3.6 V, 5 W Zener diode (3.6 V typical at 350 mA). This is not the perfect dummy load. This reverse voltage is slightly higher than the typical forward voltage of 3.42 V of a Lumileds “Luxeon Star” 1 W LED. The 1N5334B has a dynamic impedance of 2.5 Ω, which is higher than the Luxeon Star’s 1-Ω impedance. The impedance will have an effect on some switching LED drivers that have a feedback loop. For simple buck circuits, the impedance only has a small effect.
Figure 3.9 Zener Diode Dummy Load.
An active (electronic) load is more precise. A constant voltage load will have (in theory at least) zero impedance, so simply adding a small value series resistor will give the correct impedance. Commercial active loads can be set to have constant current or constant voltage—a constant voltage setting is required to simulate an LED load.
A constant voltage load built using a low-cost discrete solution is shown in Fig. 3.10. This is a self-powered load and so can be isolated from ground. The Zener diode can be selected to give the desired voltage (add 0.7 V for the emitter–base junction of the transistor). The transistor should be a power device, mounted on a heat sink.
Figure 3.10 Active Dummy Load.
The circuit in Fig. 3.10 has low impedance. Although the Zener diode does have a few ohms impedance, the current through it is very small and the effect of the transistor is to reduce the impedance by a factor equal to the gain HFE. Suppose the transistor HFE = 50 at 1 A and the Zener diode impedance Zd = 3 Ω. Changing the collector current from 500 mA to 1 A will cause the base current to rise from 10 to 20 mA. A 10 mA change in current through the Zener diode will cause 30 mV voltage rise. This change at the transistor collector is equivalent to an impedance of 30 mV/0.5 A = 0.06 Ω. In other words, the circuit impedance is equal to the Zener diode impedance divided by the transistor gain.
An impedance of 0.06 Ω is unrealistically low, but a power resistor can be added in series to give the desired load impedance. Due to the potentially high load current, both the transistor and series resistor should be rated for high power. The transistor should be mounted on a large heat sink.
3.4. Common Mistakes
The most common mistake is to use expensive high power LEDs when testing a prototype circuit. Instead, a dummy load should be used, using either an electronic load or a number of Zener diodes connected in series. If using Zener diodes, 3.6 V 5 W types should be used in place of each LED. Only once the circuit has been tested under all conditions should LEDs be used.
Another mistake is to connect LEDs in parallel. It can be done if the LEDs are very well matched, both thermally and their forward voltage drops, but this is very difficult to achieve. A current regulator/current mirror in series with each LED string is by far the best option.
3.5. Conclusions
A constant voltage–regulated LED driver is preferred when there are a number of LED modules that can be connected in parallel. Each module will have its own linear or switching current regulator, so the LED current is constant. An example would be channel lighting, as used in shop name boards.
A current-regulated LED driver is preferred when it is desirable to have a number of LEDs connected in series. A series connection ensures that all the LEDs have the same current flowing through them and the light output will be approximately equal. If the load voltage (of the LEDs) is just below the source voltage, a linear current regulator is a good cost-effective solution. A switching current regulator is needed if the LED load voltage is higher than the source voltage, or if the LED load voltage is significantly lower than the source voltage. A switching current regulator is also the favored option while driving high power LEDs, for reasons of efficiency. An efficiency of 90% is normally achievable.
..................Content has been hidden....................
You can't read the all page of ebook, please click here login for view all page.