Global Sources
EE Times-Asia
Stay in touch with EE Times Asia
EE Times-Asia > Optoelectronics/Displays

Hysteretic converters for multiple LED lighting

Posted: 16 Jul 2008 ?? ?Print Version ?Bookmark and Share

Keywords:PMOS? LED? converters? controllers?

Hysteretic converters are widely used to drive LEDs in many lamp replacement and emerging lighting applications. The ease of use and inherent stability of the topology makes it a good first choice for efficient inductive switching regulator solutions. This simple topology can be used in many different configurations, sometimes outside their intended use. However, there are difficulties to overcome, and understanding their limitations can help maximize the system performance.

Figure 1: Control is based around a comparator with a predetermined hysteresis.

This article explores the topology in-depth, using examples of different circuit configurations and discusses some inherent problems and their impact on particular applications.

The hysteretic converter is essentially an on-off topology. It can be used in buck, boost or buck-boost configurations, but its inherent stability makes it most suitable for step-down LED driving applications, since hysteretic converters stabilize within one oscillation, while pulse-width modulation (PWM) controllers, for example, usually take many tens of cycles to regulate. The hysteretic converter can be described in terms of control mechanism, accuracy, frequency, duty cycle and propagation delay.

With reference to Figure 1, control is based around a comparator with a predetermined hysteresis. The current in the LED is usually measured with a resistor and ramps up and down to the limits set by the comparator. Setting the level is a trade-off between measurement accuracy/noise immunity and efficiency. Typically a hysteresis voltage between 50mV and 250mV is chosen.

Frequency of oscillation is dependant on many factors, inductor choice being the most important. One of the key points about hysteretic converters is they are self oscillating. This means the frequency will vary due to input voltage, LED current and the number of LEDs being driven. They are however always running in continuous mode, which means the inductor never saturates or is completely drained of current. This inherent stability means hysteretic converters operate over a wide voltage range and do not require compensation with external components. They are also not limited to duty cycle ranges as with many PWM topologies.

Duty cycle however does affect accuracy. The duty cycle is essentially governed by the ratio of the input voltage to the output voltage, which in turn is usually determined by the number of LEDs being driven compared to the input voltage. For example, a high input voltage e.g. 30V driving a single 3V LED will produce a duty cycle of 10 percent. Whereas nine 3V LEDs at (27V forward voltage) on a 30V supply produce a 90 percent duty cycle. The second situation produces a more efficient solution. The problem with both of these extremes is the current in the LED is averaged from the hysteretic (ripple) sense voltage which at 50 percent duty cycle is approximately an equal triangle. At the extremes of the duty cycle, effects such as propagation delay and overshoot result in the current deviating from the demand, as shown in Figure 2. Tight current control is not usually possible at duty cycles less than 20 percent and greater than 80 percent.

Propagation delay and rise time also affect the maximum frequency of operation as well as self-heating and accuracy. As the frequency increases, transition losses begin to dominate power loss in the switching element above DC losses, although this is true of any switching topology.

Accuracy considerations
PWM is the preferred method for dimming LEDs, in order to avoid changing the color of the LEDs and to provide dimming over a wide brightness range. However, to maintain accuracy across the full resolution when using inductive, hysteretic converters there are a number of factors to be considered.

Figure 2: Effects such as propagation delay and overshoot result in the current deviating from the demand.

A simplified circuit for driving white LEDs is shown in Figure 1. In this type of converter, no output smoothing capacitor is required and the LEDs are in series with the inductor. This gives an advantage in terms of speed of start-up, solution size and cost. However, the lack of an output capacitor means that energy can only be stored in the inductor. When dimming, all the energy is discharged during off-cycles and must be restored during the on-cycles.

Figure 3a shows the current in the LEDs. When the supply voltage is applied, the internal MOSFET switch conducts and the current through the sense resistor, LEDs, inductor and switch ramps up from zero to an upper threshold. On reaching this threshold the current then begins to ramp down to the lower threshold ILTH and then starts to ramp up again towards IUTH. The thresholds are determined by the sense resistor and an internal reference voltage.

Figure 3: The current in the LEDs is shown. In 3a, the current ramps up and down within upper and lower thresholds.

PWM dimming
The PWM waveform shown in Figure 3b could be the least significant bit of an 8bit signal used to control the brightness of the LEDs. For ideal dimming, driving of the PWM signal high should result in an immediate oscillation with an average value equal to IAVG and the current should drop to zero as soon as the PWM signal is taken low. The trace in Figure 3a shows there are two major contributions to error in the output current, as indicated by the shaded areas. During the initial ramp-up (shaded blue) the current should be equal to IAVG, for this period of time the average current is too low. Similarly, during the final ramp-down the current should be zero but the area shaded green shows that this is not the case. If the duty cycle of the LED current is 50 percent the rising/falling slew rates are the same and these two errors will cancel, but the duty cycle will often vary from 50 percent. If the converter performs many cycles of oscillation during the on-period of the PWM then the effect of these errors will become negligible.

Figure 4: Looking at the 25kHz trace, the error is +0.63bits, so the error in the output current would be 2.46mA. This means the output current is 6.37mA rather than 3.91mA.

Figure 5: The output current accuracy changes with the ratio of the PWM to converter oscillation frequencies at low PWM duty cycles.

At high PWM duty cycles, small errors may not be perceptible due to the response of the LED and the human eye, but at very low PWM duty cycles the errors may become significant. Figure 4 and Figure 5 show how the output current accuracy changes with the ratio of the PWM to converter oscillation frequencies at low PWM duty cycles. Each trace represents a different converter oscillation frequency; the PWM frequency is 100Hz and the PWM duty cycle is represented by the x-axis. The y-axis shows the error in the average output current in terms of the bit-resolution. So for example, with a full-scale output current of 1A and 8bit dimming, each bit would represent (1/28)*1A = 3.91mA. Looking at the 25kHz trace, the error is +0.63bits, so the error in the output current would be 0.63*3.91mA = 2.46mA. This means the output current is 6.37mA rather than 3.91mA.

Consider as an example, the ZXLD1362 LED driver driving 3.5W white LEDs from a 48V supply using a 100?H inductor. If it is PWM dimmed at 200Hz to 10bit resolution, then the output current accuracy will be as shown in Table1.

Table 1: The impact of PWM frequency and resolution on output current accuracy are listed.

When PWM dimming hysteretic converters the ratio of the PWM frequency to converter frequency determines the accuracy at low output currents. For best accuracy, it is recommended that this ratio is much greater than the number of dimming steps, i.e. the period of one PWM bit should be much greater than the period of one converter cycle. A rule of thumb is that for nbit dimming, the LED hysteretic switching frequency should be greater than 2n times the PWM frequency, ideally greater than 2(n+2). One of the key compromises is avoiding stroboscopic effects with low frequency PWM dimming and the amount of accuracy that is needed, particularly at low brightness levels or as the PWM frequency is increased relative to the converter switching frequency.

One method to improve the accuracy of PWM dimming is to use a bypass element across the LED(s), for example a PMOS, as shown in Figure 6. In this way, the inductor current always flows and hence the ramp-up and ramp-down errors are removed which will improve the accuracy, however efficiency is compromised.

Figure 6: One method to improve the accuracy of PWM dimming is to use a bypass element across the LED.

DC dimming
DC dimming is not usually used in controlling high brightness LEDs. This is due to changes in the color temperature of the LED. White LEDs however generate their color from a phosphor excited by a blue led and in this case the color is less affected by the LED current. For architectural and mood lighting color rendering may not be too important, even if the color changes a little as the intensity is reduced. In any event for white LEDs the color change during dimming will be much less than that occurring with the equivalent incandescent lamp when similarly dimmed.

Many switching controllers do not offer good dimming rangestypically 10:1 reduction from a maximum value. Because the eye responds in logarithmic fashion a 10:1 dimming in current would not produce a pleasing reduction in brightness and would still appear to be only half the full brightness. The circuit in Figure 7 shows a method which takes advantage of the simplicity, inherent stability and flexibility of the hysteretic topology to produce a DC dimming range of around 50:1.

In some architectural applications, dimming the light by reducing the input voltage would be an advantage. A simple circuit with a resistor in series with an LED would have the desired effect, but if a 12V supply is used to drive a 5W led there would be about 10W dissipated in the resistor at full brightness. The circuit in Figure 7 produces the desired effect of efficiently reducing the current whilst maintaining current control as the voltage across the two input terminals is reduced.

The converter controls the current by maintaining an average of 100mV between the terminals Vin and Isense. There is normally a single resistor in this position. A degree of current adjustment is available by over or under driving the ADJ pin. The circuit works by combining this with a P-channel MOSFET whose small signal resistance adds to the normally fixed resistance between the terminals Vin and Isense. At low voltages the Rds(on) of the MOSFET dominates the effective resistance. At higher voltages the total current is boosted by raising the ADJ pin voltage, which maximizes the dynamic range.

Figure 7: The circuit shows a method which takes advantage of the simplicity, inherent stability and flexibility of the hysteretic topology to produce a DC dimming range of around 50:1.

The MOSFET Rds(on) could have a variation between devices of about 20 percent. In practice this is likely to be around 10 percent for the overall sense resistance. This means there will be a variation between different lamps driven from the same reduced voltage. The LEDs will also vary in their intensity versus current characteristics. The effect of Rds(on) variation depends on its proportion of the total sense resistance.

The operating frequency will rise at the lower currents causing a loss of efficiency but this is not serious as the LED power is low. In this way a much smoother dimming control is possible and it does not require any more than the standard 2 pins normally fitted to the LED lamp.

The measured characteristic for two values of sense resistor is shown in Figure 8 and the circuit diagram in Figure 7.

Figure 8: The measured characteristics for two values of sense resistor are shown.

Common anode connection
For Buck LED controllers high sided current sensing is preferred which naturally positions the LED after the current sense resistor and inductor. The simplicity of the hysteretic converters provides the means to drive LEDs with a common anode scheme.

The common anode circuit is shown in Figure 9, and consists of connecting the anode of the LED directly to the supply voltage. The LED string is still in series with the sense resistor and the inductance, allowing the normal operation of the hysteric converter. The Common Anode name usually refers to a configuration with a single LED (or a set of LEDs in parallel), but the concept could be extended to a series of LEDs, or several chains of LED that share the same V+ rail.

Figure 9: The common anode circuit shown consists of connecting the anode of the LED directly to the supply voltage.

This configuration has several advantages related mainly to the circuit performance, but also to installation convenience and component count in the system. From the performance point of view, the circuit shows an improved load regulation compared to the standard buck topology. Moreover, the circuit has a lower switching frequency which reduces the switching power losses and improves efficiency.

Figure 10: Thermal management becomes simpler for multiple LED chain systems when the anodes can all sit on one heat sink at the same potential.

Thermal management also becomes simpler for multiple LED chain systems as all the anodes can all sit on one heat sink at the same potential see Figure 10. Finally, since the voltage variation on the input is reduced, the common anode configuration allows for a smaller input capacitor.

The common anode topology simplifies the installation in signage and light wall applications, where drivers are usually remotely separately from the LED chains. In such cases, the first anode of each chain is directly connected to the power supply, so a single wire is needed to connect all the chains. Nonetheless a separate wire is still needed to connect the cathodes of each chain.

Finally, the common anode enables savings not only on the wiring side, but also on the component side. There is usually a capacitor in parallel with the string of LEDs to reduce the ripple voltage across the LEDs. This is not necessary in the common anode connection as the input capacitor already fulfils that requirement. It should be noted that the supply current to the hysteretic converter passes through the LEDs but has negligible effect on efficiency.

The main disadvantage with the common anode connection in hysteretic converters is that the available LED output voltage has to be lower than the minimum operating input voltage of the hysteretic converter. This reduces the maximum number of LEDs that can be driven compared to the standard buck configuration.

Hysteretic converters can be used over extensive voltage ranges and with a wide number of LED loads. The topology is suitable for PWM or DC dimming, however to maximize circuit performance limitations must be understood. The inherent simplicity and stability can be of benefit to an increasing number of LED lighting applications.

- Alan Dodd
Senior IC Designer

Silvestro Russo
Systems Application Engineer

Kit Latham
Senior Applications Engineer

Colin Davies
Global Applications Manager

Zetex Semiconductors

George Goh
Field Application Executive
Zetex GmbH

Article Comments - Hysteretic converters for multiple L...
*? You can enter [0] more charecters.
*Verify code:


Visit Asia Webinars to learn about the latest in technology and get practical design tips.

Back to Top