The Transition from Driving LED Backlights to Driving Solid-State Lighting

The secret to getting the most from LED solid-state lighting is to understand that different topologies must be implemented for different applications.

by Graham Upton, Michael Keene, and Michael Kretzmer

THE LIGHT-EMITTING DIODE (LED) has come a long way from its humble beginnings as a simple indicator light to the basis of the plethora of applications for which it is used today. Market analysis suggests that it is relatively early in the LED's development, and that as it continues to mature over the next few years, we will see numerous industries greatly transformed by the introduction of today's higher-efficacy, lower-cost LEDs. Those of us in the display industry have watched as LEDs slowly infiltrated some of the smaller display backlights and recently spread to the medium- and larger-sized LCDs.

As the LED is adopted into different mediums, special attention should be paid to the way in which LEDs are electrically controlled in order to maximize the total value of the design. Although great advances have been made in IC and LED driver design, there is no one-size-fits-all LED driver. Today's savviest designers and integrators let their application needs and functional requirements dictate the best type of driver to install in their application. Prior to selecting an LED driver for an original-equipment-manufacturer (OEM) display, for example, Endicott Research Group encourages integrators to prioritize the features that will influence the selection. Some of the key parameters include size, cost, efficiency, power, noise, dimming, and flexibility. Often the driver design will be partially dictated by decisions made higher up in the supply chain (for example, LED string configurations). This phenomenon is in no way unique to the display industry and factors into other areas of solid-state lighting (SSL). This article will highlight some of the standard topologies currently employed to drive OEM LED-backlit displays and move on to discuss some of the major hurdles that inhibit a generic LED driver from being implemented directly into general lighting.

AC/DC LED Drivers for SSL

The International Electrotechnical Commission (IEC) and American National Standards Institute (ANSI) have defined requirements for power-factor and input-current harmonics generated by power supplies fed by AC line voltage. These requirements vary depending on application, and both agencies have specific requirements for lighting equipment. Power factor is defined as the ratio of real to apparent power being consumed by an electronic device, which speaks to the phase relationship between current drawn from the line and line voltage. When the input current consumed by an electronic device is linearly proportional to and in phase with the input voltage (as with a purely resistive load), the power factor is said to be unity (or 1) [Fig. 1(a)]. A power factor of unity is ideal because it reduces distribution losses and contributes little noise and distortion to the line.

When a load is composed of either non-linear or reactive circuit elements, the power factor deviates from unity. If a load is linear, capacitive, and resistive, the input current will lead the input voltage [Fig. 1(b)]. If a load is linear, inductive, and resistive, the input current will lag the input voltage. In either case, the apparent power (the instantaneous product of voltage and current) will not equal the actual power. One reason why agencies such as IEC and ANSI have imposed limits on the difference between real power and apparent power is that many power utility meters measure apparent power.

Bridge Rectifier and Bulk Capacitor in AC/DC Converters

Before power factor and current harmonic limitations were adopted, the majority of single-phase power supplies consisted of a bridge rectifier and bulk capacitor, which would convert the incoming AC into DC with some ripple. That ripple would then be filtered out by additional passive filters or active regulators, creating smooth DC power. While this circuitry is inexpensive and simple to imple-ment (and still commonly used in low-wattage supplies today), the combination of bridge rectifier and bulk capacitor consumes nearly all of the input current during a brief period when the sinusoidal input voltage waveform is at its peak, creating a highly distorted input-current waveform [Fig. 1(c)]. The reason this brief current draw occurs is that the bridge rectifier is not forward-biased and conducting until the instantaneous line voltage exceeds the DC voltage across the bulk capacitor. (As stated, this current draw only occurs at the peak of the input voltage waveform.) Consequently, these input-current spikes reduce the efficiency of the power distribution network because the instantaneous power lost in the line is equal to the square of the input current multiplied by the line resistance. In addition to this reduction in efficiency, the line voltage becomes distorted near the peak of the sinusoidal waveform [notable in Figs. 1(c) and 1(d)] when the distribution network is excessively loaded with devices utilizing this configuration of AC/DC conversion. In order to control the current draw from the line, an active power-factor-correction (PFC) circuit may be used. Figure 1(d) shows the typical input current draw for an active PFC circuit.

Power-Factor-Correction Circuit

In order to reduce the input-current harmonics that the bridge rectifier and bulk capacitor AC/DC conversion circuit create, and to keep the current drawn by the power supply in phase with the input voltage, a circuit very similar to that shown in Fig. 2 can be utilized. This circuit, based upon a pulse-width-modulating (PWM) boost converter, is popular and well documented, although it is not the only currently utilized active PFC topology.

 

Fig_1a (a)    Fig_1b (b)


Fig_1c (c)    Fig_1d (d)

Fig. 1: In charts (a)–(d), the input voltage is represented by the blue line and input current by the pink line. (a) Unity power factor. (b) Non-unity power factor. A 90° phase shift between input current and input voltage is the result of a capacitive load. (c) Input voltage and current waveforms typical of the bridge rectifier and bulk capacitor combination. (d) Typical input voltage and current waveforms for a PFC circuit.

 

Fig_2

Fig. 2: This typical resistive network configuration shows the non-isolated PFC stage.

 

The main difference between a "boosting" PFC and the boost converter is that the PFC circuit modulates its input-current consumption in order to create an input-current waveform that approximates that of a resistive load. In Fig. 2, this input-current shaping is achieved by sampling the input voltage (in the U.S., a 120-Hz rectified version of the input voltage) and modulating Q1 on time so that the circuit consumes a nearly sinusoidal input current that matches the input voltage in frequency and phase. The resistive network seen in Fig. 2 is not always the scheme used for determining the correct modulation of input current, but is a popular one nonetheless. The main reason a boost converter is the basis of most active PFC circuits utilized today is that the boost can draw input current from the line from just above 0 V to the peak of the AC waveform, limiting input-current distortion.

PFC circuits generally require a large amount of output filtering capacitance (C1) because they pass a large amount of low frequency to the output when modulating the input current. These circuits also generate very high output voltages, typically 200–400 V or more (they can only boost the 110 or 220 V coming in!), requiring the addition of a conversion stage between the PFC output and the load circuitry. The high output voltage of the boost converter is not generally desired; it is merely a by-product of the architecture. In addition to the current-control circuitry an active PFC incorporates, nearly all switching converters connected to the line require a complex filter at their input to reduce the noise they generate back to the utility and to filter the noise coming off the line.

The Two-Stage Approach to AC/DC Conversion

The typical loads that an AC/DC converter driving a series string of LEDs will power often run on 48 V or less (nowhere near the 200 V a non-isolated PFC converter will output in many cases). Therefore, a DC/DC converter (usually one that isolates the output from the input via a transformer) generally follows the PFC and is used to convert the high-voltage output from the power-factor corrector to a voltage or current suitable to drive the load. The advantage of this configuration is that the converter driving the load directly is supplied with a mostly DC input voltage, allowing the supply to have high-input rejection and high-load regulation. If properly configured, the power stage is capable of driving multiple LED strings and providing an isolated output. One drawback to this configuration is that the losses from the PFC and DC/DC converter are in series, typically limiting the efficiency to the high 70% and low 80% range. Because LEDs are a disruptive technology for general lighting – and because expectations about the energy savings inherent in switching to SSL are so high – driving at a high efficiency is critical. This configuration also requires a significant number of components and board layout space for two power converters, which results in a large footprint and high cost.

A Single-Stage Approach to AC/DC Conversion

One way to limit power-supply footprint and part count is to change the configuration of the PFC from a boost to a flyback converter and to add a stage of isolation to the feedback network. This circuit (Fig. 3) achieves AC/DC conversion and enables the PFC stage to create an output voltage that is greater or less than the input voltage. The circuit also achieves output isolation, which is necessary in many general lighting applications. The flyback transformer (T2), in which power is transferred from line to load, operates nearly the same as the boosting inductor in Fig. 2, with current flowing in the primary during Q2 on-time and in the secondary during Q2 off-time. Considering input current, the main difference between the boost and flyback configuration is that the flyback circuit does not draw current when Q2 is off. The rest of the circuit operates in the same way as a non-isolated boost PFC. This configuration, if properly implemented, can achieve efficiencies in the high 80% to low 90% range and can reduce cost when compared with the two-stage approach. Poor line rejection and mediocre load regulation are inherent to this circuit's operation; therefore, it is only recommended for driving loads such as LEDs that are not highly susceptible to moderate ripple and voltage variations. This circuit forms the basis of a single-string current source, as well as a multiple-string current source, which will be discussed shortly.

Those with experience in driving LEDs know that it is not enough simply to have clean DC power because LEDs need to be driven by a constant-current source in order to avoid thermal runaway and other failure-mode effects. Indeed, the low-voltage output described above does require additional hardware to provide the constant current for the LEDs. Regardless of how it is labeled, this style of low-voltage output is a popular mode of powering LEDs be-cause it allows the integrator a great deal of flexibility with the load and reduces the number of different power supplies that have to be stocked. For example, a lighting designer may configure from 1 to 11 strings of up to 6 or 7 LEDs to a single 100-W 24-V single-output power supply. Of course, each string requires a constant-current provider. The alternate method is to use a power supply that includes a constant-current source onboard, but one is needed for each string of LEDs that must be driven. Both methods have several advantages and disadvantages, as shown in Table 1.

 

Fig_3

Fig. 3: This PFC incorporates isolated output.

 

Power-Factor-Correction Circuit with Current Output

With a simple modification that converts the output-voltage feedback network to an output-current feedback network, the power-factor- correction circuit shown in Fig. 3 will maintain a constant output current as opposed to a constant output voltage. This change in the feedback configuration is identical to that between the PWM voltage boost and PWM current boost (both driven by DC as opposed to AC sources, and neither requiring output isolation). As with the PWM boost, this circuit will only drive one LED string. As discussed for the previous circuit, the single-stage approach will not achieve the input rejection or load regulation that the two-stage approach is capable of, but because an LED string is a static, moderately insensitive load, only marginal output regulation is required. Because calling an LED load "moderately insensitive" is a qualitative assessment, it is necessary to note that LEDs do not require an extremely low ripple supply to create a uniform and consistent perceived brightness. Moderate PWM dimming ratios are achievable, although LED current-amplitude modulation as opposed to PWM may be required in many situations. The drawbacks of PWM dimming are currently the subject of debate among many engineering and also healthcare professionals. Dimming via PWM requires that all dimming circuits be locked together (or synchronous) in frequency; otherwise, beat frequencies, or perceivable "flicker," can occur. When dimming is synchronous, the frequency must be high enough that the human eye does not perceive any "strobing" (the human eye can perceive frequencies well above 120 Hz in some cases). This "strobing" can create health issues for those with epilepsy and also pose a hazard for oper-ators of heavy machinery in industrial environ-ments where the perception of motion is key.

The Linear Current Source

Linear current sources are the most widely used current sources in low-power applications (including many OEM LED displays) due to their small size and lack of a costly magnetic component. Linear current sources are relatively easy to implement and many fully integrated adjustable voltage regulators can be easily configured as a linear current source. While there are many different linear-current-source topologies, one of the most widely used is seen in Fig. 4. This circuit is the design upon which many popular IC voltage and current regulators are based. U3 acts as the "brains" of the circuit, continuously comparing the amount of current flow through the load to a reference level and making adjustments (via Q3) if necessary. The main advantage of a linear current source is the low parts count, low cost, and relatively small size. The main disadvantage to linear current sources is that they require greater headroom (and therefore dissipate more power) than switching current sources, such as the hysteretic and PWM current sources. This power dissipation inherent to linear current sources limits them to lower power applications or applications that can tolerate large heat sinks and higher device junction temperatures.

 


   

    Input   Output  
    Two Stage
AC to DC
Single Stage
AC to DC
Constant Voltage Constant Current
   
  Pros Relatively simple circuits: boost non-isolated PFC & isolated DC to DC, high-load regulation, flexible Fully isolated flyback PFC, Highly efficient ~90%, economical, smaller sizes possible Allows for a great deal of flexibility - can "drive" several strings in parallel, fewer part numbers need to be stocked Driver is optimized for the load → Highest possible efficiencies
  Cons Inefficient (two stages in series) → ~80%, requires three stages for multiple output driver → ~70% relatively large, potentially more expensive More complex circuit, poor line rejection & load regulation → Loads should not be susceptibleto ripple/voltage variations, requires second stage for multiple output driver Requires additional hardware to provide constant current, stocking 1 part number → May be more power than you need Driver is specific to LED string configuration –lack of flexibility
AC to DC Apps Applications requiring tight output regulation with quick transient response General lighting, very-high-efficiency requirements (for energy savings payback: street lighting, parking, high bay, refrigeration, etc.), smaller package requirements, cost-sensitive applications For lighting integrators handling a large variety of lighting configurations with multiple strings High-volume applications, applications requiring the highest efficiency

 

Fig_4

Fig. 4: This type of basic op-amp-based linear current source is widely used.

 

Adaptive Power-Factor-Correction Circuit with Linear Outputs

In order to drive multiple strings with an AC input, one might consider integrating an adaptive PWM boost-circuit feedback scheme with the PFC controller. The PFC portion will work the same as the low-voltage circuit, but will regulate its output voltage based on linear-current-source headroom levels, as shown in Fig. 5. This circuit will work best in lower-power applications and applications where there is adequate heat management because the power dissipated in the linear-current-regulating devices will cause those devices to heat exces-sively unless they have an adequate junction to ambient thermal resistance. Excellent dimming ratios are achievable with this circuit utilizing PWM. The discrete linear current sources, with fast transient response, adequately reject PFC output variations and regulate the load.

The Hysteretic Current Source

Hysteretic current sources are switching current sources. A hysteretic current source will turn on for a period of time in which a rising current is delivered to the load (the inductor will determine the rate of rise). During this period, the extra "headroom" power that is normally dissipated in a linear current source is stored in the inductor. Once the rising current in the load reaches a pre-determined level (typically 10–20% above the desired output current), the circuit is disconnected from the supply and the energy stored in the inductor is delivered to the load. Once the current passing through the series combination of storage inductor and the load falls to a second pre-determined level (typically 10–20% below nominal output cur-rent), the circuit is connected to the supply again, restarting the cycle. Even though the load current is at the desired level for only two very brief moments during the ramp-up and ramp-down cycle, the average current is held to the desired level. Hysteretic current sources hold the advantage of lower power dissipation over linear current sources, thus increasing their efficiency and giving them the ability to drive higher power loads with fewer thermal considerations than linear sources. Hysteretic current sources are typically more expensive than linear current sources due to their need for a fast-switching field-effect transistor (FET), fast-clamping diode, and large inductor for energy storage. Both hysteretic and linear current sources exhibit fast transient response, which can allow for extremely high dimming ratios (Fig. 6).

Power-Factor-Correction Circuit with Hysteretic Outputs

For higher-power multi-string applications, a variation on the previous circuit with hystere-tic converters and a fixed-voltage output is feasible. This PFC circuit, as shown in Fig. 7, maintains the high efficiency and excellent thermal characteristics of the hysteretic output stages, while achieving extremely high dimming ratios. The hysteretic current source exhibits exceptional rejection of PFC output variations and excellent load regulation.

EMI Filter

EMC norms dictate maximum conducted and radiated emissions that electronic devices can create within the range of hundreds of kilohertz to tens of megahertz. In addition to the limitations on emissions, electronic devices are required to meet defined levels of immunity to conducted and radiated interference signals. Devices must also repeatedly survive transient surges of energy in order to work reliably when powered off-line. To achieve compliance with these norms and ensure reliable operation over device lifetime, a passive EMI filter is generally utilized as an interface between the wall source and AC/DC converting power supply. The design of this filter is a complex and iterative process, beyond the scope of this article.

Conclusion

As LEDs continue to gain acceptance in more and more applications, the demands on the driver continue to increase. This presents unique challenges and opportunities for driver manufacturers working to keep pace with lighting designers and integrators. The key to getting the best value from SSL is to understand that no one solution is the answer for all applications. There are a variety of different topologies, each providing the best solution for certain applications, and the desired results of the application should be considered at the outset when choosing an LED driver design. •

 

Fig_5

Fig. 5: This adaptive PFC has multiple linear-current-source outputs.

 

Fig_6

Fig. 6: The hysteretic current source is a type of switching source.

 

Fig_7

Fig. 7: A PFC can incorporate multiple hysteretic-current-source outputs.

 


Michael Keene is the Head Design Engineer for Solid-state Lighting Projects; Michael Kretzmer is Director of New Business Development, Solid State Lighting; and Graham Upton is Vice President, Engineering and New Business Development, at Endicott Research Group. They can be reached at SSL@ergpower.com.