Patent Application
20080018261
Since their commercial appearance in the 1960's, light emitting diodes (LED) have become ubiquitous in electronic devices. Traditionally, LED light output was ideal for indicator applications but insufficient for general illumination. However, in recent years a great advance in the development of high-intensity LEDs has occurred. These new LEDs operate at much higher current levels than their predecessors (350 milliamps to several amperes compared to the 10-50 milliamp range for traditional LEDs). These new power LEDs produce sufficient output to make them practical as sources of illumination.
Presently, the high cost of the new power LEDs renders them best suited for applications where the unique characteristics of LEDs (ruggedness, long life, etc.) offset the extra expense. However, the cost of these high power LEDs continues to fall while efficiency (light output per unit of electrical energy in) continues to rise. Predictions are that in the near future, LEDs will be the source for general illumination, preferred over incandescent, florescent, and other arc-discharge lamps.
LEDs are a type of semiconductor device requiring direct current (DC) for operation. For optimum light output and reliability, that direct current should have a low ripple content. Since the power grid delivers alternating current (AC), a line-powered device must convert the AC to DC to power the LEDs. This conversion is called rectification. The rectifying device, or rectifier, must also operate without modification or adjustment under multiple input conditions, such as the 50- or 60-Hz utility power frequency provided in different geographic areas.
Further, LEDs are current driven rather than voltage driven devices. The driving circuit must regulate the current more precisely than the voltage supplied to the device terminals. The current regulation requirement imposes special considerations in the design of LED power supplies; most power supplies are designed to regulate voltage. Indeed, the design of the majority of integrated circuits (IC) commercially available for controlling power supplies is for voltage regulation.
Another increasingly common requirement for line-operated equipment is power factor correction (PFC). PFC devices maximize the efficiency of the power grid by making the load “seen” by the power grid “look” resistive. The efficiency of resistive loads arises from the unvarying proportionality of the instantaneous voltage to the instantaneous current at any point on the AC sinusoidal voltage waveform. Since most of Europe presently requires all new electrical equipment to be power factor (PF) corrected, the requirement is expected to be mandated in the near future within the US.
AC utility power, while always sinusoidal, is provided to the point of use in a variety of RMS voltages. In the United States, 120 VAC single-phase is the most common, although in some circumstances 240 VAC or 277 VAC single-phase and 208 VAC or 480 VAC three-phase voltages are used. In Europe, 125 and 250 VAC single-phase is prevalent and in Japan, 100 VAC. “Universal input voltage” LED power supplies must accept input voltages over some portion of this voltage range (and optimally over this entire voltage range), widened by a tolerance (typically 10% less than the minimum and 10% above the maximum). Sensing the voltage and automatic adjustment without intervention or loss of performance is another design factor.
For safety, it is desirable for the output of the power circuit (connected to the LEDs) to include galvanic isolation from the input circuit (connected to the utility power grid). The isolation averts possible current draw from the input source in the event of a short circuit on the output and should be a design requirement.
Another design requirement is for the conversion from the incoming AC line power to the regulated DC output current to be accomplished through a single conversion step controlled by one switching power semiconductor. A one-step conversion maximizes circuit efficiency, reduces cost, and raises overall reliability. Switching power conversion in the circuit design is necessary but not sufficient to satisfy the one-step conversion requirement while capitalizing on the inherent efficiency.
For increased versatility, the LED driver circuit should allow dimming the LEDs' light output. The dimming circuit should incorporate galvanic isolation from both the primary (utility input side) and secondary (LED output side) of the LED driver circuit, and should operate from a separate low-voltage power supply. This architecture increases overall system safety, allows dimming of multiple LEDs, and permits the use of low-voltage wiring techniques to lower installation costs.
Typically, the color of high-output LEDs changes when the current supplied to them changes. To satisfy the requirement of no discernable color change as the LEDs are dimmed, the dimming circuit must employ an alternate to reducing the current through the LEDs, such as pulse-width modulation.
Regulatory standards, imposed through various European governmental directives (CE Mark) and in the US by the Federal Communications Commission (FCC), must be met by all new line-powered electronic equipment. These regulations center on electromagnetic interference (EMI) both radiated through the air and conducted through the input power connection. The circuit design must be compliant to all regulations in effect in all geographic localities where the device is sold.
While the primary application of this LED driver circuit is to drive a single series string of power LEDs, it should also have the capability for driving several strings at the same or different current levels. This will allow it to work in special applications as a driver for color-changing LEDs.
Most power-factor-corrected (PFC) line-powered power supplies use boost topology because of its simplicity, low cost, and efficiency. For example, U.S. Pat. App. 20060022214 to Morgan, et al. and U.S. Pat. App. 20050231133 to Lys, and U.S. Pat. No. 6,441,558 to Muthu, et al. (2002) use such a PF correction.
In a boost PFC circuit such as this, the DC output voltage must be greater than the maximum peak input voltage, under all conditions. For example, for a PF corrected circuit designed to operate from 240 VAC mains voltage, the output voltage must be set to be greater than 340 VDC (roughly the peak voltage from the 240 VAC waveform). Typically, 400 VDC is the chosen output voltage.
LEDs are nearly constant voltage devices. That is, their forward voltage drop changes very little as their forward current fluctuates. There may also be a significant amount of variation in the forward voltage drop from one LED to another. For these reasons, current regulation must be included in circuits that drive LEDs. For low power LEDs, it is common to start with a constant voltage source, and use a series (ballast) resistor to set the current through the LED(s), for example U.S. Pat. No. 6,949,889 to Bertrand (2005); and as shown in
These reasons reflect that using a ballast resistor is not practical to drive high-power LEDs. A circuit designed to drive-high power LEDs should include a circuit that actively monitors the current in the LED string and adjusts the drive accordingly. For increased efficiency, a significant concern in high-power LED driver circuits, switching (rather than linear) power supply topologies must be used.
One traditional way to drive high-power LEDs efficiently from AC line input is to cascade a boost-PF stage with a buck current regulator stage. For example, U.S. Pat. No. 7,178,941 to Roberge, et al. (2007) uses this approach and
This approach is not an ideal for several reasons. First, the circuit requires two switching stages to convert the incoming AC line power to regulated DC LED current. There are greater switching losses and the circuit is more complex and expensive. Second, the DC output voltage from the PFC stage is typically much higher than the total series LED string voltage, resulting in a less than optimum buck LED current regulator stage. It must operate at a higher frequency than needed if the DC rail and LED string voltages were more closely matched, or a larger inductor must be used. Either alternative adds to circuit cost, complexity, and losses.
It is often desirable to have galvanic isolation between the input of a switching power supply and the output for example, U.S. Pat. No. 7,135,966 to Becattini (2006). Using a transformer to transfer the energy from the input (primary) side to the output (secondary) side is common. When regulation of the output voltage is required, a feedback signal is typically sent from the secondary side to the primary side through an optically coupled isolator. One of numerous circuit topologies used to accomplish this isolated transfer of energy is the isolated flyback topology, for example U.S. Pat. No. 5,513,088 to Williamson (1996), and shown in
In an isolated flyback circuit, the transformer doubles as the energy storing inductor; energy from the primary circuit is stored in the magnetic field of the flyback transformer via one winding during the charge time interval, and is subsequently extracted to the secondary circuit via another winding during the discharge time interval. Note that one advantage to the isolated flyback topology is that the output voltage can be matched more closely to the required load voltage during the conversion process.
Isolated flyback circuits are generally designed to produce a regulated output voltage. The conventional method of building an isolated LED driver with LED current regulation would be to cascade two switching stages, for example U.S. Pat. No. 7,178,971 to Pong, et al. (2007), and as shown in
The goal of this design is to create an AC line powered LED string driver to power the LED string at a regulated current, while using only one switching/conversion stage. It must do this over a wide range of input voltages. Additionally, the circuit must do so while providing galvanic isolation between the primary and secondary circuits while presenting a power-factor-corrected (resistive) load to the incoming utility power.
The primary of T1 “looks” like a simple inductor when Q1 is on and primary current flows because secondary rectifier D2 is reversed biased when Q1 is turned on. Consequently, T1 charges like a standard simple inductor in a typical non-isolated boost PF correction circuit (such as shown in
In a typical isolated voltage-output flyback circuit, the voltage stored on C1 is sampled using a voltage divider, and the proportional signal would be sent back across the galvanic barrier via an optocoupler to provide the controller IC (U1) with a voltage feedback signal. Regardless of whether the controller IC includes a PFC function, it would modulate the drive intervals of switch Q1 in an attempt to regulate the voltage stored on secondary storage capacitor C1. If U1 includes a PFC function, it would also modulate the conduction intervals of Q1 such that the current drawn from the line during each short conduction interval is proportional to the instantaneous line voltage during that conduction interval.
PFC control integrated circuits (as well as other power converter circuits) are available in several types, including discontinuous, continuous, and critical conduction modes. Discontinuous conduction mode PFC circuits are the simplest. The circuit typically runs at a constant frequency. It is designed to allow the inductor current to decay to zero and remain at zero for some period while the switch is off. After this delay period, the switch is turned back on to start the next cycle. The peak inductor current flow is naturally modulated by the rectified line voltage, as shown in
Continuous conduction mode PFC circuits do not allow the inductor current to decay to zero while the switch is off before the next cycle. The current in the inductor ramps up and down in a saw-tooth waveform, modulated by the rectified line voltage, as shown in
The invention described herein is applicable to all three conduction mode PFCs in addition to other power conversion circuit designs.
One key purpose for the circuit described herein is to drive a string of LEDs at a constant current level, as shown in
A primary point of departure from traditional designs in the circuit described in this patent application involves the signal fed back to the controller IC. This design does not use the voltage across the bulk capacitor, as in a traditional circuit, for the feedback to the controller IC. Instead, the current in the LED string, measured as the proportional voltage drop across a sensing resistor, is used for the feedback signal.
The design departure provides several notable differences from traditional voltage controlled output circuits:
Bulk capacitor C1 acts as an energy reservoir to buffer the conflicting requirements of power-factor-corrected input and constant-current output of the circuit design. By definition, the input power to the PFC circuit varies as the input voltage passes through complete cycles. In fact, the instantaneous input power at any phase angle along the sine wave is proportional to the square of the voltage at that phase angle. Conversely, since the LEDs are nearly constant voltage devices, driven at an essentially constant current, the output power is fixed. Hence, C1 absorbs energy when the incoming AC voltage is near its maximum magnitude, and releases energy when the incoming AC voltage is near its minimum value.
C1 also reduces the ripple in the LED string current. The LEDs are most efficient when run at a constant current. Some ripple in the current will exist, however, corresponding to the charging and discharging of capacitor C1. The greater the value of C1, the less relative ripple will exist in the LED string current.
One desirable feature for any light source, including a LED-based light source, is the ability to dim. The most obvious way to dim LEDs is to decrease the forward current through the LEDs. However, dimming by reducing the current can result in a shift in the color of the LEDs, which may be detrimental.
A better approach for dimming LEDs is by using pulse width modulation. The LED string is driven at a fixed, high current while they are on. With pulse width modulation, the LEDs turn on and off at a frequency high enough to avoid visible flicker but with reduced average light output, in proportion to the percentage of time (duty cycle) that the LEDs are emitting during each of the switching cycles.
Since the LEDs are operating at normal, high current levels when they are on, color is unaffected. This dimming technique takes advantage of the fact that the eye integrates the light that it receives. As long as the flashing frequency is sufficiently fast, the eye perceives no flicker. In practice, any flash rate over about 100 Hz is sufficiently fast for the eye's light integration to eliminate the perception of flicker while perceiving the reduced intensity level.
Many PWM dimming systems operate at low frequencies, 100-200 Hz. However, dimming at a rate in this range in a PF corrected circuit introduces unwanted problems because of the nearness of the dimming PWM rate to the rectified line frequency, typically 100 or 120 Hz. This closeness can cause the input power to fluctuate as the dimming frequency and the rectified line frequency beat against one another. The result can be a visible pulsation in the light intensity, an increase in harmonics in the current drawn by the circuit from the AC line, and/or a decrease in power factor.
One way to avoid these problems is to PWM dim at a sufficiently high frequency to prevent these beat frequency problems. Using a PWM frequency of 20 kHz or above also ensures any mechanical vibration due to the dimming signal is inaudible.
There may be advantages to using a lower frequency (such as 100-200 Hz) for collectively dimming multiple LED strings, in spite the apparent advantages of using a higher frequency (such as 20 kHz) for pulse width modulation. For example, wave shaping to reduce the EMI emitted by the distributed dimming signal is far simpler at lower frequencies. In that case, a circuit can be used to convert the low frequency distributed dimming signal to a high frequency PWM signal that actually controls the LED string currents. A microcontroller is ideal for this purpose.
However, if the same filtering and sensing circuit is used when the LED string is PWM dimmed, the average current will drop in proportion to the duty cycle of the dimming signal. The control IC will receive an indication of reduced LED current, and increase the switch (Q1) duty cycle in an attempt to compensate for the dimming.
One way to avoid this problem is shown in
An alternate method of regulating the current only during the PWM dimming “on” period is with sampling techniques, as shown in
In some circumstances, it is desirable to drive multiple series strings of LEDs with a single circuit (avoiding the expense of multiple circuits). For example, if color changing is desired, the circuit may need to drive strings of red, green, and blue LEDs. If more than one series string of LEDs are connected in parallel and driven from the same voltage source (the bulk cap, in this case), as shown in
One way of solving this problem is to insert a constant current regulator circuit at the base of each string, as shown in
A very simple form of constant current regulator is shown in
It is not necessary that all of the LED strings are regulated at the same current. By using different Base/Emitter bias resistor values, each of the strings may be set to regulate at a different current value without otherwise affecting the global operation of the circuit. This can be very useful when combining different colored strings of LEDs create unique colors; the current required by each LED string will not necessarily be equal.
In cases where the multiple LED strings must be driven at fixed current levels and never dimmed), the sensed current signals from each string's current sense resistor can be averaged together and then sensed (shown in
In order to maximize the efficiency of the circuit, it is important that the current regulator circuitry in these multiple string designs recognizes when all strings are operating at their maximum (regulated) current values, and provides no additional power to the bulk capacitor beyond this point. While the current regulator circuits for each string will continue to regulate current if more power is supplied, the additional power will simply be wasted in the regulator circuits, with the possible additional disadvantage of overheating and circuit damage.
One preferred method of detecting when all strings have reached their current regulation value is to monitor the current levels with a microcontroller. This is particularly applicable when a microcontroller is in place to generate the PWM dimming signals.
Dimming of each of multiple LED strings is possible, either as a group (to the same duty cycle or relative brightness levels) or independently (where each is set to its own level). Independent LED string dimming is particularly useful when the LED strings are of different colors, and use of differential dimming allows changing the color that results from mixing the LED strings' light outputs. When dimming multiple strings, it is still desirable to keep the “on” current of each string at the desired, pre-established level. The current measuring techniques described above (refer to
In the interest of simplifying the circuitry, the same semiconductor switch can be used to both PWM dim and regulate the current in each series LED string, as shown in
In order to limit the radiated and conducted EMI from the circuit, it is necessary to employ both line filters (for conducted noise) and shielding (for radiated noise). In many instances, these noise-limiting components can account for a large portion of both the cost and physical size of the circuit. Any circuit design features yielding a reduction of the generated EMI (and reducing the size and expense of filtering components) is very desirable.
In recent years rectifiers made from a new semiconductor material, silicon carbide (SiC) have been developed. One great advantage to SiC rectifiers is their lack of reverse recovery time. In a switching power supply circuit such as the one described herein, this lack of reverse recovery time reduces EMI generation (in this case, by the secondary rectifier). This can deliver significant reduction in the size and cost of the EMI filtering components, providing a significant cost advantage. This advantage will increase significantly as the cost of power LEDs drops and as they become the preferred solution for general illumination.
In the actual working circuit, two separate isolated low-voltage power supplies are required, to operate the circuitry on both sides of the galvanic barrier. A two-winding inductor is required by the design: two additional windings can be added to this inductor to provide the low voltage DC bias supply needed, at little additional cost.
Utility AC power, at 50 or 60 Hz and 80-310 VAC, enters the circuit at the upper left corner of the schematic. Incoming power passes though an EMI filter composed of X-capacitor C1, common mode choke L1, X-capacitor C2, and Y-capacitors C3 and C4 (which shunt noise to ground). The voltage passes though the rectifier bridge (D1, D2, D3, and D4) to filter capacitor C5, a low value ceramic capacitor serving as a short-term energy reservoir for the high frequency switching circuitry that follows.
The output from the bridge rectifier and filter capacitor passes to the primary of multi-winding inductor/transformer T1. MOSFET Q1, controlled by Power-Factor Correction IC U1, controls the current flow through T1's primary winding.
While many different PFC ICs are available, the International Rectifier part IR1150 was chosen for use in a preferred embodiment. The IR1150 offers multiple advantages, such as not needing to sample the input voltage directly and constant current mode operation without the circuit complexity usually associated with it.
U1 monitors instant incoming line voltage, measured at sensing resistor R1. A low-pass filter composed of resistor R2 and capacitor C6 remove high frequency components of the signal from R1 before presentation to the input of U1. The value of R3 sets the operating frequency of U1. Capacitors C7 and C8 and resistor R4 are compensation components that set the frequency response and establish the stability of the circuit. U1 drives the gate of MOSFET Q1 through gate resistor R5, which limits ringing on the gate of the MOSFET.
U1 uses the information from R1 and secondary LED string current information fed back via an optocoupler, to modulate the MOSFET drive signal. This dual functionality regulates secondary LED current to the correct value while the input power from the utility is drawn in a PF corrected (resistive) fashion.
T1's primary side auxiliary winding Paux provides power for the primary side bias circuitry. Diode D5 rectifies the output of this winding, and resistor R6 limits the surge current from the winding in the event of a transient. Zener diode D6 clamps the voltage at filter and bulk capacitors C9 and C10. Resistor R7 provides a low level of leakage current to charge C9 and C10 when the circuit is first energized, before power being provided by winding Paux. Regulator U2 provides a regulated 15 volts for use by the primary side circuitry. Capacitor C11 is an output capacitor required for regulator stability as well as a bypass filter for U1.
Similarly, T1's secondary side auxiliary winding Saux provides power for the secondary side bias circuitry. Diode D7 rectifies the output of this winding, and resistor R8 limits the surge current from the winding in the event of a transient. Zener diode D8 sets the voltage limit at filter and bulk capacitors C12 and C13. Regulator U3 provides a regulated 5 volts for use by the secondary side circuitry. Capacitor C14 provides required regulator stability.
The output from T1's secondary winding is fed to rectifier D9. When Q1 is on current builds through the primary winding of T1, diode D9 is reverse biased and no secondary current flows. When Q1 turns off, the polarity of T1's primary and secondary windings suddenly changes as primary current tries to continue flowing. Rectifier D9 is suddenly forward biased, and the energy stored in the primary (having no primary conduction path) transfers to the secondary, causing flow of current through D9 and charging bulk capacitor C15.
D9 must have a very short reverse recovery period. When MOSFET Q1 first turns on, reversing the polarity of the transformer windings, D9 looks like a short until the charge is swept from D9's junction. During the reverse recovery period, D9 looks like a short, reflected to the primary of T1. Because of this apparent short, very large current flows when the MOSFET first turns on, imposing high stress on the MOSFET and generating a large EMI signature. Silicon carbide rectifier D9, having no recovery period, was chosen to avoid these problems caused by conventional rectifiers.
The positive rail voltage rail stored on bulk capacitor C15 connects to the series LED strings at the output of the driver. Although only three series LED strings are shown, any reasonable number of LED strings may be employed, provided the circuit can supply sufficient power to drive them all.
Once bulk capacitor C15 has charged to a voltage greater than the minimum series LED sting voltage, that string will begin to conduct current (when its associated control transistor is turned on). As the rail voltage continues to rise, the other series LED strings will also begin to conduct as the potential exceeds the series voltage of each string (again, assuming the associated control transistor is turned on).
Transistors Q2, Q3, and Q4 are the control transistors for the three separate series LED strings shown. No control transistors are required if the circuit is driving a single LED string and dimming is not needed. The base of each of these control transistors connects to an open collector output on the microcontroller.
The microcontroller controls the individual LED strings in the following manner: If an open collector output transistor in the microcontroller turns on, the associated control transistor's base is pulled toward ground, and the control transistor (along with the connected series LED string) will be turned off.
When a microcontroller's open collector output turns off, the associated control transistor is free to operate normally. A resistor (such as R14 for Q2) pulls up the base of each control transistor but not above voltage clamp set by two series-connected diodes (D10 and/D11 for Q2). This biases the base of the transistor at two diode forward voltage drops (about 1.4 volts) above circuit ground.
One of these two diode drops compensates for the control transistor's Base-Emitter junction voltage drop, leaving approximately 0.7 volts across the current setting resistor (R15 for Q2). The value of the current setting resistor sets the control transistor's emitter current. Since the collector current (and therefore the series LED string current) is nearly the same as the emitter current, this resistor sets the LED string current for that branch.
In order to have the needed current flow in all of the series LED branches, bulk capacitor C15's charge must be to a potential greater than voltage than the highest series LED string voltage requirement. The current in each of the branches is determined by measuring the voltage across the associated current set resistors (R15 for Q2).
These current signals, filtered by a low pass filter (composed of R23 and C18 for Q2), are monitored by the microcontroller (U4), using an internal analog to digital converter (A/D). The microcontroller senses all of the connected series LED channels and sends a signal indicating the lowest channel's current back to the PFC control IC located in the primary circuit (U1). The PFC uses this signal to adjust the current to the correct value.
The LED strings are dimmed by pulse width modulation (PWM). During the on portion of the PWM cycle, the LEDs are at full intensity; eliminating current based color shift. Since it is desirable to regulate the current only during the on period (rather than averaging over the entire on/off cycle), the microcontroller only samples during the period when it has a channel turned on.
The microcontroller sends an analog signal representing the LED strings current back to the PFC control IC through digital optocoupler OPT1. The optocoupler's duty cycle is proportional to the measured LED string current. A low-pass filter, composed of R10 and C16 on the PFC side of the optocoupler, reconstructs the analog voltage corresponding to the LED string current. R9 is a pull-up resistor required by the output of the optocoupler.
The over-voltage and shutdown pin on the PFC controller IC (pin 4) is held within a nominal range by the voltage divider formed by R26 and R27. If the bulk capacitor charges up to a sufficiently high voltage (presumably due to a failure in some other portion of the circuit), the inverting input on comparator US will exceed the voltage of the reference connected to the non-inverting input. R20 and R21 divide the voltage down, and capacitor C17 is a noise filter to prevent false trips).
When an over-voltage occurs, the output of the comparator will go low, turning on optocoupler OPT2. This will pull U1's OVP pin below 0.6 volts, disabling the PFC IC's output and preventing bulk cap C15's voltage from rising any higher. Adding a latch function (if desired) will insure the circuit remains disabled after an over-voltage fault until power is cycled.
Having an external PWM dimming input to the circuit may be desirable. If so, the PWM signal would drive optocoupler OPT3. A voltage of sufficient magnitude, of either polarity, turns on optocoupler OPT3. Its output of OPT3 feeds into the microcontroller. Resistor R11 limits the current through the optocoupler's LEDs, and resistor R12 keeps noise from turning on the optocoupler. This circuit is designed such that the lack of an input from the dimming optocoupler indicates “full brightness”, and the circuit can be present without an external dimmer or further modification.
| Number | Date | Country | |
|---|---|---|---|
| 60796583 | May 2006 | US |
