FIELD OF THE INVENTION

This invention relates to tapped linear driver circuits, for example for driving an LED lighting load.

BACKGROUND OF THE INVENTION

Tapped-linear drivers (TLDs) for LEDs are well known. Unlike switch mode drivers, they do not contain high-frequency switching elements that deteriorate the electromagnetic interference (EMI) performance of the driver.

In a stacked arrangement, the LEDs to be driven are arranged in a series string, with sections of the string selectively bypassed by respective bypass switches. Each section typically comprises a series of individual LEDs. The switching operations aim to match the mains voltage at different points in during with the mains period with the required LED forward voltage. By doing so the TLD maintains high efficiency that is comparable to a switch mode driver.

Many topologies for TLDs are known. Typically, a rectifier is provided at the input for providing a fully rectified input to TLD.

The basic stacked topology uses a single current source, and has bypass switches around respective LEDs or groups of LEDs.

A disadvantage of this configuration is that all of the power dissipation occurs in the current source. A TLD has a typical efficiency of around 90%. In this way, the dissipation in a 40W TLD will be around 4W. Using surface-mounted components is highly preferred so that that all of the required heat dissipation is transferred to the copper regions of a PCB which carries the TLD.

Excessive dissipation concentrated locally at one spot causes overheating. Typically a field effect transistor (FET) or a bipolar junction transistor (BJT) is used as an element of the current source. In this way, all the power dissipation must be handled by the single transistor. In practice, the high levels of power dissipation typically require an additional heatsink and require the use of a through-hole component or several surface-mounted components in parallel or in series to distribute the heat. For example, a through-hole transistor with a metal heatsink attached is typically used. From a manufacturing point of view it means that this is a hand-mounted component.

Automated mounting in a surface mount device (SMD) pick-and-place machine is highly preferred due to the lower manufacturing cost. However, the dissipation is too high to be handled by a single SMD package on a standard printed circuit board (PCB) such as a glass reinforced epoxy laminate board (e.g. FR4), even with a large copper heat dissipating polygon around the component.

A typical feasible heat dissipation value is 1 W per package mounted on an FR4 PCB with enough copper around the components.

To overcome the heat dissipation problem, multiple FETs or BJTs may be connected in parallel and distributed over the PCB for together implementing the current source function. This solution has high cost due to over-dimensioning of the current source because of the thermal limitation of preferred PCBs.

Another issue is how to provide switching of the bypass switches. Voltage-based control of the switches involves measuring the instantaneous input voltage and controlling the switches so that the input voltage and the LED voltages are matched. By doing so the losses in the current source are minimized. However, determining the exact switching instant requires knowledge of the voltages at different points along the LED string and the tolerance of those voltages.

Current-based control of the switches is based on the principle that the switches are normally closed, and only open when a current exceeds a certain pre-set level. In this way the voltage information is not required and the switching occurs based on the current that flows through the switch. Current and voltage matching is again required to determine the correct current levels to be used.

Both methods thus have the disadvantage that the exact switching instant is always shifted with respect to the time when input voltage is in fact equal to the LED voltage.

US2009/230883 shows an example wherein several current sources are distributed over the PCB for powering the LED and comprising some additional transistors for bypassing the current sources. Such implementation corresponds to several combination of the previously indicated prior art and it needs a lot of components with an oversizing of the needed power. In addition even if the current source are distributed, one current source is working much more than the others. Another aspect of a TLD, and the stacked TLD in particular, is light flicker due to the feeding the LEDs with pulsating DC current. The frequency of that pulsation is twice the frequency of the mains voltage, so generally 100Hz or 120Hz.

This pulsation is seen as light flicker and should be be eliminated. Electrolytic capacitors are usually added in parallel with each LED section. In this way, both a high power factor and a low ripple of the LED outputs are achieved.

There is a need for a TLD architecture which addresses some of all of the issues explained above.

SUMMARY OF THE INVENTION

The invention is defined by the claims.

According to examples in accordance with a first aspect of the invention, there is provided a tapped linear LED driver arrangement, comprising:

an LED string comprising a set of LED sections mounted in series;

for a plurality of the LED sections a respective parallel current source mounted in parallel with said LED section, and a series current source mounted in series with the LED string, each parallel or series current source operating either as a short circuit, an open circuit or a current regulator,

wherein the current sources are adapted to regulate a current one at a time, while the other current sources are either in open circuit or in short circuit, with each current source being adapted to regulate the current through the preceding LED sections.

This aspect of the invention provides a stacked tapped linear driver (TLD) with distributed power dissipation by having multiple current sources, a maximum of power dissipation being made in regulation mode. The current sources are in parallel with a respective LED section, and thus take the place of a bypass switch in a more conventional arrangement. The distribution of the power loss may for example allow only surface mount device (SMD) components to be used, and may therefore be well suited for automatic assembling using a pick-and-place machine.

Each of the LED sections apart from a first LED section may have a respective parallel current source mounted in parallel, and the series current source is placed after the last LED section. The first parallel current source (which is in parallel with the second LED section) is for driving current through the first LED section, so the first LED section does not have a parallel current source. A set of sense resistors may be provided, with each sense resistor respectively mounted in series between each adjacent pair of LED sections. These sense resistors are able to sense the current flowing (as the voltage across their terminals) and this information may be used to configure the current sources.

For example, each parallel current source may comprise a current source transistor, a biasing resistor for biasing the base of the current source transistor and a control transistor which switches in dependence on the voltage across the sense resistor. In this way, the parallel current source circuit combines the functions of a bypass switch and a current source. The base voltage applied to the current source transistor is used to turn it on (for a bypass mode), turn it off, or operate it in a linear mode as a linear current source.

The series current source may be of the same type as the parallel current sources. However, alternatively, the series current source may comprise a current source transistor and a biasing zener diode for biasing the base of the current source transistor. This use of different current source designs enables the overall circuit to have improved temperature stability. For example, the series current source may have a positive temperature coefficient of current and the parallel current sources each may have a negative temperature coefficient of current.

The arrangement may further comprise a set of capacitors, with a respective capacitor in parallel with each LED section. These are used to reduce flicker.

This aspect also provides a method of driving an LED string comprising a set of LED sections mounted in series using a tapped linear LED driver arrangement, comprising:

regulating a current with current sources one at a time and setting the non-regulating current sources to a short circuit mode or to an open circuit mode, wherein the current sources comprise parallel current sources each mounted in parallel with a respective LED section, and a series current source mounted in series with the LED string, wherein each current source is adapted to drive the current through the preceding LED sections.

By using current sources one at a time, the power dissipation is spread over time. The circuit is kept simple by operating the current sources so that they also perform the function of bypass switches, having a short circuit mode and an open circuit mode.

According to examples in accordance with a second aspect of the invention, there is provided a tapped linear LED driver arrangement, comprising:

an LED string comprising a set of LED sections in series;

a respective parallel capacitor in parallel with each LED section; a series chain of further capacitors, each in parallel with a pair of adjacent LED sections;

for a plurality of the LED sections, a respective switch in parallel with the LED section, with a respective decoupling diode between one or both sides of the switch the LED section terminals; and

a series current source in series with the LED string.

The circuit enables one electrolytic capacitor to be shared between LED sections.

The switches are preferably adapted to operate in response to a voltage difference between a rectified input voltage and the cathode voltage of an associated one of the LED sections.

This enables the switches to close substantially exactly at the time when both the input voltage and the LED string voltage match, and thus has the optimal efficiency. This is achieved by sensing the voltage difference between the input voltage and the LED-string voltage.

Each of the LED sections apart from a first LED section may have a switch in parallel.

A decoupling diode is preferably connected to the cathode of each LED section, apart from a first LED section.

The parallel capacitors preferably comprise ceramic capacitors. These are low cost capacitors, able to provide the required flicker reduction. They are combined with the further capacitors which preferably comprise electrolytic capacitors.

BRIEF DESCRIPTION OF THE DRAWINGS

Examples of the invention will now be described in detail with reference to the accompanying drawings, in which:

Figure 1 shows a first known tapped linear driver architecture;

Figure 2 shows a second known tapped linear driver architecture;

Figure 3 shows a third known tapped linear driver architecture;

Figure 4 shows a first example of tapped linear driver architecture in schematic form;

Figure 5 shows a circuit design for implementing the architecture of Figure 4;

Figure 6 shows timing diagrams for explaining the operation of the architecture of Figure 4; Figure 7 shows a second example of tapped linear driver architecture in schematic form; and

Figure 8 shows a circuit design for implementing the architecture of Figure 7.

DETAILED DESCRIPTION OF THE EMBODIMENTS

According to a first aspect, the invention provides a tapped linear LED driver arrangement in which multiple current sources are provided each in parallel with a respective LED section. The current sources operate one at a time, with each current source being adapted to drive current through the preceding LED sections. This distributes the required heat dissipation over the substrate area. According to a second aspect, the invention provides a tapped linear LED driver arrangement in which a parallel capacitor is provided for each LED section, and further capacitors are provided each in parallel with a pair of adjacent LED sections. Decoupling diodes are provided at one or both sides of LED section parallel bypass switches.

Figure 1 shows the basic stacked topology. It receives as input a full wave rectified mains signal. The rectifier is not shown, but typically comprises a diode bridge circuit. A single current source 10 draws current through the LED string, which has sections Dl, D2, D3, D4. Each section comprises one or more LEDs in series, and possibly also resistors.

The number of LEDs in each section may be the same or different to the other sections, and they may each be a collection of LEDs of the same color (with different sections being different colors). Alternatively, all LEDs may be the same color, or the sections may have LEDs of different colors.

The second section D2 to the last section D4 each have a respective parallel bypass switch SI, S2, S3. These switches are operated to match the voltage across the LED string (i.e. those sections which are not bypassed) to the mains voltage at a particular point in time during each rectified half period of the mains voltage cycle. The current source 10 supplies the LEDs with the required current (preferably a sine wave for low total harmonic distortion) corresponding to the power that is to be delivered to the LEDs.

The difference between the instantaneous rectified mains voltage and the LED voltage of the connected sections is dropped across the current source 10. In this way, the energy is dissipated across the current source 10 and is defined as the product of the voltage and the current across the current source 10. One approach for controlling the switches SI to S3 is to provide voltage-based control of the switches using a controller 12. The instantaneous input voltage is measured by the controller 12 and the switches are switched in order to match the input voltage and the summed LED section voltages. By doing so the losses in the linear current source 10 are minimized.

However, determining the exact switching instant of the switches SI, S2 or S3 requires knowledge of the voltages at the junctions between the sections Dl to D4 and the tolerance of those voltages.

An alternative approach for controlling the switches SI to S3 is to use current-based control as shown in Figure 2. The second section D2 to the last section D4 have bypass switches which are normally closed. They open when a current exceeds a certain pre-set level. In this way, voltage information is not required and the switching occurs based on the current that flows through the switch. Exact matching between the current level, the section voltages and the input voltage is however required for implementing this control approach.

Both methods of controlling the switches have the disadvantage that the exact switching instant is always shifted compared to the time when the input voltage is actually equal to the voltage across the non-bypassed LED sections.

Another aspect of a TLD, and the stacked TLD in particular, is light flicker due to feeding the LEDs with pulsating DC current. A frequency of that pulsation is double the frequency of the mains voltage, namely 100Hz or 120Hz. This pulsation is seen as a light flicker.

As shown in Figure 3 electrolytic capacitors CI to C4 may be added, with one in parallel with each LED section Dl to D4. By doing so both high power factor and low ripple of the LEDs are maintained.

The parallel capacitors are decoupled by diodes dl to d3. These are in series with the main light emitting diode sections Dl to D4, with one decoupling diode at the junction between two adjacent LED sections.

The current charges the capacitors in parallel with the LED sections when the associated switch is active, and the capacitor is discharged to LED section when the switch is not active. In this way, the pulsation of the current and hence the light flicker is reduced.

The decoupling diodes dl to d3 that are in series with each switch prevent the capacitors discharging via the switches SI, S2, S3 when the switches are conducting, and when the LED section is being bypassed. A first aspect of this invention relates to the issue of local heating of the current source.

It is desired to distribute the power losses over multiple elements but without over- dimensioning the current source 10.

Figure 4 shows the approach for distributing the losses over multiple current sources.

There is again an LED string comprising a set of LED sections Dl to D4 in series. For a plurality of the LED sections, a respective parallel current source 101, 102, 103 is in parallel with the LED section, and a series current source 10 is in series with the LED string as in the examples of Figures 1 to 3.

The current sources operate one at a time, with each current source being adapted to drive current through the preceding LED sections.

As shown in Figure 4, the top LED section Dl does not need a parallel current source. Thus, each of the LED sections apart from a first LED section has a respective current source in parallel, and the further current source 10 is after the last LED section.

The sources 10, 101, 102, 103 are unidirectional sources that can only sink the current when a positive voltage is applied to the terminals of the current source. The parallel current sources are disabled if the current flows through the associated parallel LED section.

In the stacked TLD the first LED section Dl is driven by the current source 101 when the input voltage exceeds the voltage of the first LED section Dl and is still below the sum of the voltages across sections Dl and D2. The current sources have different current levels. In particular, each current source has a larger current level than the preceding current source (when progressing along the string in the direction of current flow). Thus, the current sources 101, 102, 103, 10 are configured in such way that Ιιοι ^< Ιιθ2 ^< Ιιθ3<Ιιο· Thus, when a current source is for driving current through more LED sections, it draws a larger current. When the current source 101 is activated, the voltage drop across the other current sources 102, 103, 10 is zero.

When the voltage at the rectified mains input reaches the sum of the voltage across the sections Dl and D2, the second section D2 is activated by the current source 102. At the same time the current source 101 is disabled.

The sequence repeats accordingly for the successive current sources 102,103,10. In this way a stepped current flows through the connected LED section.

By this scheme, the loss is spread between the linear elements of the circuit, namely the four current sources and is not concentrated in the single current source 10 as in the well- known stacked TLD. Figure 5 shows a physical implementation of the stacked TLD circuit of Figure 4 having a series connection of transistor current sources that are dimensioned as explained above.

The TLD is connected to the mains via a rectifier bridge 50. The current source 101 is formed using a current source transistor Ql, a biasing resistor Rl for biasing the base of the current source transistor Ql and a control transistor Ql 1. A sense resistor Rl 1 is in series with the LED sections, between the second and third sections D2, D3. The control transistor Ql 1 switches in dependence on the voltage across the sense resistor Rl 1.

The bias resistor Rl is connected between the rectified mains line and the base of the current source transistor Ql. The control transistor Ql 1 and sense resistor Rl 1 are in series between the base and emitter of the current source transistor. The base of the control transistor Ql 1 is connected to the emitter of the current source transistor.

The current source transistor Ql operates either as a short circuit, an open circuit or a current regulator under the control of the control transistor Ql 1.

The constant current through the collector of the current source transistor Ql is maintained by negative feedback on the control transistor Ql 1 by measuring the voltage across the sense resistor Rl 1.

If the voltage across the sense resistor Rl 1 exceeds the base emitter voltage Vbe of the control transistor Ql 1, it opens and the base current of the current source transistor Ql reduces to the value determined approximately as 0.7V/R11.

This means that when the current flowing through the LED string increases to a certain level, the current source transistor Ql is turned off. Thus, as the current increases, the current source transistors are turned off in sequence, and the current is driven through progressively more series-connected LED sections.

The parallel current sources all have the same structure, but operate at different threshold levels. The current source 102 has a current source transistor Q2, control transistor Q12, sense resistor R12 and bias resistor R2. The current source 103 has a current source transistor Q3, control transistor Q13, sense resistor R13 and bias resistor R3.

When the current source 101 is operating, the current source transistors Q2, Q3 and Q4 of the lower current sources 102, 103, 10 are in a conducting mode. This is because they are biased by their appropriate biasing resistors R2, R3, R4. The negative feedback control transistors Q12 and Q13 are completely de-activated because the voltage across the sense resistors R12, R13 does not reach the 0.7V that is needed to activate the control transistors Q12 and Q13. When the input voltage exceeds the sum of the voltages across the two LED sections Dl and D2, the current source 102 activates. The current of the current source 102 flows through the resistor Rl 1 and because it is larger than the current of the current source 101, the transistor Ql 1 reaches saturation and Ql turns off. The top current source 101 is thus driven into an open circuit mode.

Each parallel current source thus has an open circuit mode (which applies to those above the active current source because the current is higher than their designed current), a short circuit mode (which applies to those below the active current source because the current is lower than their designed current) and a current regulating mode.

The sequence repeats for the further current sources.

The current source 10 is slightly different as it is not in parallel with an LED section. It has a current source transistor Q4, a bias resistor R4 a zener diode Z4 and a tail resistor R14. This configuration of the current source instead of the two -transistor implementation of the parallel current sources is chosen for better temperature stability. The parallel current sources 101, 102, 103 have negative temperature coefficient of the current (TCI), meaning that the current of the sources decreases with the temperature due to the negative base emitter voltage of the control transistors. The current source 10 has a positive TCI that compensates for the negative TCI of the parallel current sources. Because there is only one current source conducting at a time this compensation is nearly perfect.

The electrical waveforms of the circuit of Figure 5 are shown in Figure 6.

The top plot shows input voltage 60 and input current 62.

The second plot shows the current I ₁₀₁ , 1 ₁₀₂ , 1 ₁₀₃ , Lo of the current sources 101, 102, 103, 10.

The third plot shows the currents of the LED sections I _DI , I _D2 , I _D 3, I _D4 - The fourth plot shows the loss in the transistors Ql, Q2, Q3 and Q4, labeled as PQ1,

PQ2, PQ3 and PQ4. These losses occur when the corresponding current source is acting.

A second aspect of the invention relates to the issue of control of switch timing.

As explained above, the known implementation of a stacked TLD based on parallel bypass switches uses either the input voltage or the current to activate and to de-activate the LED sections. In the ideal case the LED sections must be activated and de-activated right at the moment when the input voltage matches the voltage of the equivalent LED string as configured by the bypass switches. Early switching will result in a collapse of the input current. Late switching results in increased losses in the current source. Neither of the known control approaches described above provide optimal switching.

The known implementation of Figure 3 uses one parallel capacitance per LED section. Based on visual perception studies, the ripple of the current through the LEDs must be limited to 30% of the average current of the LEDs. The ripple is determined by the differential resistance of the LEDs and the parallel capacitor. As a result of the typical differential resistances of LEDs, the parallel capacitor is typically an electrolytic capacitor. From a cost point of view it is preferred to minimize the number of such capacitors for example to use one capacitor per two or three LED sections.

Unfortunately this is not possible using the existing configuration.

Figure 7 shows a TLD design which addresses these two issues, namely the exact switching instant of the LED sections and reducing the number and the size of the filter capacitors.

In this example, the LED string is again formed of 4 sections, namely Dl, D2, D3 and

D4. The LED sections are connected in series with each other, and each section is in parallel with a respective small ceramic capacitor CI, C2, C3 and C4. There are also two electrolytic capacitors CIO and C20 each in parallel across two adjacent LED sections. Thus, in this example the first electrolytic capacitor CIO is across the LED sections Dl and D2 and the second electrolytic capacitor C20 is across the LED sections D3 and D4. The LED sections are decoupled by diodes dl, d2 and d3.

The electrolytic capacitors CIO, C20 form a series chain, with each in parallel with a pair of adjacent LED sections. The decoupling diodes are between one or both sides of each switch and the LED section terminals.

Figure 8 shows a physical implementation of the circuit.

The switches SI, S2 and S3 are implemented by Darlington p-n-p transistors Q100, QlOl (switch SI) Q200, Q201 (switch S2) and Q300, Q301 (switch S3). The current source 10 is shown as an ideal unidirectional current source with a clamping diode D10.

The current source may for example comprise a MOSFET or BJT with current feedback as in the examples above.

As shown in Figure 7, the switches are driven by the positive voltage detectors 71, 72, 73. The circuits are shown in Figure 8. Each voltage detector comprises a transistor pair. One transistor Q71, Q72, Q73 of the pair is a high- voltage type transistor that is normally turned on from the internal power supply Vcc via a resistor arrangement. These transistors form current sources that drive the Darlington bypass switches. Driving the BJTs of the bypass transistor with a fixed base current results in better switching performance and minimizes power dissipation of the Darlington switches.

A first aspect of this design relates to determining the exact switching instant of the switches. It is implemented by observing the voltage difference between the corresponding voltage at the junction between LED sections and the mains. In this way, the actual voltage drop across the set of LED sections is used, rather than a predicted or modeled value.

When this difference is positive the bypass switch is opened. The implementation of this principle is shown in Figure 7.

The switches SI, S2 and S3 are normally closed and conducting. When the input voltage is below voltage the voltage across the section Dl (called LED1) there is no current flowing in the circuit. When the input voltage rises above the voltage LED1, the current that is determined by the current source 10 flows through the LED section Dl .

When the voltage of the mains exceeds the voltage (LED1+LED2) (i.e. the sum of the voltages across the first and second LED sections Dl, D2) the difference becomes positive and is sensed by the comparator 72 at the junction between LED sections D2 and D3. Switch SI then opens and the LED section D2 starts conducting via the diode dl that is connected between SI and S2.

In similar way the switching of S2 and S3 occurs.

A second aspect of this design relates to reducing the number of electrolytic capacitors. As explained above, this is achieved by adding an electrolytic capacitor across series connections of two LED sections while small parallel ceramic capacitors are placed across the individual LED sections for filtering the pulsations that occur during switching.

The LED string is decoupled from the ground via the corresponding decoupling diodes dl, d2, d3. The voltage at the bottom of the bottom LED section D4 is defined as Vin- (LED1+LED2+LED3+LED4). This becomes negative around the zero-crossing of the mains.

In this way, the two disadvantages of the existing circuits are addressed. This improvement has been achieved by re-arranging the decoupling diodes and sensing the voltage difference of the LED string and input voltage.

The invention is applicable for LED lamps, and integrated as well as standalone LED drivers.

The examples above all have four LED sections. This is purely by way of example. Other variations to the disclosed embodiments can be understood and effected by those skilled in the art in practicing the claimed invention, from a study of the drawings, the disclosure, and the appended claims. In the claims, the word "comprising" does not exclude other elements or steps, and the indefinite article "a" or "an" does not exclude a plurality. The mere fact that certain measures are recited in mutually different dependent claims does not indicate that a combination of these measured cannot be used to advantage. Any reference signs in the claims should not be construed as limiting the scope.