GB2466031A - LED lamp having rectifier and switching circuit - Google Patents

Abstract

A LED lamp comprises a capacitor Cd, a full wave rectifier D1 and a switching circuit to maintain a relatively constant DC voltage Vs to drive one or a number of LEDs connected in series D22. A switching "buck" voltage to current converter may be supplied by Vs to provide power for higher powered LED devices (see fig. 2). Protection against transients may be provided by Triac Q1 and sense resistor Ri as well as PTC resistor Rp. The LED lamp may also incorporate other features such as automatic switching by sensing ambient light, dimming, variable random switching to simulate occupancy and occupancy detection. The lamp may also have provision for colour filers to be fitted.

Description

Description

Light emitting diode (LED) lighting White LED's are becoming available in many different types, current requirements, power outputs, and colour temperatures, but all require the current to be controlled to a specific maximum level both to maintain their life and still give adequate light output. To achieve this, it is proposed to use a dropper capacitor Cd and full-wave rectifier Dl as shown in the three optional circuits shown in figures Ia, lb and Ic to drop the relatively high AC mains voltage to a suitable level. While such arrangements are already known, the range of the mains input can vary widely and the LED's can either be under run or overrun (and possibly damaged) on such excursions or cause annoying flicker. In the first instance of this application, a special low frequency switching circuit is used initially to maintain a relatively constant DC level of voltage Vs, which is just sufficient to power a simple linear current sink T2 to drive the LED's at their maximum rated current and at minimum power loss. This involves shorting the output from the full wave rectifier device should the DC level of Vs rise above a predetermined level, so that the current flowing through the AC dropper capacitor, Full-wave Rectifier Dl and blocking diode D2 into the smoothing capacitor Cs is diverted away from overcharging Cs when not required.

This switch may be a transistor, Darlington connected transistor or mosfet device TI, or other suitable efficient electronic switch, but is shown for simplicity in this case as a bipolar transistor.

The switching device is controlled by a non-inverting hysteretic voltage level detector (Schmidt trigger) comprising T6 and T7, or a non-inverting buffer as part of an integrated circuit Si.

Controlling the slew rate of the voltage at this relatively slow switching frequency by means of the addition of a negative feedback Miller' capacitor Cm prevents electromagnetic interference being generated without the need for complex bulky and expensive inductive filters such as are normally required in switching supplies, but with only minimal increase in switching loss in the switch. In fact when the LED's are switched off the actual power used by the rest of circuit can be a fraction of I watt, so the bulb can be left permanently connected to the Ac supply line even during the day, at little cost to the user or the environment. The AC current through the capacitor is altered only momentarily when this switching is done. In fact a small leading power factor current is permanently drawn, without consuming any appreciable real power, and the harmonic current is low. As much of the current in a home, office or other establishment has a slightly lagging power factor due to motors, refrigerators, transformers, this may be considered an advantage as it helps compensate for the lagging power factor. Should this be considered undesirable, a triac device may be placed in series with the capacitor, which can then be switched off or to a low phase angle sufficient to maintain the internal supplies just enough to maintain the light sensing section in normal operation.

* .. In order to minimise the components required for a number of individual current sources, the LED's are connected in series up to such a number as to make the total forward drop of the LED's in the chain just below that of the maximum collector voltage of suitably rated low cost transistor devices, both for the switch TI and the current regulator T2. The establishment of Vs * ;* * e at slightly higher than the total voltage of the LED series chain keeps the power dissipation in * current sink transistor T2 to a minimum and makes its job of maintaining a constant current * through the LED's easier. The voltage at the collector of T2 may also be fed via a diode D8 and * , resistor Rx so that, should it be higher than the threshold of the hysteretic switch SI then the voltage Vs is reduced slightly to a value just sufficient to keep the internal supplies functioning.

This reduces, the dissipation in T2 to minimum values should the total forward drop of the LED's be much lower than the value set by Zi, Ru and RI. T2 base drive resistor Rb4, emitter resistor Re and base bias diodes Dbl and Db2 form a classic linear current source, where the voltage drop across Re caused by the current through from T2 emitter plus the Vbe of T2 are constrained to match the sum of the forward voltage drops of Dbl and Db2. This circuit has a beneficial negative temperature coefficient of around -0.25% per degree C, which reduces the current through the LED's should the bulb temperature increases, helping to protect the LED's and prolong their useful life.

However the tolerances of the forward voltage drops of these various devices can build up such that the overall tolerance of the current controlled through the LED's is substantial. To ensure the LED's are not over run at worst case tolerances it is necessary to allow for this by under running them slightly, thus wasting the potential maximum light output capability of the devices.

An alternative current sink is shown in Fig Ic, whereby a very low power operational amplifier OA 1 is used to control the current in the sink transistor 12 against a reference voltage on its bob-inverting input set by Rx, Ry, and Rz. In the drawing shown, the voltage across Rz, which is a proportion of Vcc, sets the voltage across and therefore the current through Re at the desired value. The Vcc tolerance can be made smaller than that of the arrangements in Fig I a and I b, so that the current through the LED's can be controlled more accurately to within a few percent of their maximum rating. The Vcc temperature coefficient can be made negative to allow a reduction in running current at elevated temperatures as before. In the dimming or random modes the operation is slightly different in that the voltage at the junction of Rx and Ry is reduced progressively (when dimming) or nearly to zero (when switching off) by the action of T3 to allow dimming or random/ambient light switch off as in the previous description, thus reducing the reference voltage at the non-inverting input and therefore the LED current.

Either configuration maintains the current in all the LED's in the series connected chain to the same controlled value. In addition, small inexpensive LED's with high luminous efficacy now exist and these would be the preferred choice due both to the high efficacy and to the ease of cooling of a distributed heat load, rather than one single high powered device or a smaller number thereof. With a high powered LED the concentration of heat loss often requires special heatsinking such as metal clad PCB material which increases the cost considerably. Where a greater light output is required, each chain of LED's can be duplicated along with its current sink transistor T2 and current sense resistor Re, up to the level that the value of the dropper capacitor Cd will supply. If progressively more chains for greater light output are required then the dropper and smoothing capacitor sizes may be progressively increased to provide sufficient current, but still maintaining the low cost of components.

However a smaller number of higher powered LED devices can still be used by incorporating a switching buck' voltage to current converter supplied from the constant voltage source Vs *.. established above, as shown in fig.2. Where adequate provision for LED heat dissipation is made, this can be a viable solution with a reduced need for interference filtering than for a voltage to current supply switching direct off full-wave rectified and smoothed line voltage. This can be where the need for high frequency switching to keep size down, and diode switching causes greater interference problems than using simple Schottky diodes at the lower voltage. The : circuit shown is a simple switching buck converter, comprising series pass PNP bipolar * transistor or P-channel mosfet, which is switched on by pulling its base (gate) low via RiO and * ,* T9. RiO supplies sufficient base current (or gate voltage) to T1O to fully turn either type of *.:.. device on at the maximum current required for the LED chain shown in fig.2. Base-Emitter S. * * * S * resistor RI I.tums 110 off. 19 is held on by base current from R9. When Tl0 is on, current ramps up through inductor LI, the LED chain and current sense resistor Rc until the voltage across Rc reaches the base voltage of 18 which will just turn it on. When T8 turns on, it shorts the base-emitter of T9 and turns it off, which in turn switches off TI 0. At the same time the voltage on the collector of T9 rises and this is fed back as positive feedback by hysteresis resistor R.h3 onto the base of current sensing transistor T8, increasing the current into the base of 18 from that already provided by Rc and forcing it to turn further on. R5 sets the proportion of hysteresis applied. When T10 is off, the current in LI continues to flow via flyback Schottky diode Ds. However the current is now ramping down, and eventually the voltage across Rc summed with the positive feedback from Rh3 again reaches the tam-off threshold of T8, whereupon it turns off abruptly. This cycle continues and the result is a nearly constant current with a small sawtooth ripple impressed on it flowing through the LED's. This current also has a small negative temperature coefficient helping to protect and prolong the life of the LED's at elevated temperatures. Rbias may be provided to sum extra current into the base of 18, thus reducing the current threshold voltage from Rc and therefore the dissipation in Re. The switching frequency is set by the hysteresis as well as the value of LI. This frequency is chosen to be slightly higher than that audible so that any potentially unpleasant noise caused by magnetostriction in LI is inaudible. Since the current in 110 and therefore Cs is discontinuous, a small interference suppression inductance Lrf and capacitance Crf may be required. Where the more advanced control functions to be described later are required to be used with this form of circuit, 13 is connected across the base of T9 to switch the circuit fully off, and the linear input from the dimmer phototransistor P2 is connected to the base of T8.

The relatively low and reasonably constant voltage Vs established to power the current sources also makes it much easier to establish internal supplies for the control electronics when these are required, as the dissipation involved in dropping the voltage from full-wave rectified line is much reduced. In Fig Ia,b,c, a simple resistor or resistor chain Rsl and Rs2, and Zener diode Z2 can be used at low cost and little penalty of power loss due to the reduced supply voltage to the resistor chain. This makes it cost effective and power efficient to add more complex control functions to the LED bulb. The LED's D4, D5 and an indicator LED md' for random mode are connected is series with Rs2 which provides power for these from the current flowing in the RS chain. A tap in the resistor network at the junction of Rsl and Rs2 allows a higher voltage Vo to be available to supply the various phototransistors, to be described later.

The first function easily to be added is the ability for the bulb to detect the darkening of a room in the evening and switch on the light automatically. Likewise the bulb will switch off in the morning as the light level increases and thus save power. Such a function is often known as a night light or security light and is easily implemented by enabling or shutting down the current source. To switch off, the base drive to the current source is shorted out by another low powered transistor T3. Sensing ambient light level is done using a phototransistor, photodiode or photosensitive resistor P1 which may be either fed directly to the base of control transistor T3 in :. the simplest form as shown in fig. 1 A, or a level detector circuit S2 which drives this shutdown transistor T3 as shown in fig.3. Either level detector circuit may beneficially have hysteresis (Schmidt trigger) provided by Rh4 to avoid random switching or flicker on small light :1 fluctuations. The special constant voltage circuit controlling Tl still controls its output voltage under conditions of no load, by activating the shorting switch for longer and at very low power loss. * S. * S * S.. *S * * . S * S.

Such night lights are used for occupancy lighting, children's rooms, corridor and stairwell lights, for example. Should low powered incandescent lights be used, in such applications the extremely fine filament required for low power at line voltages is very easily broken by shock or the electromagnetic force of switching on at high cunent when the filament is cold and its resistance is low. The same is true to reduced extent where compact fluorescent lamps also suffer markedly reduced lifetimes due to frequent switching on, which damages the heater elements in the ends of the tube. However LED's show no decrease in useful life due to frequent switching; in fact it may enhance their life due to better cooling and reduced on time.

This ability to be frequently switched off and on without damage enables the LED bulb to perform the second optional function, that of a randomly switched light. Random switching gives the appearance of a home or office being occupied, and therefore enhances security. This random switching may be achieved by controlling the current source from a pseudorandom binary sequence generator timed by a low frequency clock oscillator. Such a circuit comprises shift registers SRi and SR2 clocked by this low frequency clock, with a feedback loop comprising Exclusive NOR gates, which select various outputs of the shift register to feed back to the data input as shown in fig.5. Such randomising circuits are well known to those skilled in the art. The low frequency clock oscillator is constructed using a modified Schmidt Trigger circuit. Its frequency can be increased by means of a rate' input from many minutes to fractions of a second per cycle. At the lower rate end, it simulates someone moving around their premises and switching on and off a light as they do so. At the higher frequency end the flickering light simulates someone who may be watching TV. Intermediate settings can be chosen to suit individual requirements, such as decorative and strobe lighting in pubs, clubs and so on. The rate input method will be described later.

The third function provided by the LED bulb is that of dimming. The LED lighting can be locally quite intense and for applications such as task or mood' lighting it may be useful for the user of the lamp to be able to reduce this intensity to suit his or her individual requirements. A dimming input voltage is provided through a high value resistor Rd from Phototransistor P2 to the switch off transistor 13, with heavy negative feedback applied by Rn connected from collector to base of T3 to approximately linearise it. This causes the current in the current source to reduce progressively as the dimming input voltage is increased, thus dimming the LED output.

A fourth optional function is the inclusion of a passive infra red (PIR) sensor circuit where again the availability of a simple and efficient regulated supply voltage makes it economic to add this at little extra cost, so the light would only be switched on when both the ambient light is low and the PIR has detected the presence of a human body, thus realising the maximum energy saving.

A timer circuit would be require to keep the LED's on for a period after triggering. Both such circuits are well known to those skilled in the art, and would control the base of T3 via another *. base resistor Rr2. * .* *iS'

*... Both the rate and dimming functions can be realised by a high value potentiometer connected across the Vs to internal ground lines, with its wiper connected to either the oscillator circuit or : the shutdown transistor respectively. However one aim of this application is to achieve a very long product life equivalent to that provided by the LED's themselves, which might be ten to * twenty years. Low cost miniature potentiometers rely on a sliding mechanical contact and for * * this to be maintained reliably over many cycles and years of operation is difficult. A unique *.:.. method of control of these functions is proposed, that of using a pair of optoswitches' ** * S * * * .5 comprising LED's D4 and 1)5 which direct their light across a small gap onto phototransistors P2 and P3 respectively. In either case, this light transmission, is occluded progressively by a gently tapering wedge which may be slid in or out of the gap by the user to increase or decrease the occlusion and therefore conduction of P2 or P3 to set either the random function or the dimming level. An additional function may be made possible by making the end of the dimmer wedge shaped with an extra flange in such a way as to occlude both the phototransistor P2 and the ambient light sensor P1 at the full extended position. This overrides both of these functions and turns on the LED's fully for 100% of the time. This override capability could be. useful for task light to enhance natural lighting for detailed work.

Protection of a long life light bulb's circuit components is essential to prevent damage due to excess current caused by transient voltages on the line. The dropper capacitor Cd has a very low impedance to such transients and should these occur due for example to bad contacts or other reasons such as high load switching, or even possibly an attempt to operate the LED bulb on an external Phase-controlled dimmer circuit, the current into rectifier Dl and switch Ti could be excessive. Either these devices must be rated for such transients, adding cost and size, or an alternative protection scheme provided. Such an alternative scheme is proposed, using a triac Qi connected across the AC input of rectifier Dl. A low value current sense resistor Ri is connected in series with the AC line in and between the triac Qi s gate and Main terminal 2. Should a fast high transient be applied to the AC line in, it will couple through Cd as a high current. This high current will create sufficient voltage across Ri to turn on QI very quickly, thus shorting the input to the rectifier Dl and diverting the excess current away from both Dl and switch TI. After the transient current has passed, Qi s current will drop below its holding value and it will turn off again. A second series resistor Rp is added in series wth either of the Ac lines to limit the current pulse to within the safe transient current rating of Qi. For normal short and infrequent transients Rp will absorb the energy without a problem, but should these transients be repeated continually Rp will start to heat up due to the increased dissipation. It is therefore proposed that the device be a positive temperature coefficient resistor (PTC) rather than a simple power resistor. In this way, if an excess of energy occurs due to repeated transients such as attempting to power the bulb from an external phase controlled dimmer, the PTC will heat up and switch to a high resistance, thus protecting the bulb. Once the transient source is terminated the PlC Rp will cool down again and the bulb may be operated normally. The PTC will also switch to high resistance in the unlikely event that the capacitor Cd should fail to short circuit. As a further precaution a fusible link or fusible resistor may be added in series with either line.

Mechanical construction of the proposed LED bulb comprises a printed circuit board (PCB) onto which the various electronic components are soldered. This is shown in the drawing Fig.6. The PCB fits into a clear moulded non-flammable plastic package made of two identical sections of clamshell' construction which clip together around the. PCB, with a wire connection to the terminals of the bayonet or Edison screw base light fitting from the AC terminals L and N of the *.. bulb circuit, all made as part of the two piece clamshell shapes. Such a plastic case may have surface texturing to diffuse the LED light and cover the electronic components inside. The chain of LED's used may be formed into a matrix of say 4x4, 8x2 or 5x3 on one or other (or both) sides of the PCB. However greater or lesser numbers of LED's may be used and in differing array shapes. In Fig.6 the PCB is shown fully populated with 64 LED's in four 2x8 arrays, but * smaller arrays are an option for lower power requirements. Larger number of LED's can be used in each string if the Vs voltage is increased. The function adjusting wedges are also shown in * , Fig.6 and are small moulded pieces of opaque and optionally differently coloured plastic with a *.:.. long slot in them. This slot will mate precisely with a protruded section in each side of the case S. * * . . * in' such a way. that when the two front and rear sections of the case are clipped together (with the PCB inside) the two wedges are held in exactly the right position relative to the optoswitches mounted on the PCB to slide in between the LED and Phototransistor. With the wedges fully pushed inwards towards the centre of the case, the phototransistors P2 and P3 are fully occluded from the light of LED's D4 and D5 respectively, by the plastic Dimming' and Random' wedges. As either or both of the wedges are pulled out, the occlusion of the phototransistors is decreased and they start to turn on from the increasing light from their relevant LED's. At full extension of the random function wedge, P3 is fully on and the random frequency is increased to maximum due to the increased current fed from P3 via limiting resistor RI to charge the oscillator capacitor Co. The dimmer wedge allows the P2 phototransistor to turn fully on part way through its range of movement, dimming the LED's progressively to zero light output. At this minimum light output point the power drawn from the AC line is so low (a fraction of a watt) that the bulb can usually be left connected to the line without the need for a mains switch. However at full extension of the dimmer wedge out of the optoswitch, there is an extra shield which then occludes both P2 and the ambient sensor phototransistor P1, This brings the LED's on at full brightness irrespective of the ambient light conditions, a function which may be used when it is desired to supplement ambient light, for example for task lighting involving detail work. The current, and therefore the resulting voltage from P3 is fed both to the Schmidt trigger oscillator, a Schmidt trigger gating circuit for the random pulse train, and a third Schmidt trigger in Fig.4 which controls an indicator LED via a transistor. This indicator shows when random mode has been selected. Since all three Schmidt triggers are part of the same monolithic IC, the gate thresholds are all very close to each other so that the gates enabling the pulse train and the indicator LED both switch at the same voltage. This indicator LED md is necessary because otherwise, at the slowest setting of the Random function, it could take many minutes for the random operation to be apparent.

The case is also designed in such a way that small inexpensive self adhesive coloured plastic film filters may be placed in front of the LED sections as required, thus changing the colour of the light to suit architectural or mood' lighting requirements. This is easily done due to the cool nature of the light and flat packaging.

Circuit detail.

Alternating voltage on Line-Neutral (L-N) causes current to pass from terminal L through dropper capacitor Cd, the relevant diodes in full wave rectifier Dl, blocking diode D2 to charge smoothing capacitor Cs, returning through the other conducting diode in Dl, current sense resistor Ri and PlC thermistor Rp to the neutral terminal N. Cs will charge up until it reaches a sufficient voltage Vs which can break over zener diode ZI. Once sufficient current passes through Zi, Ru and RI so that the voltage across RI reaches the base voltage of T6 (Fig. 1 a) it turns on, turning off transistor Ti and allowing base current or voltage from Rbl to turn on Ti.

Likewise in Fig. lb and c, when the voltage across RI reaches the threshold of non-inverting *:*: buffer Sl (fig. lb), its output goes high and turns on transistor Ti via current limiting resistor Rbl. Note that Ti may be a bipolar, Darlington or mosfet device. A hysteresis resistor Rhi applies positive feedback either to the base of T4 or the input to Si so that either switches on * * and off cleanly; it will not switch off again until the voltage of Vs has dropped by a certain * * * . . . *... amount so that the switching frequency is kept low. A Miller capacitor Cm is connected between ** * * * * * S. collector and. base (or Drain-gate) on Ti to slow down the rate of change of voltage on the collector (drain) of Ti to a level just slow enough that no radio frequency interference is generated. SI in this case is a spare Exclusive NOR gate from the four available in one integrated circuit package, configured as a non-inverting buffer and Schmidt trigger, the other three Exclusive NOR sections being used in the random pulse sequence generator. Each Circuit in Fig la, Fig lb or Fig lc will provide the same function of driving TI on or off depending on the DC voltage Vs. The value of Vs is chosen by selection of the breakdown voltage of Zi and values of Ru and RI such that it is slightly higher than the total forward drop of all the LED's in the series connected chain. Should the voltage at the collector of T2 be high due to low LED forward drops or low current operation, this is fed directly via D8 and Rx to the base of T4 or input of SI, overriding the seuing of Vs by ZI, Ru and RI and reduces Vs still further to save still more of the already minimal power losses.

13 is the control/switch off transistor and is operated in different modes; the first as a switch in random and ambient light control modes, and the second as a linear control device in the dimming mode. In either case it shunts away all or some of the base current for 12 respectively.

In the case where it operates as a switch the value of base current limiting resistor Rr is much lower than the value of the dimming input and causes T3 to switch on or off depending on the state of the random pulse train from gate S2 or the ambient light level. However the linearly increasing voltage from P2 is fed via a much higher value resistor Rd which acts in conjunction with linearising negative feedback resistor Rn to progressively starve 12 of base current and therefore reduce the LED current progressively, thus dimming them. In the simple circuit where the random and dimming functions are not required, fig Ia shows the ambient light sensing phototransistor connected via a resistor to the base of T3, with a hysteresis resistor Rh4 applying positive feedback to cause the switch to have an upper and lower threshold level to eliminate flicker.

In the more complex form of control, in fig. 3 the phototransistor P1 is connected with its emitter to circuit ground and its collector to the Vcc supply rail. The collector is also connected to one input of a two input Nand Schmidt trigger S2. When the ambient light reaching P1 is sufficient to turn it on enough to sink sufficient current from load resistor R12 to reduce the voltage to the lower threshold of S2, the output of S2 will go high, turning on T3 and turning off T2 and the LED's. Should the light level be too low for P1 to sink sufficient current, the voltage on S2 input goes above its upper threshold, its output goes low (depending on the digital state of the other random pulse train input) and T3 turns off, turning on 12 and enabling the LED's. So when P1 is in darkness, S2 output is low unless the pulse train exists, in which case T3 and therefore the LED's turn on and off randomly controlled by this pulse train. Rh4 applies negative feedback to the Schmidt trigger (in this case) to reduce the hysteresis level as required.

The clock oscillator shown in Fig.4 is a modification of a classic Schmidt trigger oscillator, . where capacitor Co is charged up via timing resistor Ro to the upper threshold of one input of Nand Schmidt trigger S3. The other input is disabled by being held high. At this point the Schmidt S3 switches its output to a low state, discharging Co quickly via discharge diode Dd down to the lower threshold of S3. Then the output of S3 switches high again and Co starts to :: charge back up towards the upper threshold. The cycle repeats itself continuously and produces a : low frequency clock Clock'. This clock is fed to the clock inputs of the shift register in the * random sequence generator. * .S * * * * * S * * S.

P3 is' the phototransistor in the random' optoswitch, controlled by the position of the random' control wedge. When this is occluded, the voltage on the connected inputs of Nand Schmidt triggers S4 and S5 is pulled towards ground by resistor Rt and is less than the upper threshold plus the diode forward drops of D6 and D7, so that each output is held high. As P3 is turned progressively on by the removal of the occluding wedge, the voltage fed from P3 emitter via Rf increases until the upper threshold is reached and the random pulse train from the random sequence generator circuit is enabled at output of S2 and fed to control transistor T3. At the same time, the output of S5 also goes high, turning off 15 and enabling the visible LED md' which indicates that random operation has been selected.

As P3 turns on further with its decreasing occlusion due to the further withdrawal of the wedge, the voltage on Rt increases until it reaches the forward voltage of D6 and D7 plus the threshold voltage of S3, and the rate of charge of Co then increases due to the current flowing from P3 and RL As the light falling on P3 increases with the reduction in occlusion caused by withdrawing the kiidom' wedge, the frequency of the clock oscillator therefore increases, changing the slow random mode from many minutes to fractions of a second.

Dimming is achieved by progressively removing the dimming' wedge occluding P2, so that the voltage on its emitter rises and feeds cm-rent into the base of control transistor 13 via resistor Rd, turning 13 progressively on, and reducing the current through current sink T2 thus dimming the LED's. Negative feedback resistor Rn helps to linearise this action. Both phototransistors P2 and P3 are fed from a higher voltage supply Vo appearing at the junction of Rsl and Rs2 so that a wide range of voltage change is available for both the dimming and random functions. In every case, the actual values of the resistors and other components shown throughout each of the circuits in Figs. I through Fig. 5 may easily be calculated by those skilled in the art. * *S S. * S. * ..� S *

S

* *. **S * .

S * * * S S... S. * * .. * **

In the second instance of this application, a series-connected switching control device is used to control the voltage Vs supplying the constant current supply to the LED chains, instead of the shunt device previously utilised, but employing as before the same principle of a dropper capacitor Cd supplying a full-wave rectifier Dl. This modified method allows operation with a external phase controlled dimmer and also results in higher efficiency of operation. The circuit is shown in Fig.7. Instead of TI being connected in parallel with the output of the bridge circuit, it is connected in series with the positive output of the bridge so that storage capacitor Cs is only charged when Ti is switched on. Ti may be a bipolar, Darlington, Mosfet or IGBT transistor.

Since on starting up this device is initially switched off, auxiliary supplies p' and n' are provided by capacitors Cas 1 and Cas2 rectified by diodes Dl 0 to Dl 3 and smoothed by Cfilt 1 and Cfllt2 to give positive (p) and negative (n) low voltage DC supplies without resistive power loss. This also enables a higher voltage for Vs (which would otherwise increase the resistive losses), and therefore more LED's can be run in series without the penalty of increased power loss. Vs is in this case a negative supply relative to the circuit ground and as such the current source utilises a pnp polarity transistor for 12 in place of the npn device previously used. A simple resistor Rb from supply p' provides base current or gate drive to TI to turn it on once p' reaches a sufficient voltage. The method of control of TI is to remove this switch on voltage or current by turning on transistor Th and thus switching it off. Th is fed via Rbasel from operational amplifier 0A2 once the voltage across the LED current control transistor T2 exceeds a certain voltage. This method maintains the dissipation across 12 at a minimum as the LED string voltage does not change much with varying current, thus keeping overall efficiency to a maximum. When the voltage at Th collector is pulled to the n' potential the network match i and Diatchl also pulls the voltage on the inverting input of 0A2 low, latching the 0A2 output high, Th on and therefore TI off. As in the parallel control circuit previously described, Cm limits the rate of turn on and off of TI so that no radio frequency interference is generated. As in the previous method circuit the switching frequency is low, so this does not result in an appreciable increase in switching loss.

An auxiliary circuit comprising 0A3, 4 and 5 generate a negative going pulse via Dzcd and Rzcd3 on 0A2's non-inverting input close to the zero crossing pint of the supply waveform on L and N to reset the output high and allow the next half cycle of control to take place. 0A3 is a conventional differential amplifier comprising Rdiffl through Rdiff5 which attenuates and level shifts the voltage at L,N to an identical shaped waveform referenced to the midpoint of the supply rails p and n. This voltage is full-wave rectified by resistors Rrect 1,2,3 and 4 and Diode Dfw such that a negative going full-wave rectified replica of the L,N voltage appears at the inverting input of 0A5. This is compared against a voltage set by Rzcdl and 2 at the non-inverting input such that OAS output goes low close to the zero crossing of L,N to reset 0A2 via Dzcd to its high output state, allowing Ti to turn on again. It also results in a positive-going ramp on Cramp, with a ramp rate set by Cramp and Rramp.

The negative-going full-wave rectifier output at the inverting input of 0A5 is also filtered by *. Rrefl and Cfilt3 to give a negative DC voltage at the non-inverting input of 0A6. This is .. compared against a reference set by Rref2 and 3 such that the negative output of 0A6 Iref is a.. * affected by the actual value of the L,N voltage and reduces below a certain threshold down nearly to zero to follow the input voltage. When a phase-controlled dimmer circuit supplies L,N * * and the phase angle is reduced, this means that below this threshold the current through the LED * * chain follows the reducing voltage at L,N, thus dimming the LED light output in the same manner as a conventional incandescent bulb would do.

* ** There is another section of circuitry required when phase control is used. The fast rising edges *.:. from the phase control circuit applied to Cd produce a spike of current which charges CS a. * . * I * * quickly and if* repeated frequently would otherwise overcharge it, so the fast rising edge is differentiated by Cdiff, Rbase2 and fed to an auxiliary switch off circuit Tdiff, Rbase3. This switches on Th and switches off Ti very quickly, with a slight delay set by Rdly and Cdly to allow just enough charge from Cd to charge CS. The emitter of Tdiff also feeds a latching signal via Rlatch2 and Dlatch 2 to charge Cramp at the non-inverting input of 0A2, force its output high and latch off Ti via Th until it is reset by zero crossing detector OAS. If Cramp is charged sufficiently, this does not reset every cycle and limits the overcharge to CS and allows normal operation as the phase control angle changes.

Control of the LED switch-off from the random function and automatic ambient light-sensing circuitry (which remains unchanged), is achieved by inverting the digital signal from S2 using the remaining Exclusive NOR gate from the quadruple package (previously used for SI). This is configured as an inverter by tying its other input low, and its output is fed via Rref3 to control the Iref signal to either the value set by the value of L,N or to close to zero volts for switch off.

The internal dimming function utilising photo-interrupter P2 controlled by an occluding wedge is now fed into part of the LED current reference chain Rx, Ry and Rz so that as the occluding wedge is removed the voltage at the emitter of P2 increases from its otherwise negative value close to a' towards to zero thus reducing the LED current towards zero and dimming them. * S. * . * * bSI

S

S

*.* .* * S

S

*o.s4S S ** * 4 S S. * .* * 5*

Claims (9)

1. Claims: 1. A long life high efficiency light bulb comprising one or a plurality of light emitting diodes powered direct off line and utilising a capacitor dropper and full wave rectifier to provide a constant current supply preceded by a switching constant voltage arrangement.
2. 2. A light bulb as in claim 1 which incorporates protection against line transients and operation on a phase-controlled dimmer 3. A light bulb as in claim I and 2 where the constant current supply may be either a linear or switched current source fed from the preceding constant voltage supply.
3. 3. A light bulb as in any preceding claim wherein there is incorporated a light sensitive device to switch off the light bulb automatically when the ambient light conditions are sufficiently high.
4. 4. A light bulb as in any preceding claim which incorporates adjustable control of the level of light output.
5. 5. A light bulb as in any preceding claim which incorporates random switching of its light output at variable rate.
6. 6. A light bulb as in any preceding claim which is incorporated in a clear or textured or diffuse plastic case.
7. 7. A light bulb as in any preceding claim which incorporates potentiometers to adjust brightness and random operation.
8. 8. A light bulb as in any preceding claim which incorporates occluding wedges in an optical photo switch to adjust brightness and random operation.
9. 9. A light bulb as in any preceding claim which has provision for colour filters to be fitted. * sO 0 * * *, *OS 0.... * .U..... * S. * 0U**' S. * * * ..