1/9 AN1234 application note march 2001 introduction the circuits to drive cfl have usually the following block diagram: figure 1. block diagram there are three sections: an emi filter, a diode bridge rectifier that gives the rectified mains, (then smoothed by a filtering capacitor), and a half bridge inverter. there is usually no voltage pre regulator for the high voltage bus (hvb), so the bus voltage will depend on the mains. a key point of l6567 is the load current regulation according to the bus value, that means that the power in the lamp is constant, not depending on mains value. the high frequency inverter topology is a very efficient one, because of the zero voltage switching principle, that let the mosfet switching losses to be held to the minimum: just the turn on one. the capacitive mode protection implemented in l6567 helps preventing mosfet hard switching. l6567 is able to control a preheat time to make lamp ignition easier and lamp life longer. the main phases of circuit working are described in the following sections: start up, preheat, ignition and steady state condition. the circuit schematic we will refer to is shown below: input filter half bridge inverter d00in1101 l6567: design hints an integrated ballast design has been made with l6567 ic. the chosen topology is an half bridge in- verter. l6567 provides all the necessary functions for driving the external power mosfets and for pre- heat, ignition and steady state operations control of the lamp. the minimum part count required makes l6567 optimal for compact fluorescent lamp driving. the design is intended for 15w cfl (or similar one) and for 220v20% mains.
AN1234 application note 2/9 figure 2. application schematic the way to get the right components value will be shown in the last paragraph. start up as soon as the mains is applied a high voltage appears across the filtering capacitor c and the half bridge in- verter (q 1 and q 2 ). l6567 is powered through rhv: the current flows from hvb to c vcc through r hv and l6567. when c vcc voltage reaches v slow1 (max. 6v) the low side mosfet q2 is turned on while the high side mosfet q1 is turned off, in order to charge the bootstrap capacitor (c boot ). when c vcc voltage reaches v shigh1 (typ. 11.7v) the oscillator starts and the rhv pin is no more involved in providing c vcc charge: it is provided by the charge pump connected to the half bridge midpoint (pin 3 out). a high voltage capacitor is needed and it is used both for the charge pump and for snubber function. preheat phase a preheat sequence is done to assure a longer lamp life: a small current is delivered to the lamp cathodes to warm them, in order to make ignition easier. we refer to a very simple lamp model: before ignition no current flows in the lamp, and the only conductive paths are the electrodes, that can be seen as two small resistors (see fig. 3.a). after ignition, current flows between the electrodes, and the lamp can be seen as a resistor connected between them. the value of this resistance can be evaluated as the ratio between the nominal lamp power and the nominal voltage (squared) across the lamp. the equivalent load connected to the midpoint of the half bridge is shown in fig. 3.b (the filaments resis- tances have been disregarded). figure 3. lamp equivalent load l l c lamp lamp l6567 l choke c hb c hb c d 2 d 4 d 1 d 3 c boot c i c f c p r ref r shunt r 220v 50-60hz r hv c vcc c snubber- chargepump c charge pump d charge pump d charge pump q 1 13 2 3 1 6 7 9 8 10 11 12 14 q 2 5 d00in1102 l choke l choke l choke c lamp c lamp (a) (b) c lamp r lamp lamp r lamp filament r lamp filament d00in1103
3/9 AN1234 application note we will have two different transfer functions (= v lamp /v half_battery ) : figure 4. transfer functions the preheat phase is typically in the a part of the upper characteristic: here we have a few khz resonant fre- quency (see fig 3.a), small gain, so the voltage across the lamp is much smaller than the ignition one. the frequency of the oscillator is decided according to the current we want to flow in the lamp cathode. there is a pretty simple way to determine the needed preheat current when it is not specified by lamp charac- terization: if r 0 is the filament resistance at room temperature , we have to warm the filament so that after pre- heat time r(t pre )~3r 0 (as a rule of thumb). moreover we have to do it with a current that allow us to use reasonable preheat time : the end customer will not wait for a long time before the lamp being on , but t pre has not to be so short to be out of control. we can force a fixed dc current in the filament and we can measure the voltage across it: when it is three times the initial one, we have reached the needed preheat time. a simple set up is shown in fig. 5. figure 5. preheat time and current measurements set up v lamp /v hb b a c freq. d00in1104 + 40v - lamp filament under test ~1m w r (to set i) v z 20v d v i d00in110 5
AN1234 application note 4/9 we have measured different lamp types with the following results: typical preheat times are nearly 1s (0.5-1.5s). when we have set the right i pre - t pre values, we have to use them taking into account the model of fig.4. here we have a resonant circuit, and in a zone we are far from the resonant frequency: the current wave form is not sinusoidal, but is nearly triangular. using the rms. current value times r shunt you have the voltage that is compared to pin 9 internal threshold. setting r shunt we set the preheat current and, as a consequence, the preheat frequency. the preheat time is set by c p capacitor, connected to pin 8. ignition phase after preheat time has elapsed l6567 oscillator sweeps down toward lower frequency, using the b part of fig. 4 characteristic. in this way the gain increases, and the voltage across the lamp and across c lamp capacitor increases too. when the frequency approaches the resonant frequency the voltage gain is very high, and the voltage across the lamp will reach the ignition one: the lamp strikes on and the load will look like model b in fig.3: that means we are now in c part of fig. 4 lower characteristic, with a lower gain and not so near to the new resonant frequency. if the lamp doesnt ignite the oscillation frequency could cross the resonant frequency and go to the left side of the upper characteristic. the frequency range lower than the resonant frequency is dangerous for mosfet switching: they switch in capacitive mode, that means there is no more zero voltage switching, and the mosfet switch with the full hvb across source and drain. we dont have this problem with l6567: the ic provides a ca- pacitive mode protection, sensing r shunt voltage, and forcing the frequency towards higher values until we are at frequency higher than the resonant one. the ignition frequency sweep lasts the time needed to reach the set working frequency or , maximum, 15/16t pre . the sweep rate is set by c i capacitor (pin 14). burn phase when the lamp is properly ignited we are in the burn phase. the minimum oscillator frequency (f min ) is set by r ref and c f (pin 10 and 12). there are two main control functions performed by l6567: there is the capacitive mode protection that has already been enabled in the ignition phase, and there is the feed forward control. this second function mainly sets the working frequency in the burn phase. l6567 checks the rectified mains value (sensing r hv current) and changes the working frequency to maintain constant lamp power. there is also the filtering action of c p to avoid the 100hz mains ripple. without feed forward frequency sweep the high voltage bus voltage variations would be applied to the half bridge inverter, and as a consequence to the lamp: sudden thin cathode lamps thick cathode lamps r 0 [ohm] current ma 3r 0 time [s] philips 11w 15 200 1.4 230 1 sylvania 15w 12 250 0.7 300 0.3 light of america 27w 1.7 525 3.5 775 1.2 light of america 42w 3 300 2.5 390 1.5 510 0.42
5/9 AN1234 application note increase and decrease of lamp power could cause a shortening of lamp life. with feed forward control the lamp works at nearly the same power level regardless the mains variation. another feature of l6567 is the chance to set the dead time value with the resistor r ref at pin 10. setting components in this application there are components typical of nearly every ballast application, to which general rules apply: mosfet s have to be chosen taking care of the high voltage bus value as far as v dsmax is concerned, and using the lower r dson for thermal consideration. with 220v mains 500v mosfet class is ok, and r dson times max. current has to be a withstandable dissipated power. considering the high dv/dt due to the switching, nb mos are safer than na type. after choosing the lamp, p lamp and v lamp set a constrain to l choke value: l has to be the main components as far as i lamp setting: that means: c lamp has the aim to prevent v ignition across the lamp to be reached during preheat, so: v clamp = i pre x clamp (f = f pre ) << v ignition c hb capacitors are the half battery capacitors, the bigger they are the smaller the ripple of the voltage across the resonant load, 100nf is the commonest value. c boot capacitor has to be chosen according to the mos type: as a rule of thumb you can use: (see an994 for further details) mosfet have no big equvalent capacitors in this kind of application, and a 100nf capacitor is often used. the charge pump components have no special requirements, except the capacitor connected to the out node that has to withstand a voltage swing equal to the high voltage bus value, and so it has to be properly rated (i.e. 500v). the remaining six parts: r shunt , r hv , r f , c i , c p , c f are strictly related to the ic working. l6567 is able to set really a big deal of application parameters with a very few number of external components, namely the six key components listed below. as a logical consequence, the same component is not related to a single application characteristic, but to two or more. i lamp p lamp v lamp ------------------ i l v l x l ff working = () ------------------------------------------------ = = = v hb v lamp C x l ff working = () ------------------------------------------------ = l v hb v lamp C () v lamp 2 p f working p lamp ------------------------------------------------------------------ = c boot >>c mos_equ. ~ q tot_gate v gate -----------------------
AN1234 application note 6/9 the table below summarize these relationships (see l6567 datasheet for further details): there are key part (i.e. r ref ) that are related even to three parameters (i.e. f min , t pre , t dt ). the suggested order to set parameters is the following: ?set r hv considering start up current and dissipation problem; ?set c f to have the feed forward frequency range: f ff =i rhv /(k 1 c f ); ?set r ref to fix the minimum working frequency: f min = k 2 r ref c f ; ?set c p to fix the preheat time: t pre = k 3 c p r ref ; ? now we have two parameters that are related just to a parameter: r shunt to the preheat current (and fre- quency) and c i that is related to the frequency sweep rate. ? at the end we have two parameters that are related to parts already choosen: t dt = k 4 r ref and t ign = k 5 t pre . we can see a numerical example. r hv choice we begin from the start up current required to charge c vcc : it has to be greater than the ic consumption before start up (iq = 250 m a), and the greater it is the shorter the start up time is. the problem is the dissipation: the greater the current, the greater the dissipation on r hv . we have to make a compromise between these two settlements, starting from reasonable current value. we can start from: i start_up = 700 m a and c vcc = 100nf. we get: if we consider v low1 and v high1 (max. 6v and 12.7v) we have t 1 ~0.9 m s and t 2 ~1.8 m s, that are reasonable time for this kind of application. it means that the ic starts working after ~2 m s. we can use the max. rms. mains value to calculate the rc- value. if the mains is 220v20% we have: we have to check if this value gives dissipation problems: it is safer to use the peak mains voltage, so: it is cheaper to use 1/4watt resistors, so we can choose two 220kohm resistors. characteristic components burn phase minimum freq. f min r ref &c f feed forward freq. f ff c f & r hv preheat and ignition time t pre &t ign c p & r ref preheat freq. f pre r shunt & load dead time t dt r ref freq. sweep rate df/dt c i start up current r hv t start_up q i start_up ---------------------------- - vc vcc i start_up ---------------------------- - == r hv 310v 0.7a -------------- ~443k w = p diss v max 2 r hv --------------- ~ 370v 2 440k w ------------------ ~0.3w =
7/9 AN1234 application note c f choice we can choose c f in order to set the feed forward frequency range. a useful formula that fits pretty well the l6567 behavior is: where k 1 ~121 (see in l6567 datasheet the fitting between calculations and measurements). it is useful to use the datasheet characterization to select the proper frequency range. an example is the graph in fig. 6. we have to choose the desired frequency range vs. the i rhv range, as done in fig.6, and we have the c f characteristics. if the frequency ranges from 40 to 60 khz a good capacitor value is 100pf. when we measure the frequency on the board we have to measure the mosfets gate on-off frequency (i.e. pin 6). we dont have to put the probe on c f . on c f we will see a trian- gular waveform (see l6567 datasheet characterisation) but with a wrong frequency: the probe capacitor is some pf (i.e. 8pf), that is not negligible compared to a 100pf capacitor. r ref choice after choosing c f value we can set the f min value with the formula: f min = (8 r ref c f ) -1 if f min is 40 khz, we get r ref =31250ohm, that may mean a r ref commercial value of 30kohm. c p choice we fix c p value setting the preheat time: t pre = 224 c p r ref we look in the preheat current-time table shown before a good setting for a 15w lamp. we choose the following preheat condition: i pre ~250ma and t pre ~0.6-0.7ms. we get c p ~100nf. r shunt choice r shunt sets the preheat current and (as a consequence) the preheat frequency. we have already chosen i pre (~250ma) from the lamp filament characterization table. this value is a dc one, so it is also the rms. one. during preheat we work in a strongly inductive mode (see fig. 4, range a), so the cur- rent is nearly triangular shaped. with this approximation we are allowed to use the following formula: l6567 compares the peak current times r shunt with an internal threshold (~600mv typ.): f ff i rhv k 1 c f ----------------- = i rms i pp 12 ---------- = v r shunt i pp 2 -------- r shunt = 0 0.2 0.4 0.6 0.8 1.0 irhv(ma) 20 40 60 80 100 120 140 f (khz) d00in1106 rref=30k w cf=220pf 180pf 150pf 120pf 100pf 82pf 68pf 56pf 47pf figure 6. freq. vs. i rhv @ different c f
AN1234 application note 8/9 at the end we get: and we have a 1.3-1.4 ohm shunt resistor. c i choice c i value is the main factor that sets the frequency sweep rates during preheat, ignition and feed forward phase. the suggested value is 100nf (see datasheet characterization). dead time and ignition time there are formulas that relate t dt and t ign to external parts: t dt = 46.75 -12 r ref all these parts have already been set, as a consequence we have t dt ~1.4 m s and t ign ~0.6s. usually these are not key parameters, and these values are reasonable. if this is not the case we have to iterate the process, changing the order in which we set the external parts (starting from the most critical ones). with the above calculations we get the following part list: filtering parts r 47 w c3.3 m f 400v l 820 m h 140ma rectifier bridge df06n ballast parts c half_battery (2) 100nf 250v l3.1mh c lamp 3.9nf 400v mosfets (2) stp2nb50 r shunt 1.3 w c snubber-charge_pump 470pf 500v c charge_pump 680pf 50v charge pump diodes (2) bas16,1n4148 c vcc 100nf 50v r hv1 r hv2 220k w c p , c i 100nf 50v r ref 30k w c f 100pf ic l6567 r shunt ~v r shunt 0.577 i pre_rms -------------------------- t ign 15 16 ------ t pre 15 16 ------ 224 c p r ref ==
information furnished is believed to be accurate and reliable. however, stmicroelectronics assumes no responsibility for the co nsequences of use of such information nor for any infringement of patents or other rights of third parties which may result from its use. no license is granted by implication or otherwise under any patent or patent rights of stmicroelectronics. specifications mentioned in this publicati on are subject to change without notice. this publication supersedes and replaces all information previously supplied. stmicroelectronics prod ucts are not authorized for use as critical components in life support devices or systems without express written approval of stmicroelectro nics. the st logo is a registered trademark of stmicroelectronics a 2001 stmicroelectronics - all rights reserved stmicroelectronics group of companies australia - brazil - china - finland - france - germany - hong kong - india - italy - japan - malaysia - malta - morocco - sin gapore - spain - sweden - switzerland - united kingdom - u.s.a. http://www.st.com 9/9 AN1234 application note
|
