[0002] Lighting fixtures and other electrical systems have a low power factor because they draw current from the alternating current (AC) mains only near its peak voltage levels, rather than throughout the cycle. Since the voltage peaks occur at the same time for all users in a given distribution network, the aggregate effect is to load the network's generators with a high current at the voltage peaks and little or no current at other times. Such loading generates harmonic distortion of the mains voltage, high neutral currents in three-phase distribution networks and the possible malfunctioning of devices operating from the mains. To avoid the line distortion, regional utility companies are forced to oversize their distribution networks, which requires a large capital investment.

[0003] Some governments are trying to relieve this problem by requiring system manufacturers to incorporate power factor correction (PFC) in some electrical systems. For example, Europe's IEC1000-3-2 specification requires PFC in lighting systems as well as the power supplies of certain other electrical devices. The PFC typically is accomplished with PFC circuits that switch the mains current through a coil at a frequency much higher than the mains frequency, and then discharge the coil current through a blocking diode into a capacitor to develop a direct current (DC) supply voltage that is further regulated to power the device or system. The current switching is controlled so that the average value of the coil, current is proportional to the AC mains voltage, i. e., in-phase

and substantially sinusoidal. This method results in power factors of. 995 or more, with 1.0 being ideal.

[0004] A significant portion of previous PFC circuits operate in a continuous conduction mode, where a new switching cycle is initiated before the previous cycle's coil current discharges to zero. Continuous conduction mode PFC systems require a high performance coil and a blocking diode with a fast recovery time in order to maintain an efficient power transfer. However, the high performance coil and blocking diode have a high cost, which increases the manufacturing cost of the continuous mode PFC systems. Moreover, these systems typically operate at a fixed switching frequency, and therefore produce a high peak energy that requires a costly filter to suppress the resulting electromagnetic interference (EMI).

[0005] Other PFC systems operate in a critical or borderline conduction mode where a new switching cycle is initiated just as the coil current reaches zero. Critical conduction mode circuits provide a high power factor but they operate over a wide switching frequency range, and require complex and costly filters to suppress the EMI. Also, under low power conditions, the switching frequency is so high that propagation delays through the PFC circuit degrade the achievable power factor.

[0006] Other PFC circuits operate in a discontinuous mode in which the coil current is allowed to decay to zero for a period of time on each switching cycle. These systems can be made to switch at a fixed frequency to reduce the EMI spectrum and allow the use of narrow band EMI filters. However, like the continuous conduction mode PFC circuits, these systems generate high peak levels of radiated energy at a single frequency that can be difficult to suppress even with the narrow band filters.

[0007] Hence, there is a need for a PFC circuit and method that switches over a controlled range in order to reduce the EMI filtering cost of an electrical system.

Brief Description of the Drawings [0008] FIG. 1 is a schematic diagram of a power factor correction (PFC) circuit; FIG. 2 is a timing diagram showing waveforms of the PFC circuit; FIG. 3 is a schematic diagram of a portion of the PFC circuit including an oscillator; FIG. 4 is a schematic diagram of the oscillator in a first alternate embodiment; FIG. 5 is a schematic diagram of the oscillator in a second alternate embodiment; FIG. 6 is a circuit diagram of the PFC circuit in an alternate embodiment; and FIG. 7 is a circuit diagram of the PFC circuit in another alternate embodiment.

Detailed Description of the Drawings [0009] In the figures, elements having the same reference number have similar functionality.

[0010] FIG. 1 is a schematic diagram of a power factor correction (PFC) circuit 100 for correcting the power factor of an alternating current (AC) mains operating at a sinusoidal AC voltage VAC while supplying a load current ILOAD to a load 28.

PFC circuit 100 is controlled by a PFC control circuit 10 operating in a discontinuous mode from a supply voltage Vcc=12. 0 volts, and includes an electromagnetic interference (EMI) filter 15, a capacitor 19, a diode bridge 20, resistors 16-18 and 45, an inductor or coil 25, a blocking diode 26 and an output capacitor 27. PFC circuit 100 produces a direct current (DC) output voltage VouT at an output node 30.

[0011] In general, PFC circuit 100 provides a high power factor for the AC mains by correcting the power factor at an

input node 32 operating at an input voltage VIN that is derived by rectifying VAC. In effect, PFC circuit 100 uses feedback to produce a resistive load between node 32 and the negative terminal of bridge 20 which, in the embodiment of FIG. 1, operates at ground potential. As a result, the average value of current flowing through node 32, and therefore the AC mains, is in phase with VIN- [0012] In particular, PFC circuit 100 functions as a step up switching regulator in which resistors 16-17 function as a voltage divider to establish a value of VouT that is boosted to a level higher than the peak level of VAC. In one embodiment, where VAC has a value of about two hundred twenty volts root- mean-square (RMS) and a frequency of about fifty hertz, PFC circuit 100 produces output voltage Vour with a value of about four hundred volts DC. In some geographical regions, where VAC has a value of about one hundred ten volts RMS and VOUT a frequency of sixty hertz, PFC circuit 100 may generate VouT at a value of about two hundred thirty volts DC. The size, breakdown voltage, etc. , of PFC circuit 100 components may be selected so that systems setting VouT at about four hundred volts DC can be operated from virtually any mains in the world. Such systems are referred to as universal mains systems. In most regions, VAC has a typical range of about plus and minus twenty percent.

[0013] In an alternative embodiment, PFC circuit 100 is configured to combine a power factor correction function with a voltage regulator in a single stage that produces VouT at a lower voltage than the peak VAC voltage. For example, resistors 16-17 may be selected so that PFC circuit 100 provides VouT at a level of, say, five volts.

[0014] EMI filter 15 is a lowpass filter that passes the low frequency component of VAC while suppressing high frequency switching signals generated by PFC circuit 100. In one embodiment, EMI filter 15 is configured to suppress signal components above about one kilohertz.

[0015] Diode bridge 20 is a standard full-wave bridge rectifier that rectifies line voltage VAC and produces a rectified sine wave input voltage VIN at node 32 with a frequency of twice the frequency of VAC or about one hundred hertz and a peak value of about three hundred ten volts.

Capacitor 19 is connected across diode bridge 20 to further reduce VAC noise.

[0016] Coil 25 has a typical inductance L25=100. 0 microhenries and a low equivalent series resistance for high efficiency operation.

[0017] PFC control circuit 10 includes a transistor 29, a pulsewidth modulated (PWM) control circuit 31 and an oscillator 35.

[0018] PWM control circuit 10 receives a clock signal CLK from oscillator 35 and initiates a series of pulses referred to as a drive signal VDRIE that switch transistor 29. Resistors 16 and 17 operate as a voltage divider that divides output voltage VouT to produce a feedback signal VFB at an input 36. In one embodiment, PWM control circuit 31 compares feedback voltage VFB with an internally generated reference voltage to modulate the widths of the VDRIE pulses. Hence, as load 28 draws an increased load current ILOAD to discharge capacitor 27 and reduce output voltage VOUT, the level of feedback voltage VFB is correspondingly lower. In response, PWM control circuit 31 increases the widths of the VDRIE pulses, which increases the charge transferred to capacitor 27 from coil 25 to regulate VouT to its specified level. Accordingly, PWM control circuit 31 is configured so that the widths of the VDRIE pulses are constant throughout a cycle of VIN if load current ILOAD is constant with respect to the frequency of VIN, or about one hundred twenty hertz. In one embodiment, PFC control circuit0 is suitable for integrating on a semiconductor die to form an integrated circuit.

[0019] Transistor 29 is a high current n-channel metal-oxide-

semiconductor field effect transistor that switches coil current ICOIL through coil 25. In one embodiment, transistor 29 is a power transistor able to switch peak values of ICOIL greater than two amperes. Transistor 29 typically has a large gate capacitance greater than five hundred picofarads. Transistor 29 is shown as being integrated on a die with other components of PFC control circuit 10, but alternatively may be formed as an external discrete device.

[0020] Coil current ICOIL has a component charging current ICHG and a component discharging current IDSCHG. The time when transistor 29 is on is referred to as a charging period TCHG during which charging current ICHG flows through coil 25 and transistor 29 to store magnetic energy in coil 25. When load current ILOAD is constant, TCHG is constant throughout a cycle of VIN. When transistor 29 switches off, the stored magnetic energy flows as discharge current IDSCHG from coil 25 through blocking diode 26 to capacitor 27 to develop output voltage VOUT on node 30. The time during which discharge current IDSCHG flows is referred to as a discharging period TDSCHG, which varies in accordance with the peak value of charging current TCHG and the voltage level of VIN- [0021] Oscillator 35 is configured as a voltage controlled oscillator that has an input 39 for sensing an input current IIN derived from input voltage VIN. Input 39 operates near ground potential so that IIN is effectively equal to VIN/R18, where R18 is the resistance of resistor 18. Since VIN has the shape of a rectified sine wave, IIN also has a rectified sinusoidal shape and is therefore representative of VIN. An output provides clock signal CLK at a frequency whose variation is dependent on IIN-In one embodiment, the magnitude of IIN is selected such that clock signal CLK varies over a range of less than two to one, which is significantly less than the switching frequency range of critical conduction mode PFC circuits, whose frequency spectrum often spans a range of twenty to one or more. In one

embodiment, oscillator 35 generates CLK with a nominal frequency of about forty kilohertz and a range from about thirty kilohertz to about fifty kilohertz.

[0022] The controlled CLK switching frequency range reduces the peak EMI radiation at any single frequency while generating a limited spectrum of EMI radiated energy to allow EMI filter 15 to be configured in a less complex and costly fashion that reduces the overall cost of PFC circuit 100. The nominal operating frequency of CLK is selected so that when operating at its highest level in response to input current IIN, the period of CLK is still low enough to operate PFC circuit 100 in a discontinuous mode, i. e. , a mode in which ICOIL is zero for a nonzero portion of a switching cycle.

[0023] Switching cycles of PFC control circuit 10 are initiated by clock signal CLK which operates with a period much smaller than the period of VIN, so a substantially constant voltage VIN appears across coil 25 during any particular switching cycle. As a result, charging current ICHG increases linearly with a slope approximately equal to V N/L to reach a peak value IPEAK=TCHG*VIN/L. Similarly, the slope of discharging current IDSCHG is substantially equal to (VOUT-VIN)/L, and its duration TDSCHG=L*IPEAK/ (VOUT-VIN) Hence, the total period when ICOIL is nonzero is given by Hence coil current ICOIL flows as a triangle wave whose average value ICOILCLK over a CLK period TCLK is given by 2 COIL-CLK'2oL Clic

where DCYCLE= (TCHG+TDSCHG)/TCLK represents the duty cycle of the nonzero coil current during each CLK period TCLK. A high power factor is achieved when the average coil current ICOIL_CLK follows the rectified sinusoidal shape of VIN, which occurs if TCHG*DCYCLE is made constant.

[0024] Since charging time TCHG is constant when load current ILOAD is constant, in order to maintain the product TCHG*DCYCLE constant and achieve a high power factor, oscillator 35 varies the switching frequency Fsw of CLK to keep DCYCLE substantially constant. The average input power <PIN> over a period of input voltage VIN is given by equation 3), 2 3) (PIN) 2 where VACS is the root-mean-square value of line voltage VAC.

When load current ILOAD is constant, PFC circuit 100 operates with average input power <PIN> being constant. Since VASES and L are constant, the constant load condition results in the product ,, * ) 4 (T *D \ 2s \ w/ (TCHG ACRMS being constant as well. From these relationships, it can be shown that the switching frequency Fsw needed to achieve a high - power factor is given by As a consequence, when switching frequency FSW is made proportional to the difference between output voltage VouT and the instantaneous rectified input voltage VIN, PFC circuit 100 operates with a PFC approaching one. In fact, under the described steady state conditions, VouT is regulated, and therefore constant, so equation 5) can be simplified to 6) FSW = K1 * (K2 - VIN), where K1 is a constant, K2=VouT and the regulation arrangement adjusts TCHG SO that at a given <PIN> and VAC

operating point. To achieve a high power factor, CLK frequency Fsw effectively is modulated with VIN so that Fsw has a lower value near the VIN peaks and a higher value when ViN is near zero volts. To accomplish this, oscillator 35 has inputs operating near ground potential, one for sensing input voltage VIN with a sense current IIN developed through resistor 18 and another for sensing output voltage VouT with a current IOUT developed through resistor 45. Oscillator 35 subtracts IIN from IouT to obtain a difference current used to establish the instantaneous value of CLK period TCLK, and therefore switching frequency Fsw.

[0025] The detailed operation of PFC circuit 100 can be seen by referring to the timing diagram of FIG. 2, showing waveforms of input voltage VIN, coil current ICOIL, drive signal VDRIE and clock signal CLK over selected switching periods (T4-TO) and (T9-T5), each having a duration in the range of about fifty microseconds, with (T9-T5) > (T4-TO). FIG. 2 shows two CLK cycles or periods TCLK, a first period that runs from time TO to time T4 and a second, longer period that runs from time T5 to time T9.

Although input voltage VIN varies as a rectified sinusoid, the CLK period is much shorter than the VIN period. Consequently, in order to better describe the invention, VIN is shown as being constant during each period, but with a value VIN1 during the first period that is lower than its value VIN2 during the second period.

[0026] Assume that initially, just prior to time TO, both CLK and VDRIE are logic low and transistor 29 and blocking diode 26 are off, so that IcoIL=0. 0 amperes.

[0027] At time TO, a first switching cycle begins as clock signal CLK transitions from a logic low level to a logic high level to initiate a pulse of drive signal VDRIE. Transistor 29 turns on to charge coil 25 with charging current ICHG at a

linearly increasing rate VIN/L25, since the voltage across transistor 29 is nearly zero, and consequently the entire voltage VIN is effectively applied across coil 25. Hence, charging current ICHG increases at a rate proportional to the instantaneous value of VIN- [0028] During the interval from time TO to T1, input signal ViN has a substantially constant voltage value VIN1, so that charging current lens increases linearly until time T1, when it reaches a peak value of IPK1=VIN1 TCHG/L25- [0029] At time T1, VDRIVE makes a transition from a high logic level to a low logic level, turning off transistor 29 to allow the energy stored in coil 25 to be transferred through blocking diode 26 to capacitor 27. The voltage dropped across blocking diode 26 is small in comparison to a voltage (VOUT-VIN), so one can consider that (VOUT-VIN1) is applied across coil 25, and that IDSCHG decreases linearly at a rate (VOUT-VIN1)/L25, until it discharges to zero at time T3=T1+IPK1*L25/ (VOUT-VIN1).

[0030] At time T2, clock signal CLK is reset from a high level to a low level, which does not cause a change in the voltage level of drive signal VDRIVE- [0031] From time T3 to time T4, ICOIL remains at zero for a nonconducting period characteristic of a discontinuous mode of operation of PFC circuit 100.

[0032] At time T4, the first switching cycle ends and another switching cycle begins. Several CLK switching cycles may follow.

[0033] At time T5, the designated second cycle commences with a low to high CLK and VDRIVE transition, but with input voltage VIN operating at a higher effective voltage value Vn>Vnn. The higher VIN2 value causes charging current ICHG to increase linearly and at a faster rate through coil 25 and transistor 29, and to reach a peak value IPK2=VIN2*TCHG/L25 at time T6 that is higher than peak value IPK1. Note that TCHG= (T 1-T O) = (T6-T5) has a constant value when ILOAD is constant.

[0034] At time T6, VDRIVE makes another high to low transition to disable transistor 29 and allow magnetic energy stored in coil 25 to be transferred as discharging current IDSCHG through blocking diode 26 for storing on capacitor 27. During the interval from time T6 to time T8, a substantially constant voltage (VOUT-VIN2) is applied across coil 25, so IDSCH decreases in a linear fashion with a slope (VOUT-VIN2)/L25, until it discharges to zero at time T8=T6+IpK2*L2s/(vouT-VIN2). Since VIN2>VIN1, coil current ICOIL reaches a higher peak current IPK2, but discharges at a slower rate (VOUT-VIN2)/L25. A second nonconducting period commences at time T8 when ICOIL discharges to zero and lasts until the second switching cycle ends and another switching cycle begins at time T9.

[0035] At time T7, clock signal CLK makes a high to low transition that does not affect the level of drive signal VDRIVE.

[0036] FIG. 3 is a circuit diagram showing a portion of PFC circuit 100 including oscillator 35 in further detail along with resistors 18 and 45. Oscillator 35 includes current mirrors 57- 60, switches 62-65, a timing capacitor 68 and a comparator 69.

Oscillator 35 is configured as a voltage controlled oscillator that produces clock signal CLK as a series of pulses generated at a nominal or center frequency that is modulated in proportion to the difference (VOUT-VIN)- [0037] Timing capacitor 68 is connected between a timing node 70 and ground potential. Capacitor 68 typically is integrated on the same die as other components of PFC control circuit 10, but alternatively may be formed as an external capacitor. In one embodiment, capacitor 68 has a value of about one hundred picofarads. Capacitor 68 is sequentially charged and discharged by currents IIM2, IIM3, IOM2 and IOM3 as described below to form a triangle or ramp voltage VRAM on node 70.

[0038] Switches 62-65 are implemented with transistors that are respectively enabled or turned on either by clock signal CLK

or a complementary clock signal CLK as shown. Hence, switches 62 and 65 are enabled or closed when CLK is logic high, while switches 63 and 64 are closed when CLK is logic high and CLK is logic low.

[0039] Comparator 69 is configured as a hysteretic comparator that compares a voltage developed on timing node 70 with a reference voltage VREF to produce clock signal CLK at its output.

Comparator 69 has outputs that provide the complementary clock signals CLK and CLK, or CLK may be derived by inverting CLK with a separate inverter (not shown). When comparator 69 is producing CLK with, for example, a logic high level, an internal hysteresis circuit reduces the comparison reference by a hysteresis amount VHYST to a value (VREF-VHYST). As a result, CLK remains logic high until VRAmp discharges to a level below (VREF- VHYST), at which point CLK transitions to a logic low. The effect of the hysteresis is that Vpmp is produced as a triangle wave that cycles between VREF and (VREF-VHYST) as shown in FIG. 2. In one embodiment in which supply voltage Vcc=12. 0 volts, VREF has a value of about three volts and VHYST has a value of about one volt, so the voltage difference (VREF-VHYST) has a level of about two volts.

[0040] Current mirrors 57-58 include scaled transistors that produce mirrored currents Ism1, Iin2/IiM3 and IIM4 that are proportional to, or multiples of, input sense current IIN.

Similarly, current mirrors 59-60 include scaled transistors that produce mirrored currents IoMl, IoM2 and Ions that are proportional to, or multiples of, output sense current IOUT- [0041] Oscillator 35 operates as follows. Assume that initially, clock signal CLK is logic low, so switches 63 and 64 are closed, switches 62 and 65 are open and VRAMP is increasing with a value less than VREF, as shown in FIG. 2. At time TO, CLK transitions to a logic high, which closes switches 62 and 65 and opens switches 63-64 to discharge capacitor 68 with current Ions

while concurrently charging with current IIM2. Current mirrors 57-60 are ratioed so that Ions is higher than IIM2, so the algebraic sum of currents Ions and IIM2 results in a net difference current (IOM3-IIM2) that discharges capacitor 68 to decrease the level of Vamp.

[0042] At time T2, VRAMP reaches the level of (VREF-VHYST), at which time CLK transitions to a logic low, which closes switches 63-64 and opens switches 62 and 65. Capacitor 68 is then charged by current IOM2 while being discharged by current IIM3.

Currents IIM3 and IOM2 are scaled so that IIM3<IoM2, which results in charging capacitor 68 with an effective difference current (IoM2-IIM3). When capacitor 68 is charged to a point where VRAMp>VREF/CLK makes a low to high transition to begin another cycle.

[0043] The scaling or mirroring ratios of current mirrors 57- 60 are further selected so that capacitor 68 is charged and discharged with currents (IoM2-IIM3) =K3* (VOUT-VIN) and (IOM3- IIM2) =K4* (VOUT-VIN), respectively, where K3 and K4 are constants.

It can be shown that switching frequency Fsw has the form shown in equation 6) above, which results in a power factor approaching one.

[0044] FIG. 4 is a circuit diagram showing a portion of PFC circuit 100 including further detail of oscillator 35 in an alternate embodiment along with resistor 18. Oscillator 35 includes current sources 80-81, current mirrors 57-58, switches 62-65, a timing capacitor 68 and a comparator 69.

[0045] Current source 80 supplies a charging reference current IREF1 from supply voltage Vcc to node 70 when CLK is high and switch 64 is closed, and current source 81 supplies a scaled or mirrored discharging reference current IREF2 to node 70 when CLK is high and switch 65 is closed. The scaling or mirroring ratios of current mirrors 57-58 and current sources 80-81 are selected so that capacitor 68 is charged with a difference

current (IIREF1-IIM3) =K5* (VREF-VIN) when CLK is high, where K5 is a constant, and discharged with a difference current (IREF2- IIM2) = K7* (VREF-VIN), where K7 is a constant. It should be evident that these equations establish switching frequency FSW in accordance with equation 6) above, thereby achieving a power factor approaching one, assuming that VREF is representative of a desired value of VOUT- [0046] FIG. 5 is a schematic diagram showing further detail of oscillator 35 in yet another alternate embodiment. This embodiment has a similar operation and structure as the embodiment described in FIG. 4, except that comparator 69 is non-hysteretic and the VRAMP switching limits are established with a limit voltage VLIM that is generated with circuitry that includes resistors 83-84 and 88-89, a capacitor 85, a squaring circuit or multiplier 86, a division circuit 87 and a switch 90.

[0047] As shown above, for a constant ILOAD and TCHG, high power factors are achievable if CLK frequency Fsw is proportional to (VouT-VIN). However, as shown in equation 6), Fsw has a large variation if ViN has a high amplitude, particularly at the voltage peaks where the peak ICOIL currents flow. This embodiment provides a circuit that reduces the overall frequency variation or jitter as follows.

[0048] Resistors 83-84 operate as a voltage divider that divides input voltage VIN, and capacitor 85 cooperates with resistors 83-84 to produce a low pass filter that produces an average voltage <VIN1> whose ripple is substantially zero, or at least is small compared to the rectified sine wave shape of VIN.

In one embodiment, resistors 83-84 and capacitor 85 are selected to set the low pass corner frequency to about ten hertz, so that VR is substantially a DC voltage. As a result of this low pass filtering, <ViNi> is indicative of the average value of VIN.

[0049] Multiplier 86 is a standard analog multiplier circuit that squares average voltage VIN1 to produce a squared voltage

VgQ=K8*<VIN1>2, where K8 is a constant.

[0050] Division circuit 87 divides a reference voltage VREF by VSQ to produce a voltage VLIM=VDIN=VREF/ (K8* <VIN1>2), which is coupled through resistor 88 to set an upper limit of VRAMP at an input of comparator 69 when clock signal CLK is low. When CLK is high, switch 90 closes and VDIV is voltage divided by resistors 88-89 to establish a lower limit of VRAmp at a level VLIM=VREF/ (K8*<ViNi>) *R89/ (R88+R89), where Rgg and R89 are the resistances of resistors 88 and 89, respectively.

[0051] Hence, switching frequency FSW=K9*<VIN>2* (VREF-VIN), where Kg is a constant. This option allows oscillator 35 to limit the switching frequency variations to facilitate EMI filtering.

[0052] FIG. 6 is a schematic diagram of PFC circuit 100 in an alternate embodiment. This embodiment eliminates the need for resistor 18 and its dissipated power PR18=IIN 2*R, 8, where R18 is the resistance of resistor 18. Hence, this embodiment is suitable for applications requiring low standby power and a less than ideal power factor.

[0053] The embodiment of FIG. 6 modulates switching frequency FSW with the instantaneous value of coil current IcoiL rather than input voltage VIN. On average, ICOIL has a sinusoidal waveform in phase with VIN due to the power factor correcting operation of PFC circuit 100. ICOIL is sensed in its return path through resistor 72 to diode bridge 20, which develops a current sense voltage Vcs across resistor 72 on a node 39 to modulate Fsw. In one embodiment, resistor 72 has a resistance of about 0.1 ohms, so Vcs has a value of about-0.1 volts when ICOIL has a magnitude of one ampere. Alternatively, ICOIL can be measured using other techniques, like a current transformer rather than current sensing resistor 72. The use of coil current ICOIL, rather than input voltage VIN, to vary switching frequency Fsw is a approach suitable for either continuous mode or discontinuous mode PFC circuits or for an embodiment in which power factor

correction is combined in a single stage with a downstream voltage regulator or converter.

[0054] The power factor of this embodiment is believed to be lower than that of the previously described embodiments because the instantaneous value of ICOIL only approximates the rectified sinusoidal shape of VIN. Nevertheless, this version has a low power consumption and can be fabricated at a low cost, which make it suitable for many applications not requiring the highest achievable power factor. In one embodiment, the power factor can be improved by connecting a capacitance across resistor 72.

The capacitance is selected to filter out high frequency components, e. g. , those above the frequency of VIN, to produce a waveform at node 39 that more ideally approximates a rectified sine wave.

[0055] FIG. 7 shows a portion of PFC circuit 100 in the embodiment described in FIG. 6, including a resistor 82, a current source 78 and further detail of current mirror 57 of oscillator 35.

[0056] Transistors 76-77 are shown as being formed as a matched or scaled pair of NPN bipolar transistors, whose emitter areas are scaled in a predetermined ratio. Current source 78 supplies a current IR through transistor 77 to establish a base- emitter voltage that biases the base electrode of transistor 76 to a fixed potential.

[0057] Resistor 82 typically is formed as an external resistor to avoid deleterious effects resulting from the negative potential of current sense voltage Vcs when ICOIL is flowing. If transistor 76 and 77 have the same emitter area ratio, their respective emitters operate at substantially the same potential, so current IM1 is proportional to ICOIL since Vcs=-R72*IcoiL. IsENSE substantially equals IM1 (neglecting 57 base current) and Vcs+ (R82*IsENSE) is zero, where the resistance of resistor 82 is R82, and selected to provide a desired sampling current ISENSE through transistor 76. Then IMl=R72*Icoz/Rs2. ISENSE

is mirrored by current mirrors 58-59 to provide differential charging and discharging currents (IREF1-IM3) and (IREF2--IMS). respectively, to timing node 70 as described above.

[0058] In summary, the present invention provides a PFC circuit that operates in a discontinuous mode with a fixed switching pulsewidth. The discontinuous mode of operation allows the PFC circuit to be fabricated with low cost blocking diode, which reduces the system cost. A pulse width modulator is synchronized to transition edges of a clock signal to generate pulses that establish a charging period for a coil current. The coil current is then discharged over a discharging period to develop a PFC output voltage from an input signal. An oscillator generates the clock signal so that its clock period is longer than the sum of the charging and discharging periods, thereby ensuring discontinuous mode operation. The oscillator has an input for sensing an input signal of the PFC circuit to modify the clock period in a controlled fashion to maintain the product of the charging period and the duty cycle of the coil current constant. The PFC circuit thereby switches the coil current over a redefined frequency range to facilitate the reduction of electromagnetic interference with a low cost EMI filter.