APPLICATION NOTE

UC3842 PROVIDES LOW-COST CURRENT-MODE CONTROL

The fundamental challenge of power supply design is to simultaneously realize two conflicting objectives : good electrical performance and low cost. The UC3842 is an integrated pulse width modulator (PWM) designedwith both these objectives in mind. This IC provides designers an inexpensive controller with which they can obtain all the performance advantages of current-mode operation. In addition, the UC3842 is optimized for efficient power sequencing of off-line converters and for driving increasingly popular POWERMOS. This application note gives a functionaldescription of the UC3842 and suggests how to incorporate the IC into practical power supplies. A review of currentmode control and its benefits is included and meFigure 1 : Two-loop Current-mode Control System. thods of avoiding common pitfalls discussed. The final section presents designs of two power supplies utilizing UC3842 control. CURRENT-MODE CONTROL Figure 1 shows the two-loop current-mode control system in a typical buck regulator application. A clock signal initiates power pulses at a fixedfrequency. The termination of each pulse occurs when an analog of the inductor current reaches a threshold established by the error signal. In this way the error signal actually controls peak inductor current. This contrasts with conventional schemes in which the error signal directly controls pulse width without regard to inductor current.

AN246/1188

1/16

APPLICATION NOTE
Several performance advantages result from the use of current-mode control. First, an input voltage feed-forward characteristic is achieved ; i.e., the control circuit instantaneouslycorrects for input voltage variations without using up any of the error amplifiers dynamic range. Therefore, line regulation is excellent and the error amplifier can be dedicated to correcting for load variations exclusively. For converters in which inductor current is continuous, controlling peak current is nearly equivalent to controlling average current. Therefore, when such converters employ current-mode control, the inductor can be treated as an error-voltage-controlled-current-source for the purposes of small-signal analysis. This is illustrated by figure 2. The two-pole control-to-output frequency response of these converters is reduced to a single pole (filter capacitor in parallel with load) response. One result is that the error amplifier compensation can be designed to yield a stable closed-loop converter response with greater gain-bandwidth than would be possible with pulse-width control, giving the supply improvedsmall-signal dynamic response to changing loads. A second result is that the error amplifier compensationcircuit becomes simpler and better behaved, as illustrated in figure 3. Capacitor Ci and resistor Riz in figure 3a add a low frequency zero which cancels one of the two control-to-output poles of non-current-mode converters. For large-signal load changes, in which converter response is limited by inductor slew rate, the error amplifier will saturate while the inductor is catching up with the load. During this time, Ci will charge to an abnormal level. When the inductor current reaches its required level, the voltageon Ci causesacorrespondingerror in supply output vol-tage. The recovery time is Riz Ci, which may be milleseconds. However, the compensation network of figure 3b can be used where current-mode control has eliminated the inductor pole. Large-signal dynamic response is then greatly improved due to the absence of Ci. Figure 2 : Inductor Looks Like a Current Source to Small Signals.

Figure 3 : Required Error Amplifier Compensation for Continuous Inductor Current Designs using (a) Dutycycle Control and (b) Current-mode Control.
(a)

(b)

2/16

APPLICATION NOTE
Figure 4 : UC3842 Block Diagram.

Current limiting is simplified with current-mode control. Pulse-by-pulse limiting is, of course, inherent in the control scheme. Furthermore, an upper limit on the peak current can be establishedby simply clamping the error voltage. Accurate current limiting allows optimization of magnetic and power semiconductor elements while ensuring reliable supply operation. Finally, current-mode controlled power stages can be operated in parallel with equal current sharing. This opens the possibility of a modular approach to power supply design. FUNCTIONAL DESCRIPTION A block diagram of the UC3842 appears in figure 4. This IC will operate from a low impedance DC source of 10 V to 30 V. Operation between 10 V and 16 V requires a start-up bootstrap to a voltage greater than16 V in order to overcome the undervoltagelockout. VCC is internally clamped to 34 V for operation from higher voltage current-limited sources (ICC 30 mA). UNDER-VOLTAGE LOCKOUT (UVLO) This circuit insures that VCC is adequate to make the UC3842 fully operational before enablingthe output

stage. Figure 5a shows that the UVLO turn-on and turn-off thresholdsare fixed internally at 16 V and 10 V respectively. The 6 V hysteresis prevents VCC oscillations during power sequencing. Figure 5b showssupplycurrent requirements.Start-up current is less than 1 mA for efficient bootstrappingfrom the rectified input of an off-line converter, as illustrated by figure 6. During normal circuit operation, VCC is developed from auxiliary winding WAUX with D1 and CIN. At start-up, however, CIN must be charged to 16 V through RIN. With a start-up current of 1 mA, RIN can be as large as 100 k and still charge CIN when VAC = 90 V RMS (low line). Power dissipation in RIN would then be less than 350 mW even under high line (VAC = 130 V RMS) conditions. During UVLO, the UC3842 outputdriver is biased to a high impedance state. However, leakage currents (up to 10 A), if not shunted to ground, could pull high the gate of a POWERMOS. A 100 k shunt, as showing in figure 6, will hold the gate voltage below 1V.

3/16

APPLICATION NOTE
Figure 5 : (a) Under-voltage Lockout and (b) Supply Current Requirements.

(a)

(b)

Figure 6 : Providing Power to the UC3842.

OSCILLATOR The UC3842 oscillator is programmed as shown in figure 7a. Oscillator timing capacitor CT is charged from VREF (5 V) through RT, and discharged by an internalcurrent source. Charge and dischargetimes are given by : tc 0.55 RT CT

0.0063 RT 2.7 ) 0.0063 RT 4.0 1 frequency, then, is : f = tc + td td RT CT ln ( For RT > 5 k, td is small compared to tc, and : 1 1.8 f RT CT 0.55 RT CT

4/16

APPLICATION NOTE
During the discharge time, the internal clock signal blanks the output to the low state. Therefore,td limits maximum duty cycle (DMAX) to : td tc DMAX = =1 tc + td where = 1/f = switching period. The timing capacitor discharge current is not tightly controlled, so td may vary somewhat over temperature and from unit to unit. Therefore, when very precise duty cycle limiting is required, the circuit of figure 7b is recommended. One or more UC3842 oscillators can be synchronized to an external clock as shown in figure 8. Noise immunity is enhanced if the free-running oscillator frequency (f = 1/(tc + td)) is programmed to be ~ 20 % less than the clock frequency.

Figure 7 : (a) Oscillator Timing Connections and (b) Circuit for Limiting Duty Cycle.

(a)

(b)

tc (tH + tL) tH = 0.693 (R A + R B) C tL = 0.693 RB C DMAX =

Figure 8 : Synchronization to an External Clock.

ERROR AMPLIFIER The error amplifier (E/A) configuration is shown in figure 9. The non-inverting input is not brought out

to a pin, but is internally biased to 2.5 V 2 %. The E/A output is available at pin 1 for external compensation, allowing the user to control the converters closed-loop frequency response. Figure 10a shows an E/A compensation circuit suitable for stabilizing any current-mode controlled topology except for flyback and boost converters operating with continuous inductor current. The feedback components add a pole to the loop transfer function at fp = 1/2 Rf Cf. Rf and C f are chosen so that this pole cancels the zero of the output filter capacitor ESR in thepowercircuit. Ri andRf fix the lowfrequencygain.They are chosen to provideas much gain as possible while still allowing the pole formed by the output filter capacitor and load to roll off the loop gain to unity (0dB) at f fswitching/4. This technique insures converter stability while providing good dynamic response. Continuous-inductor-currentboost and flyback converters each have a right-half-plane zero in their transfer function. An additional compensation pole is neededto roll off loop gain at a frequency less than that of the RHP zero. Rp and Cp in the circuit of figure 10b provide this pole.
5/16

APPLICATION NOTE
The E/A output will source 0.5 mA and sink 2 mA. A lower limit for Rf is given by : VE/A OUT(max) 2.5 V 6 V 2.5 V = = 7 k Rf (MIN) 0.5 mA 0.5 mA E/A input bias current (2 A max) flows through Ri, resulting in a DC error in output voltage (Vo) given by : Vo(max) = (2 A) Ri Figure 9 : UC3842 Error Amplifier. It is therefore desirable to keep the value of Ri as low as possible. Figure 11 shows the open-loopfrequency response of the UC3842 E/A. The gain represent an upper limit on the gain of the compensated E/A. Phase lag increases rapidly as frequency exceeds 1 MHz due to second-order poles at 10 MHz and above.

Figure 10 : (a) Error Amplifier Compensation Addition Pole and (b) Needed for Continu ous Inductor-current Boost ad Flyback.

Figure 11 : Error Amplifier Open-loop Frequency Response.

(a)

(b)

6/16

APPLICATION NOTE
CURRENT SENSING AND LIMITING The UC3842 current sense input is configured as shown in figure 12. Current-to-voltageconversion is done externally with ground-referencedresistor RS. Under normal operation the peak voltage across RS is controlled by the E/A according to the following relation : VC 1.4 V VRS (pk) = 3 where : VC = control voltage = E/A output voltage. RS can be connected to the power circuit directly or through a current transformer, as figure 13 illustrates. While a direct connection is simpler, a transformer can reduce power dissipation in RS, reduce errors caused by the base current, and provide level shifting to eliminate the restraint of ground-reference sensing. The relation between VC and peak current in the power stage is given by : N VRS(pk) )= (VC 1.4) i(pk) = N ( 3 RS RS where : N = current sense transformer turns ratio. = 1 when transformer not used. Figure 12 : Current Sensing. For purposes of small-signal analysis, the controlto-sensed-current gain is : N i(pk) = VC 3 RS When sensing current in series with the power transistor, as shown in figure 13, current waveform will often have a large spike at its leading edge. This is due to rectifier recovery and/or interwinding capacitance in the power transformer. If unattenuated,this transient can prematurely terminate the output pulse. As shown, a simple RC filter is usually adequate to suppress this spike. The RC time constant should be approximately equal to the current spike duration (usually a few hundred nanoseconds). The inverting input to the UC3842 current-sense comparator is internally clamped to 1 V (figure 12). Current limiting occurs if the voltage at pin 3 reaches this threshold value, i.e. the current limit is defined by : N.1V iMAX = RS

7/16

APPLICATION NOTE
Figure 13 : Transformer-coupled Current Sensing. Figure 14 : Output Cross-conduction.

TOTEM-POLE OUTPUT The UC3842 has a single totem-pole output. The outputtransistors can be operated to 1 A peak current and 200 mA average current. The peak current is self-limiting, so no series current-limiting resistor is needed when driving a power MOS gate. Cross-conduction between the output transistors is minimal, as figure14 shows. The average addedpower due to cross-conduction with Vi = 30 V is only 80 mW at 200 kHz. Figures 15-17 show suggested circuits for driving POWERMOS and bipolar transistors with the UC3842 output. The simple circuit of figure 15 can be used when the control IC is not electrically isolated from the powerMOS. Series resistor R1 provides damping for a parasitic tank circuit formed by the power MOS inputcapacitanceand any series wiring inductance. Resistor R2 shunts output leakage currents (10 A maximum) to ground when the under-voltage lockout is active. Figure 16 shows an isolated powerMOS drive circuit which is appropriate when the drive signal must be levelshifted or transmitted across an isolation boundary. Bipolar transistors can be driven effectively with the circuit of figure 17. Resistors R1 and R2 fix the on-state base current.Capacitor C1 providesa negative base current pulse to remove stored charge at turn-off. PWM LATCH This flip-flop, shown in figure 4, ensures that only a single pulse appears at the UC3842 output in any one oscillator period. Excessive power transistor dissipation and potential saturation of magnetic elements are thereby averted.

SHUTDOWN TECHNIQUES Shutdown of the UC3842 can be accomplished by two methods ; either raise pin 3 above 1 V or pull pin 1 below 1 V. Either method causes the output of the PWM comparator to be high (refer to block diagram, figure 4). The PWM latch is reset dominant so that the output will remain low until the first clock pulse following removal of the shutdown signal at pin 1 or pin 3. As shown in figure 18, an externally latched shutdown can be accomplished by adding an SCR which will be reset by cycling VCC below the lower under-voltage lockout threshold (10 V). At this point all internal bias is removed, allowing the SCR to reset. Figure 15 : Direct POWERMOS Drive.

8/16

APPLICATION NOTE
Figure 16 : Isolated POWERMOS Drive.

Figure 17 : Bipolar Drive with Negative Turn-off Bias.

AVOIDING COMMON PITFALLS Current-mode controlled converterscan exhibit performance peculiarities under certain operating conditions. This section explains these situations and how to correct them when using the UC3842. SLOPECOMPENSATION PREVENTS INSTABILITIES It is well documented that current-mode controlled converters can exhibit subharmonic oscillations when operated at duty cycles greater than 50 %.

Fortunately, a simple technique (usually requiring only a single resistor to implement) exists which corrects this problem and at the same time improves converter performance in other respects. This slope compensationtechnique is described in detail in Reference 6. It should be noted that duty cycle here refersto output pulse width dividedby oscillator period, even in push-pull designs where the transformer period is twice that of the oscillator. Therefore, push-pull circuits will almost always require slope compensation to prevent subharmonic oscillation.

9/16

APPLICATION NOTE
Figure 18 : Shutdown Achieved by (a) Pulling Pin 3 High (b) Pulling Pin 1 Low.
(a)

(b)

most easily derived from the control chip oscillator as shown in figure 20a. The sawtooth slope in figure 19b is m = m2/2. This particular slope value is significant in that it yields perfect current-modecontrol ; i.e. with m2/2 the average inductor current follows the control signal so that, in the small-signal analysis, the inductor acts as a controlled current source. All current-mode controlled converters having continuous inductor current therefore benefit from this amount of slope compensation, whether or not they operate above 50 % duty. More slope is needed to prevent subharmonic oscillations at high duty cycles. With slope m = m2, such oscillations will not occur if the error amplifier gain (AV(E/A)) at half the switching frequency (fs/2) is kept below a threshold value (reference 6) : 2 CO AV (E/A) < 4 m = m2 f = fs/2 Co = sum of filter and load capacitance = 1/fs Slope compensation can also improve the noise immunity of a current-mode controlled supply. When the inductor ripple current is small compared to the average current (as in figure 19a), a small amount of noise on the current sense or control signals can cause a large pulse-width jitter. The magnitude of this jitter varies inversely with the difference in slope of the two signals. By adding slope as in figure 19b, the jitter is reduced. In noisy environments it is sometimes necessary to add slope m > m2 in order to correct this problem. However, as m increases beyond m = m2/2, the circuit becomes less perfectly current controlled. A complex trade-off is then required ; for very noisy circuits the optimum amount of slope compensation is best found empirically. Once the required slope is determined, the value of RSLOPE in figure 20a can be calculated :
3 m= VRAMP tRAMP

where :

Figure 19 illustrates the slope compensation technique. In figure 19a the uncompensated control voltage and current sense waveforms are shown as a reference. Current is often sensed in series with the switching transistor for buck-derived topologies. In this case, thecurrent sense signal does not track the decaying inductor current when the transistor is off, so dashed lines indicate this inductor current. The negativeinductor current slope is fixed by the values of output voltage (Vo) and inductance (L) : VL VF Vo (VF + Vo) diL = = = dt L L L where : VF = forward voltage drop across the freewheeling diode. The actual slope (m2) of the dashed lines in figure 19a is given by : diL RS (VF + Vo) RS . = m2 = N dt NL where : RS and N are defined as the Current Sensing section of this paper. In figure 19b, a sawtooth voltage with slope m has been added to the control signal. The sawtooth is synchronized with the PWM clock, and practice is

. AV(E/A) =

0.7 V /2

RSLOPE ZF | fs

)=

1.4

RSLOPE ZF | fs

RSLOPE =

3 m 1.4

(ZF | fs) = 2.1 . m . . ZF | fs

where : ZF| fs is the E/A feedback impedance at the switching frequency. For m = mL : RAMP Rs (VF + VO) RSLOPE = 1.7 ( ) ZF | fs NL

10/16

APPLICATION NOTE
Figure 19 : Slope Compensation Waveforms : (a) No Comp. (b) Comp. Added to Control Voltage. (c) Comp. Added to Current Sense.

Note that in order for the error amplifier to accurately replicate the ramp, ZF must be constant over the frequency range fs to at least 3 fs. In order to eliminate this last constraint, an alternative method of slope compensation is shown in figures 19c and 20b. Here the artificial slope is added to the current sense waveform rather than subtracted from the control signal. The magnitude of the added slope still relates to the downslope of inductor current as described above. The requirement for RSLOPE is now :
m= VRAMP Rf 0.7 Rf ( )= ( ) tRAMP Rf + RSLOPE /2 Rf + RSLOPE

For m = m2 : RSLOPE = Rf ( 1.4 NL 1) RS (VF + VO)

RSLOPE loads the UC3842 RT/CT terminal so as to cause a decrease in oscillator frequency. If RSLOPE >> RT thenthe frequency can be corrected by decreasing RT slightly. However, with RSLOPE 5 RT the linearity of the ramp degrades noticeably, causing over-compensationof the supply at low duty cycles. This can be avoided by driving RSLOPE with an emitter-follower as shown in figure 21.

RSLOPE =

1.4 Rf 1.4 Rf = Rf ( 1) m m

11/16

APPLICATION NOTE
Figure 20 : Slope Compensation Added (a) to Control Signal or (b) to Current Sense Waveform.
(a)

(b)

Figure 21 : Emitter-follower Minimizes Load at RT/CT Terminal.

NOISE As mentioned earlier, noise on the current sense or control signals can cause significant pulse-width jitter, particularly with continuous-inductor-currentdesigns. While slope compensationhelps alleviate this problem, a better solution is to minimize the amount of noise. In general, noise immunity improves as impedance decrease at critical points in a circuit. One such point for a switching supply is the ground line. Small wiring inductances between various ground points on a PC board can support commonmode noise with sufficient amplitudeto interfere with correct operation of the modulating IC. A copper ground plane and separate return lines for high-current paths greatly reduce common-mode noise.

12/16

APPLICATION NOTE
Note that the UC3842 has a single ground pin. High sink currents in the output therefore cannot be returned separately. Ceramic bypass capacitors (0.1 F) from VI and VREF to groundwill provide low-impedance paths for high frequency transients at those points. The input to the error amplifier, however, is a high-impedance point which cannot be bypassed without affecting the dynamic response of the power supply. Therefore, care should be taken to lay out the board in such a way that the feedback path is far removed from noise generating components such as the power transistor(s). Figure 22a illustrates another common noise-induced problem. When the power transistor turns off, a noise spike is coupled to the oscillator RT/CT terminal. At high duty cycles the voltage at RT/CT is approaching its thresholdlevel ( 2.7 V, establishedby the internal oscillator circuit) when this spike occurs. A spike of sufficient amplitude will prematurely trip the oscillator as shown by the dashed lines. In order to minimize the noise spike, choose CT as large as possible, remembering that deadtime increases with CT. It is recommended that CT never be less than 1000 pF. Often the noise which causes this problem is caused by the output (pin 6) being pulled below ground at turn-off by external parasitics. This is particularly true whendriving POWERMOS.A diode clamp from ground to pin 6 will prevent such output noise from feeding to the oscillator. If these measures fail to correct the problem, the oscillator frequency can always be stabilized with an external clock. Using the circuit of figure 8 results in an RT/CT waveform like that of figure 22b. Here the oscillator is much more immune to noise because the ramp voltage never closely approaches the internal threshold.

Figure 22 : (a) Noise on Pin 4 Can Cause Oscillator to Pre-trigger. (b) With External Sync. Noise Does not Approach threshold Level.

(a)

(b)

MAXIMUM OPERATING FREQUENCY Sinceoutput deadtimevaries directly with CT, therestraint on minimum CT (1000 pF) mentioned above results in a minimum deadtime varies for the UC3842. This minimum deadtime varies with RT and therefore with frequency, as shown in figure 23. Above 100 kHz, the deadtime significantly reduces the maximum duty cycle obtainable at the UC3842 output(also show in figure 23). Circuits not requiring large duty cycles, such as the forward converter and flyback topologies, could operate as high as 500 kHz. Operation at higher frequencies is not recom-

mended because the deadtime become less predictable. The speed of the UC3842 current sense section poses an additional constraint on maximum operating frequency. A maximum current sense delay of 400 ns represents 10 % of the switching period at 250 kHz and 20 % at 500 kHz. Magnetic components must not saturate as the current continues to rise during this delay period, and power semiconductorsmust be chosen to handlethe resulting peak currents. In short, above 250 kHz, may of the advantages of higher-frequency operation are lost.

13/16

APPLICATION NOTE
Figure 23 : Deadtime and Maximum Obtainable Duty-cycle vs. Frequency with Minimum Recommended CT.

CIRCUIT EXAMPLES 1. OFF-LINE FLYBACK Figure 24 shows a 25 W multiple-output off-line flyback regulator controlled with the UC3842. This regulator is low in cost because it uses only two magnetic elements, a primary-side voltage sensing technique, and an inexpensive control circuit. Specifications are listed below. SPECIFICATIONS : Line Isolation : Switching Frequency : Efficiency @ full load : InputVoltage OutputVoltage : 3750 V 40 kHz 70 % 95 VAC to 130 VAC (50Hz/60Hz) A. + 5 V, 5 % : 1 A to 4 A load Ripple voltage : 50 mV P-P Max. B. + 12 V, 3 % : 0.1 A to 0.3 A load Ripple voltage : 100 mV P-P Max C. 12 V, 3 % 0.1 A to 0.3 A load Ripple voltage : 100 mV P-P Max

14/16

APPLICATION NOTE
Figure 24 : 25W off-line Flyback Regulator.

15/16

APPLICATION NOTE

Information furnished is believed to be accurate and reliable. However, SGS-THOMSON Microelectronics assumes no responsibility for the consequences 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 SGS-THOMSON Microelectronics. Specifications mentioned in this publication are subject to change without notice. This publication supersedes and replaces all information previously supplied. SGS-THOMSON Microelectronics products are not authorized for use as critical components in life support devices or systems without express written approval of SGS-THOMSON Microelectronics. 1995 SGS-THOMSON Microelectronics - All Rights Reserved SGS-THOMSON Microelectronics GROUP OF COMPANIES Australia - Brazil - France - Germany - Hong Kong - Italy - Japan - Korea - Malaysia - Malta - Morocco - The Netherlands - Singapore Spain - Sweden - Switzerland - Taiwan - Thaliand - United Kingdom - U.S.A.

16/16