DC-to-DC Design Guide - Vishay Intertechnology

1 AN607 Vishay SiliconixDocument Number: Design GuideSergeJaunay,JessBrownINTRODUCTIONM anufacturers of electronic systems that require powerconversion are faced with the need for higher-density dc-to-dcconverters that perform more efficiently, within a smallerfootprint, and at lower cost despite increasing output meet these demands, Siliconix has combined advancedTrenchFETRand PWM-optimized process technologies,along with innovative new packages, to provide:Dlowest on-resistance for minimum power dissipationDlowest gate charge for minimum switching lossesDdV/dt shoot-through immunity1 Dimproved thermal in thermal management for increasing powerdensity are being achieved with Vishay Siliconix packagingtechnologies such as the PowerPAKt(Si7000 Series), thethick leadframe D2 PAK(SUM Series),and ChipFETt(Si5000 Series).DThe PowerPAK SO-8 offers the steady-state thermalresistance of a DPAK in an SO-8 PowerPAK 1212-8 is approximately half the size of aTSSOP-8 while decreasing the thermal resistance by anorder of SUM Series reduces thermal resistance by 33% overstandard D2 PAK is 40% smaller than a TSOP-6 package whileoffering lower on-resistance and lower thermal should be noted that lower thermal resistance results inhigher possible maximum current and power complete array of Vishay Siliconix MOSFET-packagedproducts ranges from the D2 PAK (SUM or SUB series), DPAK(SUD Series), and PowerPAK (Si7000 Series) types ofpackages to the LITTLE FOOTR packages.

2 These smalloutline devices range from the SO-8 down to the tiniestMOSFET available - the LITTLE FOOT MATERIALS witching CharacteristicsThe basic characteristics of a MOSFET are key tounderstanding how these devices work in switchmode reality the freewheel diode will have some form of reverserecovery effect ( a and b, Figure 2), and as a result, thecurrent through the drain source of the MOSFET (Q1, Figure1) will increase. To accommodate the extra drain-sourcecurrent, VGSmust increase above the value necessary is carrying the combined load and recovery current(period ). Therefore, the recovery current of the freewheeldiodeaddstotheloadcurrentseenby thecontrollingMOSFET(Q1). At the end of period a, the reverse recovery currentfalls, along with the gate-source voltage. This is because thediode has recovered. The recovery current in turn will decay tozero, resultingin thegate voltagereducing tothe originalvaluerequired to support the load current (period b). During thisperiod, the freewheel diode starts to support voltage, and theVDSvoltage falls, and the Miller Plateaubegins.

3 As with theideal-recovery diode explanation, this continues until thevoltage falls to its on-state value (end of ) and thegate-source voltage is unclamped and continues to theapplied gate-voltage is effectively the reverse of turn-on, apart from thatthere is no limitation by the freewheeling diode (in thisparticular circuit). For turn-off the Miller Plateau indicates thestart of the rise of the drain-source voltage, and the voltage ofthe Miller Plateau willrepresent therequired VGSto sustaintheload current. The turn-off delay is the period from when thegate voltage falls from its on-state value to when it reaches theMiller Plateau value ( load-current value).A simple buck converter, shown in Figure 1, shows thebehavior of the MOSFET during turn-on and turn-off whenswitching an inductive load. During these periods, a positivestepinputis appliedto turnthe deviceon, anda steptransition,from positive to zero, is applied to turn the MOSFET a positive step-input voltage on the gate, the voltageacross the gate-source of the MOSFET (VGS) ramps upaccording to the time constant formed by the gate resistance(Rg) and input capacitance (Ciss), as shown in Figure 2a(period ).

4 Once VGSreaches the threshold voltage (Vth), thechannel is turned on, and the current through the device startsto ramp up (period ). At the end of period , there are twopossible switching transients that VGScould follow. In the firstcase, the freewheel diode (D1, Figure 1) is assumed to havean ideal reverse recovery, represented bythe solidwaveformsin Figure 2. Once the channel is supporting the full-loadcurrent, the voltage across the device can begin to decay (theend of point ) because the diode is now able to supportvoltage. As the drain-source voltage falls, the gate-sourcevoltage stays approximately constant. This phenomenon iscalled the Miller Plateau, and it continues until the voltagea)Specifically designed to prevent spurious turn-on during high rates of dV/dtb)SUM is an improved D2 PAK package, with lower rDS(on)and thermalresistanceAN607 Vishay Number: 7191710-Oct-02falls to its on-state value. At the end of period (Figure 2), thegate-source voltage is unclamped and continues to theapplied gate-voltage value.

5 This additional gate voltage fullyenhances the MOSFET channel and reduces the rDS(on).In reality the freewheel diode will have some form of reverserecovery effect ( a and b, Figure 2), and as a result, thecurrent through the drain source of the MOSFET (Q1, Figure1) will increase. To accommodate the extra drain-sourcecurrent, VGSmust increase above the value necessary is carrying the combined load and recovery current(period ). Therefore, the recovery current of the freewheeldiodeaddstotheloadcurrentseenby thecontrollingMOSFET(Q1). At the end of period a, the reverse recovery currentfalls, along with the gate-source voltage. This is because thediode has recovered. The recovery current in turn will decay tozero, resultingin thegate voltagereducing tothe originalvaluerequired to support the load current (period b). During thisperiod, the freewheel diode starts to support voltage, and theVDSvoltage falls, and the Miller Plateaubegins. As with theideal-recovery diode explanation, this continues until thevoltage falls to its on-state value (end of ) and thegate-source voltage is unclamped and continues to theapplied gate-voltage is effectively the reverse of turn-on, apart from thatthere is no limitation by the freewheeling diode (in thisparticular circuit).

6 For turn-off the Miller Plateau indicates thestart of the rise of the drain-source voltage, and the voltage ofthe Miller Plateau willrepresent the required VGSto sustaintheload current. The turn-off delay is the period from when thegate voltage falls from its on-state value to when it reaches theMiller Plateau value ( load-current value).D1 VOUTCLQ1 VINFIGURE circuit of a buck converterFIGURE waveforms for a typical MOSFET in a buck converterNote: The solid line shows an idealized curve with no recoveryof the anti-parallel diode. The dotted line shows the effect ofreverse recovery of the freewheel diode on the gate waveformand the corresponding switching SiliconixDocument Number: MOSFETs: N- and are two fundamental types of MOSFETs: n-channelandp-channel. An n-channel device needs a positive gate voltagewith respect to the source voltage, whereas a p-channelMOSFET requires the gate voltageto benegative withrespectto the source. Due to these criteria, each device sometimesappears to be geared for specific applications, such asp-channels for load switches and n-channels for low-sideswitches.

7 In reality it is only the drive circuits that need to of an n-channel MOSFETDGSSVVGSFIGURE of a p-channel MOSFETFor the high-side switch portrayed in the buck converter ofFigure 1, it would be possible to use either a p- or n-channeldevice; however, the operating conditions of the buckconverter will determine which is to be , consider the high-side MOSFET as shown in Figure it is turned on, the source voltage will tend towards thedrain voltage (minus the voltage across the device Vs Vd).Therefore, if the gate drive is generated from the input voltage(Vin) as the MOSFET turns on, VGSwill reduce as the sourcepin (Vs) goes to Vin(Vd).Forann-channeldevice,thismeansth atthegatevoltagemustbe higher than the drain voltage to maintain VGSabove theMiller Plateau voltage to ensure that the MOSFET stays fullyon. To achieve this there are three common strategies orcircuits:a) use an isolated supply with the 0 V referenced to thesource voltage to ensure that the applied VGSis thesame as the voltage driving the gate;b) use a charge-pump circuit that generates a voltage high-er than the dc-link voltage to drive the gate; orc) use a bootstrap circuit that again generates a voltagehigher than the dc-link voltage, but which requires aswitching circuit to charge up the bootstrap capacitorafter the top device is turned method is to use a p-channel device in situations inwhich the gate voltage does not need to be higher than thedc-link voltage.

8 This is appropriate when the drain voltage ofthe MOSFET is less than 20 V because the gate signal can bederived directly fromthe inputsignal. Withdc-link voltages>20V and with limitations of 20 V on maximum gate voltages, itis necessary to level-shift the applied gate voltage to ensurethat the gate voltage does not exceed the maximum , with a dc-link voltage of 50 V, the applied gatevoltage must be level-shifted to at least 30 V. It should also benoted that the performance characteristics of a p-channelgenerally are inferior to those of an n-channel due to thephysical structure of the low-side devices, it is generally accepted that n-channeldevices are used because the source connection of theMOSFET is connected to power ground. As such, then-channel MOSFET only will require a positive signalreferenced to power ground, whereas a p-channel devicewould require a negative signal to ground to keep the deviceturned RectificationImprovements in efficiency can be made by replacing therectifying diodes, or freewheel diodes, with MOSFETs.

9 This isbecause the MOSFET has the capability to conduct current inboth directions, and reductions in conduction loss can beachieved due to the I2R losses of the MOSFET being lowerthan the IV losses associated with the diode. However, thecircuit and load conditions will determine whether the increasein efficiency offsets the extra cost, and sometimes additionalcircuitry, demanded by synchronous should be noted that the freewheel diode (D1, Figure 1), orrectifying diode, is still required to prevent both MOSFET sconducting at the same time -- the necessity of dead timebetween Q1 and Q2 results in a short period of diodeconduction -- and causingshoot-through ,with the inclusion of a MOSFET, it is possible to use theinherent body diode, though this typically demonstratesperformance inferior to that of an external Schottky diode. Asa result, it is sometimes beneficial to use a Schottky diode asthe anti-parallel diode bypassing the inherent body diode andresulting in an improvement of the conduction and recoveryperformance of the freewheel Number: 7191710-Oct-02 NON-ISOLATED TOPOLOGIESNon-isolated Buck ConverterBasic operationThe buck (or step-down) converter, shown in Figure 5, is usedto convert a positive dc voltage to a lower positive dc can be a bi-directional converter, but for simplicity s sake,consider only the power flow from the higher voltage to thelower circuit schematic for a buck converterNote: Q2 is the MOSFET channel, D2 is the body diode of theMOSFET, and Sch2 is an external Schottky input voltage has to be greater than the output voltage forenergy to flow from the input through to the of Q1 Figure 6 shows the turn-on of Q1.

10 Because there is a positivevoltage difference between Vinand Vout, there is a currentbuild-up in the inductor according to:diLdt=Vin VoutL[1]VOUTLVINQ2 ILQ1D2 Sch2 FIGURE of Q1 OnceQ1isturnedoff(Figure7),thecurrentflo wingthroughtheinductor cannot be reduced to zero instantaneously. Rather,the current requires a freewheel path, which will be Q2, D2, orSch2, depending on the circuit topology. The current decaysthrough the freewheel path according to:diLdt=VoutL[2]Table 1 shows the approximate voltage and current stressesforthe buckconverter basedon and current stresses for the buckconverterControlling SwitchFreewheelElementVoltage (ideal)VinVinVoltage (practical)Vin IrDS(on)Vin+VoutVin+IrDS(on)orCurrent pk (ideal)IoIoCurrent pk (practical)Io+ Io2Io+ Io2 Current rmsIo Io1 Figure8showstheidealizedwaveformsforthe buckconverterunder continuous-current-mode SiliconixDocument Number: and voltage waveforms for the buckconverter in constant-current operationAs shown in Figure 8, the input current will be the same as thecurrent though the MOSFET Q1.