LINFINITY Application Note AN-10 Design Procedure for Microprocessor Buck Regulators Copyright © 1998 Rev. 0.2.1 07/98 AN-10 Design Procedure for Microprocessor Buck Regulators INTRODUCTION This document is intended to help designers of computer motherboards and other circuits using Linfinity’s LX166x family of buck regulator controllers. Linfinity’s family includes adjustable and programmable controllers, some with additional linear regulator functions. A buck (step-down) switching regulator is commonly used in applications such as powering microprocessors. They are ideal for converting a 5V-system voltage to the 2V or so, at up to 20A, that processors require. The main advantages of a buck regulator are high efficiency; relatively simple design; no transformer; low switch stress and small output filter. The main disadvantage is possible over voltage if the main switch shorts. VIN +5V CIN Q1 LX166x An alternate configuration is the non-synchronous buck regulator, shown in Figure 2, where the lower MOSFET is replaced by a free-wheeling diode. This usually results in lower efficiency, particularly at low output voltages. The LX166x family devices are designed to operate in either synchronous or non-synchronous modes. See datasheets for details. BUCK REGULATOR VCC +12V body diode of Q2 during the deadtime. Typical waveforms are shown in Figure 14 on page 9. MODULATED CONSTANT OFF-TIME ARCHITECTURE Switching Frequency The Modulated Constant Off-Time Architecture is described fully in AN-9. The off-time is kept constant under normal conditions, but is modulated as a function of the output and input voltage in order to keep the operating frequency constant. The architecture is shown in Figure 3. VOUT L VIN Error Comp VREF Q2 R Q COUT VOUT S GND VPEAK Figure 1: Synchronous Buck Regulator CT IDIS VVAL VCC +12V Off-Time Comp Figure 3: Modulated Constant Off-Time Architecture VIN +5V CIN Duty Cycle LX166x Q1 L D1 VOUT As with any buck regulator, the inductor volt-second balance conditions must hold. (V IN COUT ) Where VIN is the input voltage (to upper MOSFET drain), VOUT is the output voltage, TON is the on-time (of the upper MOSFET) and TOFF is the off-time of the upper MOSFET. GND Figure 2: Non-Synchronous Buck Regulator Figure 1 shows a typical synchronous buck regulator. The current through the inductor is always continuous. When the upper MOSFET, Q1, is on, power is passed to the output through the inductor, L. After Q1 is switched off, there is some deadtime before Q2 turns on. Since the inductor current is continuous, current flows through the Copyright © 1998 Rev. 0.2.1 07/98 −V OUT ) × T ON = (V OUT × T OFF The duty cycle, D, equals the ratio of output to input voltage. D = V OUT V IN The off-time for a particular switching frequency, fSW, is LinFinity Application Note Page 2 AN-10 Design Procedure for Microprocessor Buck Regulators T OFF = (1 − D ) f SW The switching frequency of the LX1668/1669 is designed to be fixed at 200kHz in steady state conditions. T OFF = 5µs × (1 −V OUT V CC 5 ) Switching Frequency Calculation (LX1660 – 65) The switching frequency of the LX1660 – 65 devices is controlled differently, and can be calculated using the following equation. (1 −V OUT V IN ) × I DIS C T × (1.52 − 0.29 ×V OUT ) ......................(1) Where IDIS is fixed at 200µA and CT is the timing capacitor. For a 5V input, this can be simplified to: f SW = 0.621 × I DIS CT (V IN − V OUT ) V OUT × ......................... (5) f SW × L V IN The input current is a square wave, whose rms value is: I INPUT = I OUT D (1 − D ) ................................ (6) Where VCC5 is the 5V supply voltage to the IC f SW = I RIPPLE = Where IOUT is the output current. The ripple current through the input capacitors will be a function of the impedance of the capacitors (ESR) and the impedance of an input inductor. Usually the capacitors will be much lower impedance, so the capacitors will have to handle a large ripple current. This can result in heating of the capacitor and can be a reliability concern if the ripple current rating of the capacitor is exceeded. The effect of different inductors is shown in Figure 4 – a lower inductance results in faster transient response but greater ripple. ........................................................... (2) INDUCTOR SELECTION A microprocessor such as the Pentium® II processor requires the power supply to be able to supply a very rapid step demand in current (20 – 30A/µs) as the processor comes out of a stop grant or other low activity state. The inductor value is the main factor in determining how fast the current will increase. L = 5µH A lower value inductor will increase the ripple current and so require lower ESR (Equivalent Series Resistance) capacitors on the output, but will allow a much faster current change. Selection of the inductor value is a compromise between reducing ripple current, IRIPPLE and improving response time. An inductor cannot change current instantly. The voltage across an inductor is the product of the inductance and rate of change of current: V = L × dI dt ..............................................(3) The inductance required to get a specific response time, TR, to a step load current change of ∆I can be approximated by: L = 2.2µH L = (V IN −V OUT ) × T R ∆I ................................(4) The peak-to-peak output ripple current is given by the following formula: Copyright © 1998 Rev. 0.2.1 07/98 Figure 4: Effect of Different Inductor Values. Trace 1 = Output voltage; Trace 3 = Input Current; Trace 4 = Input Ripple Voltage. LinFinity Application Note Page 3 AN-10 Design Procedure for Microprocessor Buck Regulators A small input inductor (~1µH) can be used to reduce ripple and noise which might affect other 5V blocks. V DYN − ≥ (I RIPPLE + ∆I ) × ESR ........................... (7) Where VDYN- is the lower limit of the dynamic voltage tolerance (usually 100mV for under 2µs). FILTER CAPACITORS The output capacitors serve to filter the output voltage. Although a certain amount of bulk capacitance is required, the primary parameter of concern when selecting capacitors is ESR. A model of a capacitor is shown in Figure 5. VOUT COUT IRIPPLE + ∆I ESR GND Figure 5: Equivalent Series Resistance Equation (3) shows that during a heavy load transient, an inductor cannot respond instantaneously. The transient current step, ∆I, has to be provided by the capacitor. The current flowing through the ESR of the capacitor causes a voltage droop whose worst case magnitude is (IRIPPLE + ∆I) × ESR. It is important that the voltage droop does not exceed the dynamic voltage specification, VDYN-, of the processor manufacturer. The requirement for ESR is as follows: ADAPTIVE VOLTAGE POSITIONING The LX166x family (except LX1660) incorporates a 40mV offset into the regulation feedback loop in order to help transient performance. This is shown in Figure 7. The controller regulates point ‘A’ – at no load, the output will have a peak output 40mV above the set point voltage. As current increases, the output voltage will fall, due to the voltage drop in the sense resistor. The benefit of adaptive voltage positioning is increased margin to deal with transient voltage undershoots as shown in Figure 6. Note that the LX166x series uses peak voltage detection, so the dc voltage offset, VOFFSET (as measured with a digital volt meter) will be approximately 25mV instead of 40mV (see Figure 6). Input R Q Figure 7: Adaptive SVoltage Positioning RS Output IL 40mV VREF Adaptive Voltage Positioning Offset VOFFSET (40mV) Output Voltage VOUT (50mV/Div) Nominal set point voltage, VSET (2.0V) Steady state voltage at high current is approximately VSET + VOFFSET - IOUT×RSENSE Dynamic voltage tolerance, VDYN(100mV for 2µs) Initial voltage drop is mainly due to the product of the load current step and ESR of the capacitors. ∆V = ∆I × ESR. (ESL effects are ignored) "A" Output Current transient step, ∆I = 0 to 14A (5A/Div) L = 2.5µH; COUT = 6 × 1500µF Sanyo MV-GX; RSENSE = 2.5mΩ Figure 6: Transient Response with Adaptive Voltage Positioning Copyright © 1998 Rev. 0.2.1 07/98 LinFinity Application Note Page 4 AN-10 Design Procedure for Microprocessor Buck Regulators The dc offset voltage, VOFFSET, modifies equation (7) as shown below: V DYN − + V OFFSET ≥ (I RIPPLE + ∆I ) × ESR ..............(8) Controllers without adaptive voltage positioning (VOFFSET = 0) will require a lower ESR (i.e. extra capacitors). See design calculation 7 on page 11. At higher current levels, such as those demanded by the Pentium II processor, it is desirable to have a lower RSENSE to minimize voltage droop. However, a lower sense resistor will result in a higher over-current trip point, unless the comparator trip voltage is also lowered. The LX1662A – 65A and the LX1668/69 have a 60mV comparator voltage, whereas the LX1660 – 65 have a 100mV voltage. This is shown in Table 1. Table 1 : Current Sense Comparator Trip Voltages CURRENT LIMIT Device VTRIP The LX1662 – 65 have a current limit function which will hold the current to a maximum limit when the current sense comparator detects an over-current situation. The LX1668/69 have the additional protection of hiccup-mode current limit, whereby the controller goes into low dutycycle operation in an over-current situation and can reset itself as soon as the short-circuit is removed. Please see application note AN-8 for further details. LX1660/61 100mV Pentium, PowerPC LX1662/63/64/65 100mV Pentium LX1662A/63A/64A/65A 60mV Pentium II LX1668/69 60mV Pentium II Sensing the voltage drop caused by the output current through a resistive element performs current sensing – this can be a sense resistor or the parasitic resistance of the inductor. VSET + 40mV ID VOUT Sense Resistor The sense resistor method can use a surface mount power sense resistor; a PCB trace resistance or the parasitic resistance of the inductor (for details of this method, please see application note AN-7). The three alternatives are contrasted in Table 2. Table 2: Current Sense Resistor Elements Method Advantages Disadvantages Surface mount power resistor n Highly accurate n Exposure to air to dissipate heat n Low cost n Flexible resistance value n Low cost n Low heat dissipation n Small space n n n n n PCB trace resistor RSENSE Parasitic inductance of the inductor Current limit comp + - VTRIP Figure 8: Current Limit Circuit The sense resistor should be chosen so that the current limit level is not excessively high, nor so low that it interferes with normal operation. The sense resistance should also be as small as possible to reduce voltage droop across it (so that the output voltage does not fall below static voltage limits and also to help reduce power losses). Current limiting will occur when the output current exceeds the current limit level, ICL: I OUT ≥ I CL = VTRIP R SENSE ............................................. (9) Application Expensive Heat dissipation Few values <5mΩ Heat dissipation Accuracy ~20% n Accuracy ~20% n Dependent on type of inductor A PCB trace resistor can be constructed, as shown in Figure 9. By attaching directly to the relatively large pads for the capacitor and inductor, heat is dissipated more effectively by the larger copper masses. Connect the sense lines as Kelvin connections, to avoid any errors in measurement. Recommended PCB trace dimensions are given in Table 3. An alternate method for current sensing uses the RDS(ON) of the upper MOSFET. This is much less accurate since RDS(ON) can vary 50 - 100% with temperature. This method also relies on peak current sensing, and so is inflexible for different output voltages. Table 3 PCB Resistor Dimensions (for 30°C rise at 20A) Copper Weight Copper Thickness Resistor Value Dimensions (w x l) mm inches Where VTRIP is the comparator trip voltage. Copyright © 1998 Rev. 0.2.1 07/98 LinFinity Application Note Page 5 AN-10 Design Procedure for Microprocessor Buck Regulators Inductor pad mal resistance are specified semiconductors and heatsinks. Sense resistor in datasheets for Junction R1 Case R2 PCB/Heat sink Sense lines (Kelvin Connections) Output capacitor pad C1 Ambient Figure 9 : PCB Trace Sense Resistor Construction 2oz/ft² 68µm 2.5mΩ 2.5 x 22 0.1 x 0.85 5mΩ 2.5 x 43 0.1 x 1.70 R3 C2 Figure 10: Thermal Model of a Power Semiconductor THERMAL ISSUES Management of heat becomes increasingly important at higher power levels. The following are important factors in managing heat: 1. Airflow – the thermal performance of heatsinks, and even PCB copper is affected greatly by airflow (or lack of). 2. Board layout - neighboring components (such as the processor cartridge) can prevent or reduce significantly the air flow around the voltage regulator. They can also contribute to heat generation – the processor is a significant heat source. 3. Copper and PCB material – outside and thicker copper layers will be better able to dissipate heat. 4. Component selection – a synchronous buck regulator is normally more efficient than a non-synchronous regulator and so creates less heat. Using MOSFET’s with lower RDS(ON) will lower heat generation. 5. Ambient temperature – a higher ambient temperature will be reflected in higher silicon junction temperatures (potentially lower reliability). The temperature can be calculated using the electrical equivalent model shown in Figure 11. The counterpart of the temperature in electrical model is the voltage and the heat power in the thermal model is equivalent to a current source in the electrical model. Figure 11 also gives the typical values for the thermal resistors, where R3 = 50Ω (the unit of thermal resistance is °C/W) is the thermal resistance of the PCB with one square inch of copper. C1 is neglected and C2 is selected to be 0.02 because the thermal time constant of the heat sink is in the order of one second. The ambient temperature is usually constant therefore is represented by a voltage source V1. When a steady heat is generated in the junction, the junction temperature can be found as, T j = Ph ⋅ Thermal Model of a Semiconductor A power semiconductor device, such as a MOSFET, can be modeled as shown in Figure 10. The top block is the semiconductor junction – the source of heat in the device. Heat generated in the device is dissipated to ambient through a series of thermal resistances and blocks as shown. Its connection to the case can be represented by a resistor, R1. The unit of thermal resistance, Rθ, is °C/W, meaning that for every watt dissipated, there will be a temperature rise of Rθ °C. Values of therCopyright © 1998 Rev. 0.2.1 07/98 Figure 11: Electrical Equivalent of the Thermal Model 3 ∑ Rθ i + T A ................................. (10) i =1 where Tj is the junction temperature, Ph is the heat power, and TA is the ambient temperature. For example, if 1W heat is generated in the junction, the ambient temperature is 50°C, then the temperature at the junction is Tj = 1W×(2+0.5+50)°C/W +50°C = 102.5°C. LinFinity Application Note Page 6 AN-10 Design Procedure for Microprocessor Buck Regulators Selecting a Heatsink From equation (10), the required heat sink to ambient thermal resistance, RθSA can be calculated as follows: Rθ SA ≤ (T J −T A ) − (Rθ JC − Rθ CS PD ) ..................(11) RθJC (junction-to-case) can be found from the FET or LDO manufacturer’s datasheet. RθCS, the case-to-heatsink thermal resistance, depends on the mounting method to attach the IC to the heatsink but is usually 0.5 – 1°C/W. TA is normally assumed to be around 55°C for most computer applications. Having calculated the required RθSA, and knowing the airflow and power dissipation, a suitable heatsink can be found in the catalog of a heatsink vendor, such as Aavid, Wakefield or Thermalloy. See the design calculations later in this application note for an example. Power lost in the MOSFET is a combination of the resistive loss due to the RDS(ON) of the FET and switching losses. ( PD = I Internal Linear Regulator (LX1668 only) The LX1668 has an internal 2.5V fixed LDO. The LDO can have a 3.3V or a 5V power supply, but heat generation is an important consideration – the SOIC or TSSOP package has a high thermal resistance, and cannot dissipate much heat. The heat dissipation is shown in calculation 10 on page 12. Power Dissipated in Controller IC Excluding the internal LDO, the power dissipation will be approximately: PD = IOPERATING × VCC .................................. (15) The LX1660 – 65 devices operate from a 12V VCC, whereas the LX1668/69 operates from 5V. Power Dissipated in Upper MOSFET 2 Where ILIN is the current and VIN and VOUT are the drain and source voltages of the linear transistor respectively (i.e. the input and output of the linear regulator). ) ⋅ R DS (ON ) ⋅ D + (0.5 ⋅ I ⋅V IN ⋅ T SW ⋅ f SW ) (12) Where PD is power dissipated, D is the duty cycle and TSW is the switching time (~100ns). PROGRAMMING THE OUTPUT VOLTAGE The LX1662 – LX1669 devices have 5 VID inputs which read the 5-bit voltage identification (VID) code to program the output voltage. The VID pins on the LX1662/63/64 and 1665 are not digital compatible – pull up resistors should not be used on the VID bus, or errors may occur. Power Dissipated in Lower MOSFET Power dissipated is due to current flowing through the RDS(ON) of the lower MOSFET. ( ) PD = I 2 × RDS (ON ) × (1 − D ) .........................(13) Care should be taken in the selection of MOSFETs – buying a lower cost FET can result in much higher heat dissipation. It may become necessary to use a heatsink with the FET, so increasing total costs. The lower RDS(ON) FET could be a surface mount device, dissipating heat to the PCB copper. External Linear Regulator (LX1664/65/68) The heat generated in the MOSFET used as the regulator’s pass element is as follows: PD = I LIN × (V IN − VOUT ) ...............................(14) Copyright © 1998 Rev. 0.2.1 07/98 LOW DROPOUT REGULATORS The LX1664 and LX1665 have an external linear regulator driver; the LX1668 incorporates a similar linear regulator driver and also has an internal low-power low dropout regulator. External Linear Regulator Driver Connecting an external MOSFET to the controller IC can make an adjustable low dropout linear regulator (LDO). The adjust (LFB) pin has a feedback voltage of 1.5V, meaning that no resistors are required for setting the output to 1.5V for GTL+ Bus termination. The dropout voltage is determined by current and the RDS(ON) of the transistor – for most applications, it is not important to have low RDS(ON). LinFinity Application Note Page 7 AN-10 Design Procedure for Microprocessor Buck Regulators Power Traces and Ground Planes VCC3 +3.3V LDRV Q3 LX166x LFB Ensure that power traces on the PCB are as wide as possible, to minimize resistive voltage drops at high currents. CIN R1 VOUT3 COUT Connect ground points together on a separate plane, as shown in Figure 13. VCC R2 VIN +5V CVCC CIN GND L LX166x Figure 12: Linear Regulator Driver COUT The output voltage, VOUT3, is calculated as follows: V OUT 3 = V FB × (1 + R 1 R 2) .............................(16) Where VFB is the feedback voltage (1.5V). For VOUT3 = 1.5V, R2 = ∞ and R1 can be shorted. The linear regulator can be disabled by pulling the feedback pin, LFB, up to 3.3V or 5V. See datasheet for details. Internal Fixed LDO (LX1668 only) The internal LDO has a fixed output voltage of 2.5V. Although the LDO can handle transient currents as high as 400mA during start-up, normal operation should be limited to 200mA or below. The LDO can use a 5V or 3.3V input – beware of excessive heat dissipation if using 5V input. See equation (14) on page 7. VOUT GND Figure 13: Power Traces Ensure that the decoupling capacitor, CVCC, in Figure 13 is placed as close to the IC as possible, to isolate the controller from any noise on the VCC rail. Note that in the LX1662 – 65, an under-voltage lockout function can “shut down” the IC during momentary undervoltage situations when the capacitor is too small or too far from the device. Use at least 1µF. FURTHER INFORMATION Please see Linfinity’s web site at http://www.linfinity.com for the latest datasheets and application notes. LAYOUT CONSIDERATIONS As with any power device, careful layout is essential. Linfinity’s devices are tolerant of noise, but basic precautions should be taken. Copyright © 1998 Rev. 0.2.1 07/98 LinFinity Application Note Page 8 AN-10 Design Procedure for Microprocessor Buck Regulators Figure 14: Typical Buck Regulator Waveforms Gate Drive Voltage Q1 Gate VSOURCE 100ns non-overlap 100ns non-overlap Q2 Gate 0V MOSFET Current IQ1 IQ2 Inductor Current IRIPPLE IL 0A Output Capacitor Current Q+ ICO tON /2 Qt OFF /2 Output Voltage VOUT 0V Copyright © 1998 Rev. 0.2.1 07/98 LinFinity Application Note Page 9 AN-10 Design Procedure for Microprocessor Buck Regulators Current Sense Threshold (mV) Production • • 100 Now • • 100 Now • • 60 Now • • • 100 Now • • • 60 Now • • • • 100 Now SO-16 • • • • 60 Now LX1665 SO-18 • • • • • • 100 Now LX1665A SO-18 • • • • • • 60 Now LX1668 SO-20 TSSOP •(TTL) • • • • • • 60 8/98 LX1669 SO-16 •(TTL) • • • • • 60 8/98 Package SO-16 LX1661 SO-16 LX1662 SO-14 • LX1662A SO-14 • LX1663 SO-16 • • LX1663A SO-16 • • LX1664 SO-16 LX1664A Copyright © 1998 Rev. 0.2.1 07/98 • LinFinity Application Note Internal LDO Now LX1660 External LDO Synchronous Rectification 100 Device OVP Driver • Power Good • 5-bit VID Hiccup Mode Current Limit Adaptive Voltage Positioning Table 4 Switching Regulator Selection Guide • Page 10 AN-10 Design Procedure for Microprocessor Buck Regulators Design Calculations This design example is used to calculate the components required for a Pentium® II processor power supply with the following characteristics: VIN=5V; VOUT=2.0V (programmable, but assumed to be 2.0V for worst-case analysis); IMAX=15A maximum steady state load current; ∆IMAX=14A worst case transient load step; ICL=20A current limit activation level; TA=55°C ambient temperature (with 100 linear ft/min air flow) 1. 2. Select Controller IC Select controller IC from Table 4 on page 10. See also Linfinity Application Note AN-6 “Power Solutions for Flexible Motherboards” for reference designs. For a Pentium II supply, select a controller with a 60mV current sense comparator (i.e. LX1662A – 65A or LX1668/69). If GTL+ Bus and clock circuit loads are far from the processor, it may be better to use a single output controller (LX1662/63 or LX1669) with low dropout regulators to power VCLOCK and VTT. 6. Select Input Capacitors Select input capacitors based on the input current calculated in the previous step (this will be the absolute worst case capacitor current- see page 3). From the capacitor datasheet, the maximum rating for the Sanyo MV-GX 1500µF capacitor is 2A. Therefore, the design requires 7.4/2, i.e. 3 – 4 capacitors for greatest reliability. 7. Select Output Filter Capacitors The output capacitors have to be selected to meet the ESR specification – see equation (8). Select Timing Capacitor (Not applicable to LX1668/69). ESR ≤ Choose a switching frequency of 200kHz by selecting the appropriate timing capacitor value. From equation (2): CT = 3. 0.621 × 200 ⋅ 10 −6 200 ⋅ 10 3 Number of capacitors ≥ 44/7.6 ≥ 6 capacitors. = 621pF If a controller without adaptive voltage positioning were used, (5 − 2) × 12 ⋅ 10 −6 14 8. 200 ⋅ 10 3 × 2.5 ⋅ 10 −6 Copyright © 1998 Rev. 0.2.1 07/98 2 × = 2.4 A 5 Select Current Sense Resistor To ensure current limiting does not happen until the current exceeds 20A, RSENSE is selected according to equation (9). We can use a 2.5mΩ resistance. 9. Output Ripple Current From equation (5), the input current will be: I RIPPLE = ≥7 R SENSE ≤ 0.06 20 ≤ 0.003Ω Input Current From equation (6), the input current will be: (5 − 2) 0.044 (0.1 + 0 ) (2.4 + 14 ) Adaptive voltage positioning results in the elimination of one capacitor! I INPUT = 15 0.4 × 0.6 = 7.4 A 5. Number of caps ≥ = 2.57 µH An inductor in the range of 2.5 to 3µH will give a sufficiently fast transient response. Suitable inductors include the surface mount HM00-97713 and the through-hole version HM00-98637 from BI Technologies. 4. Sanyo MV-GX capacitors have a maximum ESR of 44mΩ (at 20°C and 100kHz). Select Inductor For most applications, a 12µs inductor response time will give adequate performance. From equation (4): L= 0.1 + 0.025 ≤ 0.0076Ω 2.4 + 14 Construct Resistor If using a surface mount sense resistor, the lowest commonly available value is 5mΩ, so use two in parallel. The lowest cost solution is to construct a sense resistor using a PCB trace. From Table 3, using 2oz/ft² LinFinity Application Note Page 11 AN-10 Design Procedure for Microprocessor Buck Regulators copper, suitable dimensions are 2.5mm wide by 22mm long. See Figure 9. 10. Thermal Analysis Upper MOSFET Using an IRL3102S (RDS(ON) = 13mΩ), from equation (12), the heat dissipated is: ( PD = 15 2 × 0.013 × 0.4 ( ) −9 + 0.5 × 15 × 5 × 100 ⋅ 10 × 200 ⋅ 10 3 PD = 1.17 + 0.75 = 1.92W ) This can be dissipated using the TO-263 surface mount package soldered to a copper pad for heatsinking. Lower MOSFET Using IRL3102S, equation (13) gives the heat as: ( ) PD = 15 2 × 0.013 × 0.6 = 1.755W Again, a surface mount package can be used for this transistor. If an IRL3303 (26mΩ RDS(ON)) is used, ( 2 ) 11. Heat Sink Requirements Maximum junction temperatures are 150°C for Linfinity LX166x and 175°C for International Rectifier IRL series MOSFET’s (see datasheets). Since MTBF decreases with increasing temperature, calculations should ideally use a lower maximum junction temperature such as 125°C. Assume RθCS is 0.5°/W. Remember that heatsink performance improves with additional air flow. Upper MOSFET The IRL3102S (RθJC = 1.4°C/W) could be used surface-mounted with the copper PCB pad as heatsink. If a heatsink is desired, it is specified by equation (11): Rθ SA ≤ A suitable heatsink would be the Aavid 577002 which has a thermal resistance of 32°C/W. Lower MOSFET From equation 2.7°C/W): Rθ SA ≤ PD = 15 × 0.026 × 0.6 = 3.51W 125 − 55 − (1.4 + 0.5 ) ≤ 34.6 °C/W 1.92 (11), using IRL3303 (RθJC = 125 − 55 − (2.7 + 0.5) ≤ 16.7 °C/W 3.51 A suitable heatsink would be the Aavid 530613, which has a thermal resistance of 16.7°C/W. This will require the use of a heatsink. External Linear Regulator Assuming VIN = 3.3V; VOUT = 1.5V and 3A steady state current, equation (14) gives the heat: PD = 3 × (3.3 − 1.5) = 5.4W Internal Linear Regulator Assuming 200mA steady state current, equation (14) gives the heat: Linear Regulator MOSFET From equation (11), using 1.4°C/W): Rθ SA ≤ IRLZ44N (RθJC = 125 − 55 − (1.4 + 0.5) ≤ 11.1 °C/W 5.4 The 563202 (11.0°C/W will be sufficient to dissipate the heat. For 3.3V input: PD = 0.2 × (3.3 − 2.5) = 0.16W For 5V input: PD = 0.2 × (5 − 2.5) = 0.50W Power Dissipated in Controller IC LX1668/69 (VCC = 5V), equation (15) gives the heat: PD = 0.024 × 5 = 0.12W (excluding LDO) LX1660-65 (VCC = 12V), equation (15) gives the heat: PD = 0.027 × 12 = 0.324W Copyright © 1998 Rev. 0.2.1 07/98 12. Temperature Rise in Controller IC The LX1668 has two package options, SO-20 and TSSOP-20, with different thermal resistances. The IC will have a different junction temperature rise, depending on the package and whether the internal LDO uses 5V or 3.3V input. Table 5 shows the temperature rise in the IC for different LDO input voltages and package options. This assumes 200mA steady state current from the LDO, as well as operating current for the IC. Table 5: Temperature Rise in LX1668 LDO Input Voltage LinFinity Application Note Total Power Dissipated Package Page 12 AN-10 Design Procedure for Microprocessor Buck Regulators SO-20 85°C/W TSSOP-20 110°C/W 3.3V 0.28W 23.8 30.8 5.0V 0.62W 52.7 68.2 Copyright © 1998 Rev. 0.2.1 07/98 LinFinity Application Note Page 13