LTC3766 High Efficiency, Secondary-Side Synchronous Forward Controller Description Features n n n n n n n n n n n n n Direct Flux Limit™ Guarantees No Saturation Fast and Accurate Average Current Limit Clean Start-Up Into Pre-Biased Output Secondary-Side Control for Fast Transient Response Simple, Self-Starting Architecture Synchronous MOSFET Reverse Current Limit PolyPhase® Operation Eases High-Power Design True Remote Sense Differential Amplifier Remote Sense Reverse Protection High Voltage Linear Regulator Controller Internal LDO Powers Gate Drive from VOUT Overtemperature/Overvoltage Protection Low Profile 4mm × 5mm QFN and Narrow 28-Lead SSOP Packages Applications n n n n The LTC®3766 is a PolyPhase-capable secondary-side controller for synchronous forward converters. When used in conjunction with the LTC3765 active-clamp forward controller and gate driver, the part creates a complete isolated power supply that combines the power of multiphase operation with the speed of secondary-side control. The LTC3766 has been designed to simplify the design of active clamp forward converters. Working in concert with the LTC3765, the LTC3766 forms a robust, self-starting converter that eliminates the need for the separate bias regulator that is commonly used in secondary-side control applications. A precision current-limit coupled with clean start-up into a pre-biased load make the LTC3766 an excellent choice for high-power battery charger applications. The LTC3766 provides extensive remote sensing and output protection features, while Direct Flux Limit guarantees no transformer saturation without compromising transient response. A linear regulator controller and internal bypass LDO are also provided to simplify the generation of the secondary-side bias voltage. Isolated 48V Telecommunication Systems Isolated Battery Chargers Automotive and Military Systems Industrial, Avionics and Heavy Equipment L, LT, LTC, LTM, PolyPhase, Linear Technology and the Linear logo are registered and Direct Flux Limit is a trademark of Linear Technology Corporation. All other trademarks are the property of their respective owners. Protected by U.S. Patents, including 7200014 and 6144194. Other patents pending. Typical Application 36V-72V to 5V/15A Active Clamp Isolated Forward Converter +VIN 36V TO 72V • 6:2 1.4µH +VOUT 5V 15A • EFFICIENCY: 94% AT 48VIN/15AOUT 2.2µF 100V ×3 FDMS86201 15mΩ 0.5W –VIN BSC0901NS 100nF 200V 33nF 200V 220µF 6.3V ×2 SiR414DP 3mΩ 2W 168Ω –VOUT Si3440DV Si3437DV (SOT23) 200k 2.2nF 250VAC FG SW SG RUN VIN NDRV VCC FS/SYNC VS+ VS– 1Ω 365k NDRV PG IS+ IS– 15.0k LTC3765 SSFLT RCORE 33nF ISMAG 0.1µF IN+ RUN VCC 4.7µF AG 10.5k DELAY 14k 1.0µF • IN– 2:1 • IS– IS+ 18.2k ITH PT– SS FS/UV SGND PGND LTC3766 PT+ GND PGND 33nF IPK SGD 26.1k 4.42k FB 15k FGD 470pF MODE 22.1k 604Ω 47pF 17.8k 3766 TA01 3766fa For more information www.linear.com/LTC3766 1 LTC3766 Absolute Maximum Ratings (Note 1) VCC Voltage..................................................–0.3V to 12V VIN Voltage.................................................. –0.3V to 33V RUN Voltage............................................... –0.3V to 33V SW Low Impedance Source............................. –5V to 40V Current Fed........... 2mA DC or 0.2A for <1μs Into Pin* VAUX, VS+, VS –, VSOUT, NDRV Voltages....... –0.3V to 16V ITH, IS+, REGSD Voltages............................. –0.3V to 6V PHASE Voltage.............................................. –0.3V to 6V IS –, SGD, FGD Voltages................................–0.3V to 12V FS/SYNC, FB, MODE Voltages......................–0.3V to 12V VSEC Voltage................................................. –0.3V to 3V IPK, SS Voltages............................................ –0.3V to 4V Operating Junction Temperature Range (Notes 2,3) LTC3766E, LTC3766I.......................... –40°C to 125°C LTC3766H........................................... –40°C to 150°C LTC3766MP........................................ –55°C to 150°C Storage Temperature Range................... –65°C to 150°C Lead Temperature (Soldering, 10 sec ) GN Package ...................................................... 300°C *The LTC3766 contains an internal 50V clamp that limits the voltage on the SW pin. Pin Configuration TOP VIEW MODE 4 25 PT– PHASE 5 24 VAUX FB 6 23 SW ITH 7 22 VIN RUN 8 21 NDRV SS 6 SS 9 20 FGD IPK 7 IPK 10 19 SGD VSOUT 8 28 27 26 25 24 23 MODE 1 22 PT– PHASE 2 21 VAUX FB 3 20 SW ITH 4 18 IS+ VS+ 12 17 IS– VS– 13 16 REGSD 19 VIN 29 GND RUN 5 VSOUT 11 18 NDRV 17 FGD 16 SGD 15 IS+ IS– REGSD FS/SYNC GND VS– VS+ 9 10 11 12 13 14 15 FS/SYNC GN PACKAGE 28-LEAD NARROW PLASTIC SSOP TJMAX = 125°C, θJA = 95°C/W PT+ 26 PT+ PGND 27 PGND 3 VCC 2 SG FG VSEC FG 28 VCC VSEC 1 GND 14 2 TOP VIEW SG UFD PACKAGE 28-LEAD (4mm × 5mm) PLASTIC QFN TJMAX = 125°C, θJA = 43°C/W EXPOSED PAD (PIN 29) IS GND, MUST BE SOLDERED TO PCB 3766fa For more information www.linear.com/LTC3766 LTC3766 Order Information LEAD FREE FINISH LTC3766EGN#PBF LTC3766IGN#PBF LTC3766HGN#PBF LTC3766MPGN#PBF LTC3766EUFD#PBF TAPE AND REEL LTC3766EGN#TRPBF LTC3766IGN#TRPBF LTC3766HGN#TRPBF LTC3766MPGN#TRPBF LTC3766EUFD#TRPBF PART MARKING* LTC3766GN LTC3766GN LTC3766GN LTC3766GN 3766 PACKAGE DESCRIPTION 28-Lead Narrow Plastic SSOP 28-Lead Narrow Plastic SSOP 28-Lead Narrow Plastic SSOP 28-Lead Narrow Plastic SSOP 28-Lead (4mm × 5mm) Plastic QFN TEMPERATURE RANGE –40°C to 125°C –40°C to 125°C –40°C to 150°C –55°C to 150°C –40°C to 125°C LTC3766IUFD#PBF LTC3766IUFD#TRPBF 3766 28-Lead (4mm × 5mm) Plastic QFN –40°C to 125°C LTC3766HUFD#PBF LTC3766HUFD#TRPBF 3766 28-Lead (4mm × 5mm) Plastic QFN –40°C to 150°C LTC3766MPUFD#PBF LTC3766MPUFD#TRPBF 3766 –55°C to 150°C 28-Lead (4mm × 5mm) Plastic QFN Consult LTC Marketing for parts specified with wider operating temperature ranges. *The temperature grade is identified by a label on the shipping container. Consult LTC Marketing for information on non-standard lead based finish parts. For more information on lead free part marking, go to: http://www.linear.com/leadfree/ For more information on tape and reel specifications, go to: http://www.linear.com/tapeandreel/ Electrical Characteristics The l denotes the specifications which apply over the specified operating junction temperature range, otherwise specifications are at TA = 25°C. (Note 2) VIN = 15V, GND = PGND = 0V, unless otherwise noted. SYMBOL PARAMETER Main Control Loop Regulated Feedback Voltage VFB Feedback Input Current IFB Feedback Voltage Line Regulation ΔVFB(LINREG) ΔVFB(LOADREG) Feedback Voltage Load Regulation Average Current Sense Threshold VISAVG VISADJ Current Sense Ripple Compensation VISOC Overcurrent Shutdown Threshold ISIN gm REA ISOFT(C) ISOFT(D) VRUNR VRUNF IRUN tON(MIN) DMAX ΔVSEC(TH) IS+ and IS– Input Current Error Amplifier gm Error Amplifier Output Resistance Soft-Start Charge Current Soft-Start Discharge Current RUN Pin On Threshold RUN Pin Off Threshold RUN Pin Hysteresis Current Minimum Controllable On Time Maximum Duty Cycle Volt-Second Limit Threshold Accuracy RVSDN VSWCL ΔVFB(OV) Volt-Second Discharge Resistance SW Clamp Voltage Output Overvoltage Threshold CONDITIONS (Note 4) ITH = 1.2V (Note 4) VIN = 5V to 32V, ITH = 1.2V Measured in Servo Loop, ITH = 0.5V to 2V Resistor Sense (RS) Mode Current Transformer (CT) Mode RS Mode CT Mode VSW = 10V, VS+ = 5V, FS/SYNC = VCC, RIPK = 23.7k RS Mode: VIS–= 0V CT Mode: VIS– = VCC l MIN TYP MAX UNITS 0.592 0.600 2 0.001 –0.01 55 0.73 10 140 0.608 50 V nA %/V % mV V mV mV 100 1.33 280 2.7 5 5 3 1.22 1.17 3.0 200 79 113 1.44 500 3.2 l 47 0.66 86 1.22 2.2 (Note 7) VSS = 2V VSS = 2V VRUN Rising VRUN Falling VRUN = 0.5V FGD = SGD = GND 2V ≤ VSW < 5V 5V ≤ VSW ≤ 40V ISW = 1mA VFB Rising 4 l l 1.18 1.13 2.2 77 –6 –4 43 15 75 51 17 –0.1 63 0.80 6 1.26 1.21 3.6 81 6 4 60 19 mV V nA mS MΩ μA μA V V μA ns % % % Ω V % 3766fa For more information www.linear.com/LTC3766 3 LTC3766 Electrical Characteristics The l denotes the specifications which apply over the specified operating junction temperature range, otherwise specifications are at TA = 25°C. (Note 2) VIN = 15V, GND = PGND = 0V, unless otherwise noted. SYMBOL PARAMETER Drivers and Control FG, SG Driver Pull-Up On-Resistance FG, SG RUP FG, SG RDOWN FG, SG Driver Pull-Down On-Resistance PT+, PT– Driver Pull-Up Resistance PT+, PT– RUP + – PT ,PT RDOWN PT+, PT– Driver Pull-Down Resistance FGD Delay tFGD tSGD SGD Delay VSW(REV) SG Reverse Overcurrent SW Threshold ISW(REV) SG Reverse Overcurrent Adjust Current VCC Supply VCCOP ICC VUVLOR VCC Operating Voltage Range Supply Current Normal Mode Shutdown UV Lockout Rising VUVLOF UV Lockout Falling VREGSD IREGSD(C) IREGSD(D) VAUX Supply VAUXOP VCCVAUX REGSD Threshold Voltage REGSD Charge Current REGSD Discharge Current VAUXLR VAUXSWP VCC Load Regulation VAUX Switchover Voltage Rising VAUXSWN VAUX Switchover Voltage Falling RAUX RPSL VIN Supply VINOP VINCL ICLMAX VCCVIN VAUX Dropout Resistance VAUX Pre-Switchover Load IIN Supply Current Operating Shutdown VIN Undervoltage Lockout VINUVLO 4 VAUX Operating Voltage Range Regulated VCC Output Voltage VIN Operating Voltage Range VIN Clamp Voltage VIN Clamp Current Limit Regulated VCC Output Voltage CONDITIONS RFGD = 10kΩ RFGD = 100kΩ RSGD = 15kΩ RSGD = 50kΩ LV MODE HV MODE LV MODE HV MODE MIN TYP 50 436 60 195 66 140 –86 –34.5 1.5 1.0 1.5 1.5 65 545 75 230 73 148 –103 –42 5 VFS/SYNC = VCC = 7V (Note 5) VRUN = GND VCC Rising, LV MODE VCC Rising, HV MODE VCC Falling, LV MODE VCC Falling, HV MODE VREGSD Rising VREGSD = 0.7V VREGSD = 0.7V VAUX = 15V, LV MODE VAUX = 15V, HV MODE ICC = 0mA to 120mA, VAUX = 8V, LV MODE VAUX Ramping Positive, LV MODE VAUX Ramping Positive, HV MODE VAUX Ramping Negative, LV MODE VAUX Ramping Negative, HV MODE ICC = 120mA, VAUX = 4.9V VAUX = 4V IVIN = 2mA, VRUN = GND VIN = 33V, VRUN = GND LV MODE (Note 6) HV MODE (Note 6) VFS/SYNC = VCC VRUN = GND VIN Rising l l l l 4.6 7.7 3.8 6.7 5 6.7 8.1 4.50 7.65 4.30 7.35 MAX UNITS 80 654 90 265 79 156 –120 –49 Ω Ω Ω Ω ns ns ns ns mV mV µA µA 10 5 210 4.7 7.9 3.9 6.9 1.21 13 3 7.0 8.5 0.8 4.70 8.00 4.50 7.70 1.7 920 5 28 3.8 6.7 8.1 30 5.5 7.2 8.5 2.6 900 450 3.2 4.8 8.1 4.0 7.1 V mA µA V V V V V μA μA 15 7.3 8.9 2 4.88 8.35 4.70 8.05 2.5 V V V % V V V V Ω Ω 32 32 7.2 7.3 8.9 V V mA V V 1200 µA µA V 3.8 3766fa For more information www.linear.com/LTC3766 LTC3766 Electrical Characteristics The l denotes the specifications which apply over the specified operating junction temperature range, otherwise specifications are at TA = 25°C. (Note 2) VIN = 15V, GND = PGND = 0V, unless otherwise noted. SYMBOL PARAMETER Oscillator and Phase-Locked Loop FS/SYNC Pin Sourcing Current IFS/SYNC Oscillator High Frequency Set Point fHIGH Oscillator Resistor Set Accuracy Δf (RFS/SYNC) PLL Sync Frequency Range fPLL(RANGE) Differential Amplifier Gain ADA Common Mode Rejection Ratio CMRRDA VS+ Input Resistance RINP VS– Input Resistance RINM Output Sourcing Current IOH Output High Fault Threshold VIN-VOHST CONDITIONS MIN VFS/SYNC = VCC 18.75kΩ < RFS/SYNC < 125kΩ 234 –12 100 1.5V ≤ VSOUT ≤ 15V, VIN = 20V VIN = 20V VIN = 20V VIN = 20V VIN = 20V, VS+ = 5V, VSOUT = 2.5V VS+ Rising Note 1: Stresses beyond those listed under Absolute Maximum Ratings may cause permanent damage to the device. Exposure to any Absolute Maximum Rating condition for extended periods may affect device reliability and lifetime. Note 2: The LTC3766E is guaranteed to meet specifications from 0°C to 85°C with specifications over the –40°C to 125°C operating junction temperature range assured by design, characterization and correlation with statistical process controls. The LTC3766I is guaranteed over the –40°C to 125°C operating junction temperature range, the LTC3766H is guaranteed over the –40°C to 150°C operating junction temperature range, and the LTC3766MP is tested and guaranteed over the –55°C to 150°C operating junction temperature range. High junction temperatures degrade operating lifetimes; operating lifetime is derated for junction temperatures greater than 125°C. Note that the maximum ambient temperature consistent with these specifications is determined by specific operating conditions in conjunction with board layout, the rated package thermal impedance and other environmental factors. l 0.99 l 0.8 TYP 20 275 1 75 120 160 3.0 1.2 MAX UNITS 316 12 500 μA kHz % kHz 1.01 1.5 V/V dB kΩ kΩ mA V Note 3: TJ is calculated from the ambient temperature, TA, and power dissipation, PD, according to the following formula: TJ = TA + (PD • θJA°C/W) where θJA is 95°C/W for the SSOP and 43°C/W for the QFN package. Note 4: The LTC3766 is tested in a feedback loop that servos VFB to a voltage near the internal 0.6V reference voltage to obtain the specified ITH voltage (VITH = 1.2V). Note 5: Operating supply current is measured in test mode. Dynamic supply current is higher due to the internal gate charge being delivered at the switching frequency. See Typical Performance Characteristics. Note 6: The VIN Regulator employs an external pass device to produce the regulated VCC output voltage. The LTC3766 is tested using a 2N3904 NPN transistor as an external pass device. Note 7: Guaranteed by design. 3766fa For more information www.linear.com/LTC3766 5 LTC3766 Typical Performance Characteristics VCC Regulator Output Voltage vs Temperature 9.0 12 11 10 9 HV MODE 8.0 7.5 LV MODE 7.0 6.5 8 7 2.90 VIN REGULATOR USING 2N3904 8.5 VCC VOLTAGE (V) VCC SUPPLY CURRENT (mA) 13 Error Amplifier Transconductance vs Temperature TRANSCONDUCTANCE (mS) 14 VCC Supply Current vs VCC Voltage 5 6 9 8 10 VCC INPUT VOLTAGE (V) 7 11 6.0 –55 –25 12 65 35 95 5 TEMPERATURE (°C) 125 0.745 55.5 97 155 125 CT MODE THRESHOLD (V) 2.5 98 65 35 95 5 TEMPERATURE (°C) 2.0 1.5 0.5 1.216 1.214 65 35 95 5 TEMPERATURE (°C) 125 155 0.720 –55 –25 155 65 35 95 5 TEMPERATURE (°C) 3766 G04 Oscillator Frequency vs RFS VRUN = 0.5V 3.05 500 3.00 400 2.95 2.90 2.80 –55 –25 53.0 155 125 600 2.85 3766 G07 6 53.5 0.725 FREQUENCY (kHz) RUN CURRENT (µA) RUN VOLTAGE (V) 1.222 1.218 54.0 3766 G05 3.10 125 CT MODE 0.730 0 –55 –25 RUN VOLTAGE RISING 65 35 95 5 TEMPERATURE (°C) 54.5 RUN Hysteresis Current vs Temperature 1.220 55.0 0.735 1.0 RUN Threshold vs Temperature 1.212 –55 –25 RS MODE 0.740 3766 G04 1.224 Average Current Sense Threshold vs Temperature 102 99 1.315 1.310 –55 –25 3766 G03 56.0 RESISTANCE (Ω) CT MODE THRESHOLD (V) RS MODE 155 0.750 100 1.320 125 3.0 101 1.325 95 5 35 65 TEMPERATURE (°C) RS MODE THRESHOLD (mV) 1.330 –25 103 RS MODE THRESHOLD (mV) CT MODE 2.75 2.70 –55 155 VAUX Drop-Out Resistance vs Temperature Overcurrent Shutdown Threshold vs Temperature 1.335 2.80 3766 G02 3766 G01 1.340 2.85 300 200 100 65 35 95 5 TEMPERATURE (°C) 125 155 3766 G08 0 0 25 50 100 75 RFS (kΩ) 125 150 3766 G09 3766fa For more information www.linear.com/LTC3766 LTC3766 Typical Performance Characteristics Oscillator Frequency vs Temperature FB Voltage vs Temperature SGD Delay vs Resistance 250 599.8 1.5 200 599.4 0.5 DELAY (ns) 599.2 1.0 599.0 598.8 FS = 18.7k FS = 124k FS = VCC –0.5 –1.0 –55 –25 65 35 95 5 TEMPERATURE (°C) 125 155 598.4 –25 5 35 65 95 TEMPERARTURE (°C) 125 0 155 30 20 RSGD (kΩ) 40 50 3766 G12 SGD Delay vs Temperature FGD Delay vs Temperature 250 600 600 RSGD = 49.9k 200 RFGD = 100k 500 500 400 400 300 150 100 DELAY (ns) DELAY (ns) DELAY (ns) 10 0 3766 G11 FGD Delay vs Resistance RSGD = 15k 200 0 20 40 80 60 RFGD (kΩ) –25 65 5 95 35 TEMPERATURE (°C) 125 3766 G13 155 EFFICIENCY (%) PT+, PT– PULL-DOWN ALL GATES PULL-UP 1.00 FG,SG PULL-DOWN 0.75 5 6 7 9 8 10 VCC VOLTAGE (V) 125 155 3766 G15 Efficiency (Figure 39 Circuit) Load Step (Figure 39 Circuit) VOUT 200mV/DIV 94 1.25 65 35 95 5 TEMPERATURE (°C) 96 2.00 1.50 0 –55 –25 3766 G14 Gate Driver On-Resistance vs VCC Voltage 1.75 RFGD = 10k 100 0 –55 120 100 300 200 50 100 RESISTANCE (Ω) FG FALLING 100 50 598.0 –55 700 0.50 SG RISING 598.2 3766 G10 2.25 150 598.6 0 0 MODE = 100k TO GND SW TIED TO PT+ 599.6 VFB (mV) CHANGE IN FREQUENCY (%) 2.0 11 IOUT 5A/DIV 92 90 88 12 3766 G16 86 VIN = 48V 20µs/DIV VOUT = 5V LOAD STEP = 5A TO 15A VIN = 36V VIN = 48V VIN = 72V 3 5 7 9 11 LOAD CURRENT (A) 13 3766 G18 15 3766 G17 3766fa For more information www.linear.com/LTC3766 7 LTC3766 Pin Functions (SSOP/QFN) SG (Pin 1/Pin 26): Gate Drive for the Synchronous MOSFET. FG (Pin 2/Pin 27): Gate Drive for the Forward MOSFET. VSEC (Pin 3/Pin 28): Volt-Second Limit. Connect a resistor from SW to VSEC, and a capacitor from VSEC to GND to set the maximum volt-second product that is applied to the main power transformer. The PWM on-time is terminated when the VSEC voltage exceeds the internally generated threshold. Tie to GND if not used. MODE (Pin 4/Pin 1): For normal isolated applications using the LTC3765, tie to either GND or VCC to set the operating voltage to either low voltage or high voltage modes respectively, as needed to drive the gates of the synchronous and forward MOSFETs. For nonisolated applications, tie to ground through either a 100k or 50k resistor to activate standalone mode (for low voltage or high voltage operation respectively). In this mode, the PT+ pin may be directly connected to the gate of a primary-side MOSFET, and a reference clock signal is generated on the PT– pin. In standalone mode, the FGD pin is ignored and the associated delay is set adaptively. PHASE (Pin 5/Pin 2): Control Input to the Phase Selector. This pin determines the phasing of the internal controller CLK relative to the synchronizing signal at the FS/SYNC pin. FB (Pin 6/Pin 3): The Inverting Input of the Main Loop Error Amplifier. Tie to VCC to enable slave mode in PolyPhase applications. ITH (Pin 7/Pin 4): The Output of the Main Loop Error Amplifier. Place compensation components between the ITH pin and GND. RUN (Pin 8/Pin 5): Run Control Input. Holding this pin below 1.22V will shut down the IC and reset the soft-start and REGSD pins to 0V. SS (Pin 9/Pin 6): Soft-Start Inputs. A capacitor to ground sets the ramp time of the output voltage. IPK (Pin 10/Pin 7): Peak Current Limit Inductor Ripple Cancellation. This pin is used to adjust the peak current limit based on the amount of inductor current ripple, thereby providing a constant average output current during current limit. Place a resistor to GND that is proportional to the main output inductor. Leave this pin floating for constant peak current limit. Minimize parasitic capacitance on this pin. 8 VSOUT, VS+, VS– (Pins 11, 12, 13/Pins 8, 9, 10): VSOUT is the output of a precision, unity-gain differential amplifier. Tie VS+ and VS– to the output of the main DC/DC converter to achieve true remote differential sensing. Also, VS+ is used for directly sensing the output voltage for inductor ripple cancellation. See the Applications Information section for details. GND (Pin 14/Pin 11, Exposed Pad Pin 29): Signal Ground and Kelvin Sense for SG Reverse Overcurrent. Connect to power ground at the source of the synchronous MOSFET. The exposed pad must be soldered to PCB ground for rated thermal performance. FS/SYNC (Pin 15/Pin 12): Combination Frequency Set and Sync Pin. Tie to VCC to run at 275kHz. Place a resistor to ground at this pin to set the frequency between 75kHz and 500kHz. To synchronize, drive this pin with a clock signal to achieve PLL synchronization from 100kHz to 500kHz. Sources 20μA of current. REGSD (Pin 16/Pin 13): Regulator Shutdown Timer. Place a capacitor to ground to limit the time allowed for the high voltage linear regulator controller to operate. When the REGSD voltage exceeds 1.21V, the linear regulator is shut down. This pin sources 13μA of current when the linear regulator is active. IS– (Pin 17/Pin 14): Negative Input to the Current Sense Circuit. Connect to the negative end of a low side current sense resistor. When using a current sense transformer, tie this pin to VCC for single-ended sensing on IS+ with a higher maximum trip level. IS+ (Pin 18/Pin 15): Positive Input to the Current Sense Circuit. Connect to the positive end of a low side current sense resistor or to the output of a current sense transformer. SGD (Pin 19/Pin 16): Synchronous Gate Rising Edge Delay. A resistor to GND sets the delay from primary gate turnoff (PT+ falling) to SG rising (and FG falling). This delay is used to optimize the dead time between the turn-off of the primary-side MOSFET and the turn-on of SG. Tie SGD to GND to set this delay adaptively based on the falling edge of the SW pin voltage. See Setting the Gate Driver Delays in the Applications Information section. For more information www.linear.com/LTC3766 3766fa LTC3766 Pin Functions (SSOP/QFN) FGD (Pin 20/Pin 17): Forward Gate Rising Edge Delay. A resistor to GND sets the delay from PT+ rising to FG rising (and SG falling). This delay is used to optimize the dead time between the turn-off of SG and the turn-on of the primary-side MOSFET. In standalone mode (100k or 50k resistor on MODE), this dead time is set adaptively and the FGD pin can be grounded. See Setting the Gate Driver Delays in the Applications Information section. NDRV (Pin 21/Pin 18): Drive Output for the External Pass Device of the High Voltage Linear Regulator Controller. Connect to the base (NPN) or gate (MOSFET) of an external N-type device. Tie to VCC pin if only using the internal LDO (VAUX pin). VIN (Pin 22/Pin 19): Connect to a higher voltage bias supply when using the linear regulator controller. The VIN pin supplies bias to the internal standby and monitoring circuits, the linear regulator controller, and the differential amplifier. Tie to VAUX pin if only using the internal LDO. SW (Pin 23/Pin 20): Connect (Kelvin) to the drain of the synchronous MOSFET. This input is used for adaptive shoot-through prevention and leading-edge blanking, monitoring the high level SW node voltage and SG reversecurrent protection. When SW is high, the voltage on this pin is internally measured for use in the inductor ripple cancellation and volt-second limit circuits. When SW is low and SG is high, this pin sources a small current and is used for SG reverse overcurrent protection. A resistor can be placed between the SW pin and the drain of the synchronous MOSFET to adjust the SG reverse-overcurrent threshold. The SW pin is internally clamped to 50V. VAUX (Pin 24/Pin 21): Auxiliary Power Input. This is the power input to an internal LDO that is connected to VCC. Whenever VAUX is greater than 4.7V (or 8V for high voltage mode), this LDO will supply power to VCC, bypassing the main linear regulator that is powered from VIN. See VAUX Connection in the Applications Information section. Do not exceed 16V on the VAUX pin. PT –, PT+ (Pin 25, 26/Pin 22, 23): Pulse Transformer Driver Outputs. For most applications, these connect to a pulse transformer through a series DC-blocking capacitor. The PWM information is multiplexed together with DC power and sent through the pulse transformer to the primary side. The PWM signal is then decoded by the LTC3765 active clamp forward controller and gate driver. In standalone mode (100k or 50k resistor on MODE), the PT+ pin has a standard PWM signal and may be directly connected to the gate of a primary-side MOSFET, while a reference clock is generated on the PT– pin. PGND (Pin 27/Pin 24): Gate Driver Ground Pin. Connect to power ground at the source of the synchronous MOSFET. VCC (Pin 28/Pin 25): Main VCC Input for All Driver and Control Circuitry. 3766fa For more information www.linear.com/LTC3766 9 LTC3766 Block Diagram VCC SG 2.2x PEAK CURRENT COMPARATOR IS+ IS– + 29.3x – + + 2V + EA – 0.60V FB – + + C – ITH + C – 0.305V RESET DOMINANT ITRP + C – 0.2V PGND WAIT OVP RQ S SKIP DMAX ADAPTIVE BLANKING AND DELAY OVP PWM DRIVER ENCODING AND LOGIC OVERCURRENT gm = 2.7mS ERROR AMPLIFIER SWHI WAIT + SS 2.93V + C – OC BLANKING OSC AND PLL MODE FG VCC PT+ PGND VCC 1.9V • • PT– BLANK PULSE XFMR BLANK VSEC(TH) DMAX C PT+/PT–DRIVE TYPE (PULSE ENCODED/STANDARD) DRIVE/VCC CONTROL MAIN XFMR VCC + – PHASE • 50V FS/SYNC SWHI • 1.4V PGND + SLOPE COMP SW + – VSW VSEC SWHI FGD HV/LV MODE SGD 4VSB RUN SD FB SS UVLO VIN(UV) 30V 5.5mA LIMIT VREF 4VSB REG WAIT SOFTSTART 4VSB VS+ VS– 80k + AMP – 80k 80k DIFFERENTIAL AMPLIFIER 10 + – VCC VIN VIN 275k UVLO SD VCC(UV) 1.22V LV: 4.7V/4.5V HV: 8V/7.7V HV LINREG DISABLE RIPPLE CANCELLATION VOUT SENSE VCC GND EN VAUX 5V TO 15V DC 12µA + – C LV: 4.7V/3.9V HV: 7.9V/6.9V IPEAK ADJUST + A – 5V TO 32V DC NDRV EN RQ S OVERCURRENT 80k – +A LV: 58k HV: 46k SSLO VSOUT 1.22V VIN SENSE REGSD 1.21V SW IPK 3766 BD 3766fa For more information www.linear.com/LTC3766 LTC3766 Timing Diagram PULSE ENCODED PWM VPT+ – VPT– PWM ON TIME LTC3765 AG LTC3765 PG ~ VIN 1 – DUTY CYCLE VIN SWP NODE 0V LTC3766 SG LTC3766 FG ~ VOUT 1 – DUTY CYCLE SWB NODE 0V VIN • SW NODE NS NP 0V 3766 TD01 SET BY LTC3766 FGD PIN SET BY LTC3766 SGD PIN SET BY LTC3765 DELAY PIN FIXED 180ns DELAY SW VIN+ • SWP PG AG VIN– VOUT+ • SWB PG FG FG LTC3765 AG IN+ IN– • • PT+ SW LTC3766 SG SG PT – 3766 F01 VOUT– Figure 1. Reference Schematic for Timing Diagram 3766fa For more information www.linear.com/LTC3766 11 LTC3766 Operation The LTC3766 is a secondary-side PWM controller designed for use in a forward converter with active clamp reset and synchronous rectification. When used in conjunction with the LTC3765 active-clamp forward controller and gate driver, it forms a highly efficient and robust isolated power supply with a minimum number of external components. By making use of a secondary-side control architecture, the LTC3766 is able to provide exceptional transient response while directly monitoring the load to ensure that both output voltage and output current are precisely controlled. This architecture provides superior performance and greater simplicity, and is particularly well suited to high power battery charger applications. Self-Starting Start-Up In most applications, the LTC3766 will be used with the LTC3765 to create a self-starting forward converter with secondary-side control. Since there is initially no bias voltage available on the secondary side, the LTC3765 must manage the start-up in an open-loop fashion on the primary side. When power is first applied on the primary side, the LTC3765 begins an open-loop soft-start using its own internal oscillator. Power is supplied to the secondary by switching the main primary-side MOSFET with a gradually increasing duty cycle from 0% to 70%, as controlled by the rate of rise of the voltage on the SSFLT pin. On the secondary side, bias voltage can be generated directly from the main transformer using a peak charge circuit, or other technique as appropriate. When the LTC3766 has adequate voltage to satisfy its start-up requirements, it provides duty cycle information through the pulse transformer as shown in Figure 2. The LTC3765 detects this signal and transfers control of the gate drivers to the LTC3766, which continues the soft-start of the output voltage. Typically, this hand-off from primary to secondary occurs when the output voltage is less than one half of its final level. The LTC3765 then turns off the linear regulator and, through an on-chip rectifier, extracts bias power for the primaryside MOSFETs from this signal. Linear Regulators In general, the bias voltage generated on the secondary side is higher than the level desired for operation of the forward and synchronous MOSFETs. Consequently, the 12 LTC3766 contains a high voltage linear regulator controller as well as a 15V VAUX bypass regulator with an internal PMOS, either of which can be used to regulate the voltage on the VCC pin. The linear regulator controller is used by tying the NDRV pin to the base or gate of an external N‑type pass device. The LTC3766 VIN pin provides bias to the linear regulator controller as well as to internal standby and monitoring circuitry. If adequate voltage is detected on the VAUX pin, then the VAUX bypass regulator will be activated and the high voltage linear regulator controller will be shut down to reduce power loss. Alternatively, if only the VAUX regulator is needed, then the NDRV pin can be tied off to VCC, while VIN is tied to VAUX. This flexible arrangement of two linear regulators allows for the convenient and efficient generation of VCC bias voltage for a wide array of applications. Using the MODE pin, the output voltage of both linear regulators can be set to either 7V or 8.5V, depending on the level needed to drive the gates of the forward and synchronous MOSFETs. Note that the undervoltage lockout (UVLO) set points as well as VAUX switchover levels are adjusted along with the VCC regulation levels. This ensures that the MOSFETs are only switched when there is adequate gate drive voltage. Run Control and Soft-Start The main on/off control for the LTC3766 is the RUN pin. This pin features precision thresholds with both internal and externally adjustable hysteresis. This pin can be used to monitor the secondary-side bias voltage or main output voltage, thereby controlling the point at which hand-off from primary to secondary side occurs. Alternatively, it can be driven directly with a control signal. In nonisolated applications when the LTC3766 is used standalone, this pin can be used as an undervoltage lockout by monitoring the main power supply input voltage. See Nonisolated Applications in the Applications Information section for details. The LTC3766 will begin a soft-start sequence when the RUN pin is high, adequate voltage is present on both the VIN and VCC pins, and switching is detected on the SW pin. Note that the LTC3766 must see switching on the SW pin prior to initiating a soft-start sequence to ensure that the LTC3765 is ready for control hand-off. The soft-start sequence begins by first measuring the voltage on the FB 3766fa For more information www.linear.com/LTC3766 LTC3766 Operation pin and then rapidly pre-setting the soft-start capacitor voltage to a level that corresponds to the output voltage, VOUT. This is done to provide a smooth ramp on the output voltage as control is transferred from primary to secondary, as well as to avoid any unnecessary start-up delay. Once the soft-start capacitor has been pre-set to the appropriate level, the LTC3766 then sends a brief sequence of pulses through the pulse transformer to establish a communication lock between the LTC3766 and the LTC3765. At this point, the LTC3766 assumes control of the primary-side MOSFETs, and the soft-start capacitor begins charging with a constant current of 5μA, continuing the soft-start of the main output voltage. Note that the soft-start voltage is used to limit the effective level of the reference into the error amplifier. This technique maintains closed-loop control of the output voltage during the secondary-side soft-start interval. Gate Drive Encoding Since the LTC3766 controller normally resides on the secondary side of an isolation barrier, communication to the primary-side gate driver must be done through a small pulse transformer. A common scheme for communicating gate drive (PWM) information makes use of short pulses and relies on receiver latches to “remember” whether power MOSFETs should be either on or off. However, this system is prone to get into the wrong state, and has difficulty distinguishing a loss of signal from a legitimate zero duty cycle signal. To alleviate these concerns, the LTC3766 uses a proprietary gate drive encoding scheme that reliably maintains constant contact across the isolation barrier without introducing any delay. The LTC3766 encodes PWM information onto the PT+ and PT – outputs, which are in turn connected to a small pulse transformer through a DC-blocking capacitor. These outputs are driven in a complementary fashion, with a constant 79% duty cycle. This results in a stable voltsecond balance, so that the signal amplitude transferred across the pulse transformer is constant. As shown in Figure 2, the beginning of the interval when (VPT+-VPT–) is positive approximately coincides with the turn-on of the main primary-side MOSFET. Likewise, the beginning of the interval when (VPT+-VPT–) is negative coincides with the maximum duty cycle (forced turn-off of main primary- side MOSFET). At the appropriate time during the positive interval, the end of the “on” time (PWM going low) is signaled by briefly applying a zero-volt differential across the pulse transformer. In the event that a zero duty-cycle signal needs to be sent, this is accomplished naturally by placing the zero-voltage differential at the beginning of the positive interval. In this manner, any duty cycle from 0% to the maximum of 79% can be sent across the pulse transformer without delay. Figure 2 illustrates the operation of this encoding scheme. 150ns 150ns +VCC VPT+ – VPT – –VCC 1 CLK PER 1 CLK PER 3766 F02 Figure 2: Gate Drive Encoding Scheme (MODE = GND or MODE = VCC) On the primary side, the LTC3765 receives the signal from the pulse transformer through a DC restoring capacitor. After communication lock has been established between the two parts, the LTC3765 extracts clock and duty cycle information from the signal and uses it to control its gate driver outputs. Note that, except for a tiny pulse, this scheme is constantly applying a differential voltage across the pulse transformer. Therefore, the LTC3765 can almost instantly detect a loss of signal and shut off the power MOSFETs. Forward Converter and Main Loop Operation Once communication lock has been established between the LTC3766 and the LTC3765, the LTC3766 will have control over the switching of the primary-side MOSFETs. During normal operation, the main primary-side MOSFET (connected to PG on the LTC3765) is turned on somewhat after the forward MOSFET on the secondary side. This applies the input voltage across the transformer, causing the SW node on the secondary side to rise. Since the SW node voltage is greater than the output voltage, the inductor current ramps upward. When the current in the inductor 3766fa For more information www.linear.com/LTC3766 13 LTC3766 Operation has ramped up to the peak value as commanded by the voltage on the ITH pin, the current sense comparator trips, turning off the primary-side MOSFET. After a short delay, the forward MOSFET is turned off and the synchronous MOSFET is turned back on, causing the inductor current to ramp back downwards. At the next rising edge of the LTC3766 internal clock, the cycle repeats as the synchronous MOSFET is turned off and the forward and main primary-side MOSFETs are again turned on. The LTC3766 error amplifier senses the main output voltage, and adjusts the ITH voltage to obtain the peak inductor current needed to keep the output voltage at the desired regulation level. comparator false trip due to the MOSFET turn-on current spike. The LTC3766 uses the voltage on the SW pin (tied to the drain of the synchronous MOSFET) to implement an adaptive leading-edge blanking of approximately 180ns. The blanking of the current comparator begins only after the voltage on SW has risen above 1.4V. This adaptive blanking is essential because of the potentially long delay from the time that PT+ rises to the time that the SW node rises, and current begins ramping up in the output inductor. This blanking also minimizes the need for external filtering. In some applications, there can be considerable resistive voltage drops between the main output voltage and the load. To address this, the LTC3766 contains a precision differential amplifier, which can be used to remotely sense a load voltage as high as 15V. As in all forward converters, the main transformer core must be properly reset so as to maintain a balanced voltsecond product and prevent saturation. This job is handled on the primary side by the LTC3765, which features an active clamp gate driver. The active clamp MOSFET works together with a capacitor to generate an optimal reset voltage for the main transformer. This optimal reset voltage minimizes voltage stress on the main primary-side MOSFET and maximizes the utilization of the power transformer core by reducing the magnetic flux density excursion. Current Sensing, Slope Compensation and Blanking The LTC3766 supports current sensing either with a current sense resistor or with an isolated current transformer. When using a current sense resistor, the IS+ and IS– pins operate differentially, and the maximum peak current threshold is approximately 75mV. Normally, the current sense resistor is placed in the source of the forward MOSFET to minimize power loss. If a current transformer is used to sense the primary-side switch current, then the IS– input should be tied to VCC and the IS+ pin to the output of the current transformer. This causes the gain of the internal current sense amplifier to be reduced, so that the maximum peak current threshold is increased to approximately 1V. As with any PWM controller that uses constant-frequency peak current control, slope compensation is needed to provide current-loop stability and improve noise margin. The LTC3766 has fixed internal slope compensation. The amount of slope has been chosen to be adequate for a wide range of applications. Normally, the use of slope compensation would have a negative impact on the accuracy of the current limit, but the LTC3766 uses a proprietary circuit to nullify the effect of slope compensation on the current limit performance. Since the LTC3766 current loop is sensing switch current, leading edge blanking is needed to avoid a current 14 Gate Driver Delay Adjustment In general, the active clamp MOSFET is switched in a complimentary fashion to the main primary-side MOSFET. Since the active clamp MOSFET is a PMOS, the active clamp gate driver (AG) and the main primary-side gate driver (PG) voltages are therefore “in-phase,” with a programmable overlap time set by the LTC3765 DELAY pin. The delay time between the active clamp PMOS turn-off and the primary switch NMOS turn-on is critical for optimizing efficiency. When the active clamp is on, the drain of the primary NMOS, or primary switch node (SWP), is driven to a voltage of approximately VIN/(1–D) by the main transformer. When the active clamp turns off, the current in the magnetizing inductance of the transformer ramps this voltage linearly down to VIN. Power loss is minimized by turning on the primary switch when the SWP voltage is at a minimum. A resistor from the LTC3765 DELAY pin to ground sets a fixed time for the PG turn-on delay. The delay time between the primary switch turn-off and the active clamp turn-on is substantially less critical. When the primary switch turns off, the main transformer leakage inductance is biased with the peak current of the 3766fa For more information www.linear.com/LTC3766 LTC3766 Operation inductor reflected through the transformer. This current drives the voltage across the active clamp PMOS quickly to 0V. Turning on the PMOS after this transition results in minimal switching power loss. The LTC3765 active clamp turn on delay is internally fixed to 180ns, which normally achieves zero voltage switching on the active clamp PMOS. On the secondary side, the turn-on delay of the forward gate (FG) and synchronous gate (SG) MOSFETs are adjusted by the FGD and SGD pins respectively. These delays are set using resistors to GND so as to minimize the dead time (when the load current is being carried by MOSFET body diodes) while avoiding shoot-through with the primary-side MOSFETs. A shoot-through condition exists if either the PG and SG gates, or the AG and FG gates are high at the same time. Note that the SG MOSFET turn-on delay has a minimum limit that is established by the falling edge of the SW node. The SG pin will not go high until SW has falling below 0.5V. Refer to Delay Resistor Selection in the Applications Information Section for more detailed information. In standalone mode (100k or 50k resistor on MODE) the dead time between PG and SG is set adaptively to prevent shoot-through. Frequency Setting and Synchronization The LTC3766 uses a single pin to set the operating frequency or to synchronize the internal oscillator to a reference clock using and on-chip phase-locked loop (PLL). The FS/SYNC pin sources a 20μA current, and it may be tied to VCC for fixed 275kHz operation or have a single resistor to GND to set the switching frequency to fSW = 4RFS. If a clock signal (>2V) is detected at the FS pin, the LTC3766 will automatically synchronize to the falling edge of this signal using an internal PLL. Current Limit and Inductor Ripple Cancellation Since the LTC3766 utilizes peak current control, the peak inductor current is limited when the load current demand increases above the current limit set point. The peak current limit is established by an internal clamp on the maximum level of the ITH voltage. The average current, however, will be less than the peak current by an amount equal to one-half of the inductor ripple current. During current limit, this ripple current will change significantly with variations in VIN, VOUT and switching frequency. Without inductor ripple cancellation, this variation in ripple current would also result in an average output current that changes significantly, even though the peak current is held at a constant value. In order to keep the average current approximately constant during current limit, the LTC3766 cancels the effect of the ripple current by adjusting the value of the peak current limit (or ITH clamp level) in proportion to the amount of inductor ripple current. This is achieved by generating an internal ramp that mimics the inductor current ramp, and then adding the amplitude of this internal ramp to the ITH clamp voltage on a cycle-by cycle basis. During the on time, the slope of the inductor current is given by: dIL VSW – VS = dt L + The LTC3766 establishes a voltage on the IPK pin of (VSW – VS+)/15, which is one-fifteenth of the voltage across the output inductor during the on-time when SW is high. By choosing a resistor RIPK that is proportional to the value of the output inductor (RIPK = KL), the current flowing in RIPK becomes proportional to the slope of the inductor current: IRIPK = VSW – VS + VSW – VS = 15RIPK 15KL + During the time when SW is high, the LTC3766 uses the RIPK current to create an internal ramp by charging an on-chip capacitor CRIP. The slope of this internal ramp voltage is given by: dVRAMP IRIPK VSW – VS + = = dt CRIP 15KLCRIP The amplitude of this internal ramp is then added to the ITH clamp level dynamically. By choosing the appropriate value of RIPK, therefore, the average current during current limit will be essentially independent of changes in ripple current. As is the case with all DC/DC converters that maintain constant frequency operation, a cycle by cycle current limit is only effective at duty cycles where the on time is 3766fa For more information www.linear.com/LTC3766 15 LTC3766 Operation greater than the minimum controllable on-time. Under short-circuit conditions, for example, the LTC3766 limits the current using a separate overcurrent comparator. When this overcurrent comparator is tripped, the LTC3766 generates a fault followed by a soft-start retry. This hiccup mode overcurrent protection is highly effective at minimizing power losses under short-circuit conditions. Direct Flux Limit In active clamp forward converters, it is essential to establish an accurate limit to the transformer flux density in order to avoid core saturation during load transients or when starting up into a pre-biased output. Although the active clamp technique provides a suitable reset voltage during steady-state operation, the sudden increase in duty cycle caused in response to a load step can cause the transformer flux to accumulate or “walk,” potentially leading to saturation. This occurs because the reset voltage on the active clamp capacitor cannot keep up with the rapidly changing duty cycle. This effect is most pronounced at low input voltage, where the voltage loop demands a greater increase in duty cycle due to the lower voltage available to ramp up the current in the output inductor. Traditionally, transformer core saturation has been avoided either by limiting the maximum duty cycle of the converter or by slowing down the loop to limit the rate at which the duty cycle changes. Limiting the maximum duty cycle does help the converter avoid saturation for a load step at low input voltage, since the duty cycle maximum is clamped; however, transformer saturation can also easily occur at higher input voltage where the maximum duty cycle clamp is ineffective. Limiting the rate of duty cycle change such that the active clamp capacitor can sufficiently track the duty cycle change also helps to prevent saturation in many situations, but results in a very poor transient response. Neither of these traditional techniques is guaranteed to prevent the transformer from saturating in all situations. For example, saturation can easily occur using these traditional techniques when starting up into a pre-biased output, where the duty cycle can quickly change from 0% to 75%. Moreover, neither of these traditional techniques is able to prevent saturation in the negative direction, which can result from sudden decreases in duty cycle. 16 The LTC3765 and LTC3766 implement a new unique system for monitoring and directly limiting the flux accumulation in the transformer core. During a reset cycle, when the active clamp PMOS is on, the magnetizing current is directly measured and limited through a sense resistor in series with the PMOS source. This prevents saturation in the negative direction. When the PMOS turns off and the main NMOS switch turns on, the LTC3765 generates an accurate internal estimate of the magnetizing current based on the sensed input voltage on the LTC3765 RUN pin and transformer core parameters customized to the particular core by a resistor from the LTC3765 RCORE pin to ground. The magnetizing current is then limited during the on-time by this accurate internal approximation. Unlike previous methods, the direct flux limit directly measures and monitors flux accumulation and guarantees that the transformer will not saturate in either direction, even when starting into a pre-biased output. This technique also provides the best possible transient response, as it will temporarily allow very high duty cycles, only limiting the duty cycle when absolutely necessary. Moreover, this technique prevents overcurrent damage to the active clamp PMOS, which is a potentially significant weakness in many active clamp forward converter designs. Additional Protection Features The LTC3766 contains a wide array of protection features, which protect the DC/DC converter in the event that abnormal conditions persist. In general, protections features are either classified as a fault or a limit. When a fault is detected, all switching stops and the LTC3766 initiates a soft-start retry. Faults of this nature include overcurrent, overtemperature, differential amplifier miswire and communication-lock fault. An overcurrent fault occurs if the peak current exceeds approximately 133% of its normal value during current limit. Note that when inductor ripple cancellation is used, the value of the peak current during current limit will vary with inductor current ripple. The overtemperature fault is set at 165°C, with 20°C of hysteresis. This is helpful for limiting the temperature of the DC/DC converter in the event of some external device failure or other abnormal condition. The differential amplifier wiring fault is gener3766fa For more information www.linear.com/LTC3766 LTC3766 Operation ated if the inputs on the differential amplifier are reversed, or if there is not enough voltage on the VIN pin to support the voltage needed on VSOUT. This is important to avoid an overvoltage condition on the output. Finally, since it is essential that the LTC3766 be in constant communication with the LTC3765, a loss of communication lock will also generate a fault. A lock condition is detected by monitoring the SW node voltage, and ensuring that it is both rising and falling as it should in response to the PWM signal being sent to the primary side. If the SW node voltage is not rising and falling in an appropriate manner, than a lock fault is generated. In addition to the four protection features that generate faults, there are also four protection features that establish a clamp or limit, without generating a fault. First, the LTC3766 contains a precision volt-second clamp. This feature is not needed when the LTC3766 is used in conjunction with the LTC3765, which incorporates the direct flux limit feature. If the LTC3766 is used standalone, however, the volt-second limit can be used by placing a resistor from the SW node to the VSEC pin and a capacitor from VSEC to GND. When the SW node is low, the capacitor is discharged by an on-chip NMOS. When the SW node is high, the capacitor on VSEC is charged. If the capacitor voltage exceeds an internally generated threshold, then the main primary switch will be turned off, thereby limiting the volt-second product applied to the main transformer. To compensate for the exponential nature of the RC charging circuit, the LTC3766 adjusts the threshold of the volt-second comparator according to: VSEC(TH) = 0.6 – 0.16 VSW(HI) where VSW(HI) is the voltage on the SW pin during the on-time of the primary switch. This keeps the volt-second limit essentially constant for SW node voltages in the range of 2V to 40V. Second, in the event that the main output voltage exceeds its regulation target by more than 17%, the LTC3766 will detect an overvoltage condition. If this happens, the LTC3766 will immediately turn off the main primary MOSFET and turn on the synchronous MOSFET. This has the effect of pulling down the output voltage to protect the load from potential damage. Overvoltage protection is not latched, and normal operation is restored when the output voltage has been reduced to within 15% of its regulation level. Third, the LTC3766 contains an adjustable synchronous MOSFET reverse overcurrent. This is accomplished by monitoring the SW voltage when the synchronous MOSFET is on (SG pin is high). If the voltage on SW exceeds a pre-determined threshold, then the synchronous MOSFET will be turned off, protecting it from potentially damaging current levels. This SW threshold for reverse overcurrent detection can be reduced by placing a resistor in series with the SW pin, which sources a current when the SG pin is high. Note that the SG reverse overcurrent threshold and the SW pin source current are adjusted based on the state of the MODE pin. This is done to accommodate the use of either high voltage or low voltage MOSFETs, which normally have significantly different on resistances. In an overvoltage condition, the SG reverse overcurrent will override the overvoltage protection and force SG low, essentially regulating the reverse SG MOSFET current at a high level while the overvoltage condition persists. However, the SG reverse overcurrent is only active after the LTC3766 has achieved communication lock. Finally, the REGSD pin can be used to limit the amount of time that the high voltage linear regulator controller is active. This is particularly useful when the LTC3766 is used standalone in a nonisolated forward converter. In this application, the pass device of the linear regulator controller may be dissipating considerable power. When the linear regulator controller is active, the REGSD pin sources a 13μA current. If a capacitor from REGSD to GND charges to a voltage greater than 1.21V, then linear regulator controller is disabled. Gate Driver Mode Control In addition to being used in conjunction with the LTC3765, the LTC3766 can also be used standalone in a nonisolated forward converter application. In this case, the MODE pin can be used to disable gate drive encoding by tying MODE to GND through either a 100k (for VCC = 7V operation) or 50k (for VCC = 8.5V operation) resistor. This causes a normal PWM signal to appear on PT+ and a reference clock to appear on PT–. 3766fa For more information www.linear.com/LTC3766 17 LTC3766 Applications Information Secondary-Side Bias and Start-Up In most applications, the LTC3766 will receive its bias voltage from a supply that is generated on the secondary side. The manner in which the secondary bias is generated depends upon the output voltage as well as the variation in the input voltage of the DC/DC converter. In all applications, however, the secondary bias must always come up before the output reaches the regulation level. This is essential to avoid an overvoltage condition on the output, since the initial start-up is performed from the primary side in an open-loop fashion. See Generating the Secondary-Side Bias for more information. Note that the LTC3766 will not begin a soft-start sequence and initiate switching until the RUN pin is high, adequate voltage is present on both the VIN and V CC pins, and switching is detected on the SW pin. The LTC3766 looks for switching on the SW pin to ensure that the LTC3765 is active and ready for control hand-off. For switching to be detected, the SW node waveform must have at least eight consecutive pulses in the range of 50kHz to 700kHz. The SW node waveform must also have a peak that is greater than 1.4V and a valley that is less than 0.5V. In standalone mode, the LTC3766 begins the soft-start sequence without waiting for a switching waveform to be detected on the SW pin. Linear Regulator Operation The LTC3766 contains two linear regulators that are used to regulate the available bias voltage down to a level suitable for driving MOSFETs. If the bias supply voltage is greater than 15V, then the high voltage linear regulator controller may be used. This makes use of an external N-type pass device. Place a capacitor of 0.22μF or greater on VIN and 1μF or greater on VCC. If the bias supply connected to the VIN pin has a relatively high output impedance, it may be necessary to use a larger capacitor on VIN to prevent the VIN pin voltage from dropping when the VCC capacitor is being charged. The VCC charge rate during linear regulator start-up is set by the LTC3766 to approximately 0.5V/μs, which will create at charging current of (0.5 • 106) CVCC. Care should be taken to ensure that this charging current does not exceed the SOA of the N-type pass device, particularly when operating at higher VIN voltages. The VCC regulation level can be set to either 7V or 8.5V as desired 18 using the MODE pin. See the section on VCC and Drive Mode Selection for details. The LTC3766 also contains a 15V internal bypass LDO. If the voltage on the VAUX pin exceeds the VAUX switchover threshold, then the high voltage linear regulator is disabled, and an internal PMOS-pass LDO uses the VAUX voltage to supply power to VCC. This allows the high voltage linear regulator to be used for initial start-up and the higher efficiency bypass LDO to be used during normal operation. Figure 3 illustrates such a configuration that uses both linear regulators. If the voltage on the VAUX pin is below the switchover threshold, then the VAUX pin is internally loaded with a resistance of approximately 920Ω. This internal load is removed after the VAUX regulator is enabled, and is used to ensure that the VAUX supply is reasonably stiff before the bypass regulator is activated. In some cases, it is desirable to use the high voltage linear regulator only briefly during start-up, so as to limit the temperature rise in the external pass device. To accomplish this, place a capacitor on the REGSD pin to ground (see Figure 3) such that: CRSD = tHVREG (13µA ) 1.21V where tHVREG is the time that the high voltage regulator will operate. When the high voltage regulator is operating, a 13μA current is sourced from the REGSD pin, and when it is shut down (e.g., the bypass regulator is active), a 3μA current is sinked into the REGSD pin. If the REGSD voltage exceeds 1.21V, the high voltage regulator is disabled. Choose a time tHVREG that is greater than the normal start-up time. After start-up, if the voltage on the VAUX pin drops, the high-voltage linear regulator will be re-energized, but only for a limited time. VIN CVIN NDRV HV BIAS SUPPLY 6V TO 32V LTC3766 VCC OPTIONAL REGSD CRSD VAUX LV BIAS CVAUX SUPPLY 5V TO 15V CVCC 3766 F03 Figure 3. Typical Linear Regulator Connections For more information www.linear.com/LTC3766 3766fa LTC3766 Applications Information When used with a bias supply that is between 5V and 10V, the VCC pin can be directly connected to the bias supply as shown in Figure 4a. Note that the VIN and NDRV pins must also be connected to the bias supply for proper operation of internal circuitry. When a bias supply between 6V and 15V is available, the VAUX bypass linear regulator can be used standalone as shown in Figure 4b. In this case, proper start-up is assured by connecting the NDRV pin to VCC. Since there is no external pass device on NDRV, however, the effective UVLO levels will be dictated by the VAUX switchover thresholds instead of the VCC UVLO thresholds. Rather than relying on the VAUX thresholds, the start-up and shutdown levels are normally set by using the RUN pin to monitor the bias supply voltage as shown in Figure 4b. See the RUN Pin Operation section for details. For applications where the available bias supply is greater than 30V, the LTC3766 also contains a current-limited 30V clamp on the VIN pin. This clamp can sink up to 3.5mA to allow the VIN pin to be used as a shunt regulator. This is especially useful in nonisolated applications where the LTC3766 is used standalone. See the Nonisolated Applications section for more information. LTC3766 VIN NDRV CVCC VCC LV BIAS SUPPLY 5V TO 10V Figure 4a. No Linear Regulator Used VIN LV BIAS SUPPLY 6V TO 15V VAUX R1 CVAUX RUN NDRV VCC Normal operation is enabled when the voltage on RUN rises above its 1.22V threshold. As shown in Figure 5, the RUN pin can be used with an external resistor divider to enable the LTC3766 operation based on a sensed voltage VX. In self-starting applications, VX is normally either the converter output voltage (VOUT) or a bias voltage. In nonisolated applications, VX is normally the converter input voltage (VIN). See Nonisolated Applications for more information on the use of the RUN pin in nonisolated applications. VX R1 RUN R2 CVCC 3766 F04b Figure 4b. VAUX Regulator Used Standalone R2 LTC3766 GND 3766 F05 Figure 5. Using the RUN Pin to Determine Start-Up A 3µA current is pulled into the RUN pin when it is below its threshold that, when combined with the value chosen for R1, increases the hysteresis beyond the internal amount of 4%. When used in this manner, the values for R1 and R2 can be calculated from the desired rising and falling VX thresholds by the following equations: R1= 3766 F04a LTC3766 RUN Pin Operation VX(RISING) – 1.043 • VX(FALLING) 3µA 1.17 •R1 R2 = VX(FALLING) – 1.17 In self-starting applications where the LTC3765 performs an open-loop soft-start, the voltage VX can be tied to VOUT of the converter (VX = VOUT) to inhibit the LTC3766 startup until the output voltage is above a given level. This sets the exact output voltage at which soft-start control is handed off from primary to secondary. This hand-off output voltage should be set high enough so as to avoid pulse-skipping operation when the LTC3766 initially takes 3766fa For more information www.linear.com/LTC3766 19 LTC3766 Applications Information control. If excessive pulse skipping occurs in applications that use a peak charge circuit to generate bias voltage, this can cause the bias supply to fall, preventing proper startup. To preclude this possibility, use the RUN pin to inhibit the LTC3766 start-up until the output voltage is at least: VOUT(ON) > 300ns NS fSWVIN(MAX) NP In PolyPhase applications, synchronization can be achieved by tying the PT – pin of the master to the FS/SYNC pin of each slave. The relative phase delay of each slave is set using the PHASE pin. Any one of five preset values can be selected as shown in Table 2. Note that the phase delay is relative to the falling edge of the incoming reference clock on the FS/SYNC pin, since the falling edge of PT – corresponds to the beginning of the PWM cycle. Note that in self-starting applications, direct RUN/STOP control should be handled only on the primary side using the LTC3765. If the LTC3765 gets disabled, the LTC3766 will sense that the primary side is no longer switching and automatically shut down. To avoid a possible output overvoltage, do not manually disable the LTC3766 unless the LTC3765 is also manually disabled. Table 2 If the RUN pin function is not needed, it can be tied directly to the VIN pin. In some applications, it is desirable to start switching at a given frequency, and then synchronize to a clock reference signal at a later time. This can be accomplished by using the circuit shown in Figure 6. Setting the Switching Frequency and Synchronization The switching frequency of the LTC3766 is set using the FS/SYNC pin. This pin sources a 20μA current, and a resistor to ground on this pin sets the switching frequency to a value equal to: PHASE PIN PHASE DELAY APPLICATION GND 180° 2-Phase and 4-Phase 25k to GND 240° 3-Phase 50k to GND 120° 3-Phase 100k to GND 90° 4-Phase 100k to VCC 270° 4-Phase 50k BAT54 FS/SYNC fSW = 4RFS 10k Alternatively, the FS/SYNC pin can be tied to VCC, which sets the switching frequency to a fixed value of 275kHz. In general, a higher switching frequency will result in a smaller size for inductors and transformers, but at the cost of reduced efficiency. Although the LTC3766 can operate from 75kHz to 500kHz, the best balance between efficiency and size for a forward converter is found when operating between 150kHz and 350kHz. If a clock signal (>2V) is detected at the FS pin, the LTC3766 will automatically synchronize to the falling edge of this signal. Table 1 summarizes the operation of the FS/SYNC pin. Table 1 FS/SYNC PIN SWITCHING FREQUENCY VCC 275kHz RFS to GND fSW = 4RFS Reference Clock fSW = fREF (100kHz to 500kHz) 20 2N2222A CLK 10nF RFS LTC3766 GND 3766 F06 Figure 6. Synchronization After Free Running Once the LTC3766 has been synchronized, do not remove the external synchronizing clock unless the LTC3766 is also shut down. Removal of the external clock after synchronization will result in operation at low frequencies for a period of time, which can lead to very high currents in external power components. Setting the Output Voltage The LTC3766 output voltage is set by an external feedback resistor divider placed across the output as shown in Figure 7. The regulated output voltage is determined by: ⎛ R ⎞ VOUT = 0.6V • ⎜ 1+ B ⎟ ⎝ RA ⎠ 3766fa For more information www.linear.com/LTC3766 LTC3766 Applications Information Be careful to keep these divider resistors very close to the FB pin to minimize the trace length and noise pick-up on the sensitive FB signal. Using a low resistance (<2k) for the output voltage divider also minimizes noise on the FB pin. If the remote sense amplifier is used, then the divider should be placed between the VSOUT pin and GND. See the Remote Sensing section for details. VOUT LTC3766 RB VFB of power level, choose a family of transformers whose rated power level exceeds that of the required amount of output power. Be careful to allow for room to “grow,” as the power requirements of many electronic systems tend to increase throughout development. Once a family of transformers has been selected, the next step is to choose a suitable transformer from within that family. This mainly consists of choosing the correct number of primary and secondary turns (NP and NS). The value of NS can be calculated from: 108 VOUT NS = fSW A CBM RA 3766 F07 Figure 7. Setting the Output Voltage Selecting the Main Transformer The job of the transformer in a forward converter is to step the voltage either up or down while providing isolation between the primary and secondary grounds. Ideally, this transformer would not store any energy (it would have infinite magnetizing inductance). Note that this objective is very different from that of the transformer used in a flyback converter. The transformer used in a flyback converter is really a coupled inductor, the purpose of which is to store energy during the primary-side on time and then deliver it to the secondary during the off-time. In a forward converter, by contrast, the power is transferred during the primary-side on-time, and the off-time is used to recover the small amount of energy that was inadvertently stored in the core of the transformer. For nearly all applications, an off-the-shelf transformer can be selected. Transformers using planar winding technology are widely available and are a good choice for minimizing leakage inductance as well as component height. There are two basic items to consider in selecting an appropriate family of off-the-shelf transformers: 1) the isolation requirements and 2) the power level requirements. If the application circuit has specific isolation requirements, choose a family of transformers whose isolation level satisfies that requirement. In addition to an isolation voltage rating, the application may require a transformer with certification from a particular agency, or it may require a specific type of isolation (e.g., basic or functional). In terms where AC is the cross-sectional area of the core in cm2 (as normally given in the transformer data sheet) and BM is the maximum AC flux density desired. For the Pulse PA08xx series power transformers used in the Typical Applications section, AC = 0.59cm2. For the Pulse PA09xx series power transformers, AC = 0.81cm2. Most high frequency transformers use a ferrite core material. Consequently, selecting a maximum AC flux density of 2000 gauss is normally a good starting point, provided that the switching frequency is between 150kHz and 350kHz. This value of BM leaves headroom during transients and avoids excessive core losses. Note that the choice of BM together with switching frequency will determine the amount of core loss for a given transformer. Consult the transformer data sheet to evaluate the resulting core loss and temperature rise. In some cases, it may be necessary to increase NS somewhat in order to reduce BM and the associated temperature rise. In all cases, be sure to stay well below the saturation flux density of the transformer core. Once the value of has NS been selected, the required transformer turns ratio can be calculated from NP DMAX VIN(MIN) = NS VOUT where VIN(MIN) is the minimum input voltage and DMAX is the maximum duty cycle. Although the LTC3766 has a maximum duty cycle of 79% (DMAX = 0.79), normally a lower value of DMAX is chosen in the above equation so that there is duty cycle headroom to accommodate load 3766fa For more information www.linear.com/LTC3766 21 LTC3766 Applications Information transients when operating at minimum input voltage. A value for DMAX of 0.65 to 0.70 is appropriate for most applications. Having selected a particular transformer, calculate the copper losses associated with the transformer winding. These losses are highest when operating at maximum duty cycle and full load. However, it is better to evaluate copper losses at the nominal operating point of 50% duty cycle, where the losses are approximately: PCU 2 ⎞ ⎛ NS ⎞ + R ⎟ ⎜⎝ SEC ⎜⎝ NP ⎟⎠ PRI ⎟⎠ 2 IMAX ) ⎛ ( = ⎜R 2 where RPRI and RSEC are the primary and secondary winding resistances respectively, and IMAX is the maximum output current. An optimal transformer design has a reasonable balance between copper and core losses. If they are significantly different, then adjust the number of secondary turns (and recalculate the needed turns ratio) to achieve such a balance. The selection of an output inductor is essentially the same as for a buck converter. For a given input and output voltage, the inductor value and operating frequency determine the ripple current. The ripple current ∆IL increases with higher VIN and decreases with higher inductance: VOUT ⎛ VOUT NP ⎞ 1– • fSWL ⎜⎝ VIN NS ⎟⎠ Accepting larger values of ∆IL allows the use of low inductances, but results in higher output voltage ripple and greater core losses. A reasonable starting point for setting the ripple current is ∆IL = 0.3(IOUT(MAX)) for nominal VIN. The maximum ∆IL occurs at the maximum input voltage. Inductor Core Selection Once the value for L is known, the type of inductor must be selected. High efficiency converters generally cannot afford the core loss found in low cost powdered iron cores, forcing the use of the more expensive ferrite cores. Actual core loss is essentially independent of core size for a fixed 22 Ferrite designs have very low core losses and are preferred at high switching frequencies, so design goals can concentrate on copper loss and preventing saturation. Ferrite core material saturates “hard,” which means that inductance collapses abruptly when the peak design current is exceeded. This results in an abrupt increase in inductor ripple current and consequent output voltage ripple. Do not allow the core to saturate! Active Clamp Capacitor The active clamp capacitor, CCL, stores the average reset voltage of the transformer over many cycles. The voltage on the clamp capacitor is generated by the transformer core reset current, and will intrinsically adjust to the optimal reset voltage regardless of other parameters. The voltage across the capacitor at full load is approximately given by: VCL = Inductor Value Calculation ΔIL = inductor value but it is very dependent on the inductance selected. As the inductance increases, core losses decrease. Unfortunately, increased inductance requires more turns of wire and therefore copper losses will increase. VIN2 ⎛ N ⎞ VIN – 1.15 ⎜ VOUT • P ⎟ N ⎠ ⎝ S NP/NS is the main transformer turns ratio. The factor of 1.15 accounts for typical losses and delays. When PG and AG on the LTC3765 are low, the bottom side of the clamp capacitor is grounded, placing the reset voltage, VCL, on the SWP node. When PG and AG are high, the top side of the capacitor is grounded, and the voltage on the bottom side of the capacitor is –VCL. Therefore the voltage seen on the capacitor is also the voltage seen at the drains of the PG and AG MOSFETs. As shown in Figure 8, the VCL voltage has a minimum when the converter is operating at 50%. For a given range on VIN, therefore, the maximum clamp voltage (VCL(MAX)) will occur either at the minimum or maximum VIN, depending on which input voltage causes the converter to operate furthest from 50% duty cycle. The maximum VCL voltage can be determined by substituting the maximum and minimum values of VIN into this equation and selecting the larger of the two. In order to leave room for overshoot, 3766fa For more information www.linear.com/LTC3766 LTC3766 Applications Information cause oscillations under certain conditions. To avoid the problems associated with this resonance, always use an RC snubber in parallel with the clamp capacitor as shown in Figure 9. The values for this RC snubber can then be calculated using: ACTIVE CLAMP VOLTAGE NORMALIZED TO 50% DUTY CYCLE 1.6 1.5 1.4 1.3 1.2 RCS = 1.1 1.0 0.9 20 30 40 50 60 DUTY CYCLE (%) 70 80 3766 F08 Figure 8. VCL Voltage vs Duty Cycle choose a capacitor whose voltage rating is greater than this maximum VCL voltage by 50% or more. Typically, a good quality (X7R) ceramic capacitor is a good choice for CCL. Also, be sure to account for the voltage coefficient of the capacitor. Many ceramic capacitors will lose as much as 50% of their value at their rated voltage. The value of the clamp capacitor should be high enough to minimize the capacitor ripple voltage, thereby reducing the voltage stress seen by the MOSFETs. However, a larger clamp capacitor will ultimately result a slower transient response to avoid transformer saturation during load transients. While Direct Flux Limit will automatically limit the PWM on-time only as needed to prevent saturation, a larger clamp capacitor will require a longer time to charge or discharge in response to a load transient. Consequently, the value of the clamp capacitor represents a compromise between transient response and MOSFET voltage stress. A reasonable value for the clamp capacitor can be calculated using the following: 1 ⎛ 4 ⎞ CCL = • 2LM ⎜⎝ 2πfSW ⎟⎠ 2 An additional design constraint on CCL occurs because of the resonance between the magnetizing inductance LM of the main transformer and the clamp capacitor CCL. If the Q of this resonance is too high, it will result in increased voltage stress on the primary-side MOSFET during load transients. Also, a high Q resonance between LM and CCL complicates the compensation of the voltage loop, and can 1 ⎛ VO N ⎞ 1– ⎜ • P⎟ ⎝ VIN(MIN) NS ⎠ LM CCL and CCS = 6CCL Figure 9 shows a typical arrangement of the active clamp capacitor with an RC snubber. Be careful to account for the effect of voltage coefficient for both CCS and CCL to ensure that the above relationship between CCS and CCL is maintained. LM VIN SWP CCL RCS CCS SNUBBER ACTIVE CLAMP PMOS 3766 F09 Figure 9. Active Clamp Capacitor and Snubber Direct Flux Limit In active clamp forward converters, it is essential to establish an accurate limit to the transformer flux density in order to avoid core saturation during load transients or when starting up into a pre-biased output. The LTC3765 and LTC3766 implement a new unique system for monitoring and directly limiting the flux accumulation in the transformer core. Unlike previous methods, the direct flux limit directly measures and monitors flux accumulation and guarantees that the transformer will not saturate in either direction, even when starting into a pre-biased output. This technique also provides the best possible transient response, as it will temporarily allow very high duty cycles, only limiting the duty cycle when absolutely necessary. Moreover, this technique prevents overcurrent damage to 3766fa For more information www.linear.com/LTC3766 23 LTC3766 Applications Information the active clamp PMOS, which is a potentially significant weakness in many active clamp forward converter designs. Since the direct flux limit functionality is implemented in the LTC3765 on the primary side, there is nothing to adjust on the secondary side. See the LTC3765 data sheet for details on using this feature. Note that if the LTC3765 terminates the PG MOSFET on-time prematurely to limit flux accumulation, the LTC3766 will sense a premature falling on the SW node. In response, the LTC3766 will automatically terminate the FG on-time, thereby allowing the transformer core to reset. A premature falling on the SW node will also occur whenever the LTC3765 has shut down for any reason. Consequently, if the LTC3766 detects 19 consecutive premature SW node falling edges on the SW pin, it will generate a lock fault and shut down. Once the required BVDSS of the N-channel MOSFET is known, choose a MOSFET with the lowest available on-resistance (RDS,ON) that has been optimized for switching applications (low QG). In most applications, the MOSFET will be used at a drain current that is a fraction of the maximum rated current, so this rating is not normally a consideration. The total losses associated with the N‑channel MOSFET at maximum output current can be estimated using: ⎛ N ⎞ V (I ) PPG = ⎜ P ⎟ OUT MAX (1+ δ )RDS(ON) V ⎝N ⎠ 2 Primary-Side Power MOSFET Selection On the primary side, the peak-to-peak drive levels for both the N-channel main switch and the P-channel active clamp switch are determined by the voltage on the VCC pin of the LTC3765. This voltage is normally provided through the pulse transformer, and is typically set in the range of 8.5V to 12V. Note that even in applications where a logic-level MOSFET may be used on the primary side, the VCC voltage on the LTC3765 must still be in this range for proper operation. Selection of the N-channel MOSFET involves careful consideration of the requirements for breakdown voltage (BVDSS) and maximum drain current, while balancing the losses associated with the on-resistance and parasitic capacitances. In an active clamp topology, the maximum drain voltage seen by this MOSFET is approximately: VDS(PG) = 1.2 • VCL(MAX) where VCL(MAX) is the maximum active clamp voltage given above in the Active Clamp Capacitor section. The factor of 1.2 has been added to allow margin for ringing and ripple on the clamp capacitor. It is important to select the lowest possible voltage rating for this MOSFET in order to maximize efficiency. Note that the RC snubber on the active clamp capacitor (see Figure 10) reduces the peak voltage stress on the primary-side MOSFET without adding to operating losses. Also, the leakage inductance of the 24 main transformer at full load can cause considerable ripple on the active clamp capacitor, pushing up the peak voltage stress seen by the primary-side MOSFET. This ripple can be reduced by using a larger active clamp capacitor and a proportionally larger RC snubber capacitor. See Active Clamp Capacitor section for more information. S IN ⎛N ⎞ V I R Q f + ⎜ S ⎟ CL MAX DR GD SW +QGTOT VCC fSW 2VMILLER ⎝ NP ⎠ where δ is the temperature dependence of the on-resistance and VCL is the active clamp voltage (see Active Clamp Capacitor section). RDR (approximately 1.7Ω for the LTC3765) is the gate drive output resistance at the MOSFET’s miller plateau voltage, VMILLER. The values of QDG, QGTOT and VMILLER can be taken from the VGS versus QG curve that is typically provided in a MOSFET data sheet. QGD is the change in gate charge (QG) during the region where the VGS voltage is approximately constant and equal to miller voltage, VMILLER. The total gate charge (QGTOT) is the gate charge when VGS = VCC. The three parts in the above equation represent conduction losses, transition losses and gate drive losses respectively. Highest efficiency is obtained by selecting a MOSFET that achieves a balance between conduction losses and the sum of transition and gate drive losses. Note that the above equation for PPG is an approximation that includes assumptions. First, it is assumed that the turn-on transition losses are relatively small because of the leakage inductance in the main transformer. Also, it is assumed that the energy stored in this leakage inductance at primary-switch turn-off is 3766fa For more information www.linear.com/LTC3766 LTC3766 Applications Information completely recovered by the active clamp capacitor. For most applications, these assumptions are valid and the above equation is a good approximation. The active clamp P-channel MOSFET has the same BVDSS requirement as that of the N-channel MOSFET. Since the P-channel MOSFET only handles the magnetizing current, it is normally much smaller (typically a SOT package). To accommodate abnormal transients, use a P-channel MOSFET that has a pulsed drain current rating of 2A or higher. Also, note that when the N-channel MOSFET turns off, the leakage inductance will momentarily force the reflected load current into the body diode of the P-channel MOSFET. Consequently, the body diode should be rated to handle a pulsed forward current of: ⎛N ⎞ ID(MAX) = ⎜ S ⎟ IMAX ⎝ NP ⎠ In some cases, it may be more practical to add a separate diode in parallel with the body diode of the P-channel MOSFET. The primary-side P-channel MOSFET may be driven by a simple level-shift circuit that shifts down the drive voltage on the LTC3765 AG pin. Alternatively, the level-shift circuit can be omitted if the source of the P-channel MOSFET is returned to the VCC pin of the LTC3765. Refer to the LTC3765 data sheet for details. In nonisolated applications where the LTC3766 is used standalone, it is necessary to use a resonant reset technique instead of the active clamp reset. As a result, there are special considerations in selecting the primary-side MOSFET. See the Nonisolated Applications section for additional information. Secondary-Side Power MOSFET Selection On the secondary side, the peak-to-peak drive level for the N-channel MOSFETs is determined by the VCC pin on the LTC3766. Assuming that one or both of the linear regulators in the LTC3766 are used, the VCC regulation voltage can be set to either 7V or 8.5V as needed for driving the gates of the MOSFETs. The first step in selecting the secondary-side MOSFETs is to determine the needed breakdown voltage. The maximum voltage seen by the synchronous MOSFET is approximately: ⎛N ⎞ VDS(SG) = 1.2 • ⎜ S ⎟ VIN(MAX) ⎝ NP ⎠ where the factor of 1.2 has been added to allow for ringing and overshoot. This assumes that a snubber has been used on the secondary side of the main transformer (see the RC Snubbers section). If no snubber is used, the ringing and peak overshoot will be considerably higher. The maximum voltage seen by the forward MOSFET is approximately: VDS(FG) = 1.2 • VOUT ⎛N ⎞ V 1– OUT ⎜ P ⎟ VIN(MIN) ⎝ NS ⎠ where the factor of 1.2 has again been added to allow for ringing and overshoot. Having determined the BVDSS requirement for the forward and synchronous MOSFETs, the next step is to choose the on-resistance. Since both secondary-side MOSFETs are zero-voltage switched, choose MOSFETs that have a low RDS(ON) and have been optimized for use as synchronous rectifiers, including a body diode with a fast reverse recovery if possible. In most applications, the nominal input voltage will correspond to approximately 50% duty cycle, so the forward and synchronous MOSFETs will be selected to have the same RDS(ON). The power loss associated with the forward MOSFET can be approximated by: ⎛ N ⎞ V (I ) PFG = ⎜ P ⎟ OUT MAX (1+ δ )RDS(ON) VIN ⎝ NS ⎠ 2 +QGTOT VBIAS fSW where δ is the temperature dependence of the on-resistance and VBIAS is the input to the LTC3766 linear regulator (if used). The value for QGTOT can be taken from the VGS versus QG curve given in the MOSFET data sheet. QGTOT is the value of QG when VGS = VCC, where VCC is the 3766fa For more information www.linear.com/LTC3766 25 LTC3766 Applications Information voltage on the VCC pin of the LTC3766. For the synchronous MOSFET, the power loss is approximately: ⎛ N V ⎞ 2 PSG = ⎜ 1– P OUT ⎟ (IMAX ) (1+ δ )RDS(ON) ⎝ NS VIN ⎠ +QGTOT VBIAS fSW The power losses for the synchronous and forward MOSFET are generally dominated by conduction losses. For both of the above power loss equations, it is assumed that the dead time (when the MOSFET body diode is conducting) has been minimized. See Setting the Gate Drive Delays section for details on minimizing the dead time. VCC and Drive Mode Selection In order to accommodate various operating gate voltages that may be required by the secondary-side MOSFETs, the MODE pin can be used to set the LTC3766 for either LV mode or HV mode operation. In LV mode, the VCC regulation point for both linear regulators is set to 7V, while the VCC UVLO and VAUX switchover rising thresholds are adjusted to 4.7V. In HV mode, the VCC regulation point is set to 8.5V, while the VCC UVLO and the VAUX switchover rising thresholds are set to 7.9V and 8V respectively. Use LV mode for MOSFETs that are rated for 4V to 5V operation, and use HV mode for those rated with 7V to 10V operation. The VCC regulation levels, as well as the UVLO and switchover voltages have been optimized to ensure that both types of MOSFETs are operated safely and efficiently. In general, MOSFETs with a higher VDS rating also have a higher operating gate voltage rating. As a result, applications with output voltages of approximately 12V and higher will generally use MOSFETs that are rated for 7V to 10V gate operation. In addition to changing the VCC regulation, UVLO and VAUX switchover levels, the selection of HV mode or LV mode also changes the behavior of the SG reverse overcurrent. In LV mode, the reverse overcurrent threshold on the SW pin is 73mV and the adjust current is 103μA. In HV mode, these levels are changed to 148mV and 42μA to account for the fact that high voltage MOSFETs have larger on-resistance than low voltage MOSFETs. For details, see the Setting the SG Reverse Overcurrent. 26 In applications where the LTC3766 is used in conjunction with the LTC3765, the signals on the PT+ and PT – pins contain encoded PWM information with amplitude equal to the VCC voltage. This encoded gate drive signal is received by the LTC3765 and decoded into PWM and clock information that drives the primary-side MOSFETs. However, the LTC3766 can also be used standalone in nonisolated forward converter applications. In such applications, the MODE pin can be used to disable the PWM encoding on the PT+ and PT – pins. As a result, the LTC3766 will generate a normal PWM gate drive signal on the PT+ pin and a reference clock on the PT – pin. Also, in standalone mode the FGD pin is ignored and the dead time between SG falling and PT+ rising is set adaptively. The MODE pin has four possible states. Tying MODE to GND or VCC will provide encoded gate drive signals with either LV mode or HV mode operation respectively. Tying MODE to GND through either a 100k or a 50k resistor will provided standard PWM gate drive signals with either LV mode or HV mode operation respectively. Table 3 Summarizes the use of the MODE pin for setting the operating voltage and gate drive encoding modes, and Table 4 summarizes the effect of low voltage and high voltage gate drive operating modes. Table 3 MODE PIN DRIVE LEVEL PT+/PT– MODE INTENDED APPLICATIONS GND LV Encoded PWM Low VOUT Isolated Apps with LTC3765 VCC HV Encoded PWM High VOUT Isolated Apps with LTC3765 100k to GND LV Standard PWM Low VOUT, Nonisolated Apps, Standalone 50k to GND HV Standard PWM High VOUT, Nonisolated Apps, Standalone Table 4 DRIVE LEVEL VCC VCC UVLO THRESHOLD (RISE/FALL) VAUX SWITCHOVER THRESHOLD (RISE/FALL) SG OVERCURRENT VTH ISW LV 7.0V 4.7V/3.9V 4.7V/4.5V 73mV 103μA HV 8.5V 7.9V/6.9V 8.0V/7.7V 148mV 42μA 3766fa For more information www.linear.com/LTC3766 LTC3766 Applications Information Input Capacitor/Filter Selection In applications with a low impedance source, or where there the input voltage is relatively low, a simple capacitive input filter is generally suitable. This capacitor needs to have a very low ESR and must be rated to handle a worst-case RMS input current of: ⎛ N ⎞ IOUT(MAX) IC(RMS) = ⎜ S ⎟ 2 ⎝ NP ⎠ Note that capacitor manufacturers’ ripple current ratings are often based on only 2000 hours of life. This makes it advisable to further derate the capacitor, or to choose a capacitor rated at a higher temperature than required. Several capacitors may be paralleled to meet size or height requirements in the design. Due to the high operating frequency of the LTC3766, ceramic capacitors can also be used for CIN. Always consult the manufacturer if there is any question. For higher input voltage applications, however, it can be very costly to use bulk capacitance that is rated to handle the required RMS current. Also, if a simple capacitor is used as an input filter, it is hard to know exactly where the AC input current will flow when a power supply is placed into a larger system. To avoid these issues, an LC filter can be used on the power supply input as shown in Figure 10. This keeps the large AC currents contained in relatively small and inexpensive capacitors whose RMS current rating is known to be adequate. Choosing an LC filter such that: ⎤ ⎡⎛ N ⎞ V R LF < 2.9 • ⎢⎜ S ⎟ RIPPLE + ESR ⎥ CF 2 ⎥⎦ ⎢⎣⎝ NP ⎠ IOUT(MAX) where VRIPPLE is the desired ripple voltage at the output of the input filter and RESR is the ESR of capacitor CF. A reasonable target for VRIPPLE 3% of nominal VIN. When using an LC input filter, the output impedance of the LC filter must never be greater in magnitude than the input impedance looking into the power stage of the DC/DC converter. This is necessary to avoid loop instabilities. In most applications, this condition is naturally satisfied because the ESR of the bulk input capacitance, CBULK, is high enough to lower the Q of the LC input filter, thereby reducing the peaking in its output impedance to a safe level. Also, using a larger value for CF reduces the Q, although this can be expensive in high VIN applications. In some situations, a series damping network must be added as shown in Figure 10. LD VIN+ RD OPTIONAL MAIN TRANSFORMER LF ZOUT CBULK VIN– f 1 < SW 5 2π LFCF ZIN • • CF 3766 F10 Figure 10. Input Filter with Optional Damping Network will attenuate the AC content of the RMS input current by a factor of approximately 5×. This greatly alleviates the RMS current requirements of the bulk input capacitor. The filter inductor should have a saturation current of at least: ISAT(LF) ≥ 1.3 • In order to keep the ripple voltage at the filter output to a reasonable level, choose a value of LF and CF that also satisfies: VOUTIOUT(MAX) In order to provide critical damping, choose LD and RD according to: LD = LF L and RD = 0.8 F 5 CF VIN(MIN) 3766fa For more information www.linear.com/LTC3766 27 LTC3766 Applications Information The damping inductor LD does not carry the DC input current. However, to ensure adequate attenuation during large transients, choose an inductor whose saturation current is at least: ⎛ VOUTIOUT(MAX) ⎞ ISAT(LD) ≥ 0.6 ⎜ VIN(MIN) ⎟⎠ ⎝ Output Capacitor Selection The selection of COUT is driven by the effective series resistance (ESR) and the resulting output voltage ripple. Typically, once the ESR requirement is satisfied, the capacitance is adequate for filtering. The output ripple (∆VOUT) is approximated by: ⎛ ⎞ 1 ΔVOUT ≈ ΔIL ⎜ ESR + 8fSWCOUT ⎟⎠ ⎝ This is because the current loop does not see the magnetizing current, and will not provide its own safeguard against saturation. Note that in nonisolated applications, however, the current sense resistor is placed in series with the primary-side switch, so the current loop will be monitoring magnetizing current. When using a current sense resistor, the IS+ and IS– pins operate differentially and the maximum peak current threshold is approximately 75mV. Normally, the current sense resistor is placed in the source of the forward MOSFET, as shown in Figure 11. Depending on PCB layout and the shielding of the traces going to the IS+ and IS– pins, it is sometimes necessary to add a small amount of filtering as shown in Figure 11. Typically, values of RFL = 100Ω and CFL = 200pF to 1nF will provide adequate filtering of noise pickup without substantially affecting the current loop response. FORWARD MOSFET where fSW is the operating frequency, COUT is the output capacitance and ∆IL is the ripple current in the inductor. The output ripple is highest at maximum input voltage since ∆IL increases with input voltage. Compared to a current transformer, a current sense resistor is less expensive and somewhat simpler to apply than a current transformer. When current sensing on the secondary side of an active clamp forward converter, direct flux limit is required to prevent transformer saturation and possible damage of the primary-side MOSFET. 28 LTC3766 RFL RSENSE RFL Current Sensing and Average Current Limit The LTC3766 supports current sensing either with a current sense resistor or with an isolated current transformer. A current transformer is generally more efficient and has the advantage of sensing current on the primary side in isolated applications. This can be important because it provides an additional safeguard against saturating the main transformer during load transients. In addition, a current transformer can generate a much larger current sense signal than a sense resistor, resulting in a vastly superior signal to noise ratio. This eases board layout concerns for noise pickup and reduces jitter as well. Also, the accuracy of LTC3766 current limit is significantly better in current transformer mode than in current sense mode. FG IS– CFL IS+ 3766 F11 Figure 11. Using a Current Sense Resistor This filter is also helpful in correcting for the effect of the ESL (parasitic inductance) of the sense resistor, which can be important for RSENSE values less than 2mΩ. The effect of the ESL is cancelled if the RC filter is chosen so that: RFLCFL = ESL RSENSE Since the LTC3766 implements an average current limit architecture, choose the value of RSENSE based upon the desired average current limit: RSENSE = 55mV ILIM(AVG) Alternatively, if a current transformer is used to sense the primary-side switch current, then the IS– pin should be tied to VCC and the IS+ pin to the output of the current 3766fa For more information www.linear.com/LTC3766 LTC3766 Applications Information transformer. This causes the gain of the internal current sense amplifier to be reduced, so that the maximum peak current voltage is increased to approximately 1V. The current transformer connections are shown in Figure 12. CURRENT TRANSFORMER + VIN • • SF = DCT IS+ RCAL 1k RSENSE CFL VCC LTC3766 GND NP • • IS– 3766 F12 NS voltage and comparing it to the output inductor current. Figure 13 shows an example of a well-calibrated current sense transformer, where the RSENSE voltage has been scaled by a factor of: MAIN TRANSFORMER NP RSENSEK CT NS Because of the magnetizing current, the slope of the scaled RSENSE voltage will not exactly match that of the inductor current. Choose RCAL so that the scaled RSENSE voltage and the inductor current are identical at the peak. PG VIN– Figure 12. Using a Current Sense Transformer Typically, the current transformer is placed in series with the power supply feed to the main transformer. This reduces common mode noise, and is generally convenient for the PCB layout. Use a small filter capacitor, CFL, between 1nF and 3.3nF, or a time constant for RSENSE • CFL of less than 75ns, to eliminate high frequency noise. The diode DCT is needed to allow the core of the current transformer to properly reset. When using a current transformer, set the value of RSENSE using: RSENSE = 0.73V N • P K CTILIM(AVG) NS where NP/NS is the turns ratio of the main transformer, KCT is the current gain of the transformer, and ILIM(AVG) is the average current limit desired. For most applications, a current transformer turns ratio of 1:100 is suitable (KCT = 0.01). The resistor RCAL is added to compensate for the effects of the magnetizing current in the main and current sense transformers, both of which cause the voltage on RSENSE to be somewhat higher than expected (2% to 8%). Typically, RCAL is in the range of 1.5k to 5k. For the highest possible accuracy, the value of RCAL should be adjusted to calibrate the current sensing at full load and nominal input voltage by carefully measuring the RSENSE 20A INDUCTOR CURRENT 2A/DIV SF • VRSENSE 2A/DIV 500ns/DIV 3766 F13 Figure 13. Properly Calibrated Current Transformer In order to maintain a constant average current while in current limit, the LTC3766 automatically adjusts the value of the peak current limit to cancel the effect of the inductor ripple current. This is accomplished by creating an internal ramp that mimics the inductor current ripple. The amplitude of this ramp is determined by the resistor on the IPK pin, which must be set to be proportional to the output inductor. The LTC3766 establishes a voltage on the IPK pin of (VSW – VS+)/15, which is one-fifteenth of the voltage across the output inductor during the ontime when SW is high. Therefore, it is imperative that the SW and VS+ pins be connected as shown in Figure 14 or Figure 15 so that the LTC3766 can properly sense the inductor voltage. If the differential amplifier is not needed, tie VS– and VS+ together to VOUT as shown in Figure 14b. For high VOUT applications where the SW node plateau voltage is greater than 40V, it is necessary to add a resistor divider on both the SW and VS+ pins, as shown in Figure 15. This divider will limit the voltage at the SW pin and also impact the SG reverse overcurrent trip threshold. See Setting the SG Reverse Overcurrent for details on selecting the resistor divider on the SW pin. 3766fa For more information www.linear.com/LTC3766 29 LTC3766 Applications Information • For resistor sense mode, place a resistor on the IPK pin that is chosen using: VOUT VSW • LTC3766 VS+ MAIN XFMR SW VS– IPK VSOUT RIPK RLOAD 3766 F14a Figure 14a. Setting the Average Current Limit (RIPK) • LTC3766 MAIN XFMR VS+ SW VS– IPK VSOUT RIPK 3766 F14 Figure 14b. Setting RIPK with No Differential Amplifier • MAIN XFMR R1 LTC3766 VSOUT SW R2 160k VS– IPK RIPK 120k VS+ KRLIPK N • P (1.32nF )KCT RSENSE NS When the LTC3766 is in current limit and the output voltage is very low, the control of the output current will be limited by the minimum on-time of the converter. Once this minimum on-time has been reached, further decreases in output voltage during current limit will result in an inductor current that continues to rise, until the overcurrent limit is reached. This will cause the LTC3766 to shut down and attempt a restart, resulting in a hiccup mode of operation. Typical average current limit performance is illustrated in Figure 16. Note that the average current delivered to the load is held substantially constant as the output voltage is decreased down to a low level, at which point the converter will enter hiccup mode. Depending upon the particular application, hiccup mode is entered either due to the loss of secondary-side bias voltage (UVLO) or due to an overcurrent fault. VOUT VSW • RIPK = VOUT VSW KRLIPK (17.6nF )RSENSE where LIPK is the inductance of the output inductor at I = ILIM(AVG). For low VOUT applications where no SW node divider is needed, KR = 1. For current transformer mode, use: RB VFB RA • RIPK = R3 R4 GND 3766 F15 6 Figure 15. Setting RIPK for High VOUT Applications KR = R2 69k •R4 = R1+R2 69k •R4+R3 ( 69k +R4) where the 69k accounts for the internal resistance on the VS+ and VS– pins. OUTPUT VOLTAGE (V) Note that the ratios of the resistor dividers on the SW and VS+ pins must be the same for ripple cancellation to operate properly. This requires that: 5 4 3 2 1 0 VIN = 72V VIN = 36V 0 5 10 15 20 LOAD CURRENT (A) 25 30 3766 F16 Figure 16. Typical Current Limit Performance 30 3766fa For more information www.linear.com/LTC3766 LTC3766 Applications Information Estimating the Average Current Limit Accuracy The accuracy of the average current limit depends on the LTC3766 specifications together with a number of application circuit parameters as well as parasitics. Consequently, it is very difficult to precisely calculate the average current limit accuracy. This accuracy can be estimated, however, by carefully considering the three primary sources of error: 1. The accuracy of the current sense resistor and/or current sense transformer. For resistor sensing, the accuracy of the current sense resistor is normally 1%. For sense resistors less than 2mΩ, however, the parasitic inductance can cause a significant error in the sensed current. This error can be eliminated by adding an RC filter as shown in Figure 11. When using a current transformer, the accuracy of the sense resistor on the current transformer secondary and the turns ratio of the transformers (both KCT and NP/ NS) are generally 1% or better. Depending on where the current sense transformer is placed, however, there can be an additional 1% to 4% error due to the magnetizing current of both the main and current sense transformers. Generally, this error is in the form of a relatively constant offset, and it can be adjusted out for a particular design for nominal input voltage and maximum load current. The resulting tolerance due to the variation in magnetizing current effects is generally less than 2%, resulting in an overall accuracy of approximately 3% for current transformer sensing. 2. The accuracy of the average current sense threshold VIS(AVG). The accuracy of the LTC3766 current sense threshold is given in the Electrical Characteristics table and depends on the current sense mode chosen. Current transformer mode provides an accuracy of 10% and is more accurate than the resistor sense mode accuracy of 15%. 3. The accuracy of the compensation for inductor ripple current. The accuracy of the inductor ripple compensation depends both on the internal adjustment of VITH as well as the tolerance of the output inductor itself. For most application circuits, the ripple compensation accuracy will be 25% or better for current transformer mode, and 35% or better for resistor sense mode. Note that the inductor current ripple is typically 30% to 60% of the average current limit, and only one-half of this peak-to-peak ripple is being compensated. As a result, the effect of the ripple compensation accuracy on the average current limit is attenuated by a factor of: FR = R R+ 2 where R is the ratio of the peak-to-peak inductor ripple current to the average current limit. For 30% to 60% ripple, for example, the value of FR varies from 0.13 to 0.23. Considering each of the above factors, the worst-case tolerance of the average current limit can be estimated as: ∆IAVG = 3% + 10% + 0.23 (25%) = 18.5% for current transformer mode and: ∆IAVG = 1% + 15% + 0.23 (35%) = 24% for resistor sense mode. Since the three sources of error are statistically independent, the current limit tolerance for current transformer and resistance sense modes can be calculated using the RSS method as approximately 12% and 17% respectively. Setting the Gate Drive Delays The forward switch gate driver (FG) and the synchronous switch gate driver (SG) operate with make-before-break timing on the FG rising edge, and with simultaneous timing on the SG rising edge. The delays for these transitions relative to the switching of the primary-side MOSFETs are critical for optimizing efficiency, and can be configured independently using the SGD and FGD pins. The SG rising delay should be adjusted to minimize the switch node (SW) body diode conduction. At full load, the power loss in the body diode is significant, and the SG rising delay can have a substantial impact on efficiency. By minimizing the dead time between PG falling and SG rising (while avoiding shoot-through), this power loss is also minimized. Similarly, the dead time between SG falling (set by the FG rising delay) and PG rising should also be minimized. 3766fa For more information www.linear.com/LTC3766 31 LTC3766 Applications Information In addition to being set to minimize the dead time between SG falling and PG rising, the FG rising delay should also be adjusted to ensure that the drain of the forward switch (SWB) is close to 0V when the switch is turned on, which minimizes switching loss. When the LTC3765 active clamp switch turns off, the drain voltage of the primary switch (SWP) decreases linearly from VIN/(1 – D) to VIN, where D is the duty cycle. On the secondary side of the transformer, SWB ramps from VOUT/(1 – D) to 0V. Switching power loss is minimized when FG and PG MOSFETs are switched with minimal drain-to-source voltage across them. The FG and PG rising delays should be adjusted to ensure that the SWB and SWP nodes are at their minimums when the switches are turned on. Keep in mind the following set of relationships when setting the delays (refer to the Timing Diagram and Figure 1): 1. The forward gate (FG) always turns on with makebefore-break timing relative to the synchronous gate (SG). This ensures that negative inductor current does not create excessive voltage on the synchronous switch drain. 2. Shoot-through is caused when the synchronous gate (SG) and the LTC3765 primary gate (PG) are simultaneously high, or when the forward gate (FG) is high and the LTC3765 active gate (AG) is low. The leakage inductance of the main transformer will prevent significant power loss due to shoot-through for a few tens of nanoseconds; however, if the PG and SG or FG and AG gates are on simultaneously for a longer period of time, the shoot-through will cause power loss, excessive heat, and potentially rapid part displacement. 3. The primary side turn-off of either AG or PG should occur before FG and SG switch, and the primary-side turn-on should occur after FG and SG switch. For example, on a particular cycle, AG goes high first (turning the PMOS off), then FG goes high, then SG goes low, then PG goes high. On the PG turn-off edge, PG goes low first, then FG goes low and SG goes high, then AG goes low (turning the PMOS on). 32 Delay Resistor Selection: PG Turn-Off Transition In general, the PG turn-off delays are relatively simpler to set and less critical than the PG turn-on delays. At the end of the PWM on-time, the LTC3766 will assert a falling edge on the PT+ pin, which in turn causes the LTC3765 to immediately turn off the PG MOSFET. After a fixed 180ns delay, the LTC3765 will then turn on the AG MOSFET. Consequently, the only delay adjustment to be made for this transition is on the secondary side using the SGD pin of the LTC3766. This pin is used to set the delay from PT+ falling to FG falling/SG rising, which must occur after PG turn-off and before AG turn-on. The first consideration in setting the SGD delay is to reduce the dead time between PG and SG, during which the body diode of the synchronous MOSFET is carrying the load current. After PG turn-off, the SW node on the secondary side will rapidly fall until being clamped by the body diode of the SG MOSFET. The objective is to turn on the SG MOSFET as the SW node crosses through 0V. The LTC3766 makes this easy to achieve by directly sensing the SW node and inhibiting SG turn-on until SW has fallen through 0.5V. In other words, minimum dead time between PG and SG can be achieved by setting the SGD delay to any value less than or equal to the delay time from PT+ falling to SW falling through 0V. In general, this delay time is in the range of 50ns to 100ns. The resistor from SGD to ground that gives a particular delay tSGD can be computed using: RSGD = ( t SGD – 12ns ) • 1kΩ 4.3ns A 10k resistor from SGD to ground sets the FG falling/SG rising delay to approximately 50ns, which is generally a good starting point. To prevent damaging cross conduction between the FG and AG MOSFETs, do not set the SGD delay to be longer than the 180ns fixed turn-on delay of the AG MOSFET. Always start low when setting the SGD delay. This is safe because of the adaptive limit that inhibits premature SG turn-on. 3766fa For more information www.linear.com/LTC3766 LTC3766 Applications Information Another important consideration in setting the SGD delay is the prevention of SWP collapse due to excessive FG turn-off delay. After PG turn-off, the SWP node is quickly driven high by the transformer leakage to a level of approximately VIN/(1 – D). Ideally, it should remain at this voltage as FG turns off, SG turns on, and then AG turns on. However, if the delay to FG turn-off is too long, the SWP voltage will momentarily fall towards VIN, and it will not rise again until being forced high by AG turn-on. This collapse of the SWP node is illustrated in Figure 17, and is more prominent at lighter loads. It can significantly degrade efficiency as the SWP node is discharged and recharged every cycle, but it is easily avoided by further shortening the SGD delay. Although the LTC3766 inhibits the SG turn-on until SW < 0.5V, this is not true of the delay to FG turn-off. The delay from PT+ falling to FG turn-off can be decreased beyond the adaptive ~ VIN 1–D SWP NODE VIN 0V AG PT + FG tSGD TOO LONG ~ VIN 1–D SWP NODE VIN 0V AG PT + FG tSGD OK 3766 F17 Figure 17. Avoiding SWP Collapse from Long Delay limit of SG turn-on, so that the FG and SG edges can be separated with a small dead time between them. This is important to allow the FG turn-off to be separately optimized based upon circuit parasitics. In most applications, a peak in full load efficiency is normally found with the SGD delay set so that there is no SWP collapse and there is a small dead time between FG turn-off and SG turn-on. In applications where efficiency is less critical, this delay can be set adaptively by tying SGD to GND. In this case, FG falling and SG rising will both be inhibited until SW < 0.5V. For fixed delay mode, always use a resistor of 8k or greater on the SGD pin to avoid activating the adaptive delay mode. Delay Resistor Selection: PG Turn-On Transition The delays associated with the PG turn-on transition are set by the DELAY pin on the LTC3765 and the FGD pin on the LTC3766. At the beginning of the PWM on-time, the LTC3766 will assert the PT+ pin high, and will then turn FG on and SG off after a delay set by the resistor on the FGD pin. On the primary side, the LTC3765 will immediately turn off the AG MOSFET in response to PT+ rising, and it will then turn on the PG MOSFET after a delay determined by the resistor on the LTC3765 DELAY pin. The FGD delay resistor on the secondary side must be selected in careful coordination with the delay on the primary side; therefore, the following procedure outlines how to choose components for both the LTC3765 and LTC3766. The first objective in setting the PG turn-on delays is to minimize switching loss by turning on the PG and FG MOSFETs at minimum drain voltage. After the AG MOSFET has turned off, the PG and FG drain voltages (SWP and SWB) will naturally ramp down to approximately VIN and 0V respectively. These voltages take 100ns to 500ns or longer to fall, depending on the main transformer magnetizing inductance and the parasitic capacitance on the MOSFET drains. Choosing the delay settings correctly can significantly impact the power loss due to switching the MOSFETs. For a particular design, the most effective procedure is to set the PG and FG delays based on the resulting waveform on SWP and SWB. In order to evaluate these waveforms, 3766fa For more information www.linear.com/LTC3766 33 LTC3766 Applications Information the delays should initially be selected so that they are long, while keeping in mind that the FG delay must be less than the PG delay to prevent potentially damaging PG/SG cross-conduction. As a first pass, use a 75k resistor from FGD to ground for a 415ns delay and 60k resistor from DELAY to ground for a 622ns delay. The SWP and SWB waveforms should appear as shown in Figure 18. The ramp rate on SWB and SWP is to a first order independent of duty cycle; however, the starting point of the ramp is a function of the duty cycle. Therefore, the longest delay time will be at high duty cycle when VIN is at a minimum. For the lowest switching losses over the range of input voltage, the delays should be chosen based on the waveforms when VIN is at its minimum operating voltage. The resistor value from the FGD pin to ground should be selected first. This should be chosen to give a delay equal to the time from PT+ rising until SWB ramps down to approximately 0V. The FGD resistor value can be determined from the following equation: RFGD = ( tFGD – 18ns) • 1kΩ 5.1ns Note that if the FG turns on before the SWP and SWB voltages have naturally fallen to their minimums, they will be instantly pulled to their minimum by the FG MOSFET PT+ 0V ~ VIN 1–D SWP NODE VIN 0V V ~ OUT 1–D SWB NODE 0V 3766 F18 tFGD Figure 18. SWP and SWB Waveforms 34 turning on. This can give the appearance that FG is turning on after SWB has ramped to 0V, although it is actually premature. Turning on FG prematurely will slightly degrade efficiency due to increased switching loss; however, if the fall time of SWP and SWB exceed a maximum FGD delay of 600ns, it is acceptable to have premature FG turn-on at low input voltage. Generally, the delay will be adequate at higher VIN to allow a complete ramp down. In rare cases, the LTC3765 and LTC3766 will be in delay phase-out mode when operating at minimum VIN voltage. This will be apparent because the measured delay will be smaller than the programmed delay on either or both chips. This feature allows the LTC3765 and LTC3766 to operate at duty cycles up to a maximum of 79% by reducing the programmed delays when they would otherwise limit the maximum duty cycle. If this mode is evident, increase VIN until delay phase-out is no longer active, and then set the FGD delay as described above. Having set the FGD delay to optimize for low voltage switching, the PG delay is next chosen to minimize the dead time between SG turn-off and PG turn-on. The delay for the primary gate can be determined by taking the delay set tolerance and rise/fall times into account. The FG delay setting on the LTC3766 and the PG delay setting on the LTC3765 are both accurate to within 15% for a range of resistance values. Given this accuracy, a reasonable choice for the LTC3765 delay time is to set the PG delay time to 1.22 • tFGD. Be aware that the fall time of SG and the rise time of PG cannot be neglected. For example, if SG is driving a MOSFET with high input capacitance, and PG is driving a MOSFET with low input capacitance, then SG will fall slowly and PG will rise quickly. This increases the potential for shootthrough. Moreover, since SG will not turn off until FG turns on (make before break), the rise time of FG is also a factor. A final consideration is that the LTC3765 experiences a delay in PT+ rising due to the pulse transformer. All of these considerations can be accounted for in the delay resistor selection by the following equation, in which tD(PT) is the delay time from PT+ rising to IN+ rising on the LTC3765, tR(FG) is the rise time of FG to 2V, tF(SG) is the fall time of 3766fa For more information www.linear.com/LTC3766 LTC3766 Applications Information SG to 1V and tR(PG) is the rise time of PG to 1V. The delay time can then be chosen such that: tPGD = 1.22 • tFGD + tR(FG) + tF(SG) – tR(PG) – tD(PT) The resistor from the LTC3765 DELAY pin to ground can be selected to give this delay by using the following equation: RDELAY = ( tPGD – 45ns ) • 1kΩ 9.5ns In practice, the LTC3765 PG turn-on delay should be optimized by monitoring the PG and SG waveforms. A conservative approach is to set the PG delay to create a dead time between SG falling and PG rising that accounts for the delay set tolerances (typically 22% of the total delay). A more aggressive approach takes into account the fact that transformer leakage inductance will delay the effect of PG turn on (i.e., SW node rising) by 75ns to 150ns or more at full load. Also, transformer leakage inductance mitigates the effect of a small amount of shoot-through by slowing the rise time of the transformer current. Higher full-load efficiency can be achieved by setting the PG turnon closer to SG turn-off. In addition, a shorter dead time at PG turn-on can reduce the overshoot and ringing on the switch node, thereby reducing the size of the required RC snubber and its associated power loss. While a shorter dead time at PG turn on can improve full-load performance, care must be taken to ensure that the worst case shoot-through at no load is well within safe limits. Maximum Duty Cycle and Delay Phase-Out While the PG turn-on delay time is important for reducing turn-on switching losses, no power is transferred from the input supply to the output load during this delay time. In most forward converter systems, the maximum available duty cycle is artificially limited by this delay, which then forces a trade-off between the optimal delay time and the maximum available duty cycle. The LTC3765 and LTC3766 implement a unique delay phase-out feature in which the PG and FG turn-on delays are gradually reduced as the demanded duty cycle approaches the maximum value of 79%. This feature allows a forward converter to be designed with an optimal delay at nominal input voltage, but still approach the maximum duty cycle at low input voltage, thereby making better utilization of the power transformer. Generating Secondary-Side Bias There are five items to consider when determining the best way to generate bias for the LTC3766 in an isolated application: 1. The required operating current. This includes the gate drive current for both primary and secondary MOSFETs as well as the operating supply currents of both the LTC3765 and the LTC3766. 2. The operating voltage needed for the MOSFET gates. Depending on whether logic-level or standard threshold MOSFETs are used, the VCC operating voltage and undervoltage lockout (UVLO) levels can be set accordingly using the MODE pin. The bias supply must provide adequate voltage to keep the LTC3766 VCC pin above its UVLO level and keep the overall supply operating at peak efficiency. 3. Current limit operation at low output voltage. The minimum required VOUT during current limit relative to the normal operating VOUT has a major impact on the design of the bias supply. The bias supply must provide adequate voltage over this range of VOUT voltages. 4. The variation in input voltage. At minimum input voltage, the bias supply must still provide enough voltage for proper operation. At maximum voltage, the bias supply must not generate a voltage that exceeds maximum ratings or dissipates excessive power. 5. The potential need for a rapid hand-off from primary to secondary control. In PolyPhase applications, it is important to quickly transfer control to the secondary side during start-up so that current sharing and proper phasing can be established before the full load current is seen at the output. By contrast, some applications may not need to have control handed off to the secondary until just prior to the output reaching its regulation value. In all applications, however, the secondary bias must always come up and control must be transferred before the output reaches the regulation level. The current that must be supplied by the secondary bias supply can be estimated using IVCC ≈ (QGPRIfSW + 3mA)NPT + QGSECfSW + 18mA For more information www.linear.com/LTC3766 3766fa 35 LTC3766 Applications Information where QGPRI is the total gate charge of all primary-side MOSFETs, QGSEC is the total gate charge of all secondaryside MOSFETs, and NPT is the turns ratio of the pulse transformer. Note that the primary-side current is scaled by the turns ratio of the pulse transformer. The 18mA constant in the above equation includes typical gate drive switching current as well as losses associated with the pulse transformer. Using VOUT Directly for Secondary-Side Bias The simplest method of generating secondary-side bias is to directly use the output voltage of the converter. This is only practical when VOUT is in the range of 5V to 15V. When VOUT is in the range of 5V to 10V, it can be directly connected to VCC as shown in Figure 4a. When VOUT is in the range of 6V to 15V, it can be used as a bias input to the VAUX regulator as shown in Figure 4b. For output voltages higher than 15V, this method is generally not practical due to high power dissipation. This simple method also does not provide constant current limit operation at lower output voltage. It also does not provide a quick hand-off to the secondary and is not recommended for PolyPhase applications. Using a Peak Charge Circuit for Secondary-Side Bias A common way to generate a bias voltage on the secondary side is by using a peak charge circuit connected to the transformer secondary, as shown in Figure 19. This circuit is useful for generating an unregulated bias voltage that can be directly tied to the VIN pin of the chip and used as an input to the high voltage linear regulator. The peak charge circuit is capable of providing bias even at low output voltages, so it is a good choice when constant current limit operation is needed over a wide VOUT range. FROM TRANSFORMER SECONDARY LTC3766 VIN NDRV VCC VBIAS = 6V TO 32V DPK RPK CPK Q1 CVCC 3766 F19 Figure 19. Peak Charge Circuit for Secondary Bias 36 Since it provides a bias voltage even when the converter is operating at tiny duty cycles, the peak charge is also a good choice for PolyPhase applications where a quick hand-off to secondary is important. However, since the output of a peak charge circuit directly follows changes in the converter input voltage, it is should only be used in applications where the input voltage varies by 2:1 or less. Note that for bias voltages on the VIN pin of 28V or greater, the internal 30V clamp will draw between 3.5mA and 7mA. This will result in 100mW to 200mW of additional power dissipation in the LTC3766. To limit the initial charging current out of the peak charge circuit, use a series resistor RPK in the range of 1Ω to 4Ω. A schottky diode DPK with a peak surge current rating of 5A or higher should also be used, and the pass transistor Q1 should have a minimum beta of at least 200. Capacitor CPK should be a ceramic capacitor with a value of at least 2.2μF or greater. During open-loop start-up, it is imperative that the peak charge bias come up and control is transferred to the secondary before an output overvoltage can occur. Since a peak charge circuit is not directly coupled to the output voltage of the converter, care must be taken to ensure that the primary-side soft-start is not too fast relative to the rise time of the peak charge bias on the secondary side. The time required for the peak charge bias voltage to rise to a level that allows control to be handed off to the secondary can be approximated using: tBIAS ≈ 103 • REQCPKCSSP +150µs where REQ is the sum of RPK and the series resistance of diode DPK, and CSSP is the LTC3765 soft-start capacitor. During open-loop soft-start, the time required for the converter output voltage to reach a given level VHO can be approximated using ⎞ ⎛ ⎟ ⎜ 2 2 4 ⎜ CSSP ( VHO ) LCOUT fSW ⎟ tOUT ≈ 10 • 2 ⎟ ⎜ ⎛ NS ⎞ ⎟ ⎜ ⎜⎝ VIN(MIN) • N ⎟⎠ ⎟⎠ ⎜⎝ P 1/3 The above equation assumes that there is no load current, which is the worst-case condition for output voltage rise. 3766fa For more information www.linear.com/LTC3766 LTC3766 Applications Information When calculating tOUT, use a value for VHO that corresponds to the target output voltage for control hand-off, typically one-half the normal regulation level or less. If tOUT is less than tBIAS, then the LTC3765 soft-start capacitor value should be increased. Note that these equations are approximations and the actual times will vary somewhat with circuit parameters. Peak Charge Bias Configurations When the peak voltage on the SW node is in the range of 7V to 32V, the peak charge can be taken directly from the SW node as shown in Figure 20. In practice, this condition only holds when the output voltage of the converter is approximately 5V. In most applications, it is necessary to add an additional auxiliary winding on the secondary for use in generating an SW NP • • NS DPK MAIN XFMR RPK LTC3766 VIN NDRV VBIAS = 6V TO 32V CPK Q1 VCC CVCC adequate bias voltage. For low VOUT applications (VOUT < 5V), this winding can be configured as shown in Figure 21 to provide a higher voltage for bias generation. This configuration is advantageous because it achieves a higher voltage on the transformer secondary with a minimum number of additional turns. The number of turns on the auxiliary winding for this configuration should be approximately: ⎛ VB(MIN)DMAX ⎞ NAUX ≈NS ⎜ – 1⎟ VOUT ⎝ ⎠ where DMAX is the maximum operating duty cycle (typically 0.65 to 0.70) and VB(MIN) is either 7V for low voltage or 10V for high voltage drive mode operation. These values for VB(MIN) are approximately 2V higher than the UVLO levels on the LTC3766 to allow for drops in the peak charge circuit. As an example, for VOUT = 1.5V, DMAX = 0.65V and NS = 1 turn, use NAUX = 2 turns, assuming low voltage drive mode. For high VOUT applications (VOUT > 6V), this winding can be configured as shown in Figure 22 to provide a reduced voltage for generating bias. In this case, choose an auxiliary winding with turns ⎛ VB(MIN)DMAX ⎞ NAUX ≈NS ⎜ ⎟⎠ VOUT ⎝ 3766 F20 Figure 20. Peak Charge Directly from SW for VOUT ≈ 5V • • NAUX MAIN XFMR SW NP • • NAUX NP NS • SW • NS DPK DPK MAIN XFMR RPK LTC3766 VIN NDRV VCC VBIAS = 6V TO 32V VIN NDRV CPK Q1 RPK LTC3766 VCC CVCC VBIAS = 6V TO 32V CPK Q1 CVCC 3766 F22 3766 F21 Figure 21. Peak Charge for Low VOUT Applications Figure 22. Peak Charge for High VOUT Applications 3766fa For more information www.linear.com/LTC3766 37 LTC3766 Applications Information At maximum VIN, there may be considerable power dissipation in the linear regulator pass device Q1. This power can be calculated using PQ1 = (VBIAS – VCC)IVCC For Figure 24 the output is given by: VBUCK = VOUT NAUX – 0.5 NS In applications where the peak charge and high voltage linear regulator must operate continuously, transistor Q1 must be capable of dissipating this power without excessive temperature rise. In such applications, use a transistor with a suitable package (SOT89) and connect the thermal tab of the transistor to an adequately large island of copper on the PCB. For a buck bias supply, inductor LBK must be rated to carry the required VCC bias current and should have an inductance value that will provide continuous current operation at one-fourth of the required bias current load or less. Choose and inductor LBK to according to: High Efficiency Secondary-Side Bias Techniques A value of 1mH for LBK is adequate for most applications. A high-efficiency alternative to using a peak-charge circuit to generate secondary-side bias is to connect a buck output to the transformer secondary. This buck output is normally combined with a peak charge circuit as shown in Figures 23 and 24. The bias voltage from this buck output can be fed directly into the VAUX pin. This arrangement combines the quick start-up and flexibility of a peak charge circuit with the higher operating efficiency of a buck bias supply. The output voltage of the buck bias supply (VBUCK) should be set to optimize efficiency during normal operation. This will typically require a somewhat higher number of auxiliary turns than is ideal for a peak charge output. As a result, the buck supply and the peak charge circuit are sometimes driven from separate auxiliary windings. Also, note that the output voltage of the peak charge circuit will increase somewhat when the VAUX bypass regulator is activated and the high voltage linear regulator is disabled. Care must be taken not to exceed the maximum voltage rating on the VIN pin of the LTC3766. For Figure 23, the output voltage of the buck bias supply is given by: ⎛ N ⎞ VBUCK = VOUT ⎜ 1+ AUX ⎟ – 0.5 NS ⎠ ⎝ LBK > VCC ICC fSW • • DBK NAUX LBK SW NP • • DBK NS MAIN XFMR VBUCK CBK DPK OPTION TO LIMIT Q1 POWER REGSD NP • VAUX • NS VIN NDRV VCC CPK Q1 OPTION TO LIMIT Q1 POWER CVCC 38 DBK CBK DPK RPK REGSD VAUX VIN NDRV VCC VBIAS = 6V TO 32V CPK Q1 CVCC 3766 F24 3766 F23 Figure 23. Buck Bias Supply for Low VOUT Applications VBUCK LTC3766 CRSD VBIAS = 6V TO 32V DBK LBK SW RPK LTC3766 CRSD NAUX MAIN XFMR Figure 24. Buck Bias Supply for High VOUT Applications 3766fa For more information www.linear.com/LTC3766 LTC3766 Applications Information The buck bias winding can also be used standalone without the peak charge supply, as shown in Figure 25. This is sometimes done in applications where the peak charge circuit is impractical, such when the VIN voltage has a wide range. When using the buck bias supply standalone, particular care must be taken to ensure that the bias output comes up more quickly than the main output, and that there is adequate bias voltage immediately after control handoff. This is made more difficult by the presence of some load on the VCC pin during start-up whereas there may be no load on the main output. In general, a clean startup with a standalone buck bias supply can be achieved by observing the following guidelines: 1) set the turns ratio of the auxiliary winding so that the operating VAUX will be at least 3V above the rising VCC UVLO voltage, 2) use a smaller value for LBK, typically one-half of that calculated in the above equation, but always large enough for continuous current in LBK during normal operation 3) use the high-voltage linear regulator to minimize the load on VCC during start-up, 4) use the RUN pin to monitor the bias voltage and set the start-up voltage to 2V above the rising VCC UVLO voltage with a hysteresis of 1.5V, 5) use a shorter soft-start time, less than 10ms if possible, 6) use a small VCC capacitor (typically CVCC = 0.22μF) and a capacitor CBK given by: CBK = 20 ⎡⎣(QGPRIfSW + 3mA )NPT +18mA ⎤⎦ fSWVHYST where VHYST is the hysteresis set by the RUN pin (1.5V). Note that this value for CBK is as small as possible so that VBUCK rises quickly, but large enough to support the bias current until control is handed off to the secondary and the duty cycle increases. Once control is handed off, both the buck supply and the main converter will be operating in continuous current mode, so their outputs will track. Another high efficiency option for generating bias is to make use of an inductor overwinding, as shown in Figure 26. This supply is created by adding a second winding on the main output inductor. During the on-time of the synchronous MOSFET, the VOUT voltage is scaled and coupled through diode DOW to capacitor COW, so that the resulting bias voltage is: VOW = VOUT NL2 – 0.5 NL1 This is similar to the buck supply in that it is highly efficient and fairly well regulated. However, it is simpler in that it does not require the use of an additional inductor to generate the bias voltage. Another advantage of this technique is that the bias voltage always tracks VOUT, so there is no concern about the bias voltage potentially lagging the output voltage during start-up. Like the buck bias supply, the inductor overwinding can be used either stand alone (as shown in Figure 26) or together with a peak charge bias supply. Use a schottky diode DOW with a peak surge current rating of 5A or higher. Capacitor COW should be a ceramic capacitor with a value of 2.2μF or greater. NL1 SW NP • • DBK LBK VBUCK NAUX DBK NP CBK MAIN XFMR MAIN XFMR RR1 RR2 VOUT NS NL2 DOW VOW COUT COW LTC3766 LTC3766 RUN • • VAUX VIN NDRV VCC VIN Q1 NDRV CVCC VCC 3766 F25 Figure 25. Using the Buck Bias Supply Standalone CVCC 3766 F26 Figure 26. Inductor Overwinding Bias Supply 3766fa For more information www.linear.com/LTC3766 39 LTC3766 Applications Information A useful variant of the inductor over-winding bias supply is shown in Figure 27, where a discrete transformer TOW has been used instead of an additional winding on the main inductor LF . This is often more convenient because standard parts can readily be used. in Figures 23 and 24). This enables a low power pass transistor to be used. See Linear Regulator Operation for more information on using the REGSD feature. In the circuit of Figure 27, a second diode DOW2 has been added to prevent DC bias current from being carried in the transformer TOW. This transformer can be either a gate-drive or flyback-style transformer, which are widely available in a range of turns ratios. Note that transformer TOW requires only functional isolation and can be physically very small. This circuit produces a bias voltage given by: The soft-start ramp time on the LTC3766 is set by placing a capacitor between the SS pin and GND. This secondary-side soft-start capacitor only controls the output voltage ramp after control hand-off has taken place. Consequently, its effect on the overall output voltage start-up will depend on the primary to secondary hand-off voltage in the particular application. Choose a soft-start capacitor using: VOW = ( VOUT – 0.5) NL2 – 0.5 NL1 CSS = During an output overload condition, the voltage generated by a either a buck supply or inductor overwinding supply will drop as the converter output voltage decreases. If this happens and there is no peak charge bias supply, then the LTC3766 will have a UVLO fault that will cause both the LTC3765 and LTC3766 to shut down and attempt a restart. If a peak charge supply is used together with a buck or inductor overwinding supply, then the LTC3766 will automatically re-energize the high voltage linear regulator when the VAUX pin gets too low. If continuous operation of the peak charge and high voltage regulator is not needed, then the REGSD pin can be used to limit the total time that this regulator is allowed to operate (shown as an option LF SW NL1 NP MAIN XFMR NS VOUT DOW2 • • NL2 DOW1 VOW COUT COW LTC3766 VAUX VIN NDRV VCC CVCC 3766 F27 Figure 27. Inductor Overwinding Using Standard Parts 40 Soft-Start Ramp Time and Control Hand-Off (5µA ) t SS ( 1.83 0.6 – VFB(HO) ) where tSS is the soft-start time after control hand-off to the secondary and VFB(HO) is the voltage on the FB pin at control hand-off. The total soft-start time will be the sum of tSS and the open-loop soft-start time prior to control hand-off set by the LTC3765. Note that during the openloop soft-start time, the output voltage ramp will vary significantly with load, since the synchronous MOSFET is not enabled and the converter may operate in discontinuous current mode. If precise control over the soft-start time is desired, use a secondary-side bias scheme that provides control hand-off at the lowest possible output voltage. See above sections on generating secondary-side bias for details. Just prior to control hand-off, the LTC3766 rapidly presets the soft-start capacitor so that the internal soft-start voltage is equal to VFB(HO), ensuring a smooth transition from primary to secondary control. Due to the dielectric absorption of the soft-start capacitor, however, the voltage on the soft-start capacitor may droop somewhat following the initial preset. This can result in a small step down in the output voltage ramp after control hand-off, and an associated negative current transient in the output inductor. To minimize this effect, use a soft-start capacitor with a low dielectric absorption, such as an NPO ceramic capacitor. Pulse Transformer Selection The pulse transformer that connects the LTC3766 PT+/PT− outputs to the LTC3765 IN+/IN− inputs functions 3766fa For more information www.linear.com/LTC3766 LTC3766 Applications Information as the communication link between the secondary-side controller and the primary-side gate driver, as shown in Figure 28. In addition, LTC3765 contains a bridge rectifier that extracts bias power from the pulse transformer, which it then uses to drive the gates of the primary-side MOSFETs. The designs have been coordinated so that the transformer turns ratio should be set to: NPT = NLTC3765: NLTC3766 = 2:1 for low voltage mode operation on the LTC3766 (VCC = 7V), and: NPT = NLTC3765: NLTC3766 = 1.5:1 for high voltage mode operation on the LTC3766 (VCC = 8.5V). The resulting VCC voltage on the LTC3765 is approximately: VCC(3765) = VCC(3766)NPT – 1.3 Using the above turns ratios will provide a primary-side VCC voltage of approximately 12V for the LTC3765 to drive the gates of the primary-side MOSFET. Note that the primary-side VCC voltage provided by the pulse transformer must also be greater than the LTC3765 UVLO threshold for proper operation. Care must also be taken not to exceed the maximum voltage rating on the LTC3765 VCC pin. The pulse transformer must also have a minimum voltsecond rating as required by the 79% duty cycle signal on PT+/PT − and the lowest frequency of operation. The required volt-seconds rating can be calculated from the minimum frequency as: Volt-Sec = 0.33 • VCC fSW(MIN) Since the pulse transformer is used for transmitting PWM information as well as bias power, choose a pulse 1µF 0.1µF IN+ PT+ 100Ω LTC3765 • • LTC3766 220pF IN– NLTC3765:NLTC3766 PT– 3766 F28 Figure 28. Pulse Transformer Connection transformer with a leakage inductance of 1μH or less. This reduces ringing and distortion of the PWM information so that a solid communication link is always maintained. For low voltage (7V) mode on the LTC3766, transformers that meet the above requirements include the PA2008 from Pulse Engineering and the DA2320 from Coilcraft. For high voltage (8.5V) mode on the LTC3766, transformers that meet the above requirements include the PA3290 from Pulse Engineering. The 1µF and 0.1µF capacitors in series with the pulse transformer of Figure 28 are for blocking and restoring the DC level of the signal. The 220pF/100Ω RC snubber shown at the IN+/IN– inputs of the LTC3765 is required to minimize ringing due to the leakage inductance of the pulse transformer. The values shown for each of these four components are appropriate in nearly all LTC3765/ LTC3766 applications. Voltage Loop Compensation The voltage loop of the LTC3766 is compensated in much the same way as a standard buck converter, by placing a compensation network on the ITH pin. It is important to note, however, that the speed and stability of the voltage loop is heavily dependent upon several factors apart from the design of the ITH compensation. Common PCB layout errors, for example, often appear as stability problems. Examples include the distant placement of the input decoupling capacitor, connecting the ITH compensation to a ground track carrying significant switch current, and routing the FB signal over a long distance such that noise pick occurs. Refer to the PCB Checklist section for additional information. Another factor that affects the voltage loop is the choice of output capacitor. If the value is too low, or the ESR is too high, then it will not be possible to achieve optimum loop performance. A third factor that can impair loop response is the presence of underdamped resonances in the power stage. Examples include an underdamped LC input filter or an active clamp capacitor resonating with the main transformer magnetizing inductance. Refer to the Input Capacitor/Filter Selection and Active Clamp Capacitor sections for details on how to properly damp these LC resonances. Before attempting to optimize the loop response, carefully consider the above factors, 3766fa For more information www.linear.com/LTC3766 41 LTC3766 Applications Information because no amount of tweaking to the ITH components can cancel their effect. Also, any theoretical analysis of loop response only considers first order non-ideal component behavior. Consequently, it is important that a final stability check be made with production layout and components. Stabilizing the voltage loop of the LTC3766 is accomplished by using the error amp to provide a gain from VOUT to ITH that compensates for the control to output gain from ITH to VOUT. The DC component of the ITH to VOUT gain is approximately: ADC1 = 2LfSWROUT 1 • 29.3RSENSE 2LfSW +ROUT 2VOUTRSENSE 3V R <L < OUT SENSE 3SR@K=2 SR@K=1 for resistor sense mode, and: 2LfSWROUT NP ADC1 = • 2.2K CTNSRSENSE 2LfSW +ROUT for current transformer mode. Since the LTC3766 utilizes current mode control, the ITH to VOUT transfer function can be basically characterized by one pole and one zero. The pole is given approximately by: fP = 1 1 + 2πROUTC πfSWLC and the zero is given by: fZ = be on the order of the down slope of the inductor, which provides adequate current-loop stability without introducing excessive phase shift at the crossover frequency. For phase margin calculations, assume that two poles exist at one-half of the switching frequency. Use of an abnormally high valued inductor will produce additional phase shift due to slope compensation, thereby forcing a lower voltage loop crossover frequency to ensure stability. In order to avoid having either too little or too much slope compensation, make sure that the inductor satisfies the following inequalities: 1 2πRESRC where RESR is the ESR of the output capacitance C. Note that the frequency of this zero will vary substantially depending on the type of capacitor chosen. The LTC3766 uses internal slope compensation to stabilize the current loop. The amount of slope that is effectively seen at the current sense (IS+) input is: for resistor sense mode and: 2VOUTRSENSEK CTNS 3V R K N <L < OUT SENSE CT S 3SR@K=2NP SR@K=1NP for current transformer mode. In some cases, the LTC3766 and LTC3765 will be in delay phase-out mode at low input voltages. This cycle-by-cycle reduction of the PG and FG turn-on delays has the effect of reducing the amount of slope compensation by approximately 20% to 40%. Consequently, a higher value of inductance may be required to maintain current-loop stability during operation in delay phase-out mode. The compensation network is typically configured as shown in Figure 29. The objective of this network is to add DC gain for excellent load regulation while providing good phase margin in the voltage loop at the highest possible crossover frequency. Normally this is achieved by adding a dominant pole at very low frequency and a zero well before the crossover frequency to remove most of the phase VOUT LTC3766 SR = KfSW(26mV) for RSENSE mode and: 0.6V SR = KfSW(0.35V) for current transformer mode, where K = 1 for duty cycles less than 50% and K = 2 for duty cycles greater than 50%. For most applications, this internal slope compensation will 42 – EA + gm = 2.7mS C3 (OPT) R3 FB R2 ITH R1 C2 C1 3766 F29 Figure 29. ITH Compensation Network For more information www.linear.com/LTC3766 3766fa LTC3766 Applications Information associated with the dominant pole. A high frequency pole is also added to reduce noise and provide attenuation of the output voltage ripple. Note that significant gain at the switching frequency in this compensation network can cause instabilities. The network of Figure 29 has a DC gain of: ADC2 = R2 gmREA R2+R3 where REA = 5MΩ is the output resistance of the error amplifier and gm = 2.7mS is the transconductance. The low frequency pole and zero are given by: fP1 = 1 1 and fZ1 = 2πREAC3 2πR1C1 and the high frequency pole is given by: fP2 = 1 2πR1C2 A good target for the 0dB crossover frequency of the voltage loop is between one-tenth and one-fifth of the switching frequency and a phase margin of 60° or more. Note that the zero produced by the ESR of the output capacitor helps to stabilize the loop by providing positive phase shift at frequencies near crossover. This tends to cancel the negative phase shift associated with the high frequency current loop poles. However, if the output capacitor is purely ceramic, the ESR zero may be at too high a frequency to contribute phase lead to the overall loop response. In this case, it can be helpful to add an optional phase lead capacitor C3 as shown in Figure 29, which generates a zero at a frequency of: fZ2 = 1 2πR3C3 This zero should be placed near the crossover frequency to provide additional phase boost. When optimizing the voltage loop, bear in mind that the large signal step response may be limited by factors other than the crossover frequency. At low input voltage, for example, the maximum duty cycle limit of 79% will impair the ability of the loop to respond to a sudden increase in load. Also, in responding to a very large load step (e.g., zero to full load) the loop may demand duty cycles that cause the main transformer to saturate. Hard saturation is prevented if current in sensed on the primary side or if the volt-second clamp is used, but the large signal step response will be limited by the available excess voltseconds in the main transformer. Setting the SG Reverse Overcurrent The LTC3766 has been carefully designed to turn off the SG MOSFET as needed to prevent an overcurrent during start-up, shutdown and normal operation. Nevertheless, the LTC3766 also contains a user-adjustable SG reverseovercurrent protection circuit as an added protection feature. This feature is also useful in special applications where it may be advantageous to limit the SG reverse current to a particular value. SG reverse overcurrent is implemented by monitoring the voltage on the SW pin when SG is high, and terminating the SG on-time for the duration of the switching cycle if the SW voltage exceeds an internal threshold. If the LTC3766 is operating at zero duty cycle when the SG overcurrent occurs, then the FG MOSFET is forced on prior to SG turn-off to re-route current to the primary and prevent avalanche from occurring. If not adjusted, the internal SG overcurrent threshold has been set high enough so that it should not interfere with the operation of normal applications. Be careful to make Kelvin connections from SW and GND to the drain and source of the SG MOSFET. In addition to a fixed internal threshold on the SW pin, a current is sourced from the SW pin so that a resistor can be added to decrease the overcurrent threshold if desired. Both the SW pin threshold and the adjust current are changed depending on whether the LTC3766 is operating in HV or LV mode, so as to account for the higher on-resistance of high voltage MOSFETs. In applications where the SW node plateau voltage is 40V or less (VIN • NS/NP ≤ 40), a single resistor can be used to set the SG overcurrent threshold (Figure 30). The resulting overcurrent VDS on the SG MOSFET is given by: VOC = VREV – IREVRSW 3766fa For more information www.linear.com/LTC3766 43 LTC3766 Applications Information The SG overcurrent trip should normally be targeted at twice the maximum VDS of the SG MOSFET during normal operation. This can be estimated using: PG 0V VSW(PK) RDS(MAX)VOUT ⎛ VOUT NP ⎞ VOC = 1– • ⎟ ⎜ V fSWL ⎝ IN(MAX) NS ⎠ SW NODE where RDS(MAX) is the maximum RDS(ON) of the SG MOSFET over temperature. This equation allows for twice the reverse SG current that would normally occur due to the inductor current ripple at no load. The % error in the SG overcurrent trip can be estimated using: 2 100 ⎛ IREVRSW ⎞ ⎛ VREV ⎞ ΔVOC = ⎟ ⎜ ⎟ +⎜ VOC ⎝ 15 ⎠ ⎝ 15 ⎠ 2 If the above error is greater than 30%, then the VOC threshold may need to be increased accordingly. To ensure that the inductor doesn’t saturate prior to the SG overcurrent trip, the inductor should have a saturation current such that: ILSAT > NS V NP IN 0V 3766 F31 Figure 31. Typical SW Node Waveform The overshoot and ringing on the SW node is due to the leakage inductance of main transformer, and it is worse at full load and maximum VIN. The peak SW node voltage (VSW(PK)) also depends heavily on the gate drive timing as well as the RC snubber that is typically used on the SW node. See Delay Resistor Selection: PG Turn-On Transition and RC Snubber sections for details. Make sure that the peak SW node voltage does not cause more than 0.2A to flow into the SW pin: VOC(MAX) VSW(PK) – 50V RDS(MIN) RSW < 0.2A where VOC(MAX) is the maximum overcurrent trip based on the above error and RDS(MIN) is the minimum RDS(ON) of the SG MOSFET over temperature. The above condition is normally satisfied with reasonable values for RSW and the use of an RC snubber to limit VSW(PK). While the circuit of Figure 30 can be used whenever the SW node plateau voltage is 40V or less, care must be taken to limit the current into the 50V clamp on the SW pin due to overshoot and ringing. Figure 31 illustrates a typical SW node waveform. In applications where the SW node plateau voltage is greater than 40V, it is necessary to add a divider as shown in Figure 32. VSW • VOUT • RSW MAIN XFMR + – VOC SG MOSFET 50V SG GND • IREV = LV: 103µA HV: 42µA SW + + C – VOUT • R1 SGOC MAIN XFMR + VDS(OC) – VREV = LV: 73mV HV: 148mV 3766 F30 Figure 30. SG Overcurrent for Low VOUT Applications 44 LTC3766 VSW LTC3766 R3 SW 50V R2 SG MOSFET IREV = LV: 103µA HV: 42µA + SG GND + C – SGOC VREV = LV: 73mV HV: 148mV 3766 F32 Figure 32. SG Overcurrent for High VOUT Applications 3766fa For more information www.linear.com/LTC3766 LTC3766 Applications Information For the circuit of Figure 32, the overcurrent VDS on the SG MOSFET is given by: ⎡ ⎛ R1+R2 ⎞ ⎛ R1+R2 ⎞ ⎤ VOC = VREV ⎜ –IREV ⎢R1+R3 ⎜ ⎟ ⎝ R2 ⎠ ⎝ R2 ⎟⎠ ⎥⎦ ⎣ In addition to producing the desired VOC threshold, there are three constraints on the selection of resistors R1, R2 and R3 that must be simultaneously met: 1) R1 and R2 must divide the maximum VSW plateau voltage down to 40V or less, 2) the impedance at the SW pin must be kept as low as possible to reduce the delay in sensing the VSW voltage, and 3) the power dissipation in R1 and R2 must be kept reasonably low. The last two constraints can be met by choosing a maximum power (PR) to be dissipated in the sum of R1 and R2. Typically, setting PR = 0.25W is a reasonable compromise that keeps the time constant low while not greatly impacting converter efficiency. The selection of R1, R2 and R3 is made using the following procedure: 1. Calculate R1 and R2 based on a maximum power (PR = 0.25W) and a divider ratio that will produce exactly 40V maximum on the SW pin: N ⎛ VOUT VIN(MAX) ⎞ 40VOUT R1= S ⎜ ⎟⎠ – P NP ⎝ PR R R2 = 40 • R1 NS V – 40 NP IN 2. If the value for VOC calculated using R1 and R2 from step 1) is greater than the target VOC value, then choose R3 such that IREV • R3(R1+ R2)/R2 equals the difference between the calculated and target VOC values. 3. If the value for VOC calculated using R1 and R2 from step 1) is less than the target value, then R3 = 0. Recalculate R1 and R2 based on maximum power (PR = 0.25W) and the desired target VOC value: R1= BIREV – AVOC + For the circuit of Figure 32, the % error in the SG overcurrent trip can be estimated using: 2 2 100 ⎛ IREV (R1+K •R3) ⎞ ⎡ VREV ⎤ ΔVOC = + K ⎟⎠ ⎢ 14 ⎥ VOC ⎜⎝ 6 ⎣ ⎦ where K = (R1 + R2)/R2. RC Snubbers Most applications will make use of an RC snubber to reduce the overshoot and ringing on the SW and SWB pins, as shown in Figure 33. The snubber capacitor is chosen to limit the peak voltage overshoot on SW or SWB by absorbing the energy in the leakage inductance of the main transformer. The snubber resistor is then chosen to provide optimum damping so as to minimize ringing. A larger snubber capacitor reduces the overshoot, but at the expense of increased power dissipation in the snubber resistor. In general, the snubber on the SWB node has far less energy to absorb and can therefore be smaller than that on the SW node. In some cases, the snubber on SWB can be eliminated entirely. The precise values needed for the RC snubbers will depend upon the specifics of each application, and should be optimized in the lab. Typical values for CS1 and CS2 range from 1nF to 4.7nF, and RS1 and RS2 are typically 1Ω to 50Ω. Always use a high quality ceramic (X7R) capacitor and resistors with a high power rating (1/4W to 1/2W) for and an RC snubber. VSW NP • • VOUT CS1 NS CS2 RS2 RS1 MAIN VSWB XFMR 3766 F33 Figure 33. Using RC Snubbers ( AVOC +BIREV )2 – 4ABVREVIREV 2AIREV B – AR1 R2 = A where A = PR(NP/NS) and B = VOUTVIN(MAX). 3766fa For more information www.linear.com/LTC3766 45 LTC3766 Applications Information Remote Sensing the input stage of the differential amplifier. If the input stage is saturated, the LTC3766 forces the VSOUT pin to 0V. In applications where the differential amplifier is not needed, connect the inputs as shown in either Figure 14b or Figure 15. The LTC3766 contains a precision differential amplifier for use in remote sensing applications. As shown in Figure 14a, this is useful in eliminating the voltage drops associated with bussing the power supply output voltage to a remote load. Be aware that the differential amplifier is powered from the VIN pin of the LTC3766, and requires 1.5V of overhead on VIN above the output voltage (VSOUT). If the voltage on the VIN pin is not adequate to support the VSOUT voltage, the LTC3766 will generate a fault. This is necessary to avoid a potential overvoltage on the main output of the converter. In addition, the LTC3766 will generate a fault if the polarity of the VS+ and VS– pins are reversed by approximately 0.3V or more. Self-Starting PolyPhase Applications Figure 34 shows the PolyPhase connections for the LTC3765 and LTC3766. On the primary side, the design of one phase of the LTC3765 can be optimized and then replicated up to four times by simply tying the SSFLT pins together. The common SSFLT pins are held low until all phases have adequate voltage on their VCC supplies and RUN pins. This prevents any of the phases from switching until every phase has satisfied the requirements for startup. When start-up conditions have been met, the SSFLT pin is released and quickly charged until all phases have In rare applications, it may be useful to raise the common mode voltage of the VS+ and VS– inputs. When doing so, always ensure that VS+ < 2(VIN – 2V) to prevent saturating VIN+ R1 NDRV IN+ LTC3765 R2 RUN VCC CVCC VIN NDRV VCC • PT+ FB LTC3766 (MASTER) PT– SS • IN– ITH SSFLT CSSP VOUT+ VBIAS PHASE 1 (MASTER) FS/UV GND RFSP FS/SYNC GND RFSS CSSS VOUT– VIN– VBIAS PHASE 2 (SLAVE) VIN NDRV VCC VIN+ R1 NDRV IN+ LTC3765 R2 RUN VCC CVCC CSSP • • FS/SYNC FB LTC3766 (SLAVE) PT+ VS– VS+ PT– IN– VSOUT ITH SSFLT FS/UV GND RFSP SS PHASE GND CSSS VOUT– VIN– 3766 F34 Figure 34. PolyPhase Connections 46 3766fa For more information www.linear.com/LTC3766 LTC3766 Applications Information switched once. The SSFLT pin currents then decrease to their nominal values. This ensures that all phases begin their asynchronous, open-loop start-up at nearly the same time. On the secondary side, the SS pins from all phases are interconnected as well. This prevents any one phase from starting until all phases have adequate bias voltage and have detected switching on their respective SW pins. Once this condition is met, the master will advance the soft-start voltage to match the VOUT of the converter, and switching begins on the secondary side on all phases. After a brief lock sequence, all phases will transfer control to the secondary. The ITH pins are interconnected between the phases so that current is shared evenly between the master (which controls the ITH pin to regulate VOUT) and the slaves. The LTC3765 SSFLT connection is also used to communicate faults. If one phase has a primary-side fault (undervoltage, overcurrent, overtemperature, or communication loss), it immediately stops switching and rapidly pulls SSFLT to 6V. The other phases will detect that SSFLT is above 5V and will also stop switching. On the secondary side, the LTC3766s detect that switching has stopped and also fault, which is communicated to all phases through the common SS connection. The voltage on the primary-side SSFLT node then slowly decreases and a restart begins. Likewise, if a fault originates on the secondary side on a give phase, this fault is communicated to the other LTC3766s so that all phases stop switching. This will cause a communication fault on the primary side followed by a restart attempt. regulator pass device will be dissipating more power and may require a larger and more thermally conductive package. The design and PCB layout are generally simplified if each phase uses its own linear regulator. The secondary side follows a similar procedure; however, there is more differentiation between the master phase and the slave phases. For the master, choose components based on the above design equations. Be aware that each phase should have its own linear regulator pass device to distribute the power dissipation. Duplicate the components for each slave, with the following exceptions: 1. Connect all of the SS pins together. Instead of having multiple capacitors from the SS node to ground, the capacitors can be consolidated into one capacitor. Note that only the master charges and discharges the softstart capacitor. 2. Connect the FB pin of the slaves to VCC. This connection puts the LTC3766 into slave mode. In this mode, the ITH pin becomes a high impedance input and the SS pin is only used for fault communication. An LTC3766 slave will not perform a pre-set of the soft-start capacitor, nor will it charge or discharge it. A slave can only force the SS pin high to indicate a fault, and it also monitors the SS pin to respond to a fault in another phase. 3. For each slave, the integrated unity-gain differential amplifier is used to sense the voltage on the ITH pin of the master. Connect the VS+/VS– inputs of each slave between the ITH and signal GND pins of the master. Connect the VSOUT pin on each slave to its own ITH pin. For the LTC3765 on the primary side, choose components based on a single-phase design. Duplicate the single phase to the desired number of phases, up to the maximum of four, with the following modifications: 4. Connect the FS/SYNC pins of each slave to the PT – pin of the master. The PT – pin of the master contains the clock signal used to synchronize the slaves and master together. 1. Connect the SSFLT pins together. Instead of having multiple capacitors from the SSFLT node to ground, the capacitors can be consolidated into one capacitor with a value equal to N • CSSFLT, where N is the number of phases. For each slave, set the relative phase using the PHASE pin. Note that ripple current in the input capacitor is minimized by operating the controllers out of phase. For a 2-phase system, set the slave at 180°. For a 3-phase system, set the slaves at 120° and 240°. For a 4-phase system, set the slaves at 90°, 180°, and 270°. Refer to Setting the Switching Frequency and Synchronization for details on setting the PHASE pin. 2. If desired, the phases can share the linear regulator of one phase by shorting their VCC and NDRV pins to the linear regulator output; however, be aware that the linear 3766fa For more information www.linear.com/LTC3766 47 LTC3766 Applications Information Volt-Second Clamp When used in applications with the LTC3765, direct flux limit will guarantee that no saturation occurs on the main transformer. Consequently, there is no need to use a voltsecond clamp in applications that have the direct flux limit feature. In applications where the LTC3766 is used standalone, however, the volt-second clamp can be used as a failsafe to prevent excessive volt-seconds from being applied to the main transformer during the PWM on-time. Figure 35 illustrates the use of the volt-second clamp. As shown in Figure 35, the SW voltage is used to monitor the voltage applied to the main transformer. During the PWM on-time, the CVS capacitor is charged by the SW node through the RVS resistor. For capacitor CVS, use a 5% or better NPO-type ceramic capacitor, since accuracy is important. Typically a value of 1nF is suitable. Likewise, use a 1% resistor for RVS. In high output voltage applications where the SW node must be divided down, use the circuit of Figure 36 to set the volt-second clamp. SW NP • • R1 NS LTC3766 SW MAIN XFMR RVS R2 VSEC CVS 3766 F36 SW NP • • Figure 36. Volt-Second Clamp in High VOUT Application NS LTC3766 MAIN XFMR SW RVS VSEC CVS 3766 F35 Figure 35. Using the Volt-Second Clamp The PWM on-time is terminated when a pre-determined threshold is reached. This will limit the applied volt-second product to: (V • S)LIM = 0.605RVSCVS The above equation is accurate even when the peak voltage on the SW node is relatively low and the charging is nonlinear, such as in low VOUT applications. This is possible because the LTC3766 senses the voltage on the SW pin and adjusts the internal volt-second comparator reference so that constant volt-seconds are maintained regardless of the voltage on SW. Consequently, it is important that the LTC3766 SW pin be connected to the secondary SW node for proper sensing of this voltage to occur. The volt-second limit should normally be set approximately 10% above the operational volt-second requirement. To accomplish this, calculate RVS using: R VS = 1.10 48 VOUT 0.605fSWC VS In this case, assuming RVS >> R1||R2, RVS can be calculated using: VOUT ⎛ R2 ⎞ ⎜ ⎟ 0.605fSWC VS ⎝ R1+R2 ⎠ R VS = 1.10 Nonisolated Applications In addition to being used with the LTC3765 in isolated applications, the LTC3766 can also be used standalone to make a nonisolated resonant-reset forward converter as shown in Figure 37. In this application, the primary-side MOSFET is driven directly by the PT+ pin, and the MODE VIN MAIN XFMR SW • SWB FG SWP PG CRST VOUT • LTC3766 PT+ SG VIN NDRV REGSD MODE CRSD VCC VAUX RVIN (FOR VIN > 30V) QP 5V TO 15V LV BIAS SUPPLY CVAUX CVIN CVCC 3766 F37 Figure 37. Nonisolated Resonant-Reset Application 3766fa For more information www.linear.com/LTC3766 LTC3766 Applications Information pin is tied to GND through either a 100k or 50k resistor to select LV or HV operating mode. The bias for the VIN pin is normally taken directly from the input voltage of the converter. The LTC3766 contains a current-limited internal 30V shunt to simplify applications where VIN > 30V. In such applications, place a current limiting resistor in series with the VIN pin calculated using: R VIN = VIN(MAX) – 30V 3.5mA Note that at low VIN, there will be a maximum drop across RVIN equal to (1.2mA) • (RVIN) that is due to the VIN pin operating current. For proper operation, the voltage on the VIN pin at low input voltage must be greater than the rising VCC UVLO by at least the threshold voltage of QP. Using a MOSFET for QP instead of an NPN eliminates the base current that would otherwise add to the VIN operating current. If more margin is needed at low VIN operation, a Darlington transistor is another option for QP. To reduce power dissipation in QP, a low voltage bias supply should be fed into the VAUX pin to power the bypass LDO. This bias supply can be generated off of either the primary or secondary of the main transformer using an auxiliary buck or an inductor overwinding supply. During an output overload condition, the low voltage bias supply will collapse, causing the high voltage linear regulator controller to be re-energized. To prevent excessive power dissipation under this condition, place a capacitor on the REGSD pin to limit the operating time of the high voltage linear regulator. The RUN pin can be used as an undervoltage lockout (UVLO) on the converter input voltage. Direct RUN/STOP control can be achieved by using a small NMOS on the RUN pin as shown in Figure 38. VIN R1 RUN R2 LTC3766 RUN/STOP CONTROL GND 3766 F38 The resonant reset capacitor, CRST, serves to generate a voltage on the SWP node during the off-time of the primary MOSFET that resets the transformer flux on a cycle-by-cycle basis. This capacitor is normally sized so that the SWP voltage exactly resonates back to VIN at the end of the off time with minimum VIN: 2 V 1 ⎡ 1 ⎛ N ⎞⎤ CRST ≈ 1– OUT P ⎟ ⎥ – CPAR ⎢ ⎜ LM ⎢⎣ πfSW ⎝ VIN(MIN) NS ⎠ ⎥⎦ where LM is the main transformer magnetizing inductance and CPAR is the total parasitic capacitance on SWP: ⎛N ⎞ CPAR = COSS(PG) + ⎜ S ⎟ ⎝ NP ⎠ 2 (COSS(FG) +CSNUB ) CPAR includes the drain capacitance of both the PG and FG MOSFETs as well as any snubber capacitance on the SWB node. In reality, the presence of leakage inductance makes the SWP node rise much faster than it otherwise would. As a result, the optimum value for CRST can be 40% to 60% higher than that calculated by the above equation. The steady-state peak voltage on the primary and forward MOSFETs is given by: VDS(PG) = VIN(MAX) + VDS(FG) = VOUT 2fSW VOUT NP 2fSW NS 1 LM (CRST +CPAR ) 1 LM (CRST +CPAR ) If a larger value of CRST is used, the peak voltage stresses can be decreased, possibly allowing the use of a MOSFET with lower BVDSS rating. However, with a larger CRST the SWP voltage at low VIN will not have time to resonate back down to VIN, thereby increasing the turn-on switching losses. In practice, some truncation of the low VIN reset waveform is often tolerated to maximize the overall efficiency of the converter. Note also that the peak MOSFET voltage stress during transients can be considerably higher, so allow at least 30% margin above these calculated voltages. The volt-second clamp can be used to reduce the peak voltage stress due to load transients. Figure 38. RUN/STOP Control for Standalone Applications 3766fa For more information www.linear.com/LTC3766 49 LTC3766 Applications Information The setting of the gate drive timing for a resonant reset converter is simplified by the adaptive delays featured in the LTC3766. When standalone mode is active (100k or 50k on MODE), the FGD pin is ignored, and the associated dead time between SG turn-off and PG turn-on is controlled adaptively. In this mode, LTC3766 delays the PG turn-on until after the SG pin has fallen below approximately 0.5V. For the PG turn-off transition, the SGD resistor is chosen to minimize the dead time and also prevent collapse of the SWP node (i.e., catch the SWP voltage at its peak if possible). Note that setting the FG turn-off so as to catch the SWP voltage near its peak will improve efficiency and allow for the use of a larger resonant reset capacitor, thereby reducing the peak voltage stresses on the MOSFETs. Adaptive delay limiting on this edge ensures that SG will not go high until SW has fallen, so shoot-through is not a concern. In nonisolated applications, the inductor current is normally sensed on the input side of the power transformer, typically using a sense resistor. Note that in this situation, the values for the sense resistor (RSENSE) and the IPK resistor (RIPK) should be calculated using the above equations, but then scaled by a factor of NP/NS. For applications where the transformer is configured to step up the voltage, however, it may be more efficient to sense current on the output side of the power transformer. In this case, be careful to avoid transformer saturation by keeping the resonant reset capacitor as small as possible and making use of the volt-second clamp. Common Mode Noise Common mode noise arises in isolated converter applications due to the parasitic capacitance between the primary and secondary windings of the main transformer. When rapid voltage changes occur on the primary-side MOSFET drain, this will inject current through the inter-winding capacitance. This causes the ground reference of the secondary to suddenly jump with respect to the primary ground. As a result, current is injected across the inter- 50 winding capacitance of the pulse transformer back to the primary, and a resulting common mode voltage can appear at the IN+ and IN– inputs of the LTC3765. While the LTC3765 has been carefully designed to reject this common mode voltage, always use a common mode filter capacitor that is directly connected between the primary and secondary grounds. This capacitor shunts away the common mode noise. Typically, a value of 2.2nF is adequate. Use a high quality ceramic Y capacitor rated for 250VAC operation, or other voltage rating as needed for the isolation and safety requirements of the particular application. Thermal Considerations When designing a forward converter with an output power of 50W or more, particular attention must be paid to the thermal aspects of the design and layout. In general, it is better to use multiple elements in parallel to spread out the power dissipation and reduce temperature rise. Beneath all power MOSFETS, use thermal vias and copper islands on multiple layers to provide cooling. If excessive temperature rise occurs, both the LTC3765 and the LTC3766 contain overtemperature shutdown circuits that will help to prevent thermal damage. Both overtemperature shutdowns are set at approximately 165°C rising with 20°C of hysteresis. PCB Checklist The LTC3766 requires proper bypassing on the VCC supply due to its high speed switching (nanoseconds) and large AC currents (Amperes). Careless component placement and PCB trace routing may cause excessive ringing and undershoot/overshoot. To obtain optimum performance from the LTC3766: 1. Use a low inductance, low impedance ground planes to reduce any ground drop and stray capacitance. Remember that the LTC3766 switches at greater than 2A peak currents and the power MOSFETs can carry 50A or more. Any significant ground drop will degrade signal integrity. 3766fa For more information www.linear.com/LTC3766 LTC3766 Applications Information 2. Plan the power/ground routing carefully. Know where the large load switching current is coming from and going to. Maintain three separate planes if possible: signal ground (GND pin), power ground (PGND pin) and power stage ground. The power ground plane should be connected with a single via to the source of the SG MOSFET. The signal ground plane should be connected with a single via to the source of the SG MOSFET for accurate VDS sensing. If resistor current sensing is used for IS+ and IS–, be careful to minimize the inductance of the plane between the sense resistor and the source of the SG MOSFET. 3. Mount a bypass capacitor as close as possible between the VCC pin and the power ground plane. 6. If resistor sense mode is used, the IS+ and IS– pins must be Kelvin connected to the sense resistor. The traces to the sense resistor must run side-by-side and be shielded with signal ground on all sides. 7. Keep the switching nodes (SW, PT+, PT –, FG, SG) away from noise sensitive nodes, especially FB, ITH, IS+ and IS–. 8. The voltage divider on the output should be connected as close as possible to the load at the output terminal of the power supply. The bottom of the voltage divider should be tied to the signal ground plane. Use the differential amplifier to sense the load voltage and eliminate distribution voltage drops. 4. Keep the copper traces between the driver output pins and the MOSFET short and wide. 5. Keep the high current switching path on both the primary and secondary as short as possible, using multiple layers in parallel to further reduce parasitic inductance. 3766fa For more information www.linear.com/LTC3766 51 LTC3766 Typical Applications +VIN 36V TO 72V • 2.2µF 100V ×3 FDMS86201 33nF 200V T1 6:2 • +VOUT 5V 15A 1nF 100V BSC0901NS 100nF 200V 15mΩ 168Ω 1/2W –VIN L1 1.4µH 1µF 10V 220µF 6.3V ×2 L1: PULSE PA1392.152 T1: PULSE PA0810 T2: PULSE PA0297 SiR414DP 10Ω 1/4W 3mΩ 2W –VOUT Si3440DV Si3437DV (SOT23) 200k 100Ω 2.2nF 250V 330pF 1Ω 365k IS+ IS– NDRV PG 0.1µF ISMAG IN+ RUN 15.0k AG VCC 220pF LTC3765 4.7µF SSFLT RCORE 33nF DELAY 10.5k 100Ω • T2 2:1 1.0µF • 100Ω IN– FS/UV SGND PGND 18.2k 14k FG SW SG RUN VIN NDRV VCC FS/SYNC VS+ VS– 4.42k IS– + IS FB PT+ PT– SS ITH LTC3766 GND PGND IPK 33nF NPO SGD 26.1k FGD 15k 604Ω 470pF MODE VSEC 22.1k 47pF 17.8k 3766 TA02 Figure 39. 36V-72V to 5V/15A Active Clamp Isolated Forward Converter Efficiency vs Load Current Pre-Biased Start-Up Load Step 96 VOUT 1V/DIV VOUT 200mV/DIV EFFICIENCY (%) 94 92 IOUT 5A/DIV IL1 5A/DIV 90 88 86 VIN = 48V 200µs/DIV VPREBIAS = 4.6V VIN = 36V VIN = 48V VIN = 72V 3 5 7 9 11 LOAD CURRENT (A) 13 3766 F39c VIN = 48V 20µs/DIV VOUT = 5V LOAD STEP = 5A TO 15A 3766 F39d 15 3766 F39b 52 3766fa For more information www.linear.com/LTC3766 LTC3766 Typical Applications 9V-36V to 24V/4.2A Active Clamp Isolated Forward Converter L1 0.47µH +VIN 9V TO 36V • T3 10µF 50V ×3 10µF 50V • D3 IS+ 2T 1.2k 1/8W • T1 L2 58µH • • 24Ω 1W 8T 4T 150pF 250V D4 Q1 –VIN 4mΩ 1W ES1PD 0.33Ω 1/8W 100Ω 100Ω 2.2nF D1 100k 2N7002 1µF 1nF 6.19k 14.3k +VIN 147k IS– 15Ω 1/8W 1µF 100V L3 680µH 2.2µF 3.65Ω IS+ 1nF AG VCC 0.1µF IN+ LTC3765 DELAY IN– RCORE FS/UV 10k 0.1µF 1µF T2 2.5:2 • • VAUX NDRV +VOUT FCX491A 4.7µF IS+ VCC SS IS– REGSD MODE 274Ω FB 27.4k 10.7k PT– 16.9k 2.2nF 250VAC 15k SSFLT 60.4k 81.6k 1nF 100µF 35V: SUNCON 35HVH100M D1-D2: ZHCS506 D3-D6: BAS21 L1: VISHAY IHLP2525CZERR47M01 L2: PULSE PA2729.583 L3: COOPER SD25-681 SG VIN PT+ SGND PGND 28.7k SW FG LTC3766 33nF NPO D2 220pF Q3 100Ω 33nF RUN IS+ 0.1µF NDRV 2.8k 1/8W 3.3nF ISMAG PG 10µF 50V 3.3nF Q1: BSC057N08NS3 Q2: Si7430DP Q3: BSC320N20NS3G T1: PULSE PA0806.004NL T2: PULSE PA0510NL T3: ICE CT102-100 (1:100) +VOUT 24V 100µF 4.2A 35V –VOUT Si7309DN 100Ω +VIN 7.5k 1/2W Q2 D5 220nF 100V D6 + 15k FGD RUN SGD VS+ FS/SYNC VS– IPK ITH 68pF GND 7.5k +VOUT PHASE VSEC PGND 2.94k 3766 TA03a 3766fa For more information www.linear.com/LTC3766 53 LTC3766 Typical Applications 18V-75V to 12V/12.5A Active Clamp Isolated Forward Converter L1 1.8µH +VIN 18V TO 75V 2.2µF 100V • T3 2.2µF 100V ×3 • D3 IS+ 4T 1.2k 1/8W • T1 L2 11µH • 51Ω 1/2W 4T 1.82k 1/4W ES1PD Q2 ES1PD 0.75Ω 1/8W –VIN 100Ω 1nF D1 200k FDC2512 1µF 1nF 6.19k 13.3k +VIN 61.9k IS– SW FG 5.11Ω 1/8W 100Ω +VIN IS+ 1nF 10k LTC3766 33nF NPO D2 NDRV AG VCC 0.1µF IN+ 0.1µF LTC3767 DELAY VCC SS IS– MODE PT+ • 16.9k 2.2nF 250VAC 33nF 15k SSFLT 60.4k 604Ω FB 11.5k 10nF Q2: BSC057N08NS3 Q3: BSC190N15NS3 T1: PULSE PA0801 T2: PULSE PA0510NL T3: ICE CT102-100 (1:100) FGD RUN SGD VS+ FS/SYNC VS– REGSD IPK PHASE ITH VSEC 75k 10nF 68µF 16V: SANYO 16TQC68M D1-D2: ZHCS506 D3: BAS21 L1: VISHAY IHLP4040DZER1R8M11 L2: PULSE PA2729.113NL Q1: FDMS86201 FMMT491 PT– 27.4k 68µF 16V ×2 +VOUT 100Ω IN– RCORE FS/UV RUN • + 10µF IS+ 1µF T2 2.5:2 220pF SG VIN NDRV 3.3nF SGND PGND 4.99k IS+ ISMAG PG 22µF 16V ×2 –VOUT IRF6217 100Ω 1nF 200V Q3 75Ω 1/8W 0.22µF 250V 33nF 200V 4mΩ 1W 33k 1/2W 1.82k 1/4W 100pF 200V Q1 +VOUT 12V 12.5A 10k GND 47pF PGND 1.82k +VOUT 1.87k 3766 TA04a Efficiency vs Load Current 96 VIN = 24V EFFICIENCY (%) 94 VIN = 48V 92 90 88 86 54 0 3 9 6 LOAD CURRENT (A) 12 15 3766 TA04b 3766fa For more information www.linear.com/LTC3766 LTC3766 Typical Applications 36V-60V to 32V at 10A 320W Isolated P/A Power Supply L1 1.8µH +VIN 36V TO 60V • T3 2.2µF 100V ×3 2.2µF 100V • D6 IS+ 4T 330Ω 1/8W • T1 L2 10µH • • 2T D4 –VIN 33nF 200V 4mΩ 1W ES1PD 0.56Ω 1/8W 100Ω 100Ω 1.5nF D1 200k FDC2512 1µF 1nF 13k 8.25k +VIN 66.5k L3 680µH 4.7µF 10k 1nF AG VCC 0.1µF IN+ LTC3765 DELAY IN– RCORE FS/UV 33nF NPO D2 0.1µF • 220pF VCC SS IS– • 604Ω FB 330pF 1k PT– 27.4k FCX491A 1µF IS+ PT+ 27.4k 2.2nF 250VAC 15.0k SSFLT 60.4k 2.2nF FGD RUN SGD VS+ FS/SYNC VS– 33.2k 10nF IPK 47k ITH GND 100pF Q1-Q6: BSC190N15NS3 T1: PULSE PA0905NL T2: PULSE PA0510NL T3: ICE CT102-100 (1:100) 31.6k 205k 1.54M 8.66k +VOUT PHASE VSEC PGND 3.09k 3766 TA05a Efficiency vs Load Current 96 VIN = 48V 95 EFFICIENCY (%) 56µF 50V: SUNCON 50HV56M D1-D2: ZHCS506 D3-D6: BAS21 L1: VISHAY IHLP6060DZER1R8M11 L2: COILCRAFT SER2814H-103 L3: COOPER SD25-681 NDRV REGSD MODE 1µF T2 2.5:2 SGND PGND 2.43k VAUX 100Ω 33nF RUN SG VIN 0.1µF 50V 3.9V +VOUT LTC3766 0.1µF NDRV SW FG 3.3nF ISMAG PG Q5 Q6 I + 4.22Ω S IS+ +VOUT 32V 56µF 10A 50V ×2 –VOUT IRF6217 100Ω +VIN IS– 165Ω 1/8W 0.22µF 250V + 2.94k Q3 Q4 D5 Q1 Q2 8.66k 1/8W D3 6T 4.7µF 50V ×4 94 93 92 91 2 4 6 LOAD CURRENT (A) 8 10 3766 TA05b 3766fa For more information www.linear.com/LTC3766 55 LTC3766 Typical Applications 36V-60V to 14V at 25A 350W Isolated Bus Converter L1 1.8µH +VIN 36V TO 60V • T3 2.2µF 100V ×3 2.2µF 100V • 4mΩ 1W ES1PD 0.68Ω 1/8W 100Ω 100Ω 1nF 1µF 1nF 14k 11k +VIN 66.5k 1.00k 1/8W 3T IS– 165Ω 1/8W 0.22µF 250V 33nF 200V SW FG 1nF LTC3766 3.3nF 33nF NPO D2 NDRV AG VCC 0.1µF IN+ LTC3767 DELAY IN– RCORE FS/UV 0.1µF 27.4k 15.0k 60.4k 150k 10nF 4.7nF Q1-Q3: BSC190N15NS3 Q4-Q7: BSC057N08NS3 T1: PULSE PA0956NL T2: PULSE PA0510NL T3: ICE CT102-100 (1:100) +VOUT FCX491A IS– MODE FB 604Ω 13.7k SSFLT 68µF 16V: SANYO 16TQC68M D1-D2: ZHCS506 D5: BAS21 L1: VISHAY IHLP4040DZER1R8M11 L2: COILCRAFT SER2814L-472KL SS PT+ • 2.2nF 250VAC 33nF RUN VCC 100Ω 27.4k 10nF 200V 4.7µF IS+ 1µF T2 2.5:2 • 220pF SG VIN NDRV ISMAG PG +VOUT 14V 68µF 25A 16V ×4 D5 4.22Ω IS+ 22µF + 16V ×4 –VOUT SGND PGND 2.43k Q6 Q7 1/8W 10k 5.1k 1W ES1PD Q4 Q5 + 432Ω IS IRF6217 100Ω +VIN FDC2512 • IS –VIN 200k L2 4.7µH + Q1 Q2 Q3 D1 5T • T1 10k PT– RUN FGD VS+ SGD VS– +VOUT FS/SYNC REGSD IPK PHASE ITH VSEC GND 100pF PGND 3766 TA06a Efficiency vs Load Current 96 VIN = 48V EFFICIENCY (%) 95 94 93 92 91 0 5 10 15 20 LOAD CURRENT (A) 25 30 3766 TA06b 56 3766fa For more information www.linear.com/LTC3766 LTC3766 Package Description GN Package 28-Lead Plastic SSOP (Narrow .150 Inch) (Reference LTC DWG # 05-08-1641 Rev B) .386 – .393* (9.804 – 9.982) .045 ±.005 28 27 26 25 24 23 22 21 20 19 18 17 1615 .254 MIN .033 (0.838) REF .150 – .165 .229 – .244 (5.817 – 6.198) .150 – .157** (3.810 – 3.988) .0250 BSC .0165 ±.0015 1 RECOMMENDED SOLDER PAD LAYOUT .015 ±.004 × 45° (0.38 ±0.10) .0075 – .0098 (0.19 – 0.25) 2 3 4 5 6 7 8 .0532 – .0688 (1.35 – 1.75) 9 10 11 12 13 14 .004 – .0098 (0.102 – 0.249) 0° – 8° TYP .016 – .050 (0.406 – 1.270) .008 – .012 (0.203 – 0.305) TYP NOTE: 1. CONTROLLING DIMENSION: INCHES INCHES 2. DIMENSIONS ARE IN (MILLIMETERS) .0250 (0.635) BSC GN28 REV B 0212 3. DRAWING NOT TO SCALE 4. PIN 1 CAN BE BEVEL EDGE OR A DIMPLE *DIMENSION DOES NOT INCLUDE MOLD FLASH. MOLD FLASH SHALL NOT EXCEED 0.006" (0.152mm) PER SIDE **DIMENSION DOES NOT INCLUDE INTERLEAD FLASH. INTERLEAD FLASH SHALL NOT EXCEED 0.010" (0.254mm) PER SIDE 3766fa For more information www.linear.com/LTC3766 57 LTC3766 Package Description UFD Package 28-Lead Plastic QFN (4mm × 5mm) (Reference LTC DWG # 05-08-1712 Rev B) 0.70 ±0.05 4.50 ±0.05 3.10 ±0.05 2.50 REF 2.65 ±0.05 3.65 ±0.05 PACKAGE OUTLINE 0.25 ±0.05 0.50 BSC 3.50 REF 4.10 ±0.05 5.50 ±0.05 RECOMMENDED SOLDER PAD PITCH AND DIMENSIONS APPLY SOLDER MASK TO AREAS THAT ARE NOT SOLDERED 4.00 ±0.10 (2 SIDES) 0.75 ±0.05 R = 0.05 TYP PIN 1 NOTCH R = 0.20 OR 0.35 × 45° CHAMFER 2.50 REF R = 0.115 TYP 27 28 0.40 ±0.10 PIN 1 TOP MARK (NOTE 6) 1 2 5.00 ±0.10 (2 SIDES) 3.50 REF 3.65 ±0.10 2.65 ±0.10 (UFD28) QFN 0506 REV B 0.200 REF 0.00 – 0.05 0.25 ±0.05 0.50 BSC BOTTOM VIEW—EXPOSED PAD NOTE: 1. DRAWING PROPOSED TO BE MADE A JEDEC PACKAGE OUTLINE MO-220 VARIATION (WXXX-X). 2. DRAWING NOT TO SCALE 3. ALL DIMENSIONS ARE IN MILLIMETERS 4. DIMENSIONS OF EXPOSED PAD ON BOTTOM OF PACKAGE DO NOT INCLUDE MOLD FLASH. MOLD FLASH, IF PRESENT, SHALL NOT EXCEED 0.15mm ON ANY SIDE 5. EXPOSED PAD SHALL BE SOLDER PLATED 6. SHADED AREA IS ONLY A REFERENCE FOR PIN 1 LOCATION ON THE TOP AND BOTTOM OF PACKAGE 58 3766fa For more information www.linear.com/LTC3766 LTC3766 Revision History REV A DATE DESCRIPTION 6/13 Switch polarity between IS+ and IS– in Figure 11 PAGE NUMBER 28 3766fa Information furnished by Linear Technology Corporation is believed to be accurate and reliable. However, no responsibility is assumed for its use. Linear Technology Corporation makes no representation that the interconnection of its circuits as described herein will not infringe on existing patent rights. For more information www.linear.com/LTC3766 59 LTC3766 Typical Application 36V-72V to 3.3V/20A Nonisolated Resonant-Reset Forward Converter VIN 36V TO 72V L1 0.85µH T1 L1: PULSE PA1294.910NL L2: COOPER SD25-102 T1: PULSE PA0810.006NL 2.2µF 100V ×3 12T 1.5nF 200V NPO BSC320N20NS3 BAT54 15mΩ 1W 100Ω 100Ω 6T • • 2T 1nF 50V 2.4Ω 1/4W 100µF + 6.3V ×2 • 220µF 6.3V BSC0901NS L2 1mH 1µF VOUT 3.3V 20A BSC0901NS BAT54 330pF 210k 16.5k 1/4W IS+ PT+ IS– VAUX FG VS– VIN Si2328DS (SOT23) 1nF 7.87k NDRV 0.1µF 50V 1µF 16V VSOUT LTC3766 VCC FB FS/SYNC ITH REGSD SS 33nF GND PGND VSEC SGD 0.22µF 100Ω SW SG VS+ RUN IPK MODE 16.2k 49.9k 20.5k +SENSE –SENSE 8.62k 100Ω 2.2nF 47pF 6.2k 1.82k 3766 TA07 Efficiency vs Load Current Start-Up Load Step 94 VIN = 36V EFFICIENCY (%) 92 VIN = 48V VIN = 72V 90 VOUT 200mV/DIV IL 5A/DIV IOUT 5A/DIV VIN = 48V VOUT = 3.3V RLOAD = 0.22Ω 88 86 VOUT 1V/DIV 6 8 16 10 12 14 LOAD CURRENT (A) 18 1ms/DIV 3766 TA07c 20µs/DIV VIN = 48V VOUT = 3.3V LOAD STEP = 10A TO 20A 3766 TA07d 20 3766 TA07b Related Parts PART NUMBER LTC3765 DESCRIPTION Active Clamp Forward Controller and Gate Driver LTC3705/LTC3726 2-Switch Synchronous Forward No Opto Isolated Controller Chip Set Isolated Synchronous Forward Active Clamp LT®1952/LT1952-1 Contollers LTC3723/LTC3723-2 Synchronous Push-Pull and Full-Bridge Controllers LTC3721-1/LTC3721-2 Nonsynchronous Push-Pull and Full-Bridge Controllers LTC3722/LTC3722-2 Synchronous Isolated Full-Bridge Controllers 60 Linear Technology Corporation COMMENTS Direct Flux Limit, Supports Self-Starting Secondary Forward Control, Works in Conjuction with LTC3766 Self-Starting Architecture Eliminates Need for Bias Voltage on Primary Side Suitable for Medium Power 12V, 24V and 48V Input Applications, Adjustable Synchronous Rectification Timing High Efficiency with On-Chip MOSFET Drivers, Adjustable Synchronous Rectification Timing Minimizes External Components, On-Chip MOSFET Drivers Adaptive or Manual Delay Control for Zero Voltage Switching, Adjustable Synchronous Rectification Timing 1630 McCarthy Blvd., Milpitas, CA 95035-7417 For more information www.linear.com/LTC3766 (408) 432-1900 ● FAX: (408) 434-0507 ● www.linear.com/LTC3766 3766fa LT 0613 REV A • PRINTED IN USA LINEAR TECHNOLOGY CORPORATION 2011