LINEAR TECHNOLOGY MARCH 2006 IN THIS ISSUE… COVER ARTICLE Take the Easy Road to Digitally Managed Power ...................1 Andy Gardner Issue Highlights ..................................2 Linear Technology in the News….........2 DESIGN FEATURES Cascadable, 7A Point-of-Load Monolithic Buck Converter ..................6 Peter Guan I2C Bus Buffer Resolves Stuck Buses, Eliminates Heavy Load Limitations and Provides Level Translation .........10 George Humphrey Negative Voltage Hot Swap™ Controller with 10-Bit ADC and I2C Monitoring ............................13 Zhizhong Hou High Efficiency Nickel Charger Operating to 34V is Easily Configured to Deliver a Safe Fast Charge...............................19 James A. McKenzie Tiny Controller Makes It Easy to Charge Large Capacitors ...............22 David Ng MOSFETs Make Sense for Tracking and Sequencing Power Supplies ........24 Thomas DiGiacomo High Voltage Boost/LED Controller Provides 3000:1 PWM Dimming Ratio ..........................31 Eugene Cheung 3A Monolithic Buck Regulator in 4mm × 4mm QFN ...........................34 Theo Phillips DESIGN IDEAS ....................................................37–45 (complete list on page 37) New Device Cameos ...........................46 Design Tools ......................................47 Sales Offices .....................................48 VOLUME XVI NUMBER 1 Take the Easy Road to Digitally Managed Power Introduction by Andy Gardner Digital management of high availability q A 7-channel ADC multiplexer power supplies holds great promise, with four external differential but it often comes at the cost of compliinputs, a 12V input, a 5V VDD incated multichip circuit solutions. For put, and an input for the on-chip example, an application with voltagetemperature sensor. current monitoring and supply voltage q Two continuous time, 8-bit, curmargining can require a number of rent output DACs with voltage ICs, including a low-drift reference, buffered outputs. The outputs of a multichannel, differential input the voltage buffers can be placed ADC with at least 12 bits of resoluin a low leakage, high impedance tion, an 8-bit DAC, and a dedicated state. microcontroller. Add to this the con- q A built-in, closed-loop servo siderable software algorithm that development efadjusts the The LTC2970 fort required for point-of-load simplifies the design of margining algovoltage of a digitally managed power rithms, voltage DC/DC converter and current monito the desired supplies by incorporating tor functions, and value. The range important features into one the cost, comand resolution of easy-to-use device. plexity, spacious the voltage servo board real-estate is user adjustrequirements and blossoming designable with two external resistors. debug time can deter even the most q Extensive, user configurable dedicated power supply designer from overvoltage and undervoltage trying digitally managed power. fault monitoring. The LTC2970 simplifies the design q An I2C and SMBus compliant of digitally managed power supplies 2-wire serial bus interface, two by incorporating important features GPIO pins, and an ALERT pin. into one easy-to-use device: a dual q An on-chip, 5V, linear regulator power supply monitor and controller. that allows the LTC2970 to operFigure 1 shows a block diagram of the ate from an external 8V to 15V LTC2970, highlighting the following voltage supply. features: q Another part in the family, the q A 14-bit, differential input, ΔΣ LTC2970-1, adds a tracking ADC with a maximum total unadalgorithm that allows two or more justed error (TUE) of ±0.5% over power supplies to be ramped up the industrial temperature range and down in a controlled manner. when using the on-chip reference. continued on page 3 L, LT, LTC, LTM, Burst Mode, OPTI-LOOP, Over-The-Top and PolyPhase are registered trademarks of Linear Technology Corporation. Adaptive Power, BodeCAD, C-Load, DirectSense, Easy Drive, FilterCAD, Hot Swap, LinearView, µModule, Micropower SwitcherCAD, Multimode Dimming, No Latency ΔΣ, No Latency Delta-Sigma, No RSENSE, Operational Filter, PanelProtect, PowerPath, PowerSOT, SmartStart, SoftSpan, Stage Shedding, SwitcherCAD, ThinSOT, True Color PWM, UltraFast and VLDO are trademarks of Linear Technology Corporation. Other product names may be trademarks of the companies that manufacture the products. L EDITOR’S PAGE Issue Highlights D igital management of high availability power supplies holds great promise, but it often comes at the cost of complicated multichip circuit solutions. The LTC2970 simplifies the design of digitally managed power supplies by incorporating important features into one easy-touse device. See our cover article for more about this breakthrough device. Featured Devices Below is a summary of the other devices featured in this issue. The LTC3415 provides a compact, simple, complete and versatile point-of-load power supply with 7A capability. (Page 6) The LTC4304 is a hot swappable 2-wire bus buffer that features pro- prietary stuck bus protection circuitry to unburden the MCU from the tasks of monitoring, troubleshooting and resolving stuck I2C buses. (Page 10) The LTC4261 integrates a –48V Hot Swap controller, 10-bit ADC monitoring and I2C/SMBus communication. It monitors the real-time board parameters such as current and voltages and communicates the data to the host. (Page 13) The LTC4011 is a complete standalone nickel chemistry fast charger that operates at high efficiency with an input voltage of up to 34V, even with a battery voltage that is much less than the DC input supply. (Page 19) The LT3750 is a current-mode flyback controller optimized for charging large value capacitors to a user selected target voltage. (Page 22) The LTC2926 MOSFET-controlled power supply tracker provides exceptionally flexible control of supply tracking and sequencing. (Page 24) The LTC3783 is a current-mode boost controller optimized for constant-current True Color PWM™ dimming of high powered LEDs. (Page 31) The LTC3412A is a monolithic buck regulator in a 4mm × 4mm QFN that can deliver efficiencies to 95% and operate at 4MHz. (Page 34) Design Ideas and Cameos Design Ideas start on page 37, including a, low noise, very low drift FET amplifier; a simple, digitally tunable active RC filter; and a way to increase the dimming range of an LED driver. Cameos appear on page 46. L Linear Technology in the News… EDN Magazine Nominates Linear for Innovation Awards EDN magazine has announced nominations for its annual EDN Innovation Awards, honoring outstanding engineering professionals and products. Linear Technology’s LTM4600 high power DC/DC µModule is nominated for an award in the Power Systems and Modules category. The LTM4600 is the first in a new family µModules from Linear Technology that leverages the company’s core strengths in power management and conversion. The LTM4600 provides designers with a complete 10A switching power supply in a tiny, low profile package that can be mounted on either side of a PC board. EDN has also nominated for Best Contributed Article, Jim Williams’ article, “Minimizing Switching Regulator Residue in Linear Regulator Outputs.” Winners will be announced in the April issue of EDN. 2 Linear A/D Converters Nominated for EE Times ACE Awards EE Times announced that they have nominated Linear Technology’s family of A/D converters as an Ultimate Product Finalist in the Analog category in the 2006 EE Times ACE Awards. The publication highlights the most significant products of the past year, with EE Times’ readers selecting finalists from a list of Ultimate Products chosen by the publication’s editors. EE Times will announce the ACE Award winners in April. Linear Technology now offers a broad line of high performance A/D converters. These ADCs at 10, 12, 14 and 16 bits, and up to 185Msps, have the lowest noise and provide extremely high bandwidth for undersampling. Linear’s A/D converter family also includes 3V devices, featuring the lowest power consumption and smallest solution size, spanning sampling rates from 10 to 125Msps at 10-, 12- and 14-bit resolution. Electronic Products Selects LTM4600 µModule as Product of the Year Electronic Products magazine selected Linear Technology’s LTM4600 µModule for their Product of the Year Awards, honoring the most significant new products of the past year. The LTM4600 was highlighted in a cover article in the January issue of Electronic Products, as well as in a February feature on how the product came about. Electronic Products’ editors commented, “Representing a new level of device integration for DC/DC conversion, the LTM4600 µModule synchronous switchmode DC/DC converter from Linear Technology is a complete power supply that can be placed almost anywhere on the board—even on the bottom—as it offers a reduction in space and circuit complexity of up to 60% over solutions using multiple components.” L Linear Technology Magazine • March 2006 DESIGN FEATURES L LTC2970, continued from page 1 linear search algorithm that compares the digitized point-of-load voltage against the target. The current being sourced by IDAC0 is then adjusted as needed one LSB per servo iteration. This current develops a point-of-load ground referenced correction voltage across resistor R40 which is buffered to the VOUT0 pin. The resulting voltage differential between the VOUT0 pin and the converter’s feedback node is multiplied by a factor of –R20/R30 and added to the nominal output voltage of the DC/DC converter, thus closing Margining and Monitoring Application Figure 2 shows a typical application circuit for monitoring and margining a DC/DC converter with external feedback resistors. The LTC2970’s VIN0_A differential inputs sense the voltage directly at the point-of-load while inputs VIN0_B monitor the voltage across sense resistor R50. The DC/DC converter’s output voltage can be margined to precise, user-programmable set points by a 5V REGULATOR VIN VOUT 12VIN 10 2R the servo loop. When voltage margining is disabled, the converter’s feedback node can be isolated from the LTC2970 by placing the VOUT0 pin in a high impedance state. Figure 3 shows the LTC2970 applied to a DC/DC converter with a TRIM pin. As in Figure 2, two external resistors are required: VOUT0 connects to the TRIM pin through resistor R30 and IOUT0 is terminated at the DC/DC converter’s point-of-load ground by R40. Following power-up, the VOUT0 pin defaults to a high impedance 0µA TO 255µA IDAC0 8 BITS R 14 IOUT0 VDD VDD 9 + 12VP CMP0 12VM VDDP GND 25 + VDDM TEMP SENSOR – VBUF0 VDD TSNSP 11 VOUT0 – UVLO POR TSNSM VIN0_AP 1 CH0_AP VIN0_AM 2 CH0_AM VIN0_BP 3 CH0_BP VIN0_BM 4 CH0_BM VIN1_AP 5 CH1_AP VIN1_AM 6 CH1_AM VIN1_BP 7 CH1_BP VIN1_BM 8 CH1_BM REF 23 0µA TO 255µA + – IDAC1 8 BITS + CMP1 ADC CLOCKS VDD 3.5k 7:1 MUX RAM ADC_Results MONITOR LIMITS SERVO TARGETS SCL 18 ASEL0 22 POR CLOCK GENERATION OSCILLATOR ASEL1 21 GPIO_0 16 SERVO CONTROLLER GPIO_1 15 DAC SOFT CONNECT FUNCTION SERVO FUNCTION MONITOR FUNCTION MANAGE FAULT REPORTING WATCH DOG TRACKING CONTROL (LT2970-1) ALERT 17 GPIO_CFG 20 7 12 VOUT1 VBUF1 – REFERENCE 1.229V (TYP) 20Ω I2C BUS INTERFACE (400kHz, SMBUS COMPATIBLE) – + 6.65X (TYP) RGND 24 SDA 19 13 IOUT1 14-BIT DELTA-SIGMA A/D 18 REGISTERS I/O CONFIGURATION IDAC0 IDAC1 ADC MONITOR FAULT ENABLE INSTANTANEOUS FAULTS LATCHED FAULTS 2 POR Figure 1. Block diagram of the LTC2970 Linear Technology Magazine • March 2006 3 L DESIGN FEATURES state allowing the DC/DC converter to power-up to its nominal output voltage. After power-up, the LTC2970’s soft-connect feature can be used to automatically find the IDAC code that most closely approximates the TRIM pin’s open-circuit voltage before enabling VOUT0. Applications that need to be sequenced can be configured to hold off the DC/DC converter when the LTC2970 powers-up by tying the GPIO_CFG pin high. This causes the GPIO_0 pin to automatically pull the DC/DC converter’s RUN pin low until the SMBus compatible I2C interface releases it. The absolute accuracy of the LTC2970 is demonstrated in Figure 4. The LTC2970 is configured to servo one of the outputs of a LTC3728 DC/ DC converter to 1V if the converter’s voltage deviates by more than ±0.1%. The LTC2970 is easily able to hold the output voltage to within ±1mV of 1V while both it and the DC/DC converter are heated from –50°C to 100°C. When the LTC2970 is isolated from the LTC3728, the output voltage drifts between 1.002V and 1.0055V over the same temperature range. 8V TO 15V VIN IN OUT VIN0_BM I– VIN0_BP R30 R20 RUN/SS 0.1µF GPIO_CFG I+ DC/DC CONVERTER FB R10 ALERT VOUT0 SCL VIN0_AP SDA I2C BUS + LOAD VDC0 – GPIO_0 IOUT0 R40 VIN0_AM GND SGND REF GND ASEL0 ASEL1 0.1µF Figure 2. Application circuit for DC/DC converter with external feedback resistors 8V TO 15V 0.1µF 12VIN VIN VDD VO+ 1/2 LTC2970 0.1µF GPIO_CFG R30 TRIM DC/DC CONVERTER VOUT0 ALERT VIN0_AP I2C BUS SCL SDA VSENSE+ + ON/OFF LOAD VDC0 – VSENSE– GPIO_0 IOUT0 R40 VIN0_AM VO– The LC2970’s features offer several benefits that differentiate it from competitive solutions: 4 VDD 1/2 LTC2970 Features REF GND ASEL0 ASEL1 0.1µF Figure 3. Application circuit for a DC/DC converter with a trim attenuates ripple components that have the potential to alias to DC. The ADC’s differential inputs can monitor supply voltages at the point of load and sense resistor voltages. The differential and common mode input ranges span –0.3V to 6V. With its 500µV/LSB resolution, the ADC can resolve voltages for a wide range of load current across sense resistor values of only a few milliohms. For 1.006 LOAD CURRENT 1.004 R1 L RL RL = INDUCTOR DCR R1 = R2 =10kΩ C1 = C2 = 0.1µF C1 UNCORRECTED VOUT (V) Delta-Sigma ADC The LTC2970’s ADC is a second-order delta-sigma modulator followed by a sinc2 digital filter that converts the modulator’s serial data into a 14-bit word at a conversion rate of 30Hz. The ADC’s TUE is less than ±0.5% when using the on-chip reference. One advantage delta-sigma ADCs offer over conventional ADCs is on-chip digital filtering. Combined with a large over-sampling ratio (OSR = 512), this feature makes the LTC2970 insensitive to the effects of noise when sampling power-supply voltages. The LTC2970’s sinc2 digital filter provides high rejection except at integer multiples of the modulator sampling frequency, fs = 30.72kHz. Adding a simple RC lowpass filter at the input of the ADC 0.1µF VIN R50 R2 1.002 1 0.998 −50 DRIFT AND OFFSET CORRECTED C2 – + −25 50 0 25 TEMPERATURE (°C) 75 14-BIT DELTA-SIGMA A/D 100 Figure 4. Corrected and uncorrected DC/DC converter output voltage vs temperature Figure 5. Network for sensing load current with inductor DCR Linear Technology Magazine • March 2006 DESIGN FEATURES L disturbance to the converter’s output voltage. There is no body diode from the VOUT0 pin to the LTC2970’s VDD supply, and the VOUT0 pin goes into a high impedance state when VDD drops below the LTC2970’s undervoltage lockout threshold. So no special precautions need to be taken in the event the DC/DC converter is still active when the LTC2970 powers down. DC/DC CONVERTER OUT R20 – + 11 R30 ADJUST VOUT0 – + R10 Voltage Servo The voltage servo feature can be configured to trigger on under voltage and/or over voltage events, run continuously, or run just once. The LTC2970 relies on a simple linear search algorithm to find the IDAC code that results in an ADC input voltage that most closely corresponds to the servo target. The polarity of the servo algorithm can be programmed as inverting (default) or noninverting. SOFT CONNECT COMPARATOR IDAC0 8 BITS 14 R40 GND IOUT0 POINT OF LOAD GROUND Figure 6. DAC connections to DC/DC converter with an external feedback resistors switching power supply applications without sense resistors, measure the load current via the DC resistance of the inductor using the application circuit shown in Figure 5. The ADC inputs are also isolated from the LTC2970’s internal supply. So the user can measure differential and common mode input voltages that are greater than VDD without turning on body diodes, and no special precautions need to be taken if the LTC2970 loses power while monitoring DC/DC converter voltages powered from a different supply. Voltage Buffered IDACs Figure 6 illustrates how each of the LTC2970’s continuous-time IDACs connects to a DC/DC converter with an external feedback network. The servo’d correction voltage is set by resistor R40. Since R40 is terminated at the point-of-load ground, the correction voltage is insensitive to load induced ground bounce. The correction voltage is buffered to the VOUT0 pin by a unity-gain amplifier whose output can be placed in a low-leakage (<100nA), high impedance state. Resistor R30 connects the VOUT0 pin to the feedback node of the DC/DC converter. The range and resolution over which the correction voltage can move the converter’s output is adjustable via resistor R30. A “soft-connect” feature allows the LTC2970 to automatically find the VOUT0 voltage that most closely approximates the DC/DC converter’s feedback node voltage before enabling the voltage buffer thus minimizing any Voltage Monitor The LTC2970 is able to perform ADC conversions on any combination of seven different input channels. Overvoltage and undervoltage threshold continued on page 46 8V TO 15V 0.1µF 12VIN VDD 10k Q10, Q11: 2N7002 D10, D11: MMBD4448V *SOME DETAILS OMITTED FOR CLARITY 0.1µF GPIO_CFG 100k GPIO_0 RUN/SS D10 LTC2970-1 Q10 VOUT0 R30A R30B ALERT I2C BUS FB R10 SCL IOUT0 SDA 0.1µF R31B R20 FB R11 IOUT1 VDC0 IN VIN DC/DC CONVERTER Q11 GND OUT RUN/SS D11 R31A VIN R40 GPIO_1 VOUT1 IN DC/DC CONVERTER OUT VDC1 R21 R41 Figure 7. The LTC2970-1 enables supply tracking Linear Technology Magazine • March 2006 5 L DESIGN FEATURES Cascadable, 7A Point-of-Load by Peter Guan Monolithic Buck Converter Introduction Easy-to-use and compact point-ofload power supplies are necessary in systems with widely distributed, high current, low voltage loads. The LTC3415 provides a compact, simple, complete and versatile solution. It includes a pair of integrated complementary power MOSFETs (32mΩ top and 25mΩ bottom) and requires no external sense resistor. A complete design involves choosing an inductor and input/output capacitors, and that’s it. The result is a fast, constant frequency, current mode, 7A DC/DC switching regulator. 0.1µF 47µF 6.3V 3x 38 1 3 4 5 10pF 6 7 8 9 MODE 10 11 12 37 36 CLKOUT RUN NC PVIN 35 PVIN 34 SVIN 33 ITHM SGND 32 ITH NC TRACK PLLLPF VFB PVIN PVIN PVIN PVIN SW SW LTC3415EUHF SW SW SW SW SW SW MODE PGOOD PGND (39) CLKIN BSEL MGN PHMODE PGND PGND PGND PGND PGND PGND PGND 13 14 15 16 17 18 31 30 29 27 22pF 30.1k 26 25 24 23 22 21 10k SVIN 20 19 100µF, 6.3V 2x 0.2µH 15k SVIN 28 VOUT 1.8V/7A Figure 1. Typical application of the LTC3415 in a 3.3V to 1.8V/7A step down converter Operation Figure 1 shows a typical application of the LTC3415 in a 3.3V to 1.8V/7A step down converter. Figure 2 shows its efficiency and power loss vs load current. Figure 3 shows its transient response to a 5A load. The LTC3415 uses a constant frequency, current mode architecture to drive an integrated pair of complementary power MOSFETs. An internal oscillator sets the 1.5MHz operating 95 10000 EFFICIENCY 90 EFFICIENCY (%) 85 VOUT = 1.8V 100mV/DIV AC COUPLED 1000 80 75 100 70 65 POWER LOSS 60 VIN = 3.3V VOUT = 1.8V Burst Mode OPERATION 55 50 frequency of the device. The main P-channel power MOSFET turns on with every oscillator cycle and turns off when the internal current comparator trips, indicating that the inductor current has reached a level set by the ITH pin. An internal error amplifier, in turn, drives the ITH pin by monitoring the output voltage through an external resistive divider connected to the VFB pin. While the P-channel power MOSFET is off, the internal synchronous N-channel power MOSFET turns on until ether the inductor current 10 100 1000 LOAD CURRENT (mA) POWER LOSS (mW) 6 2 SGND Features The overall solution is extremely compact since the LTC3415’s QFN 5mm × 7mm package footprint is small while its high operating frequency of 1.5MHz allows the use of small low profile surface mount inductors and ceramic capacitors. For loads higher than 7A, multiple LTC3415s can be cascaded to share the load while running mutually antiphase, which reduces overall ripple at both the input and the output. Other features include: q Spread spectrum operation to reduce system noise, q Output tracking for controlled VOUT ramp-up and ramp-down, q Output margining for easy system stress testing, q Burst Mode® operation to lower quiescent current and boost efficiency during low loads, q Low shutdown current of less than 1µA, q 100% duty-cycle for low drop out operation, phase-lock-loop to allow frequency synchronization of ±50%, q Easily cascadable for multi-device load sharing with multiphase operation q Internal or external ITH compensation for either ease of use or loop optimization. 1Ω VIN IINDUCTOR STEP = 0A TO 5A 5A/DIV IOUT STEP = 0A TO 5A 5A/DIV 10 1 10000 Figure 2. Efficiency and power loss vs load current for the circuit in Figure 1 VIN = 3.3V 40µs/DIV L = 0.2µH COUT = 2 × 100µF Figure 3. Transient response to a 5A load for the circuit in Figure 1 Linear Technology Magazine • March 2006 DESIGN FEATURES L 10k 0.1µF RUN 100pF 1000pF 0.1µF 1Ω 1Ω VIN VIN 38 1 SGND 2 3 4 5 6 7 8 9 10 11 SVIN 12 37 36 CLKOUT RUN PVIN NC 35 PVIN 34 SVIN 33 ITHM SGND 32 ITH NC TRACK PLLLPF VFB PVIN PVIN PVIN PVIN SW SW LTC3415EUHF SW SW SW SW SW SW MODE PGOOD PGND (39) CLKIN BSEL PHMODE MGN PGND PGND PGND PGND PGND PGND PGND 13 14 15 16 17 18 38 22µF 6x 31 1 15k 30 VTRACK 29 10k SGND 2 3 4 28 27 30.1k 10pF 100pF 26 5 6 25 7 24 8 23 9 10k 22 21 10 SVIN 11 BSEL 20 MGN SVIN 12 36 35 PVIN 34 SVIN 33 ITHM SGND 32 ITH NC TRACK PLLLPF VFB PVIN PVIN PVIN PVIN SW SW LTC3415EUHF SW SW SW SW SW SW MODE PGOOD PGND (39) CLKIN BSEL PHMODE MGN PGND PGND PGND PGND PGND PGND PGND 13 19 37 CLKOUT2 RUN PVIN NC 14 15 16 17 18 31 30 29 28 27 26 25 24 23 22 21 20 BSEL MGN 19 100µF 6.3V 2x VOUT 1.8V/14A 0.2µH 0.2µH Figure 4. Dual-phase single output 3.3V to 1.8V 15A application using two LTC3415s running 180° out of phase with respect to each other. Modes of Operation: Burst, Pulse-Skip, and Forced Continuous Three modes of operation can be selected through the MODE pin. Tying it to VIN enables Burst Mode operation for highest efficiency. During low output loads, the peak inductor current limit is clamped to about a quarter of the maximum value and the ITH pin is monitored to determine whether the device will go into a power-saving Sleep mode. Quiescent current is reduced to 450µA in Burst Mode operation because most of the internal circuitry is turned off. For applications that aim to reduce output ripple and strive to maximize operating time at constant frequency, pulse-skip mode is a good solution. Pulse-skip mode is enabled by letting the Mode pin float or tying it to VIN /2 or VOUT. Inductor current is still not allowed to reverse as in Burst Mode operation, but the peak inductor current limit is no longer clamped internally. ITH has full control of output current until ITH drops so low (at low output Linear Technology Magazine • March 2006 loads) that minimum on-time of the device is reached and the LTC3415 begins to skip cycles. For applications that require constant frequency operation even at no load, the LTC3415 can be put into forced continuous mode operation by tying the Mode pin to ground. In this mode, inductor current is allowed to reverse while the internal power MOSFETs are always driven at the same frequency. Output Tracking For applications that require controlled output voltage tracking between 100 95 90 EFFICIENCY (%) starts to reverse, as indicated by the SW pins going below ground, or until the beginning of the next cycle. 85 80 1 75 2 3 4 6 70 65 60 55 50 1 10 LOAD CURRENT (A) 100 Figure 5. Combined efficiency of loadsharing in 1-phase, 2-phase, 3-phase, 4-phase, and 6-phase operation. their various outputs in order to prevent excessive current draw or even latch-up during turn-on and turnoff, the LTC3415 includes a Track pin that allows the user to program how its output voltage ramps during start-up and shutdown. During startup, if the voltage on the Track pin is less than 0.57V, the feedback voltage regulates to this tracking voltage, thus programming the output voltage to follow along. Inductor current is not allowed to reverse during tracking, ensuring monotonic voltage rise. When the tracking voltage exceeds 0.57V, tracking is disabled and the feedback voltage regulates to the internal reference voltage (0.596V). In other words, output voltage is controlled by the Track voltage until the output is in regulation. Taking the LTC3415’s Run pin to below 1.5V would normally shut down the part, but if the output also needs tracking during shutdown, then the LTC3415 must remain active even if the Run pin is low. So, during shutdown, if Track is 0.5V or more below SVIN, then even if Run is low, the LTC3415 will not shutdown until Track has fallen below 0.18V, thus allowing the output to properly 7 L DESIGN FEATURES 0° 180° 90° CLKIN CLKOUT +90° PHMODE CLKIN CLKOUT +90° CLKIN CLKOUT +90° PHMODE PHMODE CLKIN CLKOUT PHASE 4 PHASE 7 PHASE 10 120° 210° 300° (420°) 60° +90° PHMODE CLKIN CLKOUT PHMODE PHASE 5 PHASE 8 CLKIN CLKOUT PHMODE PHASE 11 +120° PHASE 3 +90° PHMODE PHASE 2 150° CLKIN CLKOUT PHMODE CLKIN CLKOUT +90° CLKIN CLKOUT +90° PHMODE PHASE 6 330° 240° CLKIN CLKOUT +90° +120° PHMODE PHASE 1 CLKIN CLKOUT (390°) 30° 270° +90° CLKIN CLKOUT PHMODE PHMODE PHASE 9 PHASE 12 Figure 6. Cascaded, 12-phase operation makes for efficient high current load sharing track the master voltage during its ramp-down. For applications that do not require tracking or externally controlled soft start, simply tie the Track pin to SVIN. Output Margining For convenient and accurate system stress test on the LTC3415’s output, the user can program the LTC3415’s output to ±5, ±10, and ±15% of its nominal operational voltage. The MGN pin, when left floating, forces normal operation. When MGN is tied to GND, it forces negative margining, in which the output voltage is below the regulation point. When MGN is tied to SVIN, then the output voltage is forced to above the regulation point. The amount of output voltage margining is determined by the BSEL pin. When BSEL is low, it’s 5%. When BSEL is high, it’s 10%. When BSEL is left floating, margin percentage is 15%. To prevent system glitches while margining, the internal output overvoltage and undervoltage comparators are disabled and thus PGOOD remains pulled high by the external resistor. Multiphase Operation For output loads that demand more than 7A of current, multiple LTC3415s can be cascaded to run out of phase while equally sharing output load 8 current. Figure 4 shows a dual-phase single output 3.3V to 1.8V 14A application using two LTC3415s running 180° out of phase with respect to each other. Figure 5 shows the combined efficiency of 1-phase, 2-phase, 3-phase, 4-phase, and 6-phase operation. The CLKIN pin allows the LTC3415 to synchronize to an external frequency (between 0.75Mhz and 2.25Mhz) and the internal phase-lock-loop allows the LTC3415 to lock on to CLKIN’s phase as well. The CLKOUT signal can be connected to the CLKIN pin of the following LTC3415 stage to line up both the frequency and the phase of the entire system. Tying the PHMODE pin to SVIN, SGND, or SVIN/2 (floating) respectively generates a phase difference (between CLKIN and CLKOUT) of 180°, 120°, or 90°, which respectively corresponds to 2-phase, 3-phase, or 4-phase operation. A total of 12 phases can be programmed by setting the PHMODE pin of each phase to different levels. For instance, a slave stage that’s 180° out of phase from the master can generate a 120° CLKOUT signal that’s 300° (PHMODE = SGND) away from the master for the next stage, which then can generate a CLKOUT signal that’s 420°, or 60° (PHMODE = SVIN/2) away from the master for its following stage. See Figure 6. A multiphase power supply significantly reduces the amount of ripple current in both the input and output capacitors. The RMS input ripple current is divided by, and the effective ripple frequency is multiplied by, the number of phases used (assuming that the input voltage is greater than the number of phases used times the output voltage). The output ripple amplitude is also reduced by, and the effective ripple frequency is increased by, the number of phases used. Output Current Sharing When multiple LTC3415s are cascaded to drive a common load, accurate output current sharing is essential to achieve optimal performance and efficiency. Otherwise, if one stage is delivering more current than another, then the temperature between the two stages is different, and that can translate into higher switch RDS(ON), lower efficiency, and higher RMS ripple. Each LTC3415 has a trimmed peak current limit such that when the ITH pins of multiple LTC3415s are tied together, the amount of output current delivered from each LTC3415 is nearly the same. Different ground potentials among LTC3415 stages, caused by physical distances and switching noises, could cause an offset to the absolute ITH value Linear Technology Magazine • March 2006 DESIGN FEATURES L Internal/External ITH Compensation During single phase operation, the user can simplify the compensation of the internal error amplifier loop (on the ITH pin) by tying it to SVIN to enable internal compensation, which connects an internal 50k resistor in series with a 50pF cap to the internal ITH compensation point. This is a tradeoff for simplicity and ease of use at the expense of OPTI-LOOP® optimization, where external ITH components can be selected to optimize the loop transient response with minimum output capacitance. In multi-phase operation where all the ITH pins of each phase are tied together to achieve accurate load sharing, internal ITH compensation is disabled and external compensation components need to be properly selected for optimal transient response and stable operation. Master/Slave Operation In multiphase single-output operation, the user has the option to run in multi-master mode where all the FB, ITH, and output pins of the stages are tied to each other. All the error amplifiers are effectively operating in parallel and the total transconductance (gm) of the system is increased by the number of stages. The ITH value, which dictates how much current is delivered to the load from each stage, is averaged and smoothed out by the external ITH compensation components. Nevertheless, in certain applications, the resulting higher gm from multiple 3415s can make the system loop harder to compensate; in this case, the user can choose an alternative mode of operation. Linear Technology Magazine • March 2006 –10 –20 –30 VIN = 5V VOUT = 1.8V RBW = 100Hz –10 –14.1dBm –20 –30 –40 AMPLITUDE (dBm) AMPLITUDE (dBm) seen by each stage. To ensure that the ground level doesn’t affect the ITH value, the LTC3415 uses a differential amplifier that takes as input not just the ITH pin, but also the ITHM pin, which doesn’t connect to any other circuitry except for the ITH differential amplifier. Therefore, the ITHM pins of all the LTC3415 stages should be tied together and then connected to the SGND pin at only one point. –50 –60 –70 –80 –50 –60 –70 –80 –90 –100 –100 a. The LTC3415’s output noise spectrum analysis in free-running constant frequency operation –37.3dBm –40 –90 1.0 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2.0 FREQUENCY (MHz) VIN = 5V VOUT = 1.8V RBW = 100Hz 1.0 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2.0 FREQUENCY (MHz) b. The LTC3415’s output noise spectrum analysis in spread spectrum operation Figure 7. Spread spectrum operation reduces EMI peaks by spreading the EMI energy over a range of frequencies The second mode of operation is single-master operation where only the error amplifier of the master stage is used while the error amplifiers of the other stages (slaves) are disabled. The slave’s error amplifier is disabled by tying its FB pin to SVIN, which also disables the internal overvoltage comparator and power-good indicator. The master’s error amplifier senses the output through its FB pin and drives the ITH pins of all the stages. To account for ground voltage differences among the stages, the user should tie all ITHM pins together and then tie it to the master’s signal ground. As a result, not only is it easier to do loop compensation, this single-master operation should also provide for more accurate current sharing among stages because it prevents the error amplifer’s output (ITH) of each stage from interfering with that of another stage. Spread Spectrum Operation Switching r egulators can be troublesome where electromagnetic interference (EMI) is a concern. Switching regulators operate on a cycle-by-cycle basis to transfer power to an output. In most cases, the frequency of operation is fixed or is a constant based on the output load. This method of conversion creates large components of noise at the frequency of operation (fundamental) and multiples of the operating frequency (harmonics). To reduce this noise, the LTC3415 can run in spread spectrum operation by tying the CLKIN pin to SVIN. In spread spectrum operation, the LTC3415’s internal oscillator is designed to produce a clock pulse whose period is random on a cycle-by-cycle basis but fixed between 60% and 140% of the nominal frequency. This has the benefit of spreading the switching noise over a range of frequencies, thus significantly reducing the peak noise. Figures 7 and 8 show how the spread spectrum feature of the LTC3415 significantly reduces the peak harmonic noise vs free-running constant frequency operation. Spread spectrum operation is disabled if CLKIN is tied to ground or if it’s driven by an external frequency synchronization signal. Conclusion With its many operational features and compact total solution size, the LTC3415 is an ideal fit for today’s point-of-load power supplies. It also has the advantage of simple upgradeability: if the load requirement of a power supply increases, no need to dread the redesign of the whole system, just stack on another LTC3415 and keep going. L for the latest information on LTC products, visit www.linear.com 9 L DESIGN FEATURES I2C Bus Buffer Resolves Stuck Buses, Eliminates Heavy Load Limitations and Provides Level Translation by George Humphrey Introduction The I2C bus is a 2-wire bidirectional communications bus primarily used for system configuration and monitoring. A bus master, typically a microcontroller unit (MCU), polls system components for information such as supply voltage and temperature, Vital Product Data (VPD) from memories on removable cards, and system configuration. The master can also use the bus to reconfigure the system through general purpose BACKPLANE input/output ports (GPIOs) and other programmable devices. Large-scale telecommunication and data storage systems are divided up into several peripheral cards. Each card plugs on to the I2C bus and communicates its system information to the MCU through the back plane. Certain considerations must be addressed to ensure a high level of reliability when designing a system using the I2C bus. Primarily, systems must BACKPLANE CONNECTOR STAGGERED CONNECTOR I/O PERIPHERAL CARD 1 VCC R1 10k C1 0.01µF R2 10k VCC ACC SDA MCU provide protection against a hung or stuck bus. If for any reason the bus is stuck low, there is the potential to bring the entire system down. Another consideration is dealing with heavy capacitive loads on the bus. The I2C standard specifies a 400pF limit on the bus. With systems becoming larger, this specification becomes problematic because of so many devices connected directly to the bus. Systems also need to be robust, allowing cards to SDAIN SCL SCLIN ENABLE LTC4304 GND R4 10k R5 10k SDAOUT I2 C DEVICES SCLOUT READY FAULT 10k I/O PERIPHERAL CARD 2 C3 0.01µF VCC ACC SDAIN SCLIN ENABLE LTC4304 GND R8 10k R9 10k SDAOUT I2 C DEVICES SCLOUT READY FAULT 10k • • • I/O PERIPHERAL CARD N C5 0.01µF ACC SDAIN SCLIN ENABLE VCC LTC4304 GND R12 10k SDAOUT SCLOUT R13 10k I2 C DEVICES READY FAULT 10k Figure 1. Typical application of the LTC4304 10 Linear Technology Magazine • March 2006 DESIGN FEATURES L be inserted or removed from a live backplane without corruption of the I2C bus data. Finally, cards must be immune to ESD events that can occur from human handling. The LTC4304 is a multifunctional device. It has a flexible architecture that allows it to be configured into almost any system. The LTC4304 features proprietary stuck bus protection circuitry that unburdens the MCU from the tasks of monitoring, troubleshooting and resolving stuck buses. Simplifying this task also saves PCB board space and connector pins. The LTC4304 provides bidirectional capacitance buffering, isolating the card bus from the backplane bus. Rise time accelerators help heavily capacitive-loaded buses meet I2C rise time specifications. The LTC4304 also enables cards to be inserted or removed from the system without corrupting the data on the I2C bus. Built-in ±15kV of human Body Model ESD protection provides a rugged front end for cards that typically have the SDA and SCL lines connected directly to the card connector. Circuit Operation Startup and Hot Swap To take full advantage of the LTC4304’s Hot Swap features, a staggered connector must be used: where VCC and GND are the longest pins, SDA and SCL pins medium length and ENABLE the shortest. After the card’s VCC and GND connect with the live backplane, the voltage on VCC starts to rise. At this time, the LTC4304’s precharge circuitry charges the capacitance on the SDA and SCL lines to 1V. Precharged SDA/SCL lines minimize disturbances on the bus when they connect to the backplane. The remainder of the LTC4304’s circuitry is disabled until the supply voltage rises above 2.5V (typical) and the voltage on ENABLE is above 1.4V (typical). Making ENABLE the shortest pin insures that SCL and SDA are firmly seated before a connection is established on the bus. When the VCC supply to the LTC4304 is valid, it assumes that the card is connected to a live backplane and looks Linear Technology Magazine • March 2006 SDAOUT 5V/DIV SDAIN 5V/DIV STUCK LOW > 30ms RECOVERS DISCONNECT AT TIMEOUT AUTOMATIC CLOCKING SCLOUT 5V/DIV FAULT 5V/DIV 200µs/DIV Figure 2. Stuck bus resolved with automatic clocking for either a stop bit or a bus idle on the backplane side on SDAIN and SCLIN. When one of these conditions is met and SDAOUT and SCLOUT are high, the connection circuitry is activated, connecting the card side bus with the backplane bus. These requirements ensure that the data on the bus is not corrupted when the card is plugged in. READY indicates the status of the connection circuitry and is high when the connection circuitry is active. Bus Stuck Low Timeout The LTC4304 monitors SDAOUT and SCLOUT on the card side of the I2C bus. When SDAOUT or SCLOUT is low, an internal timer is started, only to be reset when SDAOUT and SCLOUT are both high. If SDAOUT or SCLOUT do not go high within 30ms, the connection between the inputs and outputs is automatically disconnected, and FAULT asserts low indicating a stuck bus. The LTC4304 also detects a fault if powered up into a stuck bus condition. In either case, the LTC4304 automatically generates up to 16 clock pulses at 8.5kHz on SCLOUT. If SDAOUT and SCLOUT go high, FAULT releases and is pulled high. The LTC4304 then looks for a stop bit or a bus idle before automatically reconnecting. While there is a stuck bus condition, a connection can be forced with a rising edge on ENABLE, even if bus idle conditions are not met. Capacitance Buffering The connection circuitry contains a unique patent-pending architecture that provides electrical isolation, isolating the capacitance on the card side bus from the backplane side bus, while maintaining full I2C functionality. This means that the MCU only sees the capacitance of the backplane and the low input capacitance of the LTC4304, <10pF guaranteed by design. The LTC4304 drives the capacitance of the circuitry on the card side. The LTC4304 regulates the voltage on the opposite side from which it is being driven to a slightly higher voltage. This voltage (call it VOS) is a function of VCC, the pull-up resistor, and an internally set constant terms given by the equation: VOS = VCC • 20 + 75mV RPULLUP For example: For a 2.7kΩ pull-up and a VCC of 3.3V, VOS = 3.3mV • 20 + 75mV = 99.4mV 2.7kΩ Rise Time Accelerators Rise-time accelerators are included on all four SDA and SCL pins. Once activated, the accelerators switch in 3.5mA of current (typical at VCC = 2.7V) into the SDA and SCL lines to make them rise faster. This ensures that rise-time requirements are met and allows the use of larger pull-up resistors to reduce power consumption. When ACC is connected to ground, all four accelerators are ON. When ACC is connected to VCC, all four accelerators are OFF. And when ACC is floating, only the accelerators on SDAOUT and SCLOUT are activated. Accelerators cannot be used on pins where the pull-up voltage is less than VCC. The flexibility of the accelerators control provided by ACC allows the LTC4304 11 L DESIGN FEATURES to interface with buses not common with its supply. Applications Stuck Bus Automatically Resolved Independent of the MCU The I2C bus protocol calls for bidirectional communication between devices. A typical example is a microcontroller unit (or MCU) communicating with a slave device. As the slave device’s internal register is read by the MCU, the MCU clocks the slave device to receive each bit of data. A problem occurs when the MCU is out of sync with the slave device—the slave device is waiting for another clock and the MCU thinks it has already sent out enough clocks. If the slave device happens to be holding the data line low, all further communications are prevented and the I2C bus is stuck. Figure 1 shows a card with a resident LTC4304 that acts as the interface between the devices on the BACKPLANE CONNECTOR continued on page 18 STAGGERED CONNECTOR 5V C1 0.01µF 3.3V R1 10k card and the I2C bus, when the card is plugged into the backplane. In normal, plugged-in operation, a connection is established between SDAIN and SDAOUT, and SCLIN and SCLOUT. Internal comparators monitor the SDAOUT and SCLOUT nodes of the circuit. When SDAOUT or SCLOUT is low, an internal timer starts. The timer is only reset when SDAOUT and SCLOUT are both high. If they don’t go high within 30ms, it is determined that the bus is stuck low and the con- R2 10k ACC SDA SDAIN SCL SCLIN 5V ENABLE VCC LTC4304 GND R3 10k R4 10k SDAOUT CARD_SDA SCLOUT CARD_SCL READY FAULT R5 10k a. VCC is higher than the pull-up voltage on SDAIN and SCLIN. BACKPLANE CONNECTOR STAGGERED CONNECTOR 3.3V C2 0.01µF 5V R6 10k R7 10k ACC SDA SDAIN SCL SCLIN 3.3V ENABLE VCC LTC4304 GND R8 10k R9 10k SDAOUT CARD_SDA SCLOUT CARD_SCL READY FAULT R10 10k b. VCC is equal to the output side pull-up voltage, and less than input side pull-up voltage. BACKPLANE CONNECTOR STAGGERED CONNECTOR 2.5V 5V C3 0.01µF 3.3V R11 10k R12 10k ACC SDA SDAIN SCL SCLIN 3.3V R15 10k ENABLE VCC LTC4304 GND R13 10k R14 10k SDAOUT CARD_SDA SCLOUT CARD_SCL READY FAULT c. VCC is greater than the pull-up voltages of both sides. Figure 3. The LTC4304 can be the bridge between a backplane operating at one voltage and a card operating at a different voltage. Here are three possible scenarios and appropriate configurations. 12 Linear Technology Magazine • March 2006 DESIGN FEATURES L Negative Voltage Hot Swap Controller with 10-Bit ADC and I2C Monitoring by Zhizhong Hou Introduction High availability –48V power systems, such as telecom and AdvancedTCA systems, allow for circuit board upgrade and replacement on a live powered backplane. The primary role of a Hot Swap controller is to make this possible by limiting potentially large inrush currents, which can cause damage to the hot swapped board or create disturbances on the backplane. Traditionally, –48V Hot Swap controllers operate autonomously—the Hot Swap controller shuts down an abnormally operating board before the host processor knows why. This simplifies system design, but gives little in the way of diagnostic support to the host. A far more robust system would use the Hot Swap controller to communicate the conditions of the board to the host processor, and let the host processor take action. To this end, the LTC4261 integrates a –48V Hot Swap controller, 10-bit ADC monitoring and I2C/SMBus communication. It monitors the real-time board parameters such as current and voltages and communicates the data to the host. Features Figure 1 shows a simplified block diagram of the LTC4261. Power is derived from the –48V RTN using an external dropping resistor connected to the VIN pin. An internal shunt regulator clamps the voltage at VIN to 11.2V above VEE (chip ground). This floating architecture allows a wide operating voltage range. The device also provides a 5V linear-regulated voltage at the INTVCC pin that can source current up to 20mA for driving external circuits. Using an external N-channel pass transistor, the negative Hot Swap circuit of the LTC4261 allows a board to be safely inserted and removed from a live –48V backplane. The device features a new inrush control technique Linear Technology Magazine • March 2006 ADR0 SCL ADR1 SDAI SDAO I2C INTERFACE ADIN ADIN2 10-BIT ADC ALERT REGISTER INTVCC VIN IN 5V REGULATOR 11.2V UVL OUT 40× UVH NEGATIVE VOLTAGE Hot Swap OV VEE VEE SENSE SS TMR GATE RAMP DRAIN Figure 1. Simplified block diagram of LTC4261 that minimizes stresses on the pass transistor in all operating conditions. Turning the device on or off can be either autonomous or controlled by a host processor through the I2C interface. Auto-retry following a fault is programmable and fully controlled by the host. Configurations of the device are stored in the internal registers as shown in Table 1. The LTC4261 continuously monitors and registers board status and fault conditions. With an onboard 10bit ADC and 3-channel multiplexer, it accurately measures real-time board current (through the voltage across the sense resistor) and two external voltages. The data are stored in the ADC registers (see Table 1). When polled by a host processor, the LTC4261 reports the ADC data along with the status and fault information using the I2C interface. The real-time board current and voltages provides a means for the host to detect any early warn- ing signal and to flag the board for maintenance before it fails. With the ALERT pin, the LTC4261 interrupts the host for specific fault conditions, when configured to do so. One unique feature of the LTC4261 is that the I2C interface can be easily configured using the address pins (ADR1 and ADR0) into a single-wire broadcast mode that only uses a single I2C signal, SDAO, to report the ADC data and fault information. This mode simplifies the interface and saves component cost by eliminating two optoisolators. The LTC4261 has additional features to sequence two power good outputs, detect insertion of a board and turn off the pass transistor if an external supply monitor fails to indicate power good within a timeout period. Using the PGIO and FLTIN pins along with the ADC, the device can detect a specific fuse that is open for up to four fuses. 13 L DESIGN FEATURES –48V RTN RIN 4× 1k IN SERIES 1/4W EACH R3 453k 1% UV = 38.5V UV RELEASE AT 43V 7 8 9 10 11 19 20 26 2 25 24 R2 16.9k 1% OV = 72.3V OV RELEASE AT 71V CUV 100nF CIN 1µF R1 CVCC 11.8k 0.1µF 1% –48V INPUT VEE UVL UVH ADIN2 OV SS TMR EN ON ADR1 ADRO VEE FLTIN PGI SCL SDAI SDAO ALERT 14 15 RG 10Ω 16 + VIN+ MODULE2 27 PWRGD1 PG 23 ADIN SENSE GATE DRAIN RAMP ON VIN– CL 330µF 100V VIN+ MODULE1 ON VIN– 18 RD 1M RF 1k CR 10nF 100V 5% CF RS 33nF 0.008Ω 1% R10 10k 1% 22 1 6 5 4 3 PGIO 28 PWRGD2 CG 47nF CTMR 47nF VIN LTC4261CGN 13 CSS 220nF 21 INTVCC Q1 IRF1310NS VOUT R11 402k 1% Figure 2. A –48V/200W Hot Swap controller (5.6A current limit, 0.66A inrush) using LTC4261 with current, input voltage and VDS monitoring 10-Bit ADC Provides Accurate Measurement of Real-Time Board Current and Voltages Quantitative monitoring of real-time board level voltage and current (and thus power) provides significant benefits in high availability systems. Real time operating data can be compared to budgeted or historical data to detect whether a circuit board is using its allotted power or if is operating abnormally. By issuing early warning to system management, an abnormally operating board can be flagged for service even before it fails. This feature greatly improves the reliability of high availability systems. The LTC4261 includes a 10-bit ADC that accurately measure voltages at the SENSE, ADIN2 and ADIN pins, all referred to chip ground (VEE). With a 2.56V full scale and 2.5mV resolution, the ADIN and ADIN2 pins are uncommitted inputs that allow monitoring of any external voltages. With the sense resistor, the SENSE pin voltage is used to measure current flowing through the pass transistor. This voltage is Table 1. LTC4261 registers Register Read/ Write Description STATUS R Provides pass transistor (on/off), EN (high/low) and PGIO input conditions. Also lists five fault present conditions. FAULT R/W Latches overcurrent, overvoltage, undervoltage, power bad, FETshort faults and EN changed state. Also logs FLTIN and PGIO inputs. ALERT R/W Enables which faults interrupt the host using the ALERT pin. Defaults not to alert on faults at power-up. CONTROL R/W Controls on or off of the pass transistor and whether the part auto-retries or latches off after a fault. Also configures the PGIO pin and enables ADC register test mode. SENSE R/W ADC data for the SENSE voltage measurement ADIN2/OV R/W ADC data for the ADIN2 (on TSSOP) or OV (on QFN) pin voltage measurement ADIN R/W ADC data for the ADIN pin voltage measurement 14 internally amplified by 40 times resulting in a 64mV full scale and 62.5μV resolution. The digital codes of the three voltages after each conversion are stored in corresponding ADC registers (see Table 1) and updated at a frequency of 7.3Hz. Setting the test mode bit in the CONTROL register halts the updating so that software testing can be performed by writing to and reading from the registers. An example of using the ADC monitoring is shown in Figure 2, where current, input voltage and VDS of the pass transistor are measured at the SENSE, ADIN2 and ADIN pins, respectively. The latter two voltages can be used to derive the output voltage referred to RTN. Another application of the ADC monitoring is to detect an open fuse in a multi-feed system, which is detailed later in this article. Independently Adjustable Inrush and Overcurrent Limits Minimize Stress on Pass Transistor A typical –48V/200W Hot Swap application using the LTC4261 is shown in Figure 2. Initial turn-on of pass transistor Q1 is autonomous but I2C can take over the control after power-up. To protect the pass transistor from overstress, the LTC4261 independently controls inrush current Linear Technology Magazine • March 2006 DESIGN FEATURES L charge. This draws a current from the RAMP pin that flows through CR and causes the GATE current to drop to 0. The RAMP pin is regulated at 1.1V so the inrush is set by the ramp rate of VOUT, which leads to: –48V INPUT 50V/DIV VOUT 50V/DIV SENSE 0.5A/DIV PG 50V/DIV IINRUSH = 20µA 10ms/DIV Figure 3. The LTC4261 start-up behavior during power-up and overcurrent upon a short circuit. As indicated in Figure 2, the current limit and circuit breaker threshold is set to 5.6A by the sense resistor RS, while the inrush is set to a much lower level (0.66A) by controlling the ramp rate of VOUT with an external capacitor CR connected between VOUT and the RAMP pin. The operation theory and the benefits of this inrush and overcurrent control technique are demonstrated below. Following a start-up debounce delay, the turn-on sequence of the LTC4261 starts with charging the SS pin up with a 10µA current. The SS voltage (VSS) is converted to a GATE pull-up current. When the GATE voltage reaches the threshold voltage of the pass transistor Q1, the inrush starts to flow through Q1. The output voltage (VOUT) begins to move and CR begins to CL CR The slew rate of VSS determines dI/dt of the inrush current. Figure 3 shows the start-up behavior of the LTC4261. The LTC4261 provides 2-level overcurrent protection: an active current limit (ACL) amplifier that also serves as a circuit breaker comparator with a threshold of 50mV±10%, and a fast pull down comparator with a threshold of 150mV. In the event of an output short or a fast input step, when VSENSE exceeds 150mV, the fast pull down comparator immediately brings the GATE down with a 110mA current. Once VSENSE falls back below 150mV, the ACL starts to servo the GATE and maintains a constant output current of 50mV/RS. If the short-circuit condition lasts longer than the circuit breaker delay of 530µs, the pass transistor is turned off and an overcurrent fault is registered. The device defaults to latch off upon an overcurrent fault but can be configured to automatically re-try after a cooling delay. Figure 4 illustrates the response of the LTC4261 to the short-circuit condition. In the case of an input step, the inrush control circuit takes over following the fast GATE pull-down and the current limit loop is disengaged before the circuit breaker timer expires. The device then operates similarly to the start-up, only with a difference that the current through the pass transistor is now a sum of inrush and load current. By decoupling start-up inrush from the current limit/circuit breaker threshold, the LTC4261 makes it possible to optimize the safe operating area (SOA) of the pass transistor in all operating conditions. The short circuit breaker timer substantially reduces the stress on the pass transistor in a short-circuit condition. Startup and input step typically impose the greatest stress on the pass transistor. Setting the inrush current much smaller than the current limit relieves the SOA requirement during start-up and input step. This allows using smaller pass transistors for large load applications, making selection of pass transistors much easier. Using the dedicated RAMP pin that is regulated separately from GATE 50mV SENSE 530µs INRUSH 10V GATE VOUT 1.77V PWRGD1 DELAY OC COOLING DELAY TMR 4x 2x PWRGD2 DELAY 2x SS PG PGIO Figure 4. Response of the LTC4261 to overcurrent fault in auto-retry mode Linear Technology Magazine • March 2006 15 L DESIGN FEATURES for inrush control, the LTC4261 does not require a large capacitor between GATE and VEE that relates to CR, so the turn-off of the pass transistor upon short-circuit can be fast even for large load applications. UVL/UVH R1 11.8k 1% 0V R1 11.8k 1% UVL 0V OV TURN-ON = 71V, OV TURN-OFF = 72.3V UV TURN-ON = 43V UV TURN-OFF = 38.5V HYSTERESIS = 4.5V UV TURN-ON = 43V UV TURN-OFF = 41.2V HYSTERESIS = 1.8V UV TURN-ON = 46V UV TURN-OFF = 38.5V HYSTERESIS = 7.5V c. b. a. Figure 5. Adjustment of undervoltage thresholds and hysteresis 71V with UV turn-off at 38.5V and OV turn-off at 72.3V. The UV levels can be adjusted by connecting a resistor RH between the UVH and UVL pins, as illustrated in Figure 5b (UVL tap above UVH tap for a larger hysteresis) and Figure 5c (UVL tap below UVH tap for a smaller hysteresis). In the latter case, the LTC4261 does not allow hysteresis to drop to zero or negative values if a larger RH is used. Instead, hysteresis reaches a guaranteed minimum (15mV typical) and increases with increasing RH, preventing comparator oscillation. Versatile On/Off Control The LTC4261 provides various methods of on/off control using the ON, EN, UVH/UVL/OV, PGIO or FLTIN pins along with the I2C interface. Turning on or off the pass transistor can be either autonomous or controlled by the system host through the I2C interface. Furthermore, the LTC4261 may reside on either the removable board or on the backplane. Even when operating autonomously, the host can exercise control over the GATE output through I2C, although EN and ON could subsequently override conditions set by I2C 1M INTVCC INTVCC LTC4261 VEE UVH –48V INPUT –48V INPUT –48V INPUT INTVCC ON R2 15k 1% 0V R1 11.8k 1% LOOP OR SWITCH 47k RH 1.91k 1% UVH R2 15k 1% 5V 1k UVL RH 1.91k 1% R2 16.9k 1% The LTC4261 provides two UV pins (UVH and UVL) that can be used to precisely set the undervoltage threshold and hysteresis. Each of the two pins has an accurate threshold: 2.56V for UVH rising (turn-on) and 2.291V for UVL falling (turn-off), and both pins have a small, built-in hysteresis of 15mV. With either a rising or falling input voltage, both the UVH and UVL pins have to cross their thresholds for the comparator output to change state. If both pins fall below their thresholds, an undervoltage fault is registered. The ratio between the UVH and the UVL thresholds is designed to precisely set 43V turn-on and 38.5V turn-off UV thresholds popular in telecom applications with minimum external components, by tying UVH and UVL together. Along with the OV pin (with a threshold of 1.77V and a hysteresis of 37.5mV), the 3-resistor divider shown in Figure 5a sets an accurate operating range of 43V to R3 453k 1% R3 453k 1% R3 453k 1% Adjustable Undervoltage Comparator Offers Both Precision and Flexibility –48V RTN –48V RTN –48V RTN 100k LTC4261 EN ON –48V 10nF –48V a. Optoisolator control EN EN VEE LTC4261 VEE ON –48V c. Contact debounce delay upon insertion for use with an ejector switch or loop-through style connection sense b. Logic control –48V RTN 453k INTVCC INTVCC SDAO ON LTC4261 SDAI SCL VEE EN UVL LTC4261 ON 28.7k UVH VEE EN –48V INPUT –48V d. Short pin connection sense to RTN e. I2C-only INTVCC SDAO ON LTC4261 SDAI SCL VEE EN I2C I2C –48V control (default off) f. I2C-only control (default on) Figure 6. The LTC4261 provides versatile on/off control 16 Linear Technology Magazine • March 2006 DESIGN FEATURES L Broadcast Mode Saves Cost and Space To facilitate I2C communication between the LTC4261 and a system host that are isolated from each other, the SDA signal is split into SDAI and SDAO. Separate pins allow the device to drive optoisolators with a minimum number of external components. Still, three optoisolators (two for inputs on SCL and SDAI, and one for output on SDAO) are needed for typical I2C operations. To further reduce component count, the LTC4261 provides a special single-wire mode that only uses the SDAO pin to continuously transmit ADC data and fault information (Figure 7). This simple communication mode saves component cost and board space by eliminating two optoisolators and is useful for applications where only monitoring is needed. The single-wire broadcast mode is simply enabled by tying the ADR1 pin to INTVCC and the ADR0 pin to VEE (Figure 7). At the end of conversion of each ADC channel, a serial data stream is sent out to SDAO with a fixed data rate of 15.3kHz ±20% in the format shown in Figure 8. The data stream Linear Technology Magazine • March 2006 –48V RTN 6× 0.51k IN SERIES 1/4W EACH 1µF 0.1µF VEE 5V 5.1k INTVCC VIN ADR1 SDAI ADR0 LTC4261 SCL 5.1k SDAO V DIN DD MICROCONTROLLER –48V INPUT Figure 7. The LTC4261 in single-wire broadcast mode reports ADC data and fault information using a single optoisolator and the SDAO signal. INTERNAL CLK PRTY OV PRTY PRTY UV 0C OV OV UV UV OC 0C ADC0 .. .. .. ADC0 ADC0 ADC9 ADC9 CH0 SDAO CH0 ADC9 CH1 CH0 START DMY CH1 DATA CH1 command. Card insertion/extraction can be conveniently detected with the EN-changes-state bit in the FAULT register. After power-up, UV, OV and other fault conditions seize control as needed to turn off the GATE output, regardless of the state of EN, ON or the I2C port. Auto-retry following the faults can be enabled or disabled by the host at any time and the configurations are stored in the CONTROL register (Table 1). Although PGIO (when configured as an input) and FLTIN control nothing directly, they are useful for I2C monitoring of connection sense or other important signals. The host can then use the information detected by these two pins to take action. Figure 6 shows some examples for on/off control using the LTC4261. The circuits in Figures 6a to 6d work equally well in both backplane and board resident applications. Circuits in Figures 6e and 6f are for I2C-only control. START Figure 8. Data format of the single-wire broadcast mode consists of a start bit (STAT), a dummy bit (DMY), two bits of ADC channel labeling (CH1 and CH0), ten bits of ADC data (ADC9:0), three fault bits (OC, UV, OV) and a parity bit (PRTY). The data are encoded with an internal clock in a way similar to Manchester encoding that can be easily decoded by a microcontroller. RTNA RTNB F1 D1 F2 D2 D3 D4 R1 150k R2 301k Which of the Four Fuses is Open? Some high availability systems such as AdvancedTCA require dual feeds on both –48V and RTN lines and a fuse on each feed, resulting in four fuses. Since the plug-in card is designed to operate without interruption even if one of the fuses is open, it can be dif- R4 80.6k RIN 4 × 1000Ω in series R5 2k CIN 1µF R3 3.4k VIN ADIN2 ADIN LTC4261 INTVCC −48V A −48V B R6 100k R8 100k R7 10k R9 10k PGIO FLTIN F3 D5 F4 D6 SCL SDAI SDAO VEE I2C/SMBUS ALERT CVCC 0.1µF Figure 9. LTC4261 makes it possible to determine which of the four fuses is open in an AdvancedTCA system. 17 L DESIGN FEATURES Table 2. The voltages at ADIN and ADIN2 (referred to VEE) indicate which RTN fuse is open in the 4-fuse monitor in Figure 9 VADIN = VADIN2 = 0V Both F1 and F2 are open 0.25 • VADIN2 ≤ VADIN < 0.7 • VADIN2 F1 is open 0.7 • VADIN2 ≤ VADIN < 1.1 • VADIN2 F2 is open VADIN > 1.1 • VADIN2 Normal Operation ficult, without direct fuse monitoring, to detect a single open fuse. The integrated ADC, I2C monitoring, and general-purpose pins of the LTC4261 make it possible to monitor the individual fuses, as shown in Figure 9. Fuses on the RTN side are sensed using resistors R1, R2 and R3 with the output measured by the ADC at the ADIN pin. At the same time, the input voltage after the ORing diodes is sensed using R4 and R5 and measured at the ADIN2 pin. Diodes D3 and D4 are used to compensate the ORing LTC4304, continued from page 12 nection between SDAIN and SDAOUT, and SCLIN and SCLOUT is broken, isolating the problem device from the bus. FAULT pulls low indicating a stuck bus. At this time it is assumed that the MCU and the device it is communicating with are out of sync. The MCU sent out all of the clocks necessary for the transaction, but the device is still in the middle of the transaction. The device is waiting for more clocks to finish putting its data on the bus, and the last bit it put on the bus happened to be a low, hence it holds the bus low until it gets more clocks (bus is stuck). Because the connection is broken, the problem device is now isolated from the I2C bus, and the rest of the system is free to resume normal operation. After the connection is broken, the LTC4304 automatically generates up to 16 clock pulses at 8.5kHz on SCLOUT, enough clocks to clear the internal register on the problem device. The device could be cleared with one or any number of clocks (up to 16) depending on how the fault occurred. At anytime, if SDAOUT and SCLOUT go high, FAULT is released to go high and the automatic clocking is stopped, and a connection is automatically enabled. 18 is open, PGIO is pulled high and the PGIO input bit in the FAULT register is set. If F4 is open, FLTIN is pulled low and the FLTIN bit is set. In the latter case, if the corresponding bit in the ALERT register is also set, the LTC4261 interrupts the host using the ALERT pin. diodes. Table 2 shows how the voltages at ADIN and ADIN2 indicate which fuse on the RTN side is open. Using the input voltage as the reference ensures valid detection in the full operating range. Fuse monitoring on the –48V side is more straightforward using the INTVCC pin along with two logic input pins, PGIO (when configured as input) and FLTIN as shown in Figure 9. Fuse F3 is sensed with R6 and R7 at PGIO, and fuse F4 is sensed with R8 and R9 at FLTIN with inverted input. If F3 Conclusion When the slave that was stuck sees a START bit after connection, it will abort the stalled communication and reset. The fault is cleared independently of the MCU. If the fault cannot be cleared, the offending circuit remains isolated from the system. Figure 2 shows an example of a stuck bus being resolved with automatic clocking. enable all four rise time accelerators. Finally, in Figure 3c, VCC is greater than the pull-up voltages of both sides. ACC is connected to VCC to disable all four accelerators. Supply Independence and Level Translation The LTC4304 can be the bridge between a backplane operating at one voltage and a card operating at a different voltage. Figure 3 shows some typical configurations. Notice that VCC, the pull-up voltage on SDAIN and SCLIN and the pull-up on SDAOUT and SCLOUT are independent of each other. The rise time accelerators must be disabled when the bus pull-up voltage is lower than VCC. Figure 3a shows a situation where the VCC is higher than the pull-up voltage on SDAIN and SCLIN. In this case ACC must be floating, disabling the rise time accelerators on the inputs only. In Figure 3b, VCC is equal to the output side pull-up voltage, and less than input side pull-up voltage. ACC is connected to GND to The LTC4261 is a full-featured, intelligent Hot Swap controller that enhances the reliability and durability of high availability –48V power systems. The internal 10-bit ADC, I2C/SMBus and registers make it easy to monitor faults and real-time power and communicate with the system host. Its unique inrush and overcurrent control technique minimizes stresses on the pass transistor in all operating conditions. L Conclusion The LTC4304 is a multifunctional device with an impressive number of features designed to increase the reliability of an I2C bus. The flexible architecture allows the LTC4304 to be configured into almost any system. The LTC4304 isolates and resolves stuck buses independently of the MCU. Capacitance buffering, level translation, Hot Swap features and ±15kV human body ESD protection make the LTC4304 an ideal solution for any I2C application. The LTC4304 is offered in small 10 pin MSOP and DFN (3mm × 3mm) packages. A feature-reduced version of the LTC4304 is also available. The LTC4303 provides all of the functionality of the LTC4304 with the exception of the FAULT output flag and ACC control pin. The LTC4303 is a drop in replacement for the LTC4300A-1 and its rise time accelerators are permanently enabled. L Linear Technology Magazine • March 2006 DESIGN FEATURES L High Efficiency Nickel Charger Operating to 34V is Easily Configured to Deliver a Safe Fast Charge by James A. McKenzie Introduction Accurate, Rock-Solid Fast Charge Termination The LTC4011 implements a complete fast charge control algorithm, best suited for charge rates of C/2 to 2C 100 95 EFFICIENCY (%) SINGLE CELL VOLTAGE (V) DCIN = 5.5V 90 DCIN = 20V 85 80 75 70 1.65 36 1.60 34 1.55 32 SINGLE CELL VOLTAGE 1.50 1.45 30 28 BATTERY TEMPERATURE 1.40 26 1A 1.35 65 BATTERY TEMPERATURE (°C) The LTC4011 is a complete standalone nickel chemistry fast charger that operates at high efficiency with an input voltage of up to 34V, even with a battery voltage that is much less than the DC input supply. Typical efficiency with a 5.5V and 20V input supply is shown in Figure 1. An undervoltage lockout of 4.25V ensures stable operation with an input as low as 5V, assuming a 10% tolerance. The IC offers a full range of charge control features that are easy to program with a minimal number of external components. Its multiple levels of safety features provide reliable fault protection that is also simple to configure. The LTC4011 functions without the need for any host Microcontroller Unit (MCU) support and requires no software of firmware. 24 CHARGE CURRENT 60 0 2 4 1.30 6 8 10 12 14 16 18 20 BATTERY VOLTAGE (V) 0 20 40 60 22 80 TIME (MINUTES) Figure 1. The LTC4011 charging efficiency with 2A output current. Figure 2. A 1C NiCd charge cycle for NiMH batteries and C/3 to C, or higher, for NiCd batteries. This means fully discharged batteries can be recharged to 100% capacity in one hour or less. Reliable charge termination at these higher rates is controlled by well-proven voltage and temperature detection techniques tailored to the selected nickel chemistry to ensure minimum charge time without degradation of recharge cycle life. Various internal and external filters, both analog and digital, eliminate spurious noise in the voltage and temperature data channels used to determine both general battery health and appropriate charge termination. Monitoring Battery Voltage for –ΔV The primary fast charge termination technique employed by the LTC4011 involves detection of peak battery 10k FROM ADAPTER 4.5V TO 34V 3k R 6V 3k INFET FAULT CHRG TOC READY R VCC TGATE BGATE LTC4011 PGND 20k TO SYSTEM 20µF LOAD DCIN TIMER 49.9k SENSE 33mΩ BAT GND VCDIV CHEM VCELL VRT INTVDD VTEMP 0.1µF 10µH 10k 0.033µF R2 20µF 0.068µF 28k 9.76k NiMH PACK WITH 10k NTC (3Ahr) Figure 3. A 3A NiMH charger with full PowerPath control Linear Technology Magazine • March 2006 19 L DESIGN FEATURES voltage followed by a sufficient drop in that terminal potential (–ΔV detection). The typical NiCd charge cycle shown in Figure 2 is terminated by –ΔV. To support a wide range of battery pack configurations, the IC senses the average cell voltage of the battery pack. A simple resistive voltage divider between the positive battery terminal and the VCDIV pin, along with some capacitance to eliminate residual PWM switching noise, is all that is required to provide this average cell voltage to the VCELL input pin. This technique, combined with the 5V to 34V input operating range, allows charging of from 1 to 16 in-series nickel cells. A negative delta of only 10mV to 20mV per cell indicates the battery is fully charged. The VCELL voltage is also monitored as a measure of basic battery health and to detect catastrophic fault conditions, as described below. Monitoring Battery Temperature for ΔT/Δt The LTC4011 can also process battery temperature using information provided by a negative temperature coefficient (NTC) thermistor. The thermistor should be in good thermal contact with the cell casings located most near the center of mass of the battery pack. This thermistor is then included in a voltage divider network between the VRT pin and ground, as shown in Figure 3, to provide a linearized input to the VTEMP pin. This configuration is flexible enough to support a very wide range of NTC thermistor types. An external analog single-pole passive filter is recommended to eliminate PWM switching noise. The voltage on the VTEMP pin is then used by the LTC4011 to qualify the charge process according to an acceptable range, roughly 0°C to 45°C. In addition, the internal data acquisition subsystem uses an on-board real-time clock to monitor the rate of temperature increase of the battery (ΔT/Δt) during fast charge. Values between 1°C/min and 2°/min typically indicate a fully charged nickel battery. Figure 4 shows a NiMH charge cycle at a 1C rate with the fast charge 20 portion of the charge cycle terminated by ΔT/Δt. Temperature processing is optional on the LTC4011 for all chemistries. Simply tying VTEMP to VRT disables all temperature-based charge qualification and termination. Chemistry-Specific Battery Profiling The basic charging algorithm applied by the LTC4011 is modified for the selected battery chemistry. Tying the CHEM pin to ground selects NiMH charging parameters, while leaving the CHEM pin open or tying it to VRT selects NiCd charging parameters. While similar, these chemistries do require slightly different fast charge termination to ensure maximum recharge cycle life. The older NiCd chemistry benefits from slight overcharge. When charging NiCd cells, the LTC4011 uses threshold levels that favor –ΔV termination, resulting in a final charge of slightly more than the rated capacity (100%) of the pack. In this case, ΔT/Δt termination, if enabled, serves as a secondary termination technique for additional safety. This is shown in the example of Figure 2. Newer NiMH batteries are normally designed to accept higher charge rates than their older NiCd cousins, but manufacturers often warn against any amount of overcharge. So, for NiMH cells, the LTC4011 selects internal thresholds that favor ΔT/Δt termination at a point when the pack is charged to about 95% capacity. In order to avoid false termination on highly discharged cells that have been inactive for a long period of time, the IC can vary the ΔT/Δt limit as the fast charge cycle progresses. The –ΔV limit then serves as secondary termination (safety), and the IC applies a timed top-off charge after fast charge termination to achieve 100% capacity, as shown in Figure 4. Obviously, while optional, use of a thermistor input for NiMH batteries is strongly recommended. In addition to these chemistryspecific measures, the LTC4011 applies some generic charge profile techniques. Battery open-circuit voltage is measured at the beginning of a new charge cycle to determine the state of charge of the attached pack. If the pack is initially heavily discharged, the IC applies a smaller conditioning current for a fixed period of time to recover the battery to a point of suitable fast charge acceptance. If the pack is initially discharged, the LTC4011 applies a –ΔV termination hold-off period to allow the internal chemistry—and hence terminal potential—to stabilize after applying full charge current. This avoids premature termination. However, if the pack is already moderately charged, the initial terminal voltage is well-behaved and –ΔV processing begins immediately to avoid accidental overcharging of the battery. If enabled, ΔT/Δt detection is always active. Automatic Recharge Keeps Batteries Ready for Use Nickel batteries exhibit a high selfdischarge rate of up to 3% per day. Once a charge cycle has completed, the LTC4011 continues to monitor the open-circuit terminal voltage of the battery for as long as an input power source is connected. If the battery volt- Table 1. LTC4011 time limit programming examples RTIMER (Ω) Typical Fast Charge Rate Precharge Limit (minutes) Fast Charge Stabilization (minutes) Fast Charge Limit (hours) Top-Off Charge (minutes) 24.9k 2C 3.8 3.8 0.75 15 33.2k 1.5C 5 5 1 20 49.9k 1C 7.5 7.5 1.5 30 66.5k 0.75C 10 10 2 40 100k 0.5C 15 15 3 60 Linear Technology Magazine • March 2006 1.60 42 1.55 40 1.50 38 SINGLE CELL VOLTAGE 1.45 BATTERY TEMPERATURE 1.40 1.35 CHARGE CURRRENT 1.30 1.25 36 0 20 60 40 TIME (MINUTES) 34 2A 32 1A TOP OFF 30 80 BATTERY TEMPERATURE (°C) CELL VOLTAGE (V) DESIGN FEATURES L 28 100 Figure 4. Typical NiMH charge cycle at 1C age indicates a loss of more than about 15% capacity, a refresh fast charge cycle is initiated to bring the stored energy level back to 100%. The duration of the recharge cycle is normally only a few minutes. This technique replaces more traditional continuous trickle charge methods. Trickle charge operates the battery in a constant state of overcharge, which can reduce the cycle life of some NiMH cells, generates continuous heat, and is somewhat less efficient that the LTC4011 automatic recharge approach. Multiple Safety Features Many safety features are built into the LTC4011. It monitors important voltage and temperature parameters during all charging phases. If the VTEMP input has been enabled, the sensed temperature is required to lie between 0°C and 45°C, or charging is suspended. After fast charge begins, the battery temperature is allowed to rise to 60°C. If this limit is exceeded, however, the sensed temperature must fall below 45°C before charging can resume. The LTC4011 also tracks its own die temperature and disables charging if it rises above an acceptable limit. Charging is not allowed to begin until the necessary voltage headroom to operate the PWM (about 500mV) has been established between the VCC and BAT pins. The battery voltage is also continuously monitored for overvoltage. If the average cell voltage on VCELL exceeds 1.95V, charging is disabled and a fault is indicated. The LTC4011 also profiles the battery voltage during charging to ensure Linear Technology Magazine • March 2006 proper charge acceptance, checking open-circuit voltage at the beginning of precharge and fast charge, and in-circuit voltage after about 20% of the fast charge cycle has been completed. Finally, the LTC4011 contains a safety timer that limits the length of time any single charge can continue. This timer is easily programmed with an external resistor connected between TIMER and GND according the formula R TIMER (Ω) = tMAX (hours)/30µs. Timing Can be Everything The LTC4011 uses the programming of the safety timer to a variety of purposes by inferring a relation between the set time limit, the programmed charge current and the capacity of the battery being charged. It assumes the period of this timer is set to 150% of the time normally required to deliver 100% charge to a fully discharged battery at the programmed current of RSENSE/100mV. The safety checkpoints discussed above, along with top-off charge duration, are then determined by intermediate intervals of the safety timer. Table 1 shows suitable values for R TIMER for a range of programmed fast charge rates. A Smaller PWM Solution The LTC4011 also embodies a complete PWM controller. Its buck regulator uses a synchronous pseudoconstant off-time architecture with high-side PFET power switch. This choice yields a PWM that is extremely easy to configure with a minimum number of external parts. Simply connect the external PFET power switch, optional NFET synchronous diode, the Schottky clamp and choke as shown in Figure 3. No external loop compensation is required, and the charge current is set by a single resistor connected between the SENSE and BAT pins. This resistor is in series between the inductor output and the battery with a value determined by the equation RSENSE = 100mV/IPROG. The LTC4011 PWM uses a unique floating LV differential architecture to deliver 5% current accuracy and low cycle-to-cycle jitter with high inductor ripple current. That in turn allows Figure 5. NiCD/NiMH charger layout use of space-efficient magnetics and a smaller output filter capacitor. The pseudo-constant off-time architecture eliminates the need for cumbersome slope compensation and allows full continuous operation over a wide VIN/VOUT range without generation of audible noise, even when using ceramic capacitors. Typical operating frequency is 550KHz. An example of the LTC4011 with a 2A PWM implementation is shown in Figure 5. PowerPath Control PowerPath™ control is a vital part of proper termination when charging nickel batteries. Because the differences being sensed for –ΔV termination are so small, the series resistance of the battery can easily cause premature termination if varying load current is drawn from the battery during charging. The LTC4011 provides integrated PowerPath support for an input PFET transistor between the DC input (DCIN) and the host’s unregulated system supply (VCC). This FET then acts as an ideal rectifier with a regulated forward drop as low as 50mV, requiring less operating head room and capable of producing less heat than a conventional blocking diode. The LTC4011 can provide up to 6V of gate drive to this pass device. Select an input FET with a low enough RDS(ON) at this gate drive level so that the combination of full charge current and full application load current does not cause excessive power dissipation. As shown in Figure 4, the PFET between BAT and VCC then serves to automatically disconnect the battery from the system load as long as a DC input is present. A Schottky diode continued on page 23 21 L DESIGN FEATURES Tiny Controller Makes It Easy to by David Ng Charge Large Capacitors Introduction Emergency warning beacons, inventory control scanners, professional photoflash and many other systems operate by delivering a pulse of energy to a transducer. This energy typically comes from a large capacitor that has been charged to some predetermined voltage. The LT3750 is a current-mode flyback controller optimized for charging large value capacitors to a user selected target voltage. This target voltage is set by the turns ratio of the flyback transformer and just two resistors in a simple, low voltage network, so there is no need to connect components to the high voltage output. The charging current is set by an external sense resistor and is monitored on a cycle-by-cycle basis. The LT3750 is available in a space saving MSOP-10 package. The device is compatible with a wide range of control circuitry, being equipped with a simple interface consisting of a CHARGE command input bit and an open drain DONE status flag. Both of these signals are compatible with most digital systems, yet tolerate voltages as high as 24V. Simple 300V, 400µF Charger Figure 1 shows the schematic of a LT3750 circuit that charges a 400µF capacitor to a target voltage of 300V. The 1:10 turns ratio of T1 and the R1, R2 resistors set the target voltage to 300V, while the R4 power resistor sets the peak charging primary current to 7.5A. Operating from a 12V power source, this circuit charges the 400µF capacitor to 300V in 1.04 seconds, as shown in Figure 2. Design Considerations The architecture balances a high degree of integration with flexibility—leaving key parameters definable by the user. The important issues to 22 T1 1:10 + C1 150µF VCC 12V C3 10µF VCC DONE VTRANS RVOUT DONE C2 4, 5 22µF ×3 60.4k D1 VOUT 300V 1 • • 6, 7 + CLOAD 400µF 10 40.2k RDCM LT3750 OFF ON CHARGE GATE M1 SOURCE GND RBG 2.49k C1: SANYO 35CV150AXA C2: TDK C4532X7R1E226M (X3) C4: RUBYCON 330FK400V22X38 C3 47pF 10mΩ D1: DIODES INC. MURS160 M1: PHILIPS PH20100S T1: TDK DCT20EFD-UXXS003 Figure 1. Simple circuit charges a 400µF capacitor. consider in completing a design are input capacitor sizing, transformer design and output diode selection. Power Stage Input Capacitor Every switching cycle, the LT3750 measures the voltage at its RVOUT pin to determine the transformer, T1, flyback voltage. It also measures the signal at its VTRANS pin, which is the voltage at the input of the power switching stage. The difference of these two signals, accounting for the T1 transformer turns ratio and the D1 rectifying diode, yields the output voltage. In order to get an accurate result, it is important that the signal at the LT3750’s VTRANS input optimally reflects the DC potential of the power stage input. Consequently, the capacitance at the input of the power switching stage must be chosen such that the ripple voltage at the VTRANS input is not excessive. The capacitor bank in the circuit represented by Figure 1 is actually made of five capacitors. C1 is a single 150µF electrolytic capacitor to provide bulk energy, C2 is three low ESR 22µF ceramic capacitors to accommodate the high switching currents, and C3 is low ESR 10µF ceramic capacitor that provides local decoupling to the LT3750. For best results, place C1 and C2 as close as possible to T1, and C3 VOUT 50V/DIV VOUT 50V/DIV IIN 1A/DIV IIN 1A/DIV 200ms/DIV Figure 2. Output voltage and input current waveforms of LT3750 capacitive charging circuit 200ms/DIV Figure 3. A transformer with a 51µH primary inductance has a longer charge time and larger input current ripple than a transformer with a 10µH primary inductance. Linear Technology Magazine • March 2006 DESIGN FEATURES L as close as possible to the VTRANS pin on the LT3750. Transformer Other than the turns ratio, there are two issues to remember when selecting a transformer. The first is that the transformer secondary must be constructed to withstand potentials of both the positive and negative voltages associated with charging the capacitor. This withstand voltage is not the same as the isolation voltage rating. In the case of the circuit shown in Figure 1, there is no isolation voltage requirement, as the primary and secondary of T1 are tied to the same ground reference. The secondary winding, however, is subjected to the output potential, or 300V, and care must be taken in selecting for parameters relevant to such high voltages, such as pin spacing and wire insulation. The other transformer parameter to keep in mind is the primary inductance. The primary inductance determines the operating frequency range, input current ripple and core loss, all of which contribute to the capacitor charge time and efficiency. The charging profile shown in Figure 2 is for a circuit using a transformer with a primary inductance of 10µH. Figure 3 shows the charging profile for the same circuit, but the primary inductance is much larger, 51µH. Note that the 51µH transformer has a longer charge time than the 10µH transformer. LTC4011, continued from page 21 can be used in place of this MOSFET, if the additional voltage drop can be tolerated when running the application from the battery. Micropower Features Support Extended Battery Use The LTC4011 has a typical shutdown supply current of 3µA and normally draws much less than 1µA from the SENSE/BAT pin combination when the DC input supply has been removed. In addition, the VCDIV pin offers a means of disconnecting the Linear Technology Magazine • March 2006 Table 1. Summary of input charging current and charge times for LT3750 circuits for transformers with different primary inductances Transformer Primary Inductance Input Charging Current Input Charging Current Ripple Charge Time TDK DCT20EFD-UXXS003 10µH 2.2A 0.5A 1040ms Coiltronics CTX02-17314 15µH 2.2A 0.4A 1000ms Coiltronics CTX02-17144 51µH 2A 1.4A 1120ms Table 1 gives a summary of the input charging current and charge times for LT3750 circuits for three different T1 primary inductances, with the 15µH device giving the best result. Output Diode Finally, it is important to consider the high AC voltages when selecting the output rectifying diode. The circuit in Figure 1 has a 300V output, but the output rectifying diode must withstand the sum of the output voltage and the voltage across the transformer secondary when the MOSFET Q1 is on. In this case, that is about 500V. This is a high voltage, but there are many manufacturers that produce switching diodes suitable for this application. While it is important to minimize board space, the designer must choose a device that does not cause a violation of the spacing requirements for both safety and producibility. According to VCELL resistor divider in the shutdown state, eliminating the current drawn by that circuit when charging is not possible. These features greatly minimize the load applied to the battery by the charger when disconnected from a DC power source, increasing the useable run time of the battery in portable applications. Conclusion The LTC4011 is a nickel chemistry charger that integrates a complete high voltage PWM controller, allow- table 6-1, “Electrical Conductor Spacing,” of IPC-2221, Generic Standard on Printed Circuit Board Design (February 1998 release), the minimum spacing between conductors that have a potential up to 500V must be no less than 2.5mm on an uncoated printed circuit board operated below an altitude of 3050m. The output diode must be chosen to ensure that the minimum spacing between the diode pads is at least 2.5mm. The circuit shown in Figure 1 uses a MURS160, which is offered by a number of manufacturers such as Diodes Inc and Vishay. It is an ultrafast recovery rectifier and has a peak repetitive reverse voltage rating of 600V. The diode comes in an SMB package, which allows the edge-toedge separation between the pads to be as much as 3mm. L Authors can be contacted at (408) 432-1900 ing it to efficiently charge batteries from a 34V input without the need for additional current source control ICs. True standalone operation and flexible control greatly simplify charger design. The PWM operates at a high frequency, enabling the use of surface mount components to save space. Reliable, robust charge termination algorithms backed by solid safety features make the LTC4011 an excellent choice for a wide range of fast charge implementations, providing long life for rechargeable nickel batteries. L 23 L DESIGN FEATURES MOSFETs Make Sense for Tracking and Sequencing Power Supplies by Thomas DiGiacomo Introduction: Charge MOSFETs with the Task of Supply Control In electronic systems with multiple supplies, the need for tracking and/or sequencing is well established. On the one hand, core and I/O power might be required to ramp up and down together with less than a diode’s voltage drop between them to avoid potentially destructive latch-up. Coincident tracking, as shown in Figure 1a, solves this problem. On the other hand, in a distributed supply chain, some supplies might need to be fully operational before others. Supply sequencing, as shown in Figure 1d, solves that problem. Other systems may require simultaneous completion of supply ramps (Figure 1b), voltage offsets or time delays (Figure 1c), or combinations of such profiles. Linear’s line of no-MOSFET tracking and sequencing control products, the LTC2923, LTC2925, and LTC2927, work outstandingly well with DC/DC converters and other supply generators that allow access to the feedback nodes that set their output voltage. In many applications, however, MOSFET control of power supply tracking and sequencing is necessary. Supply modules provide no access to their feedback node, and some linear regulators also resist the current-injection control method employed by the LTC2923 family. MOSFET-controlled tracking and sequencing can improve power system segmenting and allow reuse, which reduces parts count and board area. A single-output supply generator can power different rails of the same voltage (e.g., analog power, digital power, and housekeeping power) because each rail’s tracking profile can be set independently. Multiple-output modules can replace several single-input modules without the need for complex ON/OFF pin signaling to implement 24 a. Coincident tracking b. Ratiometric tracking c. Offset tracking d. Supply sequencing Figure 1. Power supply voltage tracking profiles sequencing or tracking. Furthermore, a series MOSFET can be shut off, which guarantees that the load is disconnected when so desired. The LTC2926 MOSFET-controlled power supply tracker provides exceptionally flexible control of supply tracking and sequencing that can realize all of the profiles in Figure 1, and combinations of them. Each of two “slave” supplies can be configured independently to track a “master” ramp signal using just an N-channel MOSFET and a few resistors per supply. A single capacitor sets the slope of a voltage ramp that may be employed Linear Technology Magazine • March 2006 DESIGN FEATURES L 1.8V MODULE as the master ramp signal (Figure 2), or may be used to ramp a third supply using an external MOSFET (Figure 3). The LTC2926 is interoperable with Linear’s no-MOSFET tracking and sequencing products, and even offers no-MOSFET control itself in some applications (Figure 4); see “Direct Supply Generator Control: ¡No Más FETs, Mi Amigo!” in this article. The LTC2926 also features automatic remote sense switching that compensates for voltage drops across the controlling MOSFETs, and two I/O signals that transmit tracking status and receive control input from upstream and downstream devices. The LTC2926 is available in 20-lead DFN (4mm × 5mm) and 20-lead narrow SSOP packages. Q1 IN OUT VIN RX1 SENSE 10Ω 2.5V MODULE Q2 IN OUT VIN NC SENSE 10Ω RSGATE MGATE RAMP SGATE1 SGATE2 D1 D2 VIN 0.1µF VCC S1 RAMPBUF S2 RFB1 RTB1 TRACK1 FB1 LTC2926 RTA1 RTB2 RFA1 RFB2 TRACK2 FB2 VIN RTA2 VIN 10k FAULT FAULT ON/OFF Linear Technology Magazine • March 2006 2.5V SLAVE2 CMGATE RX2 How It Works: Injection Controls Your Ramp-age and Keeps You on Track The LTC2926 achieves supply tracking and sequencing by influencing the feedback node that sets a supply voltage, as do the LTC2923, LTC2925, and LTC2927. In all four products, a tracking cell converts the master ramp voltage into a ramping current that is injected into the aforementioned feedback node. Whereas the latter products control supply generators themselves (those with accessible feedback nodes, like DC/DC converters), the LTC2926 controls rudimentary voltage regulators whose inputs are the supply voltages and whose outputs are the tracked and sequenced supply rails. In Figure 5, the integrated gate controller cell, the external N-channel power MOSFET (QEXT), and a resistive voltage divider (RFA and RFB) form the basic voltage regulator. In regulation, the slave supply voltage equals the reference voltage times (1 + RFB/RFA). In drop out mode, the MOSFET becomes a closed switch, and the slave supply voltage equals the input supply voltage. The injection of current at the feedback node of the gate controller regulator reduces the effective value of its reference voltage. As the master ramp rises, the fixed ratio of 1.8V SLAVE1 RFA2 10k ON PGTMR GND STATUS/PGI STATUS CPGTMR Figure 2. Typical 2-supply tracking application. The master ramp signal is created by connecting a capacitor from the MGATE and RAMP pins to ground. 3.3V MODULE Q0 IN OUT VIN RX0 SENSE Q3 10Ω 1.8V MODULE VIN MASTER Q1 IN OUT 1.8V SLAVE1 RX1 SENSE 10Ω 2.5V MODULE Q2 IN OUT VIN 2.5V SLAVE2 CMGATE RX2 SENSE 10Ω RSGATE MGATE RAMP SGATE1 SGATE2 D1 D2 VIN 0.1µF VCC S1 RAMPBUF S2 RFB1 RTB1 TRACK1 RTB2 FB1 LTC2926 RTA1 RFA1 RFB2 TRACK2 VIN FB2 RTA2 VIN 10k FAULT FAULT ON/OFF ON RFA2 10k GND PGTMR STATUS/PGI STATUS CPGTMR Figure 3. Typical 3-supply tracking application. MOSFET Q0 creates a ramping master supply that doubles as the master ramp signal. 25 L DESIGN FEATURES SUPPLY MODULE VIN Q1 IN OUT SLAVE1 RX1 SENSE 10Ω SUPPLY MODULE VIN MASTER IN OUT RX0 SENSE VIN 10Ω CMGATE 0.1µF VCC MGATE S2 RAMP SGATE1 ON D1 VCC 0.1µF RAMP S1 IN RAMPBUF RFB1 RTB1 TRACK1 FB1 LTC2926 RTA1 RTB2 VIN LTC2927 D2 VIN SDO RAMPBUF FB RTB3 OUT FB SLAVE3 TRACK RFA1 GND RTA3 TRACK2 DC/DC RFA3 RFB3 FB2 RTA2 VIN ON/OFF ON GND IN PGTMR DC/DC CPGTMR OUT FB RFA2 SLAVE2 RFB2 Figure 4. LTC2926 and LTC2927 4-rail application. The second slave channel of the LTC2926 requires no MOSFET in this example. the feedback resistors multiplies the increasing reference voltage to create a rising slave supply voltage that is limited by the input supply voltage at the drain of the MOSFET. With proper selection of the feedback resistor ratio, the gate controller cell drives the SGATE pin to VCC + 5V when the slave supply reaches it maximum. The logic-level MOSFET becomes a simple closed switch, able to pass input supply voltages of from 0V to VCC. The relationship between the master ramp and the slave voltage is called the tracking profile, and it is a function of the input supply voltage, the master ramp voltage, the track resistors (R TA, R TB), and the feedback resistors (RFA, RFB). All of the profiles in Figure 1 can be realized by properly selecting the track and feedback resisQEXT INPUT SUPPLY TRACKING CELL LTC2926 GATE CONTROLLER CELL VCC + RTB RTA LOAD VCC + 5V 0.8V – MASTER RAMP 0.8V + – 10µA SGATE 10µA TRACK RFB FB ITRACK SLAVE SUPPLY IFB RFA Figure 5. Simplified tracking cell and gate controller cell combination 26 Linear Technology Magazine • March 2006 DESIGN FEATURES L tors. Combinations are also possible because each channel’s profile is set independently. MOSFETs Make Remote Sense, Too: Whip a Problem with a Switch Even with the selection of a low RDS(ON) MOSFET as the tracking control device, the load current causes a voltage difference between the supply generator and the load. Worst yet, dynamic load current produces a dynamic voltage error. Without modification, the standard solution, to let the supply generator remotely sense the load voltage and compensate for the drop, does not work. During tracking up or down, the load voltage is deliberately SUPPLY MODULE + VDS – Q0 OUT RX SENSE LOAD MGATE OR SGATE Q3 RSGATE Figure 6. Functional schematic for remote sensing the load Tracking and Sequencing Supply Rails in Three Easy Steps Any of the profiles shown in Figure 1 can be achieved by using the following simple design procedure. Figure 3 shows a basic 3-supply application circuit. 1. Set the ramp rate of the master signal. Solve for the value of CMGATE based on the desired ramp rate (volts per second) of the master ramp signal, SM, and the MGATE pull-up current. CMGATE = IMGATE SM (1) where IMGATE ≅ 10µA If the gate capacitance of the MOSFET is comparable to CMGATE, reduce the value of CMGATE to account for it. If the master ramp signal is not a master supply, tie the RAMP pin to the MGATE pin 2. Choose the feedback resistors based on the slave supply voltage and slave load. It is important that the feedback resistors are significantly larger than the load resistance. Determine the effective slave load resistance, RL (not shown), to satisfy the following equation: RFB ≥ 100 • RL (recommended), RFB ≥ 23 • RL (required) (2) The LTC2926 must be able to fully enhance the slave control MOSFET at the end of ramping. Select RFA based on the resistor tolerance, TOLR, and the absolute maximum slave supply voltage, VSLAVE(max): 1− TOL R RFA < RFB • 1+ TOL R V (max) SLAVE − 1 VFB(REF ) (min) (3) where VFB(REF)(min) = 0.784V Note: Design with a value of VSLAVE(max) that covers the maximum possible slave supply voltage by a good margin. If the slave generator exceeds that voltage during operation, an overvoltage shutdown can occur. The gate controller cell will turn off the MOSFET in an attempt to reduce the over-range supply voltage, which activates the STATUS/PGI pull-down, and thus a fault can occur if the power good timeout period has passed. Linear Technology Magazine • March 2006 If no-MOSFET operation is desired and conditions for it are met, let the generator’s design set RFA and RFB, substitute the generator’s reference voltage for VFB(REF) in step 3. 3. Solve for the tracking resistors that set the desired ramp rate and voltage offset or time delay of the slave supply. Choose a ramp rate for the slave supply, SS. If the slave supply tracks the master coincidently or with only a fixed offset or delay, then the slave ramp rate equals the master ramp rate. Calculate the upper track resistor, R TB, from: S R TB = RFB • M SS (4) Choose a voltage difference based on the type of profile to be implemented: ∆V = a voltage difference (offset tracking), or ∆V = SM • tDLY (supply sequencing), where tDLY is a delay time, or ∆V = 0V (coincident or ratiometric tracking) (5a) (5b) (5c) Be sure that the slave ramp rate and its offset or delay allow the slave voltage to finish ramping before the master ramp reaches its final value; otherwise, the slave supply voltage will be held below its intended level. Finally, determine the lower track resistor, R TA: R TA = VFB(REF ) RFB VTRACK VFB(REF ) VTRACK ∆V + − + RFA R TB R TB where VTRACK = VFB(REF) = 0.8V (6) Note that large ratios of slave ramp rate to master ramp rate, SS/SM, may result in negative values for R TA. In such cases the offset or delay must be increased, or the slave ramp rate must be reduced. 27 L DESIGN FEATURES OUT MASTER SUPPLY MODULE LOAD CGATE OUT SUPPLY MODULE SLAVE RX1 LOAD SENSE MGATE VCC RAMP CHARGE PUMP 10µA REMOTE SENSE SWITCH 10µA D1 RAMPBUF S1 1x CHARGE PUMP 10µA RSGATE + VCC + 4.9V – 1.23V – ON ON|OFF STATUS/PGI + MGATE + RAMP + 4.9V – RAMP + VCC – SGATE1 + VCC + 4.9V – SGATE2 + VCC + 4.9V – CHARGE PUMP 10µA RSGATE 10µA CHARGE PUMP VCC 0.8V 10µA + SGATE1 – RTB1 0.8V TRACK1 RTA1 + RFB1 10µA – FB1 GND RFA1 Figure 7. Simplified functional block diagram for the LTC2926 28 Linear Technology Magazine • March 2006 DESIGN FEATURES L not equal to the generator voltage, and the feedback would send the generator voltage higher and higher attempting to equalize them. The LTC2926 solves the voltage drop problem with automatic remote sense switching. In Figure 6, one of the two integrated N-channel MOSFET remote sense switches connects the load to the supply generator’s sense input. During ramp up and ramp down, the switch is open, and resistor RX provides local feedback to the sense input. After tracking has completed, the RSGATE signal closes the remote sense switch, and the supply generator dynamically compensate out the power MOSFET’s voltage drop. The RSGATE signal is available on a pin so that additional external switches may be controlled if necessary. Tracking Typical Behavior: Supplies Inclined to Marry Loads; Separation on the Decline The operation of the LTC2926 in an application can be understood by considering the simplified block diagram with external components in Figure 7. Assume that the supply generators’ outputs and the VCC supply have reached their nominal values, and that the ON input is low. In that case, the STATUS/PGI pin is pulled-down, the remote sense switch is open, and the MGATE pin is pulled to ground, which means the master load is disconnected from the master supply. Track resistor R TB1 is grounded by the ramp buffer Q1 1.8V SOURCE 10Ω 2.5V SOURCE IN OUT V2 V1 RST S2 RST S1 GND TOL 10Ω SGATE1 SGATE2 ON/OFF output, so the injected feedback current (a duplicate of the track current) is at its maximum, which forces the FB1 pin voltage above 0.8V. Thus the SGATE1 pin is pulled low, so the slave supply is also disconnected. When the ON pin voltage is brought high, the MGATE pin sources current into an external capacitor that sets the incline rate of the master ramp signal (see Figure 2). The master ramp may be used to create a master supply with the addition of an N-channel MOSFET (see Figure 3). The buffered ramp output (RAMPBUF pin) allows tracking resistors to be driven without loading the MGATE pin current, and keeps track currents from back feeding the master load. As the master ramp rises, the track current decreases, and the gate controller brings up the VIN IN OUT RFB2 SLAVE I/O an Explanation: It’s My Fault That You’ve Separated … The LTC2926 communicates with other devices in the system via the ON 6 + RAMPBUF SGATE TRACK NC FB GND RFAB RFAA 4 VSLAVE (V) FB ′ RTB CONTROL VIA FB PIN AND SPLIT RFA RESISTOR – GND ′ RTA CONTROL VIA FB PIN 3 2 MOSFET CONTROL ONLY 1 0 b. Figure 9. Slave supply control without a MOSFET (a) generator reference VFB(GEN) ≤ 0.75V, and (b) generator reference VFB(GEN) > 0.75V Linear Technology Magazine • March 2006 LOAD VOLTAGE MONITOR (10% TOLERANCE) 5 LTC2926 a. TOL slave supply voltage until it reaches the slave generator voltage, after which point the MOSFET is fully enhanced. As tracking has completed, the remote sense switch then closes, and finally the STATUS/PGI pin is asserted. When ON is brought low, the tracking profile runs in reverse. The STATUS/PGI pull-down activates, and the RSGATE pin pulls down, which opens the remote sense switches. Next, the MGATE pin sinks current, which reduces the master (supply) ramp and slave supply voltages in reverse order. As the master and slave supplies near ground, the slave supplies (and master supply if implemented) are disconnected, which completes the ramp-down process. RFB NC GND S1 GND Figure 8. External monitors control the LTC2926 via I/O pins – + RTA STATUS/PGI RST RFA2 RFA FB RFA1 V1 S2 CPGTMR LTC2926 RAMPBUF SGATE TRACK FAULT VFB(GEN) FB RFB1 FB1 V2 RST GND PGTMR FB2 VFB(GEN) RTB ON LTC2926 SUPPLY VOLTAGE MONITOR (10% TOLERANCE) RFB GND 2.5V SLAVE2 LTC2904 SUPPLY GENERATOR SLAVE Q2 LTC2904 SUPPLY GENERATOR VIN 1.8V SLAVE1 0 0.25 0.50 0.75 1.00 VFB(GEN) (V) 1.25 1.50 Figure 10. Regions of possible slave control without a MOSFET 29 L DESIGN FEATURES input signal that initiates ramp up and ramp down, and two input/output signals, FAULT and STATUS/PGI. Each of the two I/O signals reports an aspect of tracking status as its output, and each accepts a shut-down command as its input. Both the FAULT and the STATUS/PGI pins include strong N-channel MOSFET pull-down transistors, and weak pull-up currents, which facilitates wired-OR signaling. A high output at the STATUS/PGI pin indicates that tracking/sequencing and automatic remote sense switching have completed. It is typically connected to the RST inputs of downstream devices such as an FPGA, a micro-controller, or a load voltage monitor (Figure 9). The weak pull-up hangs from a charge-pumped rail, which allows the STATUS/PGI output to control external MOSFET switches, as well as become a logic signal with the addition of a pull-up resistor. The input function of the STATUS/ PGI pin allows downstream devices and the LTC2926 itself to force open the remote sense switches and bring about supply disconnect if the pin voltage is low and the power good timeout period has expired. The MGATE, SGATE1, SGATE2, and RSGATE pins that control MOSFET gates are all pulled low to effect immediate supply disconnect and open the remote sense switches. In addition, an internal fault latch is set, which keeps the loads cut off until it is reset and re-armed via the ON pin. When connected as in Figure 8, the LTC2904 supply monitor forces supply disconnect if the programmed 10% load voltage tolerance is exceeded after the timeout period (set by CTIMER) expires. The FAULT pin’s input aspect allows upstream devices to set the fault latch, open the remote sense switches, and cause supply disconnect without a timeout period when the pin is pulled low. In addition, the STATUS/PGI pull-down is activated, which informs downstream devices of the fault. Under normal conditions, a weak pull-up keeps the FAULT pin voltage within a diode drop of VCC—the internal Schottky diode allows the pin to be pulled above VCC safely. Again, 30 the loads remain cut off until the fault latch is reset and re-armed by toggling the ON pin. The FAULT pin might typically be connected to the RST output of upstream supply devices such as voltage, current, or temperature monitors (Figure 8). Automatic fault retry is possible by tying the ON and FAULT pins together. MOSFET-controlled tracking and sequencing can improve power system segmenting and allow reuse, which reduces parts count and board area. A singleoutput supply generator can power different rails of the same voltage (e.g., analog power, digital power, and housekeeping power) because each rail’s tracking profile can be set independently. Direct Supply Generator Control: ¡No Más FETs, Mi Amigo! The LTC2926 can even set a tracking profile without MOSFETs just like the LTC2923, under certain conditions. As is the case for no-MOSFET tracking with the LTC2923 family, the supply generator must allow access to the feedback node that sets its output voltage, and its reference voltage must be ground-based. (MOSFET control is required for many three-terminal regulators, for example, because their references are relative to their output node.) For tracking control when the slave generator’s reference voltage is low enough, VFB(GEN) ≤ 0.75, simply connect the LTC2926’s FB pin to the supply generator’s FB pin (Figure 9a). Choose the track resistors based on the tracking profile and the generator’s feedback resistors. When the master ramp signal is low, the tracking current is high, and it keeps the slave generator’s output low. When the master ramp signal reaches its maximum, the LTC2926’s FB pin current is zero, and it has no effect on the output voltage accuracy, transient response, or stability of the generator. A generator with VFB(GEN) > 0.75V may be controlled without a MOSFET if the slave voltage is large enough; see Figure 10. The R TA resistor must be split to create a new injection point for FB pin current, and the track resistor values must be scaled, as well (Figure 9b); consult the LTC2926 Data Sheet for details. Conclusion: No Joke, It’s a Great Product The LTC2926 solves a host of tracking and sequencing headaches and can simplify design by means of MOSFET control. MOSFET control separates supply generator start-up and shutdown details from specific tracking profile requirements, which allows for supply segmenting and generator consolidation. Because the LTC2926 creates its own regulator to ramp the rails, a multitude of supply generators can now be tracked and sequenced, including modules and 3-terminal linear regulators. The LTC2926 is interoperable with Linear’s no-MOSFET tracker/ sequencers, and even provides that functionality itself, which can keep device count down and reduce parts assortment. Its integrated automatic remote sense switching eliminates a problem associated with series MOSFET control, and intelligent I/O lets this device broadcast status as well take shut-down commands from upstream and downstream devices. All of these features and fine control of start-up and shut-down of power supply rails in a single package make the LTC2926 powerful solution for tracking and sequencing. L For further information on any of the devices mentioned in this issue of Linear Technology, use the reader service card or call the LTC literature service number: 1-800-4-LINEAR Ask for the pertinent data sheets and Application Notes. Linear Technology Magazine • March 2006 DESIGN FEATURES L High Voltage Boost/LED Controller Provides 3000:1 PWM Dimming Ratio by Eugene Cheung Performance, Accuracy, Versatility: LEDs and Beyond The LTC3783 is a current-mode boost controller optimized for constantcurrent True Color PWM™ dimming of high-powered LEDs. Proprietary techniques provide extremely fast, true PWM load switching with no transient undervoltage or overvoltage issues. High dimming ratios of 3000:1 (at 100Hz), important for such applications as video projectors and LCD backlights, can be achieved digitally, maintaining the color integrity of white and RGB LEDs. The LTC3783 also provides an analog input for an additional 100:1 dimming (300,000:1 total). This versatile part can also be used in a boost, buck, buck-boost, SEPIC, or flyback converter, and as a constantcurrent, constant-voltage regulator. No RSENSE™ operation uses a MOSFET’s on-resistance to eliminate the currentsense resistor, increasing efficiency. Applications for the LTC3783 include high voltage LED arrays and LED backlighting, as well as voltage regulators in telecom, automotive, and industrial control systems. VIN OSC VREF + VITH – CITH + VOUT VSW S – LATCH COMP SW1 COUT R LOAD RSENSE VSENSE PWMOUT PWMIN SW2 VS IOUT RSENSE Figure 1. Simplified conventional boost converter with PWM dimming The LTC3783 operates from input supplies ranging from 3V to 36V, and provides output overvoltage protection while regulating output current. When a sense resistor is used, the maximum output voltage is limited only by external components. The controller includes integrated drivers for power and PWM MOSFET switches, and a variable feedback voltage (0V to 1.23V) allows the designer full control over load current accuracy vs efficiency. These features make the part especially attractive for higher-power LED lighting applications. One resistor sets operating frequency from 20kHz to 1MHz, and, to reduce switching noise interference, the LTC3783 is synchronizable to an external clock. Programmable soft start limits inrush current during startup, preventing input current spikes. In addition to boost operation (VOUT > VIN), the controller offers an alternate constant-current, constant-voltage operating mode for buck-boost or buck applications. In these cases, the achievable PWM dimming ratios are generally lower. VIN OSC VREF + SW3 – VITH S – CITH + LATCH COMP R VOUT VSW GATE SW1 COUT LOAD VSENSE RSENSE PWMOUT PWMIN SW2 VS IOUT RSENSE Figure 2. Simplified boost converter with True Color PWM dimming Linear Technology Magazine • March 2006 31 L DESIGN FEATURES A Novel, Simple, yet Highly Effective PWM Dimming Scheme VIN 6V TO 16V (< TOTAL VF OF LEDs) 10µF ×2 1M 2.2µH ZETEX ZLL51000 LTC3783 Why PWM, Anyway? The brightness of an LED is a function of the current through it. Analog dimming simply reduces the DC current flowing through the LED, while digital, or PWM, dimming alters the duty cycle of an otherwise-constant LED current, thus varying the effective average current. The problem with analog dimming is that the chromacity of the LED also changes with current. PWM dimming avoids this problem because the on-current is constant, allowing the light intensity, i.e. average current, to be varied without a color shift. Enter True Color PWM A simplified conventional currentsource boost controller is shown in Figure 1, where IOUT is the regulated current source. When the output load is abruptly disconnected by PWMIN via SW2, the feedback loop cannot adjust the inductor current IL, controlled by ITH (cycle-to-cycle current threshold), instantaneously. Consequently, VOUT rises due to excess current being fed into COUT, causing an output overvoltage condition while the error amp slews its ITH compensation capacitor CITH down to the appropriate zero-current level. When the load is reconnected, VOUT is pulled lower by the load current as the compensation capacitor is slewed up to match output current demand. Depending on the particulars of the application 105k 0.1µF 10µF 6.04k 10k 1µF 4.7µF 20k M1 M2 12.4k 0.3Ω COUT 10µF GND *LUMILEDS LXHL-BW02 Figure 3. PWM-dimmed boost LED application (compensation, load parameters, and component values) with respect to PWMIN on-time, the load current will generally overshoot or undershoot, causing a color shift with varying light intensity. True Color PWM dimming, as implemented in the LTC3783, is depicted in simplified form in Figure 2. When PWMIN goes low, the output load is disconnected (SW2), switching is simultaneously disabled (SW1), and compensation capacitor CITH is disconnected (SW3). Disabling SW1 switching with PWMIN low prevents the VOUT overvoltage condition from occurring, and disconnecting CITH preserves the appropriate steady-state ITH value. When PWMIN goes high again and the load is reconnected, VOUT and VITH are already at their respective full-load values, and load current is restored virtually instantaneously. This new technique allows the load to be quickly connected and disconnected. This results in a higher 10µF, 50V ×2 UMK432C106MM RUN VIN PWMIN OV/FB ITH PWMOUT SS ILIM VREF GATE FBP SENSE FBN INTVCC FREQ GND SYNC 10µH SUMIDA CDRH8D28-100 PMEG6010 LTC3783 100k LED* STRING 4.7µF 0.05Ω M1, M2: SILICONIX Si4470EY 1M PWM 5V AT 0Hz TO 10Hz 237k RUN VIN PWMIN OV/FB ITH PWMOUT SS ILIM VREF GATE FBP SENSE FBN INTVCC FREQ GND SYNC VOUT <25V effective PWM dimming ratio, since, for a given PWM frequency, the dimming ratio is constrained by the shortest pulse duration (hence, lowest duty cycle) that can be delivered. Applications Boost PWM LED Driver Multi-LED systems usually connect the LEDs in series to ensure that the current through each LED is the same, regardless of the varying I-V characteristics of each LED. In such systems, the cumulative LED string voltage is often higher than the system supply, thus calling for a boost converter (VOUT > VIN). Figure 3 shows such a solution with PWM dimming. VFBP across RL sets the LED load current level. VFBP is set (via R1 and R2) to 0.1V in the interests of higher efficiency in light of load current accuracy. Setting the load current sense resistor voltage to 0.1V allows for only 35mW power dissipation in the VIN 9V TO 26V RL 0.28Ω LED STRING 1-4 EA LUMILEDS LXHL-BW02 EACH LED IS 3V TO 4.2V AT 350mA VOUT 40.2k 0V TO 1.23V FAIRCHILD FDN5630 10µF, 50V C5750X7R1H106M CERAMIC 4.7µF 0.05Ω 1k GND Figure 4. Buck-boost LED application 32 Linear Technology Magazine • March 2006 DESIGN FEATURES L CIN 10µF, 25V ×2 1M • • • • PWM 169k 1% 1k 0.01µF 10k 1% 30.1k 1% RUN VIN PWMIN OV/FB ITH PWMOUT SS ILIM VREF GATE FBP SENSE FBN INTVCC FREQ GND SYNC 0.22µF 250V ×2 • LTC3783 100pF D1 T1 VIN 8V TO 16V 1.21M 1% LED+ VOUT 130V 150mA • 100pF 200V NPO 1206 20Ω Q1 4.7µF 10V 12.4k 1% 0.05Ω LED– Q2 0.47Ω 1% CIN: TDK C3225X7R1E106K COUT: TDK C3225X7R2E224K D1: PMEG6010 Q1: Si4486EY Q2: FAIRCHILD FDC2512 (TSOP6) T1: COILTRONICS VP3-0055 Figure 5. PWM-dimmed flyback LED application resistor, as compared to 430mW for a 1.23V sense voltage. A worst-case VFB offset of <3mV ensures better than 3% load current accuracy as well. Additional analog dimming could be accomplished by replacing R1 and R2 with a potentiometer or other variable voltage source. Resistor R T on the FREQ pin determines GATE switching frequency. The 6.04kΩ value chosen sets the system up for 1MHz switching, which permits a higher PWM dimming ratio than the standard 300kHz switching frequency allows. In case the LEDs are disconnected while the supply is running, resistors R3 and R4 set the maximum output overvoltage shutdown threshold, nominally at VOV/FB = 1.32V. Overvoltage protection is essential in current-source boost applications because with an openload fault, capacitor and FET drain voltages can easily exceed maximum device ratings. Because of the True Color PWM topology, and the 1MHz clock rate, this boost application circuit can achieve a PWM dimming ratio in excess of 3000:1. Authors can be contacted at (408) 432-1900 Linear Technology Magazine • March 2006 Buck-Boost PWM LED Driver In some LED applications, the desired VIN and VOUT overlap, thus requiring buck-boost or SEPIC functionality. Figure 4 depicts such a system. In this setup, LED current is returned to VIN, and the LEDs actually see a voltage of VOUT – VIN, allowing a nominally boost configuration to function as a buck-boost. In contrast, a SEPIC configuration would require a 2-inductor- or transformer-based solution, resulting in increased complexity and lower efficiency. However, a true SEPIC would also provide for a grounded load, which may be desirable in some applications. In this mode, PWM dimming is available through the PWM input, albeit not at a dimming ratio comparable to the one in the boost configuration. Lack of a PWMOUT-controlled load switch means transient response cannot be as rapid, since the output voltage is then allowed to sag when PWMIN is low, necessitating some recovery period when PWMIN goes high again. The ILIM pin provides analog dimming, as the ILIM range (0.12V < VILIM < 1.23V) controls the (VFBP – VFBN) differential proportionally from 10mV to 100mV. This allows the LED current to be linearly varied by an additional 10:1 ratio. High Voltage (130V OVP) Flyback LED Driver The 130V application shown in Figure 5 is similar to that of Figure 3, but because of the extremely high boost ratio, a 3:1 transformer is added in order to reduce the GATE duty cycle such that the part’s maximum duty cycle is not violated. This circuit is capable of driving a string of LEDs at 150mA, which can add up to less than 130V total forward voltage, at which point the OV/FB pin is activated to stop all switching. This prevents a potential overvoltage condition at the output. As presented, the application circuit can provide a PWM dimming ratio of 500:1. Other Functionality In the 130V application above, because VFBP > 2.5V, the LTC3783 is in constant-current, constant-voltage operation, meaning the control loop seeks to regulate the load sense resistor voltage (VFBP – VFBN) at 100mV. This is distinct from the pure voltage mode of the boost application, in which VFBP = VFBN. Also, in constant-current, constant-voltage mode the OV/FB pin becomes a linear feedback input, which regulates VOV/FB at 1.23V if (VFBP – VFBN) < 100mV. L 33 L DESIGN FEATURES Introduction The LTC3412A is much more than a drop-in replacement for the popular LTC3412 monolithic buck switching regulator. Though still offered in a thermally enhanced 16-lead TSSOP, the LTC3412A is also squeezed into a smaller footprint 4mm × 4mm QFN, reducing the package height to 0.75mm from 1.1mm. At the same time, the maximum load current has risen to 3A from the LTC3412’s 2.5A, and the minimum input voltage has been pared down to 2.25V from the earlier 2.625V. The LTC3412A can still deliver efficiency as high as 95%, accept input voltages up to 5.5V, and switch at frequencies up to 4MHz, making it a compact and efficient solution for portable electronics that require low supply voltages (down to 0.8V) converted from typical battery voltages. Like its predecessor, the LTC3412A employs a constant frequency, current-mode architecture. The switching frequency can be set between 300kHz and 4MHz by an external resistor, or each switching cycle can commence with the falling edge of an external clock signal fed into the Sync/Mode pin. The LTC3412A’s high switching frequency permits the designer to use tiny, low value inductors while holding output voltage ripple to a minimum. Choice of Operating Modes The LTC3412A can be configured for Burst Mode operation, forced continuous operation, or pulse skipping. In mobile applications where battery life is of paramount importance, Burst Mode operation boosts efficiency by reducing gate charge losses at light loads, and so reducing supply current to just 64µA at no load. Forced continuous operation switches at constant frequency regardless of load current; this simplifies filtering of the switch noise. Pulse skip mode provides a good compromise between light load efficiency and output voltage ripple. During Burst Mode operation, switching cycles are skipped during light loads to reduce switching losses. The LTC3412A provides for external VIN 3.3V CFF 22pF X5R CIN3** 100µF R1 392k 1 CITH 330pF X7R RITH 17.4k 3 4 RSS 2.2M 16 CIN1 22µF PGOOD ITH VFB SW 3000 2500 2000 1500 1000 500 0 0.1 0.2 0.3 0.4 0.5 VBURST (V) 0.6 0.7 control of these cycles’ peak current (the burst clamp level) by varying the DC voltage at the Sync/Mode pin within a 0V–1V range. Figure 1 shows the relationship between this DC voltage (which is generally tapped off of the feedback resistor network) and the burst clamp level. Higher peak currents deliver more energy, so fewer bursts are required to maintain the output voltage. If the minimum peak inductor current delivers more energy than the load current demands, the control loop causes the internal power switches to skip more cycles. Lower burst frequencies can improve efficiency at light load, at the expense 5 ROSC 294k 6 SYNC/MODE 7 15 90 Burst Mode 80 OPERATION 14 SW LTC3412A 13 EFE PGND R2 69.8k CSS 1000pF X7R 8 RT PGND SW RUN SW SGND PVIN L1* 0.47µH VOUT 2.5V 3A 12 11 COUT** 100µF ×2 10 9 CIN2 22µF X5R 6.3V *VISHAY IHLP-2525CZ-01 **TDK 4532X5R0J107M Figure 2. A 2.5V, 4A step-down regulator in Burst Mode operation 34 3500 100 2 CC 47pF R3 115k PVIN 4000 Figure 1. By tying the Sync/Mode pin to a point on the feedback voltage divider, as in Figure 2, the designer can freely adjust the peak current of the bursts (See Figure 4). GND EFFICIENCY (%) PGOOD RPG 100k SVIN MAXIMUM PEAK INDUCTOR CURRENT (mA) 3A Monolithic Buck Regulator by Theo Phillips in 4mm × 4mm QFN 70 FORCED CONTINUOUS 60 50 40 30 20 10 0 0.01 VIN = 3.3V VOUT = 2.5V FIGURE 4 CIRCUIT 0.1 1 LOAD CURRENT (A) 10 Figure 3. Efficiency vs load current for the circuit of Figure 2, VIN = 3.3V, VOUT = 2.5V (Burst Mode operation). If Sync/Mode were tied to ground, the efficiency would be lower (pulse skip mode), but output voltage ripple would be greatly reduced. Linear Technology Magazine • March 2006 DESIGN FEATURES L ther reduction in the battery voltage forces the internal P-channel MOSFET to remain on for more than one cycle, and ultimately to remain on 100% of the time. This dropout state extends the useful operating input voltage over the run-time of the battery. Output voltage simply follows the input voltage as the battery continues to discharge, reduced by drops across the inductor and P-channel MOSFET. VOUT 20mV/DIV INDUCTOR CURRENT 1A/DIV 5µs/DIV Figure 4. Burst Mode operation, showing output voltage ripple associated with bursts of inductor current. The peak currents are held to 1.1A because the Sync/Mode voltage is set to 0.49V. The LTC3412A can deliver efficiency as high as 95%, accept input voltages up to 5.5V, and switch at frequencies up to 4MHz, making it a compact and efficient solution for portable electronics that require low supply voltages (down to 0.8V) converted from typical battery voltages. CIN3** 100µF C1 22pF X5R VIN 5V R1 634k 1 PGOOD RPG 100k CITH 820pF X7R RITH 7.5k 2 3 CC 47pF R2 200k ROSC 137k ITH VFB 5 6 CSS 1000pF X7R 8 PVIN PGOOD 4 7 RSS 2.2M SVIN SW 16 CIN1 22µF X5R 6.3V 15 14 SW LTC3412A 13 EFE PGND RT SYNC/MODE PGND SW RUN SW SGND PVIN Figure 2 shows a 2.5V step-down DC/DC converter that is configured for Burst Mode operation. This circuit provides a regulated 2.5V output at up to 3A from a 3.3V input. Figure 3 shows that efficiency peaks at 94%, and remains high across the load spectrum. The measured output voltage ripple is around 25mVP–P at loads from 10mA to 200mA, and drops to around 5mV at heavier loads. Lower ripple (5mV–10mV) can be maintained at the lightest loads with pulse skip mode, with some sacrifice in light load efficiency. The switching frequency for this circuit is set at 1MHz by a single external resistor, ROSC. Operating at frequencies this high allows the use of a low value (and physically small) inductor. In this application, Burst Mode operation maintains the high efficiency at light loads. This powers down the majority of the internal circuitry during the intervals between switching cycles. The peak inductor current is set by the R2-R3 voltage divider, which generates 100 L1* 0.47µH 90 VOUT 3.3V 3A 12 11 COUT** 100µF ×2 10 9 CIN2 22µF X5R 6.3V 80 70 60 50 40 30 20 GND *VISHAY IHLP-2525CZ-01 **TDK C4532X5R0J107M Figure 5. A 2MHz regulator, 5V to 3.3V at 3A, operating in forced continuous mode Linear Technology Magazine • March 2006 A High Efficiency 2.5V, 3A Step-Down Regulator with All Ceramic Capacitors EFFICIENCY (%) of a slight increase in output voltage ripple. Conversely, lowering the minimum peak inductor current increases the burst frequency and reduces the output voltage ripple. Burst Mode operation dissipates minimal power in light load applications, but sometimes noise suppression takes priority over efficiency. To reduce noise and RF interference, the LTC3412A can be set up for forced continuous operation by connecting the Sync/Mode pin to the Signal Input Voltage pin. This mode maintains a constant switching frequency regardless of output load, dovetailing with noise sensitive applications in which it is necessary to avoid switching harmonics in a particular signal band. In pulse skip mode, the burst clamp is set to zero current, which limits the minimum peak inductor current to a level set by the minimum on-time of the control loop. Pulse skip mode is implemented by connecting the Sync/ Mode pin to GND. Pulse skipping has lower ripple than Burst Mode operation and has better light-load efficiency than forced continuous mode. As the battery voltage decreases toward the output voltage, the duty cycle and the on-time increase. Fur- 10 0 0.01 VIN = 5V VOUT = 3.3V FIGURE 5 CIRCUIT 0.1 1 LOAD CURRENT (A) 10 Figure 6. Efficiency of the circuit in Figure 5 reaches well into the 90s at moderate loads. 35 L DESIGN FEATURES VOUT 100mV/DIV AC COUPLED VIN 2V/DIV 0V VOUT 2V/DIV 0V INDUCTOR CURRENT 2A/DIV VSW 5V/DIV VIN = 5V 40µs/DIV VOUT = 3.3V f = 2MHz LOAD STEP = 300mA TO 3A 0V VIN = 3.3V VOUT = 2.5V f = 1MHz RLOAD = 1.5 100µs/DIV Figure 7. The circuit of Figure 5 delivers a clean transient response when the load is stepped from 300mA to 3A. Figure 8. If the circuit of Figure 2 experiences a dip in VIN, a regulated output is maintained until the LTC3412A goes smoothly into dropout. Below that level, VOUT follows VIN anywhere above the undervoltage lockout. For output voltages programmed below the UVLO, the LTC3412A can maintain its output voltage even when the input sinks to 2.25V. a 0.49V reference at the Sync/Mode pin. This corresponds to approximately 1.1A minimum peak inductor current, as shown in Figure 1. Figure 4 illustrates how Burst Mode operation produces a burst of inductor current pulses that are repeated periodically. Each inductor current pulse increases to approximately 1.1A during each switching period before the main power MOSFET is shut off. The process repeats for multiple switching cycles until the charge on the output capacitor is refreshed. Once this is accomplished, both the main and synchronous power MOSFETs are held off while the load current is supplied solely by the charge on the output capacitor. This sleep state continues until the output voltage drops low enough to initiate another burst cycle. Varying the voltage on the Sync/Mode pin affects the amplitude of the group of current bursts as well as the frequency at which they are repeated. tion near an AM radio receiver, since it operates above the broadcast band and the switch noise can be filtered in a predictable manner. Efficiency for this circuit, shown in Figure 6, is as high as 95% at moderate loads. Ceramic capacitors offer low cost and low ESR, but many switching regulators have difficulty operating with them because the extremely low ESR can lead to loop instability. The phase margin of the control loop can drop to inadequate levels without the aid of the zero that is normally generated from the higher ESR of tantalum, niobium, or special polymer capacitors. The LTC3412A, however, includes OPTILOOP compensation, which allows it to operate properly with ceramic input 36 Lower Minimum Input Voltage Sagging input voltages are handled well by the LTC3412A. Figure 8 shows the response of VOUT to a substantial hit to VIN. Until VIN drops to VOUT, the effect on VOUT is negligible; below that point the P-channel FET turns on 100% of the time (dropout), and VOUT follows VIN right down to the guaranteed 2.25V undervoltage lockout. Output voltages set below the UVLO can be maintained while maximizing battery life, or where input rails are just loosely regulated. Conclusion The LTC3412A is a monolithic, synchronous step-down DC/DC converter that is well suited for applications requiring up to 3A of output current. OPTI-LOOP compensation allows diverse transient responses to be optimized with ceramic capacitors. For excellent thermal handling, the LTC3412A is offered in two tiny packages, each with an exposed pad to facilitate heat sinking (Figure 9). Its high switching frequency, low undervoltage lockout and internal low RDS(ON) power switches make the LTC3412A an excellent choice for compact, high efficiency power supplies. L For further information on any of the devices mentioned in this issue of Linear Technology, use the reader service card or call the LTC literature service number: 2MHz, High Efficiency 3.3V, 3A Step-Down Regulator with All Ceramic Capacitors Figure 5 shows a 3.3V step-down DC/ DC converter using all ceramic capacitors. Switching at 2MHz, this circuit provides a regulated 3.3V output at up to 3A from a 5V input voltage. This converter is an ideal choice for opera- and output capacitors. The LTC3412A allows loop stability to be achieved through a wide range of loads and output capacitors with proper selection of compensation components on the ITH pin. Figure 7 shows the stable transient response associated with the circuit of Figure 5. 1-800-4-LINEAR Figure 9. The LTC3412A is offered in two thermally enhanced packages: the 16-lead TSSOP (both sides shown) and the ultra-thin 4mm × 4mm QFN (shown in tweezers). Ask for the pertinent data sheets and Application Notes. Linear Technology Magazine • March 2006 DESIGN IDEAS L 3A Converter Drives LEDs with 500:1 Dimming by Jaino Parasseril Introduction DESIGN IDEAS 3A Converter Drives LEDs with 500:1 Dimming .........................37 Jaino Parasseril 40nVP–P Noise, 0.05µV/°C Drift, Chopped FET Amplifier .....................39 Jim Williams 100 VIN = 5V 4 LEDs fPWM = 100Hz 10 1 0.1 0.1 10 1 PWM DUTY CYCLE (%) Save Board Space with a High Efficiency Dual Synchronous, 400mA/800mA, 2.25MHz Step-Down DC/DC Regulator ..............44 Damon Lee Linear Technology Magazine • March 2006 Figure 3), resulting in a wide dimming range of 500:1 at a PWM frequency of 100Hz. The LED current can be controlled by feeding a PWM signal with a broad range of frequencies. Dimming below 80Hz is possible but not desirable due to perceptible flashing of LEDs at lower frequencies. The LED current can be controlled at higher frequencies, but the dimming range decreases with increasing PWM frequency. In high temperature applications, the leakage of the Schottky diode D1 increases, which in turn, discharges the output capacitor during the PWM “off” time. This results in a smaller with the addition of a few external components results in a wider dimming range of 500:1. The technique requires a PWM logic signal applied to the gate of both NMOS transistors (refer to Figure 2). When the PWM signal is taken high the part runs in normal operation and ILED of 100mV/ RSENSE runs through the LEDs. When the PWM input is taken low the LEDs are disconnected and turn off. This unique external circuitry produces a fast rise time for the LED current (see L1 2.0µH VIN 5V C1 3.3µF C1: TAIYO YUDEN EMK316BJ335ML C2: TAIYO YUDEN UDK325BJ106MM D1: ZETEX ZLLS1000 D2: DIODES INC. 1N4148 L1: TOKO D53LC (PN A915AY-2ROM) NMOS1: ZETEX 2N7002 NMOS2: FAIRCHILD FDG327N LED1 TO LED4: LUMILEDS LXHL-BW02 ISN1 ISP1 D1 VIN IADJ1 IADJ2 SHDN FBN 75k LT3477 ISP2 5V 0V PWM 100Hz FBP NMOS1 100k RC 2.49k CC 10nF LED1 RT GND 300mA RSENSE 0.33Ω ISN2 VREF D2 C2 10µF 1M SW VC Mayur Kenia Philip Karantzalis 100 Figure 1. The average LED current changes proportionally with PWM duty cycle. Many LED manufacturers specify PWM dimming because it offers wider dimming ranges and better LED performance than analog dimming. SC70 LED Driver Drives Six White LEDs from a Single Li-Ion Battery ..............40 A Simple Digitally Tunable Active RC Filter .................................42 PWM dimming offers several advantages over analog dimming and is the method preferred by LED manufacturers. With PWM dimming, the chromaticity of the LEDs remains unchanged and a wider dimming range is possible. ILED (mA) For LED applications where a wide dimming range is required, two competing methods are available: analog dimming and PWM dimming. The easiest method is to simply vary the DC current through the LED—analog dimming—but changing LED current also changes its chromaticity (color shift), undesirable in many applications (such as LCD backlights). The better method is PWM dimming, which switches the LED on and off, using the duty cycle to control the average current. PWM dimming offers several advantages over analog dimming and is the method preferred by LED manufacturers. By modulating the duty cycle of the PWM signal, the average LED current changes proportionally as illustrated in Figure 1. The chromaticity of the LEDs remains unchanged in this scheme since the LED current is either zero or at programmed current. Another advantage of PWM dimming over analog dimming is that a wider dimming range is possible. The LT3477 is a 3A DC/DC converter that is ideally suited for LED applications. For the LT3477, analog dimming offers a dimming ratio of about 10:1; whereas, PWM dimming SS LED2 CSS 33nF 6.81k LED3 LED4 NMOS2 Figure 2. A 5V input drives four white LEDs in boost mode with PWM dimming 37 L DESIGN IDEAS PVIN 32V PWM 5V/DIV C1 2.2µF RSENSE 0.33Ω IL 1A/DIV LED1 ILED 200mA/DIV VIN = 5V 4LEDs, 300mA fPWM = 100Hz BOOST MODE C1: NIPPON UNITED CHEMICON NTS40X5R1H225M C2: TAIYO YUDEN GMK316BJ105ML C3: TAIYO YUDEN LMK316BJ335KL L1: TOKO D53LC (PN A915AY-100M) D1: ZETEX ZLLS400 D2: DIODES INC 1N4148 NMOS1, NMOS2: ZETEX 2N7002 PMOS: SILICONIX SI2303BDS LED1 TO LED6: LUMILEDS LXHL-BW02 10µs/DIV Figure 3. Rising LED current for the circuit in Figure 2 settles in under 20µs, thus allowing short pulse widths, and high dimming ratios. effective LED dimming ratio. Consequently, the dimming range decreases to about 200:1 at 85°C. PWM dimming can be used in boost mode (Figure 2), buck mode (Figure 4) and buck-boost mode (Figure 5). For the typical boost topology, efficiency exceeds 80%. Buck mode can be used to increase the power handling capability for higher current LED applications, A buck-boost LED driver works best in applications where the input voltage fluctuates to higher or lower than the total LED voltage drop. L LED6 • • • 6-LED 300mA STRING NMOS2 ISP1 D1 VIN IADJ1 IADJ2 C3 3.3µF FBN 10k LT3477 SHDN ISP2 ISN2 VC VREF D2 5V 0V RT FBP PWM GND SS CSS 33nF NMOS1 100k 100Hz 280k SW ISN1 PWM C2 1µF PMOS L1 10µH VIN 3.3V 1k 1k 6.81k CC 0.1µF Figure 4. Buck mode converter drives six white LEDs with PWM dimming from a 32V input 1k NMOS2 PWM 1k 300mA LED1 LED2 PMOS VIN 10V C1 3.3µF ISP1 L1 4.7µH ISN1 FBN 49.9k LT3477 ISP2 VC 5V 0V FBP PWM 100Hz ISN2 VREF D2 NMOS1 100k 1M SW1 VIN IADJ1 IADJ2 SHDN RSENSE 0.33Ω D1 RT GND SS CSS 33nF 6.81k C2 10µF RC 1.5k CC 10nF C1: TAIYO YUDEN LMK316BJ335ML C2: TAIYO YUDEN UDK325BJ106MM D1: ZETEX ZLLS1000 D2: DIODES INC. 1N4148 L1: TOKO D53LC (PN A915AY-2ROM) NMOS1, NMOS2: ZETEX 2N7002 PMOS: SILICONIX Si2303BDS LED1, LED2: LUMILEDS LXHL-BW02 Figure 5. Buck-boost mode converter drives two white LEDs with PWM dimming from a 10V input 38 Linear Technology Magazine • March 2006 DESIGN IDEAS L 40nVP–P Noise, 0.05µV/°C Drift, by Jim Williams Chopped FET Amplifier Figure 1’s circuit combines the 5V rail-to-rail performance of the LTC6241 with a pair of extremely low noise JFETs configured in a chopper based carrier modulation scheme to achieve extraordinarily low noise and DC drift. This circuit’s performance suits the demanding transducer signal conditioning situations such as high resolution scales and magnetic search coils. The LTC1799’s output is divided down to form a 2-phase 925Hz square wave clock. This frequency, harmonically unrelated to 60Hz, provides excellent immunity to harmonic beating or mixing effects which could cause instabilities. S1 and S2 receive complementary drive, causing the A1based stage to see a chopped version of the input voltage. A1’s square wave output is synchronously demodulated by S3 and S4. Because these switches are synchronously driven with the input chopper, proper amplitude and polarity information is presented to DC output amplifier A2. This stage integrates the square wave into a DC voltage, providing the output. The output is divided down (R2 and R1) and fed back to the input chopper where it serves as a zero signal reference. Gain, in this case 1000, is set by the VERT = 20nV/DIV HORIZ = 5s/DIV Figure 2. Noise measures 40nVP–P in 50s sample period TO LTC201 V + PIN 5V 18.5kHz V+ 5V 74C74 ÷ 2 74C90 ÷ 10 DIV LTC1799 OUT RSET Q 925Hz 54.2k* TO Ø1 POINTS 5V Ø1 –5V TO LTC201 V – PIN 1µF + 1µF + 5V 5V R1-R2 ratio. Because the input stage is AC coupled, its DC errors do not affect overall circuit characteristics, resulting in the extremely low offset and drift noted. Figure 2, noise measured over a 50 second interval, shows 40nV in a 0.1Hz to 10Hz bandwidth.This low noise is attributed to the input JFET’s die size and current density. L Q TO Ø2 POINTS 8 INPUT 3k 909Ω** 909Ω** 6 7 S1 S2 0.47µF 1µF LSK389 Ø2 11 10 9 499Ω** Ø2 1µF A1 LTC6241HV + 1µF 1 – 10M –5V 2 100k S3 S4 3 240k 14 15 16 100k 10k 0.001µF Ø1 1µF * = 0.1% METAL FILM RESISTOR ** = 1% METAL FILM RESISTOR = LTC201 QUAD = LSK389 = LINEAR INTEGRATED SYSTEMS FREMONT, CA NOISE = 40nVP-P 0.1Hz TO 10Hz OFFSET = 1µV DRIFT = 0.05µV/°C R2 +1 GAIN = R1 OPEN-LOOP GAIN = 10 9 I BIAS = 150pA – A2 LTC6241HV OUTPUT + R2 10k R1 10Ω Figure 1. 40nV noise chopper amplifier Linear Technology Magazine • March 2006 39 L DESIGN IDEAS SC70 LED Driver Drives Six White LEDs from a Single Li-Ion Battery by Mayur Kenia Introduction The highly integrated LT3491 provides a compact, simple solution for driving backlight circuits in battery-powered portable devices, such as cellular phones, PDAs, and portable GPS devices. It features internal compensation, open-LED protection, a 32V power switch and a 32V Schottky diode all inside a tiny SC70 package. Specifically, the LT3491 is a fixed frequency step-up DC/DC converter that drives up to six white LEDs in series from a Li-Ion cell. Series connection of the LEDs provides identical LED currents resulting in uniform brightness without the need for ballast resistors. In addition, the 2.3MHz switching frequency allows the use of tiny inductors and capacitors. Figure 1 shows how easy it is to drive six white LEDs from a Li-Ion battery. Figure 2 shows the efficiency of the six white LED application circuit. SHUTDOWN AND DIMMING CONTROL VIN 3V TO 5V CTRL VIN L1 10µH CAP LT3491 SW C1 1µF RSENSE 10Ω LED GND C1: TAIYO YUDEN LMK212BJ105MD C2: TAIYO YUDEN GMK316BJ105ML L1: MURATA LQH32CN100 Figure 1. Li-Ion powered driver for six white LEDs C2 1µF side sense moves the sense resistor to the top of the LED string. Figure 3 shows the advantage of high side sense in a cell phone where the LEDs are placed some distance away from the driver. The technique eliminates the return wire thereby reducing the number of wires running through the hinge. Also, it allows the LEDs to be placed further away from the converter without compromising performance. Dimming & Shutdown Control A single pin performs both shutdown and accurate LED dimming control. The dimming range of the part extends High Side Sense The LT3491 features a unique high side LED current sense that enables the part to function as a “one wire” current source—one side of the LED string can be returned to ground anywhere, allowing a simpler “one wire” LED connection. Traditional LED drivers use a grounded resistor to sense LED current requiring a 2-wire connection to the LED string. High 80 VIN = 3.6V 6 LEDs 75 LED DRIVER EFFICIENCY (%) 70 LT3491 65 60 55 50 45 40 0 5 10 15 20 LED CURRENT (mA) Figure 2. Efficiency of the circuit in Figure 1 40 Figure 3. Traditional sense (left) and high side sense (right) implemented solutions in a cell phone. The high side solution reduces the number of connections through the hinge. Linear Technology Magazine • March 2006 DESIGN IDEAS L 100 L1 10µH SENSE VOLTAGE (mV) 200 VIN 160 C1 1µF 120 SW CAP RSENSE 10Ω LT3491 LED GND 80 CTRL C2 1µF 40 0 0 500 1000 1500 2000 VCTRL (mV) 5V 0V Figure 4. LED sense voltage versus CTRL pin voltage from 1.5V at the CTRL pin for full LED current down to 100mV. The CTRL pin directly controls the regulated sense voltage across the sense resistor that sets the LED current. Figure 4 shows the regulated sense voltage versus CTRL pin voltage. In addition to using a DC voltage at the CTRL pin, either a filtered PWM signal or a direct PWM signal can be used to control the LED current. Direct PWM dimming achieves wider dimming compared to using a filtered PWM or a DC voltage. Direct PWM dimming uses a MOSFET in series with the LED string to quickly connect and disconnect the LED string. Figure 5 presents direct PWM dimming in a Li-Ion to four white LED application. The PWM signal controls both the turn on and turn off of the part through the CTRL pin and the MOSFET. Figure 6 shows the linearity of PWM dimming. The available dimming range depends on the settling time of the application PWM FREQ NORMALIZED SENSE VOLTAGE (%) VIN 3V TO 5V 240 10 1 0.1 Q1 2N7002 and the PWM frequency used. The application in Figure 5 achieves a dimming range of 300:1 using a 100Hz PWM frequency. Figure 7 shows the available dimming range for different PWM frequencies. Torch and Flash Mode LED Control White LEDs are quickly gaining popularity as the illumination source for camera phones. White LEDs provide a simple compact solution for flash and torch lighting in cell phone applications. The LT3491 provides a small overall solution to flash and torch control. Torch and flash applications typically use a single high power white LED. The LED driver cannot be setup as a boost because the input voltage in camera phones is very close to, if not higher, than the forward voltage of the LED. A higher input voltage creates a DC path to ground that will drain the battery. High side sensing allows the LT3491 to drive a single LED from higher inputs, thus avoiding this problem. Figure 8 shows the application circuit for torch and flash control powered from two Li-Ion cells. The voltage at the control pin can be moved between two DC levels to toggle between torch and flash operation. Conclusion The LT3491 provides a highly integrated solution to drive backlight applications up to six LEDs from a Li-Ion cell input with the added advantage of high side sense in a tiny SC70 package. L C2 4.7µF VIN 6V TO 9V 10k RSENSE 1Ω PWM DIMMING RANGE PULSING MAY BE VISIBLE D1 1k CAP FLASH MODE ILED = 200mA V CTRL 1.5V 100 C1 1µF 10 1000 100 PWM FREQUENCY (Hz) 10000 Figure 7. Dimming range vs frequency LED L1 10µH VIN LT3491 CTRL VCTRL 680mV TORCH MODE ILED = 100mA 10 1 100 Figure 6. Linearity of PWM dimming for a 100Hz PWM frequency and the application of Figure 5 100k Figure 5. Li-Ion to four white LEDs with direct PWM dimming. 10 1 PWM DUTY CYCLE (%) 0.1 SW GND C1: TAIYO YUDEN LMK212BJ105MD C2: TAIYO YUDEN LMK212BJ475MG D1: AOT-2015 HPW1751B L1: MURATA LQH32CN100 Figure 8. A 2-Cell Li-Ion driver for torch and flash mode control For more information on parts featured in this issue, see http://www.linear.com/designtools Linear Technology Magazine • March 2006 41 L DESIGN IDEAS A Simple Digitally Tunable Active RC Filter Introduction It can be tuned to a few cutoff frequencies via a Serial Peripheral Interface (SPI)1. This is a simpler alternative to using switches with resistor or capacitor arrays or multiple DACs that require a large number of active and passive components. Digital control through an SPI is useful in a variety of applications to control band-limiting for sensor and transducer signals. Typical applications are: accelerometers for vibration analysis, hydrophones for sonar detection, LVDTs for linear motion measurements and microphones for sound reception and recording. The tuning of the cutoff frequency (fCUTOFF) of an active RC filter can be implemented using switched-capacitor circuits or continuous time circuits. In applications that require infinite tuning resolution of an any-order filter in a single IC package, a switched-capacitor approach is preferable (simply changing the clock frequency tunes fCUTOFF). In applications that require tuning a continuous time filter to just a few cutoff frequencies, tuning can be implemented using op amps, CMOS switches and resistor or capacitor arrays. Continuous time filters can also be tuned with high resolution over a large frequency range via digital control by using DACs to multiply the RC time constant of op-amp-based integrators (for example, an 8-bit DAC-based tuner allows for 256 frequency steps). Figure 1 shows a simple, low order and low cost continuous time filter circuit. An SPI-Tunable Second Order Filter The Figure 1 circuit is a state-variable second order filter using two ICs, a low noise CMOS quad op amp (LTC6242) and a low noise dual Programable Gain Amplifier, PGA (LTC6912-X). The gains of the two LTC6912-X by Philip Karantzalis amplifiers (GA and GB) are independently programmed via SPI control. The SPI controlled gain settings of an LTC6912-1 are 1, 2, 5, 10, 20, 50 and 100 and for an LTC6912-2 1, 2, 4, 8, 16, 32 and 64. The Figure 1 filter has three inverting outputs providing a highpass, bandpass and lowpass frequency responses. An optional inverting amplifier connected to one of the three outputs provides for a noninverting or a differential filter output. The filter’s second order transfer function is a function of the circuit’s resonant frequency, f0 and Q values. The f0 frequency is equal to the integrator’s RC constant, the dual PGA gain and the ratio of resistors R4 and R2 (if R4 = R2 and GA = GB = Gain then f0 = Gain/2πRC). The filter’s Q value is equal the ratio of resistors R3 and R2 and the two PGA gains (if R4 = R2 and GA = GB then Q = R3/R2). The filter’s passband gain is equal to the ratio of R4/R1, R3/R1 and R2/R1 R4 10k 10k R3 – 10k C 1/4 LTC6242 R + – 1/4 LTC6242 INA + R2 10k VIN R1 – 1/4 LTC6242 + OUT A LTC6912-1 OR LTC6912-2 C R –GA VLP– SPI SELECTABLE GAINS LTC6912-1: 1, 2, 5, 10, 20, 50, 100 LTC6912-2: 1, 2, 4, 8, 16, 32, 64 – 1/4 LTC6242 + INB –GB OUT B LOWPASS = R4 2 2 4π f0 VLP − R1 = − VIN s 2 + 2πf0 s + 4π 2 f0 2 Q BANDPASS = R3 2πf0 • s VBP − R1 Q = − VIN s 2 + 2πf0 s + 4π 2 f0 2 Q HIGHPASS = R2 2 s VHP − R1 = − VIN s 2 + 2πf0 s + 4π 2 f0 2 Q VBP– 3-WIRE CONTROL VHP– IF R4 = R2 AND G A = GB = GAIN, THEN f0 = OPTIONAL INVERTER FOR A NON-INVERTING OR DIFFERENTIAL OUTPUT GAIN R3 800kHz AND Q = AND MAXIMUM f0 = 2πRC R2 Q • GAIN Figure1. An SPI tunable second order active RC filter 42 Linear Technology Magazine • March 2006 DESIGN IDEAS L GAIN = 1 R = 1.58M C = 1000pF R4 = R2 = R1 = 10k R3 = 6.98k 10 100 1k 10k FREQUENCY (Hz) 100k 6 0 −6 −12 −18 −24 −30 −36 −42 −48 −54 −60 −66 −72 6 0 GAIN = 1 −6 AMPLITUDE (dB) GAIN = 64 AMPLITUDE (dB) AMPLITUDE (dB) 6 0 −6 −12 −18 −24 −30 −36 −42 −48 −54 −60 −66 −72 GAIN = 64 R = 1.58M C = 1000pF R4 = R2 = R1 = 10k R3 = 6.98k 10 100 1k 10k FREQUENCY (Hz) Figure 3. Tunable second order Butterworth highpass response using the LTC6912-2 for a lowpass, bandpass and highpass filters respectively. the f0 frequency is equal to 1.274 • fCUTOFF and the filter’s Q value is equal to 0.577. Figure 2 shows the tunable range of a Butterworth lowpass filter using an 100Hz integrator frequency (R = 1.58MΩ, ±1% and C = 1000pF, ±5%) and an LTC6912-2 to tuned the filter’s fCUTOFF from 100Hz to 6.4kHz. Figure 3 shows the tunable range of a Butterworth highpass filter that is the mirror oposite of the lowpass filter reponse of Figure 2. The output response to a step change is approximately equal to 5/fCUTOFF, (if the step A Tunable Lowpass or Highpass Filter The shape of the amplitude response of the second order depends on the f0 frequency relative to the cutoff frequency and the Q value. In a second order Butterworth highpass or lowpass response the f0 frequency is equal to fCUTOFF (f–3dB) and the filter’s Q value is equal to 0.707. In a second order Bessel highpass or lowpass response −18 −24 −30 −36 −42 −48 100 100k Figure 2. Tunable second order Butterworth lowpass response using the LTC6912-2 −12 10k 1k FREQUENCY (Hz) 100k Figure 4. Tunable second order bandpass filter using an LTC6912-1 with gains 1–8 change is to a 1kHz fCUTOFF then the filter settles five milli-seconds after a step change). The maximum tunable f0 frequency is a function of the gainbandwidth product of the op amps and the circuit’s sensitivity to the highest PGA gain that is used for tuning. For the amplifiers shown, based on empirical data, a maximum f0 of 800kHz/[Q • Gain] limits gain error to ≤2dB). For example, if only the lowest 1, 2, 5 and 10 gains of an LTC6912-1 are used for tuning, a second order Butterworth lowpass filter (f0 = fCUTOFF) continued on page 45 R4 10k 6 0 R3 10k −6 – C 1000pF INA 1/4 LTC6242 + VIN – 1/4 LTC6242 OUT A −12 −18 −24 −30 GAIN = 8 GAIN = 1 −36 R2 10k R1 10k –GA AMPLITUDE (dB) R 316k C 1000pF R 316k + −42 −48 100 LTC6912-2 – INB 1/4 LTC6242 + –GB 10k OUT B 10k 3-WIRE CONTROL 10k IF G A = GB = GAIN, THEN fNOTCH = 1k FREQUENCY (Hz) GAIN 2πRC 10k – 1/4 LTC6242 + VOUT(NOTCH) Figure 5. An SPI-tunable second order notch filter Linear Technology Magazine • March 2006 43 L DESIGN IDEAS Save Board Space with a High Efficiency Dual Synchronous, 400mA/800mA, 2.25MHz Step-Down DC/DC Regulator by Damon Lee Introduction The ever shrinking nature of cell phones, pagers, PDAs and other portable devices drives a corresponding demand for smaller components. One way to shrink DC/DC regulator circuitry is to increase the switching frequency of the regulator, thus allowing the use of smaller and cheaper capacitors and inductors to complete the circuit. Another way is to combine the switcher and MOSFETs in one small, monolithic package. The LTC3548 DC/DC regulator does both. The LTC3548 is a 10-lead MSOP/ DFN, dual, synchronous, step-down, current mode, DC/DC regulator, intended for low power applications. It operates within a 2.5V to 5.5V input voltage range and has a fixed 2.25MHz switching frequency, making it possible to use low-profile capacitors and inductors that are only 1mm high. The LTC3548 is the latest in the LTC3407 and LTC3407-2 family of dual regulators and features an improved Burst Mode ripple and two outputs of 400mA and 800mA. It is available in small MSOP and DFN packages, allowing Figure 1. Two DC/DC regulators occupy less than 0.2in2 of board space two DC/DC Regulators to occupy less than 0.2 square inches of board real estate, as shown in Figure 1. The outputs of the LTC3548 are independently adjustable from 0.6V to 5V. For battery-powered applications that have input voltages above and below the output voltage, the LTC3548 can be used in a single inductor, positive buck-boost converter configuration (see data sheet for details). Two built in 0.35Ω switch provides high efficiency at maximum output current. Internal compensation minimizes external components and board space. 120 VOUT1 100mV/ DIV 100 LTC3407-2 95 100 IL1 500mA/ DIV IL1 500mA/ DIV 80 EFFICIENCY (%) VOUT RIPPLE (mV) 90 VOUT1 100mV/ DIV 60 LTC3548 40 Burst Mode OPERATION 85 80 PULSE SKIP MODE 75 70 VIN = 3.3V VOUT1 = 1.8V LOAD = 100mA 4µs/DIV Figure 2. Comparison of ripple for Burst Mode operation of the LTC3548 and LTC3407-2. 44 Efficiency is extremely important in battery-powered applications, and the LTC3548 keeps efficiency high with an automatic, power saving Burst Mode operation, which reduces gate charge losses at low load currents. With no load, both converters together draw only 40µA, and in shutdown, the device draws less than 1µA, making it ideal for low current applications. The LTC3548 features an improved Burst Mode ripple voltage, which is only about one third of the ripple for the LTC3407 and LTC3407-2, as shown in Figure 2 and Figure 3. The LTC3548 uses a current-mode, constant frequency architecture that benefits noise sensitive applications. Burst Mode is an efficient solution for low current applications, but sometimes noise suppression is a higher priority. To reduce noise problems, a pulse-skipping mode is available, which decreases the ripple noise at low currents. Although not as efficient as Burst Mode at low currents, pulseskipping mode still provides high efficiency for moderate loads, as seen in Figure 4. In dropout, the internal 20 VIN = 3.6V, VOUT = 1.8V NO LOAD ON OTHER CHANNEL CHANNEL 1; CIRCUIT OF FIGURE 3 65 VIN = 3.6V 0 0.1 1 10 100 LOAD CURRENT (mA) 1000 Figure 3. Burst Mode operation output voltage ripple vs load current for the LTC3548 and LTC3407-2. 60 1 10 100 LOAD CURRENT (mA) 1000 Figure 4. Efficiency of Burst Mode operation and pulse skip mode Linear Technology Magazine • March 2006 DESIGN IDEAS L P-channel MOSFET switch is turned on continuously, thereby maximizing the usable battery life. A Power-On Reset output is available for microprocessor systems to insure proper startups. Internal overvoltage and undervoltage comparators on both outputs will pull the POR output low if the output voltages are not within ±8.5%. The POR output is delayed by 262,144 clock cycles (about 175ms) after achieving regulation, but will be pulled low immediately when either output is out of regulation. A High Efficiency 2.5V and 1.8V Step-Down DC/DC Regulator with all Ceramic Capacitors The low cost and low ESR of ceramic capacitors make them a very attractive choice for use in switching regulators. In addition, ceramic capacitors have a benign failure mechanism unlike tantalum capacitors. Unfortunately, the ESR is so low that it can cause loop stability issues. A solid tantalum capacitor’s ESR generates a loop zero at 5kHz–50kHz that can be instrumental in giving acceptable loop phase margin. Ceramic capacitors, on the other hand, remain capacitive to beyond 300kHz and usually resonate Digi-Tune Filters, continued from page 43 can be tuned to 110kHz (maximum f0 = 800kHz/[0.707 • 10]). A Tunable Bandpass Filter The –3dB bandwidth of a second order filter is equal to the center frequency (fCENTER) divided by the Q value (bandwidth = fCENTER/Q). The sensitivity of the second order bandpass filter to the tolerance of the integrator’s RC values is proportional to the filter’s Q. Typically with a Q ≤ 4, using a ±1% R and a ±5% C for the filter’s two integrators is practical for a second bandpass filter. The sensentivity of the second order bandpass filter with Q > 4 increases rapidly for each unit of Q increase and the filter’s two integrators should use ±1% RC components. Linear Technology Magazine • March 2006 VIN = 2.5V* TO 5.5V C1 10µF RUN2 VIN MODE/SYNC VOUT2 = 2.5V* AT 400mA C3 4.7µF L2 4.7µH C5, 68pF R4 887k RUN1 POR LTC3548 SW2 SW1 VFB1 VFB2 R3 280k C1, C2, C3: TAIYO YUDEN JMK212BJ106MG C3: TAIYO YUDEN JMK212BJ475MG GND R5 100k POWER-ON RESET L1 2.2µH C4, 33pF R2 R1 604k 301k VOUT1 = 1.8V AT 800mA C2 10µF L1: MURATA LQH32CN2R2M11 L2: MURATA LQH32CN4R7M23 *VOUT CONNECTED TO VIN FOR VIN ≤ 2.8V (DROPOUT) Figure 5. Dual output step-down application yields 1.8V at 800mA and 2.5V at 400mA. with their ESL before the ESR becomes effective. Also, inexpensive ceramic capacitors are prone to temperature and voltage effects, requiring the designer to check loop stability over the operating temperature range. For these reasons, great care is usually needed when using only ceramic input and output capacitors. The LTC3548 was designed with ceramic capacitors in mind and is internally compensated to handle these difficult design considerations. High quality X5R or X7R ceramic capacitors should be used to minimize the temperature and voltage coefficients. Figure 5 shows a typical application for the LTC3548 using only ceramic capacitors. This circuit provides a regulated 2.5V output and a regulated 1.8V output, at up to 400mA and 800mA, from a 2.5V to 5.5V input. Figure 4 shows the bandpass filter of Figure 1 tuned from 2kHz to 16kHz using a 2kHz integrator frequency (R = 205k, ±1% and C = 390pF, ±5%) and an LTC6912-2 with gain settings 1, 2, 4, and 8. The tuned center frequencies responses of Figure 4 are 2.73% lower than the design values of 2kHz, 4kHz, 8kHz and 16kHz and equal to the error of the circuit’s RC values of the two integrators (measured values of aproximatelly 206k for each R and 403pF for each C). The gain error at 16kHz is due to the filter’s f0 frequency approaching the maximum f0 for a Q = 4 and a PGA gain equal to 8 (maximum f0 = 25kHz = 800kHz/{4 • 8]). The maximum f0 frequency is a function of the gain-bandwidth product of the LTC6912-X op amps. Conclusion The LTC3548 is a dual monolithic, step-down regulator that switches at 2.25MHz, minimizing component costs and board real estate requirements for DC/DC regulators. The small size, efficiency, low external component count, and design flexibility of the LTC3548 make it ideal for portable applications. L Other Filter Options Figure 5 shows an example of a second order notch filter. The notch filter’s integrator frequency is 500Hz (1/[2π • 316kΩ • 1000pF]) and with PGA gains 1, 2, 4 and 8 the notch frequency is tuned to 500Hz, 1kHz, 2kHz and 4kHz respectively. Any of the filters discussed above can be made into SPI-tunable fourth order filters by cascading two second order circuits. L Notes 1 SPI is a synchronous communication protocol using a 3-wire interface between a microprocessor and a peripheral device 45 L NEW DEVICE CAMEOS New Device Cameos Wide Dynamic Range RF/IF Log Detector The LT5537 is a wide dynamic range RF/IF detector, which operates from below 10MHz to 1000MHz. The lower limit of the operating frequency range can be extended to near DC by the use of an external capacitor. The input dynamic range at 200MHz with ±3dB nonlinearity is 90dB (from –76dBm to 14dBm, single-ended 50Ω input). The detector output voltage slope is nominally 20mV/dB, and the typical temperature coefficient is 0.01dB/°C at 200MHz. The LT5537 is available in a tiny 8-Lead (3mm × 2mm) DFN package. Dual Output Synchronous DC/DC Controller Draws Only 80µA Quiescent Current in an Automotive System The LTC3827 is a low quiescent current, 2-phase dual output synchronous step-down DC/DC controller. The LTC3827 draws only 80µA when one output is active and only 115µA LTC2970, continued from page 5 registers allow the user to define instantaneous and/or latched faults in the event one of the input voltages deviates outside an acceptable window. The GPIO_0 and FAULT pins can be configured to assert if a fault occurs. Tracking Two or More Supplies with the LTC2970-1 The LTC2970-1 enables power supply tracking with the addition of a few external components. A special global address and synchronization command allow multiple LTC2970-1s to track and sequence multiple pairs of power supplies. A typical LTC2970-1 tracking application circuit is shown in Figure 7. The GPIO_0 and GPIO_1 pins are tied directly to their respective DC/ DC converter RUN/SS pins. When GPIO_CFG is pulled-up to VDD, the LTC2970-1 automatically holds off 46 when both outputs are active, making it ideally suited for automotive applications, such as navigation systems, where one or more supplies remains active while the engine is off. The LTC3827’s input supply range of 4V to 36V is wide enough both to protect against high input voltage transients and to continue to operate during automotive cold crank. The LTC3827 features a ±1% internal reference and can provide output voltages from 0.8V up to 10V, making it perfect for the higher voltage supplies typically required for audio systems, analog tuners, and CD/DVD players in many automobiles. Each output can deliver up to 20A of current at efficiencies as high as 95%. The LTC3827 is rated for operation from –40°C to 85°C, and has a maximum operating junction temperature of 125°C. The LTC3827’s constant frequency, current mode architecture provides excellent line and load regulation, and its 2-phase operation reduces input capacitance requirements. The LTC3827 smoothly ramps each output voltage during startup using separate adjustable soft-start and tracking input pins. It operates at a selectable frequency between 250kHz and 550kHz, and can be synchronized to an external clock from 140kHz to 650kHz using its phase-locked loop (PLL). Output overvoltage and overcurrent (short circuit) protection are provided internally. With both outputs shut down, the LTC3827 draws a mere 8µA. The LTC3827 is offered in two packages: a 28-lead SSOP (LTC3827-1) and a 32-pin 5mm × 5mm QFN (LTC3827). the DC/DC converters after power-up. N-channel FETs Q10 and Q11 and diodes D10 and D11 form unidirectional range switches around R30A and R31A while GPIO_CFG is high, which allow the VOUT0 and VOUT1 pins to drive the converter outputs all the way to/from ground through resistors R30B and R31B. When GPIO_CFG pulls low, FETs Q10 and Q11 turn off. R30A and R31A then combine in series with R30B and R31B for normal margin operation. The 100kΩ/0.1µF lowpass filter in series with the gates of Q10 and Q11 minimizes charge injection into the feedback nodes of the DC/DC converters when GPIO_CFG pulls low. power applications into one easy-touse device. A multiplexed, differential input 14-bit delta-sigma and a low drift on-chip reference deliver less than ±0.5% total unadjusted error. Two continuous-time, 8-bit, voltage-buffered IDACs can also be programmed through the I2C and SMBus compatible interface to servo power-supplies to the desired voltages. Extensive, user configurable fault monitoring and a built-in servo algorithm reduce the burden on system computing resources and shorten software development time. The LTC2970 and LTC2970-1 are available in a 24-lead QFN package. Conclusion The LTC2970 dual power supply monitor and controller combines the necessary features essential for digitally managed, high availability For further information on any of the devices mentioned in this issue of Linear Technology, use the reader service card or call the LTC literature service number: 1-800-4-LINEAR Ask for the pertinent data sheets and Application Notes. L for the latest information on LTC products, visit www.linear.com Linear Technology Magazine • March 2006 DESIGN TOOLS L DESIGN TOOLS CD-ROM Product Information The March 2006 CD-ROM contains product data sheets, application notes and Design Notes. Use your browser to view product categories and select products from parametric tables or simply choose products and documents from part number, application note or design note indexes. Linear Technology offers high-performance analog products across a broad product range. Current product information and design tools are available at www.linear.com. Our CD-ROM product selector tool, which is updated quarterly, and our most recent databook series can be obtained from your local Linear Sales office (see the back of this magazine) or requested from www.linear.com. www.linear.com Product information and application solutions are available at www.linear.com through powerful search tools, which yield weighted results from our data sheets, application notes, design notes, Linear Technology magazine issues and other LTC publications. The LTC website simplifies the product selection process by providing convenient search methods, complete application solutions and design simulation programs for power, filter, op amp and data converter applications. Search methods include a text search for a particular part number, keyword or phrase, or a powerful parametric search engine. After selecting a desired product category, engineers can specify and sort by key parameters and specifications that satisfy their design requirements. Purchase Products Online Credit Card Purchases—Purchase online direct from Linear Technology at www.linear.com using a credit card. Create a personalized account to check order history, shipment information and reorder products. Linear Express Distribution — Get the parts you need. Fast. Most devices are stocked for immediate delivery. Credit terms and low minimum orders make it easy to get you up and running. Place and track orders online. Apply today at www.linear.com or call (866) 546-3271. Applications Handbooks Linear Applications Handbook, Volume I — Almost a thousand pages of application ideas covered in depth by 40 Application Notes and 33 Design Notes. This catalog covers a broad range of real world linear circuitry. In addition to detailed, systems-oriented circuits, this handbook contains broad tutorial content together with liberal use of schematics and scope photography. A special feature in this edition includes a 22-page section on SPICE macromodels. Linear Applications Handbook, Volume II — Continues the stream of real world linear circuitry initiated by Volume I. Similar in scope to Volume I, this book covers Application Notes 40 through 54 and Design Notes 33 through 69. References and articles from non-LTC publications that we have found useful are also included. Linear Applications Handbook, Volume III — This 976-page handbook includes Application Notes 55 through 69 and Design Notes 70 through 144. Subjects include switching regulators, measurement and control circuits, filters, video designs, interface, data converters, power products, battery chargers and CCFL inverters. An extensive subject index references circuits in Linear data sheets, design notes, application notes and Linear Technology magazines. Linear Technology Magazine • March 2006 Brochures Power Management & Wireless Solutions for Handheld Products — The solutions in this product selection guide solve real-life problems for cell phones, digital cameras, PDAs and other portable devices, maximizing battery run time and saving space. Circuits are shown for LiIon battery chargers, battery managers, USB support, system power regulation, display drivers, white LED drivers, photoflash chargers, DC/DC converters and RF PA power supply and control. Automotive Electronic Solutions — This selection guide features high performance, high reliability solutions for a wide range of functions commonly used in today’s automobiles, including telematics, infotainment systems, body electronics, engine management, safety systems and GPS navigation systems. Industrial Signal Chain — This product selection guide highlights analog-to-digital converters, digital-to-analog converters, amplifiers, comparators, filters, voltage references, RMS-to-DC converters and silicon oscillators designed for demanding industrial applications. These precise, flexible and rugged devices feature parameters fully guaranteed over the –40°C to 85°C temperature range. Battery Charger Solutions — This guide identifies optimum charging solutions for single-cell batteries, multi-cell batteries and battery packs, regardless of chemistry. Linear offers a broad range of charger solutions, including linear chargers, linear chargers with regulators, pulse chargers, switchmode monolithic chargers, switchmode controller chargers, and switchmode smart battery chargers. Wireless & RF Solutions — This brochure presents high performance RF solutions for use in various transceiver architectures employed in 2G, 2.5G and 3G cellular basestations, wireless point-to-point radios, WiMAX, broadband wireless access, satellite receivers, GPS receivers, cable and VOD infrastructure equipment, RFID readers, wireless handheld transceivers and software defined radios. Software SwitcherCAD™ III/LTC SPICE — LTC SwitcherCAD III is a fully functional SPICE simulator with enhancements and models to ease the simulation of switching regulators. This SPICE is a high performance circuit simulator and integrated waveform viewer, and also includes schematic capture. Our enhancements to SPICE result in much faster simulation of switching regulators than is possible with normal SPICE simulators. SwitcherCAD III includes SPICE, macromodels for 80% of LTC’s switching regulators and over 200 op amp models. It also includes models of resistors, transistors and MOSFETs. With this SPICE simulator, most switching regulator waveforms can be viewed in a few minutes on a high performance PC. Circuits using op amps and transistors can also be easily simulated. Download at www.linear.com FilterCAD™ 3.0 — FilterCAD 3.0 is a computer aided design program for creating filters with Linear Technology’s filter ICs. FilterCAD is designed to help users without special expertise in filter design to design good filters with a minimum of effort. It can also help experienced filter designers achieve better results by playing “what if” with the configuration and values of various components and observing the results. With FCAD, you can design lowpass, highpass, bandpass or notch filters with a variety of responses, including Butterworth, Bessel, Chebychev, elliptic and minimum Q elliptic, plus custom responses. Download at www.linear.com SPICE Macromodel Library — This library includes LTC op amp SPICE macromodels. The models can be used with any version of SPICE for analog circuit simulations. These models run on SwitcherCAD III/LTC SPICE. Noise Program — This PC program allows the user to calculate circuit noise using LTC op amps, determine the best LTC op amp for a low noise application, display the noise data for LTC op amps, calculate resistor noise and calculate noise using specs for any op amp. Databooks Amplifiers (2 Books) — • Operational Amplifiers • Instrumentation Amplifiers • Application Specific Amplifiers References, Filters, Comparators, Special Functions, RF & Wireless — • Voltage References • Special Functions • Monolithic Filters • RF & Wireless • Comparators • Optical Communications • Oscillators Monolithic Switching Regulators — • Micropower Switching Regulators • Continuous Switching Regulators Switching Regulator Controllers (2 Books) — • DC/DC Controllers • Digital Voltage Programmers • Off-Line AC/DC Controllers Linear Regulators, Charge Pumps, Battery Chargers — • Linear Regulators • Charge Pump DC/DC Converters • Battery Charging & Management Hot Swap Controllers, MOSFET Drivers, Special Power Functions — • Hot Swap Controllers • Power Switching & MOSFET Drivers • PCMCIA Power Controllers • CCFL Backlight Converters • Special Power Functions Data Converters (2 Books) — • Analog-to-Digital Converters • Digital-to-Analog Converters • Switches & Multiplexers Interface, System Monitoring & Control — • Interface — RS232/562, RS485, Mixed Protocol, SMBus/I2C • System Monitoring & Control — Supervisors, Margining, Sequencing & Tracking Controllers 47 SALES OFFICES NORTH AMERICA GREATER BAY AREA Bay Area 720 Sycamore Dr. Milpitas, CA 95035 Phone: (408) 428-2050 FAX: (408) 432-6331 Sacramento Phone: (408) 432-6326 PACIFIC NORTHWEST Denver Phone: (303) 926-0002 Portland 6700 SW 105th Ave., Ste. 207 Beaverton, OR 97008 Phone: (503) 520-9930 FAX: (503) 520-9929 Salt Lake City Phone: (801) 731-8008 Seattle 2018 156th Ave. NE, Ste. 100 Bellevue, WA 98007 Phone: (425) 748-5010 FAX: (425) 748-5009 SOUTHWEST Los Angeles 21243 Ventura Blvd., Ste. 238 Woodland Hills, CA 91364 Phone: (818) 703-0835 FAX: (818) 703-0517 Orange County 15375 Barranca Pkwy., Ste. A-213 Irvine, CA 92618 Phone: (949) 453-4650 FAX: (949) 453-4765 San Diego 5090 Shoreham Place, Ste. 110 San Diego, CA 92122 Phone: (858) 638-7131 FAX: (858) 638-7231 CENTRAL Chicago 2040 E. Algonquin Rd., Ste. 512 Schaumburg, IL 60173 Phone: (847) 925-0860 FAX: (847) 925-0878 Cleveland 7550 Lucerne Dr., Ste. 106 Middleburg Heights, OH 44130 Phone: (440) 239-0817 FAX: (440) 239-1466 Columbus Phone: (614) 488-4466 Detroit 39111 West Six Mile Road Livonia, MI 48152 Phone: (734) 779-1657 Fax: (734) 779-1658 Indiana Phone: (317) 581-9055 Kansas Phone: (913) 829-8844 Minneapolis 7805 Telegraph Rd., Ste. 225 Bloomington, MN 55438 Phone: (952) 903-0605 FAX: (952) 903-0640 Wisconsin Phone: (262) 859-1900 NORTHEAST Boston 15 Research Place North Chelmsford, MA 01863 Phone: (978) 656-4750 FAX: (978) 656-4760 Connecticut Phone: (860) 228-4104 Philadelphia 3220 Tillman Dr., Ste. 120 Bensalem, PA 19020 Phone: (215) 638-9667 FAX: (215) 638-9764 SOUTHEAST Atlanta Phone: (770) 888-8137 Austin 8500 N. Mopac, Ste. 603 Austin, TX 78759 Phone: (512) 795-8000 FAX: (512) 795-0491 Dallas 17000 Dallas Pkwy., Ste. 200 Dallas, TX 75248 Phone: (972) 733-3071 FAX: (972) 380-5138 Fort Lauderdale Phone: (954) 473-1212 Houston 1080 W. Sam Houston Pkwy., Ste. 225 Houston, TX 77043 Phone: (713) 463-5001 FAX: (713) 463-5009 Huntsville Phone: (256) 885-0215 Orlando Phone: (407) 688-7616 Raleigh 15100 Weston Pkwy., Ste. 202 Cary, NC 27513 Phone: (919) 677-0066 FAX: (919) 678-0041 Tampa Phone: (813) 634-9434 ASIA EUROPE CHINA Linear Technology Corp. Ltd. Unit 2108, Metroplaza Tower 2 223 Hing Fong Road Kwai Fong, N.T., Hong Kong Phone: +852 2428-0303 FAX: +852 2348-0885 FINLAND Linear Technology AB Teknobulevardi 3-5 P.O. Box 35 FIN-01531 Vantaa Finland Phone: +358 (0)9 2517 8200 FAX: +358 (0)9 2517 8201 Linear Technology Corp. Ltd. Room 902, Peregrine Plaza 1325 Huaihai M. Road Shanghai, 200031, PRC Phone: +86 (21) 6375-9478 FAX: +86 (21) 5465-5918 FRANCE Linear Technology S.A.R.L. Immeuble “Le Quartz” 58, Chemin de la Justice 92290 Chatenay Malabry France Phone: +33 (1) 41 07 95 55 FAX: +33 (1) 46 31 46 13 Linear Technology Corp. Ltd. Room 511, 5th Floor Beijing Canway Building 66 Nan Li Shi Lu Beijing, 100045, PRC Phone: +86 (10) 6801-1080 FAX: +86 (10) 6805-4030 GERMANY Linear Technology GmbH Osterfeldstrasse 84, Haus C D-85737 Ismaning Germany Phone: +49 (89) 962455-0 FAX: +49 (89) 963147 Linear Technology Corp. Ltd. Rm. 2109, Shenzhen Kerry Centre 2008 Shenzhen Renminnan Lu Shenzhen, China Phone: +86 755-8236-6088 FAX: +86 755-8236-6008 Linear Technology GmbH Haselburger Damm 4 D-59387 Ascheberg Germany Phone: +49 (2593) 9516-0 FAX: +49 (2593) 951679 JAPAN Linear Technology KK 8F Shuwa Kioicho Park Bldg. 3-6 Kioicho Chiyoda-ku Tokyo, 102-0094, Japan Phone: +81 (3) 5226-7291 FAX: +81 (3) 5226-0268 Linear Technology GmbH Jesinger Strasse 65 D-73230 Kirchheim/Teck Germany Phone: +49 (0)7021 80770 FAX: +49 (0)7021 807720 Linear Technology KK 6F Kearny Place Honmachi Bldg. 1-6-13 Awaza, Nishi-ku Osaka-shi, 550-0011, Japan Phone: +81 (6) 6533-5880 FAX: +81 (6) 6543-2588 ITALY Linear Technology Italy Srl Via Giovanni da Udine, 34 I-20156 Milano Italy Phone: +39 (02) 38093656 FAX: +39 (02) 38093659 Linear Technology KK 7F, Sakuradori Ohtsu KT Bldg. 3-20-22 Marunouchi, Naka-ku Nagoya-shi, 460-0002, Japan Phone: +81 (52) 955-0056 FAX: +81 (52) 955-0058 KOREA Linear Technology Korea Co., Ltd. Yundang Building, #1002 Samsung-Dong 144-23 Kangnam-Ku, Seoul 135-090 Korea Phone: +82 (2) 792-1617 FAX: +82 (2) 792-1619 SINGAPORE Linear Technology Pte. Ltd. 507 Yishun Industrial Park A Singapore 768734 Phone: +65 6753-2692 FAX: +65 6752-0108 SWEDEN Linear Technology AB Electrum 204 Isafjordsgatan 22 SE-164 40 Kista Sweden Phone: +46 (8) 623 16 00 FAX: +46 (8) 623 16 50 UNITED KINGDOM Linear Technology (UK) Ltd. 3 The Listons, Liston Road Marlow, Buckinghamshire SL7 1FD United Kingdom Phone: +44 (1628) 477066 FAX: +44 (1628) 478153 TAIWAN Linear Technology Corporation 8F-1, 77, Nanking E. Rd., Sec. 3 Taipei, Taiwan Phone: +886 (2) 2505-2622 FAX: +886 (2) 2516-0702 Linear Technology Corporation 1630 McCarthy Blvd. Milpitas, CA 95035-7417 TEL: (408) 432-1900 FAX: (408) 434-0507 © 2006 Linear Technology Corporation/Printed in U.S.A./30.5K www.linear.com Linear Technology Magazine • March 2006