® AMD-K6 ® Processor Power Supply Design Application Note Publication # 21103 Rev: G Issue Date: February 1999 Amendment/0 The contents of this document are provided in connection with Advanced Micro Devices, Inc. ("AMD") products. AMD makes no representations or warranties with respect to the accuracy or completeness of the contents of this publication and reserves the right to make changes to specifications and product descriptions at any time without notice. No license, whether express, implied, arising by estoppel or otherwise, to any intellectual property rights is granted by this publication. Except as set forth in AMD’s Standard Terms and Conditions of Sale, AMD assumes no liability whatsoever, and disclaims any express or implied warranty, relating to its products including, but not limited to, the implied warranty of merchantability, fitness for a particular purpose, or infringement of any intellectual property right. AMD’s products are not designed, intended, authorized or warranted for use as components in systems intended for surgical implant into the body, or in other applications intended to support or sustain life, or in any other application in which the failure of AMD’s product could create a situation where personal injury, death, or severe property or environmental damage may occur. AMD reserves the right to discontinue or make changes to its products at any time without notice. © 1999 Advanced Micro Devices, Inc. All rights reserved. Trademarks AMD, the AMD logo, K6, and combinations thereof are trademarks, and AMD-K6 is a registered trademark of Advanced Micro Devices, Inc. Windows and Windows NT are registered trademarks of Microsoft Corporation. Winstone is a registered trademark of Ziff-Davis, Inc. Other product names used in this publication are for identification purposes only and may be trademarks of their respective companies. 21103G/0—February 1999 AMD-K6® Processor Power Supply Design Contents Revision History . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 Processor Power Requirement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 Voltage Planes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 Power Supply Specification . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5 Selecting a Power Supply Design . . . . . . . . . . . . . . . . . . . 5 Linear Regulator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 Switching Regulator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 Switching Regulator Layout . . . . . . . . . . . . . . . . . . . . . . 10 Decoupling and Layout Recommendations . . . . . . . . . . . . . . . . . . . . 11 Power Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 Contents Current Transient Response . . . . . . . . . . . . . . . . . . . . . . Examples . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Output Voltage Response Measurement Techniques . . . . . . Output Voltage Response Measurement Utility. . . . . . Decoupling Capacitance and Component Placement. . . . . . . High-Frequency Decoupling . . . . . . . . . . . . . . . . . . . . . . Power Sequencing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Power Supply Solutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 12 18 19 20 23 26 27 Digital-to-Analog Converter (DAC) . . . . . . . . . . . . . . . . Cherry CS5166 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Contact Information . . . . . . . . . . . . . . . . . . . . . . . . . . . . Elantech EL7571 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Contact Information . . . . . . . . . . . . . . . . . . . . . . . . . . . . Harris Semiconductor HIP6004 and HIP6005 . . . . . . . Contact Information . . . . . . . . . . . . . . . . . . . . . . . . . . . . Linear Technology LT1553 . . . . . . . . . . . . . . . . . . . . . . . Contact Information . . . . . . . . . . . . . . . . . . . . . . . . . . . . LINFINTY LX1664 and LX1665. . . . . . . . . . . . . . . . . . . Contact Information . . . . . . . . . . . . . . . . . . . . . . . . . . . . Maxim MAX1638 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Contact Information . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27 28 28 30 31 32 33 34 34 36 36 38 38 iii AMD-K6® Processor Power Supply Design 21103G/0—February 1999 Micro Linear ML4902 . . . . . . . . . . . . . . . . . . . . . . . . . . . Contact Information . . . . . . . . . . . . . . . . . . . . . . . . . . . . Fairchild RC5051 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Contact Information . . . . . . . . . . . . . . . . . . . . . . . . . . . . Semtech SC1182 and SC1183 . . . . . . . . . . . . . . . . . . . . . Contact Information . . . . . . . . . . . . . . . . . . . . . . . . . . . . Unisem US3004 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Contact Information . . . . . . . . . . . . . . . . . . . . . . . . . . . . Unitrode UCC3880 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Contact Information . . . . . . . . . . . . . . . . . . . . . . . . . . . . Voltage Regulator Vendor Information . . . . . . . . . . . . . . . . . . . . . . . iv 41 41 43 44 45 45 48 48 51 52 53 Contents AMD-K6® Processor Power Supply Design 21103G/0—February 1999 List of Figures Figure 1. 321-Pin CPGA VCC and Ground Pins Location. . . . . . . . . 4 Figure 2. Linear and Switching Voltage Regulators. . . . . . . . . . . . . 6 Figure 3. Basic Asynchronous Design. . . . . . . . . . . . . . . . . . . . . . . . . 9 Figure 4. Basic Synchronous Design . . . . . . . . . . . . . . . . . . . . . . . . . 10 Figure 5. Power Distribution Model . . . . . . . . . . . . . . . . . . . . . . . . . 11 Figure 6. Load Current Step versus Output Voltage Response. . . 15 Figure 7. Bulk Decoupling versus Output Voltage Response for 3.2 V @10 A . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16 Figure 8. Bulk Decoupling versus Output Voltage Response for 2.2 V @7.5 A . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 Figure 9. Bulk Decoupling versus Output Voltage Response for 2.4 V @15 A . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18 Figure 10. Via Layout For Low Inductance . . . . . . . . . . . . . . . . . . . . 21 Figure 11. Suggested Component Placement . . . . . . . . . . . . . . . . . . 22 Figure 12. 0.1 µF (c1) and 0.01 µF (c2) X7R Capacitor Impedance versus Frequency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 Figure 13. Decoupling Inductance . . . . . . . . . . . . . . . . . . . . . . . . . . . 26 Figure 14. Power Sequencing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26 Figure 15. Cherry CS5166 Switching Power Supply Design . . . . . . 29 Figure 16. Elantec EL7571 Switching Power Supply Design . . . . . . 30 Figure 17. Harris HIP6004 1.3V–3.5V Switching Power Supply Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32 Figure 18. Linear LT1553 1.8V to 3.5V Switching Power Supply Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 Figure 19. LINFINITY LX1664 Switch-Mode Power Supply Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37 Figure 20. Maxim MAX1638 Switching Power Supply . . . . . . . . . . . 40 Figure 21. Micro Linear ML4902 Switching Power Supply Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41 Figure 22. Fairchild RC5051 Power Supply Design . . . . . . . . . . . . . 43 Figure 23. Semtech SC1182 Voltage Power Supply Design . . . . . . . 46 List of Figures v AMD-K6® Processor Power Supply Design 21103G/0—February 1999 Figure 24. Unisem US3004 Dual Supply Design . . . . . . . . . . . . . . . . 50 Figure 25. Unitrode UCC3880 Switching Power Supply . . . . . . . . . 51 vi List of Figures AMD-K6® Processor Power Supply Design 21103G/0—February 1999 List of Tables Table 1. Voltage Error Budget for 0.35-Micron Processors (Model 6) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 Table 2. Voltage Error Budget for 0.25-Micron Processors (Models 7, 8, 9) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16 Table 3. Representative ESR Values. . . . . . . . . . . . . . . . . . . . . . . . 20 Table 4. Inductance Contributions of Components . . . . . . . . . . . . 21 Table 5. Decoupling Capacitor Values . . . . . . . . . . . . . . . . . . . . . . 22 Table 6. Capacitor Recommendations . . . . . . . . . . . . . . . . . . . . . . 23 Table 7. Voltage Output VID Codes . . . . . . . . . . . . . . . . . . . . . . . . 27 Table 8. Cherry CS5166 Bill of Materials . . . . . . . . . . . . . . . . . . . . 28 Table 9. Elantec EL7571 Bill of Materials . . . . . . . . . . . . . . . . . . . 31 Table 10. Harris HIP6004 Bill of Materials . . . . . . . . . . . . . . . . . . . 33 Table 11. Linear LT1553 Bill of Materials . . . . . . . . . . . . . . . . . . . . 34 Table 12. LINFINITY LX1664 Bill Of Materials . . . . . . . . . . . . . . . 36 Table 13. Maxim MAX1638 Bill of Materials . . . . . . . . . . . . . . . . . . 39 Table 14. Micro Linear ML4902 Bill of Materials . . . . . . . . . . . . . . 42 Table 15. Fairchild RC5051 Bill of Materials . . . . . . . . . . . . . . . . . . 44 Table 16. Semtech SC1182 Bill Of Materials . . . . . . . . . . . . . . . . . . 47 Table 17. LDO Voltage Selection. . . . . . . . . . . . . . . . . . . . . . . . . . . . 47 Table 18. Unisem US3004 Bill of Materials . . . . . . . . . . . . . . . . . . . 48 Table 19. Unitrode UCC3880 Bill of Materials . . . . . . . . . . . . . . . . 52 List of Tables vii AMD-K6® Processor Power Supply Design viii 21103G/0—February 1999 List of Tables AMD-K6® Processor Power Supply Design 21103G/0—February 1999 Revision History Date Rev March 1998 E Changed general reference voltage 2.x V to 2.2 V. March 1998 E Revised Example 4, “Actual 2.2 V @ 7.5A” on page 13. March 1998 E Revised Table 2, “Voltage Error Budget for 0.25-Micron Processors (Models 7 and 8),” on page 16. March 1998 E Revised Figure 8, “Bulk Decoupling versus Output Voltage Response for 2.2 V @7.5 A,” on page 17. March 1998 E Revised Figure 17, “Linear LT1575 2.2V/2.9V/3.2V Linear Power Supply Design,” on page 34. March 1998 E Revised Figure 18, “Linear LT1553 1.8V to 3.5V Switching Power Supply Design,” on page 35. May 1998 F Revised to provide information for the AMD-K6®-2 processor Model 8. May 1998 F Removed Table 1, “AMD-K6® Processor Power Specifications”. The voltage and current specifications for Models 6 and 7 are provided in the AMD-K6® Processor Data Sheet, order # 20695. The voltage and current specifications for Model 8 are provided in the AMD-K6®-2 Processor Data Sheet, order # 21850. May 1998 F Expanded information in “Power Supply Specification” starting on page 5. May 1998 F Added Example 5 “Hypothetical 2.3 V @ 15 A” on page 17 and Figure 9 on page 18. May 1998 F Added the following power supply solutions: “Cherry CS5166” on page 28, “Harris Semiconductor HIP6004 and HIP6005” on page 32, “Linear Technology LT1553” on page 34, “Maxim MAX1638” on page 38, “Fairchild RC5051” on page 43, “Semtech SC1182 and SC1183” on page 45, and “Unisem US3004” on page 48. May 1998 F Cut the following power supply solutions: Cherry CS5151/CS5156, Harris Semiconductor HIP6003, Linear Technology LT1575 and LT1430, Maxim MAX1624, Raytheon RC5036 and RC5041, Semtech SC1151, and Unisem US2075 May 1998 F Revised description of “LINFINTY LX1664 and LX1665” on page 36. May 1998 F Added Table 14, “Micro Linear ML4902 Bill of Materials,” on page 42. May 1998 F Combined and revised voltage regulator vendor information into one table. See “Voltage Regulator Vendor Information” on page 53. Feb 1999 G Added information about the AMD-K6-III processor Model 9. Feb 1999 G Added information about the 5-bit VID code on page 2. Feb 1999 G Added “Switching Regulator Layout” on page 10. Feb 1999 G Added information on determining the number of capacitors to Example 2 on page 13. Feb 1999 G Changed Example 5 to 2.4 V and changed Figure 9, “Bulk Decoupling versus Output Voltage Response for 2.4 V @15 A” on page 18. Feb 1999 G Changed the recommended utility in “Output Voltage Response Measurement Utility” on page 19. Revision History Description ix AMD-K6® Processor Power Supply Design x 21103G/0—February 1999 Revision History AMD-K6® Processor Power Supply Design 21103G/0—February 1999 Application Note AMD-K6 ® Processor Power Supply Design Introduction Unless otherwise noted, the information in this application note pertains to all desktop processors in the AMD-K6® family, which includes the AMD-K6 processor (Models 6, and 7), the AMD-K6-2 processor (Model 8) and the AMD-K6-III processor (Model 9). For information about mobile processor power supply considerations, see the Mobile AMD-K6® Processor Power Supply A pplic ation Note , o rder# 216 77 and t he Mobil e AMD-K6 ® -2 Processor Power Supply Application Note, order# 22495 Processors in the AMD-K6 family are high-performance x86-compatible processors with over 8.8 million transistors. The newer generation of processors manufactured with the CS44E 0.25-micron (µm) process uses 2.2 volts (V) to power the core circuitry of the processor while the I/O portion operates at the industry-standard 3.3 V. The previous 2.9 V and 3.2 V AMD-K6 processors were fabricated using AMD’s enhanced 0.35-µm process technology. Due to the large number of transistors that can switch simultaneously, power supply designs must meet large transient power requirements. This application note is intended to guide the board designer through the process of developing a reliable power supply that meets the low-voltage, high-current demands of the AMD-K6 Introduction 1 AMD-K6® Processor Power Supply Design 21103G/0—February 1999 processors. The goal is to design a solution that works over a wide voltage range and a 5.8 amps (A) to 14 A current range. (Previously, the suggested range was 5.8A to 10A. This change allows motherboard designers to prepare for the next generation of processors.) This application note also provides basic guidelines on circuit decoupling for reduction of noise generated by fast current transients. The core voltage for the 0.25-µm process is 2.2V/2.4V. However, AMD encourages designers to provide flexibility to support multiple voltages in their designs. This flexibility may entail a resistor-value change or changing the location of a zero-ohm resistor or a jumper. By providing flexibility in the power design, future lower voltage parts may be able to be used with little or no changes to the motherboard. As process geometries continue to shrink, the core voltages are planned to drop. An easy way to prepare for this is to use controllers that implement the 5-bit VID code. For core voltage specifications for the following AMD-K6 processors, refer to: ■ ■ ■ Models 6 and 7 — AMD-K6® Processor Data Sheet, order# 20695 Model 8 — AMD-K6®-2 Processor Data Sheet, order# 21850 Model 9 — AMD-K6®-III Processor Data Sheet, order# 21918 This document contains the following sections: ■ ■ ■ 2 Power Supply Specification on page 5—Gives an overview of power supply design considerations. This section describes the basic elements of a power supply and the constraints of different design approaches. Decoupling and Layout Recommendations on page 11— Describes the decoupling and layout recommendations of the power supply design. Proper decoupling is required in order to deliver a reliable power source across the power planes and to reduce the noise generated from the fast current transients. Power Supply Solutions on page 27—Describes several voltage regulator circuits that are designed by voltage regulator vendors. These circuits can be used to generate the proper core and I/O voltages for the processor. Because the information provided is preliminary, AMD recommends that board designers consult with the voltage regulator vendors to obtain the most up-to-date information. Introduction AMD-K6® Processor Power Supply Design 21103G/0—February 1999 Processor Power Requirement Voltage Planes Two separate supply voltages are required to support the processor—VCC2 and VCC3. VCC2 provides the core voltage for the processor and VCC3 provides the I/O voltage. The power supply pin assignments for the 321-pin CPGA package (See Figure 1) are as follows: VCC2 (Core): A-07, A-09, A-11, A-13, A-15, A-17, B-02, E-15, G-01, J-01, L-01, N-01, Q-01, S-01, U-01, W-01, Y-01, AA-01, AC-01, AE-01, AG-01, AJ-11, AN-09, AN-11, AN-13, AN-15, AN-17, AN-19 VCC3 (I/O): A-19, A-21, A-23, A-25, A-27, A-29, E-21, E-27, E-37, G-37, J-37, L-33, L-37, N-37, Q-37, S-37, T-34, U-33, U-37, W-37, Y-37, AA-37, AC-37, AE-37, AG-37, AJ-19, AJ-29, AN-21, AN-23, AN-25, AN-27, AN-29 Processor Power Requirement 3 AMD-K6® Processor Power Supply Design 21103G/0—February 1999 Figure 1. 321-Pin CPGA VCC and Ground Pins Location 4 Processor Power Requirement AMD-K6® Processor Power Supply Design 21103G/0—February 1999 Power Supply Specification For voltage and current specifications for the following AMD-K6 processors, refer to: ■ Models 6 and 7 — AMD-K6® Processor Data Sheet, order# 20695 ■ Model 8 — AMD-K6®-2 Processor Data Sheet, order# 21850 Model 9 — AMD-K6®-III Processor Data Sheet, order# 21918 ■ AMD’s processors have two pins that indicate the voltage requirements of the device. VCC2DET#, when asserted low, indicates that the core voltage is different than the I/O voltage. The VCC2DET# pin is available on 0.35 µm processors (Model 6) that operate at 2.9 V or 3.2 V. Along with VCC2DET#, the 0.25µm devices (Models 7, 8, and 9) have an additional pin— VCC2H/L#. When asserted low, VCC2H/L# indicates a 2.2V/2.4V processor core voltage. On 0.35µm devices, this pin is a No Connect. Selecting a Power Supply Design Most PC platforms today require DC-to-DC voltage conversion circuits to supply lower voltages to the processor core and I/O. Two types of regulators are used—linear and switching. A linear regulator provides excellent dynamic-load response in the low-voltage, high-current environment. It also contributes to simplified design and lower cost. However, the efficiency loss and heat generated by a linear regulator should be addressed by board designs. Although most desktop system designs can tolerate the efficiency loss, care should be taken to ensure the design can handle the heat. In a high-current model, the power dissipation from the regulator can be as much as that of the processor itself. In order for the voltage regulator thermal solution to meet the case temperature requirement, the linear regulator requires a larger heatsink. As processor voltages drop and currents increase, it becomes more difficult to implement a linear solution. Linear regulator solutions are impractical for currents above 7 A. A switching regulator meets the efficiency and size limitations of mobile board designs and is also an excellent choice for desktop designs. Switching regulators are found in most notebook computers that require both low-profile design and power dissipation reduction. Figure 2 shows linear and switching regulators. The switching regulator uses a series Processor Power Requirement 5 AMD-K6® Processor Power Supply Design 21103G/0—February 1999 switch in conjunction with the output capacitor (CO) to control the ON/OFF ratio in order to obtain an average output voltage. Because the switch turns off frequently, only a small amount of power is lost during conversion. + + VIN Control – Feedback RL VOUT – Efficiency = Linear Regulator VOUT VIN + + VIN CO RL VOUT – – Switching Regulator Figure 2. Linear and Switching Voltage Regulators As the trend toward smaller process geometries continues (0.35-micron to 0.25-micron), the processor core voltage will continue to drop. To provide maximum flexibility for upgrading a motherboard, regulator controllers with the 5-bit VID code are preferable. Using this feature, processors that have not yet been announced can be supported, as long as they do not exceed the current limit of the design. Designing a point solution (such as, 2.9 V @ 7.5 A) eliminates many design variables, however, this approach limits flexibility and upgradeability. There are two strategies for extending the life of a motherboard while retaining low cost. The first strategy entails designing the board for the maximum current anticipated. This approach increases the cost because the components used are more expensive and may be physically larger, therefore occupying more room. The second strategy entails the development of two designs — one that operates at 10 A and one that operates at 15A. The motherboard can be laid out to accept components for 6 Processor Power Requirement AMD-K6® Processor Power Supply Design 21103G/0—February 1999 either design. With this approach, a simple bill of material change is all that is necessary to upgrade to a higher-power processor. One of the key motherboard components is the power transistor. The transistor can be replaced with one that has a lower RDS(ON) (resistance-drain-to-source when the transistor is on) or two transistors can be paralleled. Another important component is the output inductor. Because an inductor that carries 15 A is physically larger than an inductor that carries 10A, the layout must allow sufficient space. Finally, a provision should be made to add extra decoupling capacitors. The calculations in the examples starting on page 12 show how many decoupling capacitors are needed for various cases. M a ny o f t h e c o m p o n e n t s a re c o m m o n , i n c l u d i n g t h e regulator/controller IC and the basic circuitry. Typically, switching transistors and the output inductor need to change. The output filter capacitance needs to be increased for the higher currents. Linear Regulator The linear regulator relies on a linear series component to continuously drive the power to a load. The series component is considered a load, and the voltage drop between the input and o u t p u t re p re s e n t s t h e p o we r l o s s . Th e h i g h e r t h e input-to-output voltage ratio, the lower the conversion efficiency. In order to meet the voltage requirement, output feedback to the control unit is commonly used to obtain an accurate (and adjustable) voltage output. For a linear regulator, converting a 5-V source to 3.3V results in a 66% conversion efficiency and a 34% power loss (See Figure 2 on page 6). The efficiency of the conversion gets worse if the output voltage is lower than 3.3V. The low dropout (LDO) linear regulator is a reasonable solution for providing the processor core voltage in systems that already support 3.3 V from the silver-box power supply or in systems converted from an existing 3.3V design to a lower voltage. Heat is an additional consideration. The voltage drop between the input and output multiplied by the current supplied is the power that must be dissipated by the regulator. For example, when converting 5V to 2.2V at 6A, the power dissipated is (5V – 2.2 V) • 6 A = 16.8 W. Therefore, linear regulators often have large heat sinks. This heat raises the ambient air temperature, Processor Power Requirement 7 AMD-K6® Processor Power Supply Design 21103G/0—February 1999 making it more difficult to cool the processor. Consider the example of converting 5 V to 2.2 V at 10 A. In this case, the power dissipated is (5V – 2.2 V) • 10 A = 28W. This heat makes using linear regulators impractical in many systems with these larger currents. To make a design that accommodates a wide range of processors, a switching design is preferable. A n o t h e r c o n s i d e ra t i o n f o r l i n e a r s u p p l i e s i n vo l ve s high-frequency decoupling on the input to the regulator. Noise from a 5-V supply can pass through a linear regulator to the processor. Generally, there is no high-frequency decoupling on the input of a power supply. A switching design seems to be less susceptible to this type of noise. Although linear regulators are good solutions at 2.2V and 7.5A, AMD does not recommend them as a desktop solution because of their lack of flexibility. Typically, a desktop motherboard should work with all available processors. A linear regulator makes such flexibility difficult to achieve while staying within heat constraints. However, a switching regulator designed for 3.2V at 14A can also accommodate a 2.2V, 7.5A processor. Switching Regulator A switching regulator varies the switch duty cycle (ON/OFF ratio) according to the output feedback. A large output capacitor (C O ) is used in the switching design to achieve a constant average output. The switching regulator delivers higher efficiency than a linear regulator, but the tradeoffs are higher ripple voltages (noise) and slower transient current response time. A series inductor is used to supply current to the load during the switch OFF time, adding complexity to the design. In addition, the inductor and the output capacitor increase the overall cost of the switching regulator design relative to a linear regulator design. The power supply design must account for a low current (ICC2 and ICC3) drain when the processor enters the Stop Grant state. The power supply must ensure the minimal current drain does not cause any adverse side effects (drift out of regulation, over-compensation, or shutdown) that could corrupt or damage the functionality of the processor. The processor voltage tolerance requirement on both core and I/O voltage pins can be handled by commonly available linear and switching regulators. This application note describes 8 Processor Power Requirement AMD-K6® Processor Power Supply Design 21103G/0—February 1999 several high-accuracy designs that provide the processor with accurate and stable voltage supplies. In the basic asynchronous circuit design shown in Figure 3, Q1 turns on to charge Cout and builds up the magnetic field in L1. When the feedback from the sense input is too high the controller turns Q1 off. Current is supplied to the load by the collapsing magnetic field in L1 and the discharge of Cout. When the sense feedback detects a drop in the load voltage, the controller turns on Q1 to recharge the circuit. CR2 supplies a return path for L1 when it is suppling current. The main reason this design is less efficient than a synchronous design is because the power dissipated in CR2 is higher than Q2 in the synchronous design. Q1 Sense Controller CR2 Cout RL Figure 3. Basic Asynchronous Design The operation of the basic synchronous circuit design shown in Figure 4 on page 10 is essentially the same as the asynchronous design. Q1 turns on to charge Cout and builds up the magnetic field in L1. When the feedback from the sense input is too high, the controller turns Q1 off. Current is supplied to the load by the collapsing magnetic field in L1 and the discharge of C out. When the sense feedback detects a drop in the load voltage, the controller turns on Q1 to recharge the circuit. Q2 supplies a return path for L1 when it is supplying current. When Q1 is on, Q2 is off and when Q1 is off, Q2 is on. The main reason this design is more efficient is because the power dissipated in Q2 is lower than the power in CR2. Processor Power Requirement 9 AMD-K6® Processor Power Supply Design 21103G/0—February 1999 Vout L1 Sense Controller Cout Q2 RL Figure 4. Basic Synchronous Design Another consideration is power dissipation in the lower MOSFET (synchronous) or diode (asynchronous). As the output voltage decreases the power dissipation in CR2 (Q2) increases. The higher power dissipation may require using a different package type or adding a heat sink to dissipate the additional power. To determine if the transistors or the diode need a heat sink use the following equation: P = I2R • duty cycle (Q1) P = I2R • (1– duty cycle) (Q2) Duty cycle ~ Vout/Vin Compare these calculations with the specifications of the device used. Switching Regulator Layout 10 Each manufacturer has example layouts. Since the layout is critical for stability and performance, AMD recommends working closely with the manufacturer. For more information on switching regulator layouts, refer to “Board Layout Boost Power-Supply Performance” by Philip Rogers in the Nov. 5, 1998 issue of EDN (www.ednmag.com). Processor Power Requirement AMD-K6® Processor Power Supply Design 21103G/0—February 1999 Decoupling and Layout Recommendations Power Distribution In order to maintain a stable voltage supply during fast transients, power planes with high frequency and bulk decoupling capacitors are required. Figure 5 shows a power distribution model for the power supply and the processor. The bulk capacitors (C B ) are used to minimize ringing, and the processor decoupling capacitors (CF) are spread evenly across the circuit to maintain stable power distribution. Power PCB Trace + VOUT CB + Plane CF Processor RL – Equivalent Circuit Power Plane PCB Trace RTRACE Processor LTRACE ESR + VOUT CB ESL CF CL RL – Figure 5. Power Distribution Model Decoupling and Layout Recommendations 11 AMD-K6® Processor Power Supply Design Current Transient Response 21103G/0—February 1999 In the power distribution model shown in Figure 5 on page 11, C B represents bulk capacitors for the power supply and C F represents high-frequency capacitors for processor decoupling. The bulk capacitors supply current to the processor during sudden excessive current demands that cannot be supplied by the voltage regulator (for example, transitioning from the Stop Grant state to normal mode). The required CB can be calculated by the following equation (ideal case): C ≥ ∆I ∆V • ∆t Where: ■ ■ ■ ∆I is the maximum processor current transient ∆V is the tolerance times the nominal processor voltage ∆t is the voltage regulator response time Examples The following examples are not the only solutions. Based on the availability of parts and the choice of controller, many correct solutions are possible. The examples, which use tantalum capacitors, are intended to give insight into the requirements, not to specify a particular solution. The use of aluminum electrolytic capacitors are acceptable as long as good quality, low-ESR parts are used. Example 1 Theoretical 3.2V @ 10A Assuming the maximum processor current transient is 10A, the voltage tolerance of the processor is less than 100 mV (3% of 3.2 V), and the voltage regulator response time is 10 µs, the minimum capacitance for the bulk decoupling is: CB ≥ (10A/0.100V) • 10µs = 1000µF ESR (equivalent series resistance) and ESL (equivalent series inductance) are introduced in the model shown in Figure 5. CB contains ESR and ESL, which cause voltage drop during current transient activity (See Figure 6 on page 15). The resistive and inductive effect of the capacitors must be taken into account when designing processor decoupling. Low ESL and ESR capacitors should be used to obtain better voltage and current output characteristics. The voltage error budget for ESL is 12 Decoupling and Layout Recommendations AMD-K6® Processor Power Supply Design 21103G/0—February 1999 shown in Table 1 on page 15. Taking into account the ESR, the following equation is used to calculate CB: C ≥ Example 2 Actual 3.2V @ 10A ∆I (∆V – (∆I • ESR)) • ∆t This example assumes the maximum processor current transient is 10A, the voltage tolerance of the processor is less than 100 mV (3.2 V ± 100 mV), and the voltage regulator response time is 10µs. Using ten tantalum capacitors with 80-mΩ ESR (the parallel resistance is 8 mΩ) as bulk capacitors, the minimum bulk capacitance is: CO ≥ ((10A/(0.100V – [10A • 8mΩ])) • 10µs = 5000µF In this example, the high current transient combined with the tight regulation specification requires significantly more decoupling capacitance than what is shown in Example 3. Therefore, ten (5000/470=10.6) 470-µF 55-mΩ capacitors are required to satisfy this curre nt transient and voltage requirement. It is possible here to use either 10 or 11 capacitors. For the worst case, the correct approach is to round up giving 11 capacitors. However, experience shows that rounding down may be sufficient as it is extremely unlikely that all capacitors will be at the maximum ESR. Example 3 Actual 2.9V @ 7.5 A This example assumes the maximum processor current transient is 7.5A, the voltage tolerance of the processor is less than 145 mV (5% of 2.9V), and the voltage regulator response time is 10µs. Using four tantalum capacitors with 60-mΩ ESR (the parallel resistance is 15 mΩ) as bulk capacitors, the minimum bulk capacitance is: CO ≥ ((7.5A/(0.145V – [7.5A • 15mΩ])) 10µs = 2300µF Five 470-µF tantalum capacitors with 55-mΩ ESR meet this requirement. However, if the brand of capacitor is changed to one with a 100-mΩ ESR, the supply is out of tolerance. Decoupling and Layout Recommendations 13 AMD-K6® Processor Power Supply Design Example 4 Actual 2.2 V @ 7.5A 21103G/0—February 1999 This example assumes a device with a maximum processor current transient of 7.5A, the voltage tolerance of the processor is less than 100 mV, and the voltage regulator response time is 10µs. Using six tantalum capacitors with 60-mΩ ESR (the parallel resistance is 10 mΩ) as bulk capacitors, the minimum bulk capacitance is: CO ≥ ((7.5A/(0.100V – [7.5A • 10mΩ])) 10µs = 3000µF Six 470-µF tantalum capacitors with 55-mΩ ESR meet this requirement. However, if the brand of capacitor is changed to one with a 100-mΩ ESR, the supply is out of tolerance. Therefore, when designing a system that supports only 2.2 V devices, the required bulk decoupling is significantly less than the bulk decoupling for the higher voltage and higher current parts. Note that the voltage tolerance is an important factor. Because of the higher Vcc2 tolerance in example 3 the decoupling requirement is slightly less than the 2.2V case. Note: The denominator of the C0 equation cannot be a negative value, which implies a negative capacitor (such as a battery). In order to achieve greater margin, the total error budget should be distributed between set point tolerance, ESL, and ESR as shown in Figure 6 and Table 1 on page 15. Although the drop from ESL is a small factor, it is not negligible. If aluminum electrolytic capacitors are used instead of tantulum capacitors, the ESL drop is larger. The high-frequency decoupling capacitors (C F ), which are typically smaller in capacitance and ESL, maintain the voltage output during average load change until C B can react. See “ H i g h -Fre q u e n cy D e c o u p l i n g ” o n p a g e 2 3 f o r m o re information. 14 Decoupling and Layout Recommendations AMD-K6® Processor Power Supply Design 21103G/0—February 1999 (Max) ICC (Min) Load Current (Max) Output Voltage ∆I = ∆V C ∆t VCC Response ESR x ∆I ESL x ∆I ∆t Voltage Regulator Response (Min) Figure 6. Load Current Step versus Output Voltage Response Allocation of the voltage error budget can be determined from Figure 6 on page 15. Given a total error budget of 100mV and using good capacitors (ten 470-µF capacitors with a 55-mΩ ESR are assumed), voltage drops for a 0.35-µm processor can be allocated as shown in Table 1. Table 1. Voltage Error Budget for 0.35-Micron Processors (Model 6) Error Budget Component V (Set Point) Calculations* 1% 5.5 mΩ x 10 A V(ESR) (5.5 mΩ = 55mΩ / 10) 0.12 nH x (10 A/10 nsec) V(ESL) {0.12 nH = (0.6 nH + 0.6 nH via) / 10} Total Budgeted Drop 0.032 V 0.055 V 0.012 V 0.099 V Note: * Calculations assume 10 capacitors Figure 7 on page 16 shows the voltage drop as a function of bulk decoupling for the 3.2V case. The graph was calculated using 55-mΩ ESR, 470-µF capacitors, and gives the designer a visual representation of how much bulk decoupling is needed. For example, at 2820 µF, the voltage is 3.1 V (10 A current transient), leaving no margin for DC-tolerance errors. At 4700 µF, the voltage is 3.144 V, allowing 0.058mV for set point tolerance, ESR, and ESL drop. Decoupling and Layout Recommendations 15 AMD-K6® Processor Power Supply Design 21103G/0—February 1999 Output Voltage vs. Capacitance 3.2 3.1 3 Voltage 2.9 2.8 Voltage 2.7 2.6 2.5 2.4 470 940 1410 1880 2350 2820 3290 3760 4230 4700 5170 5640 Capacitance in m icro Farads Figure 7. Bulk Decoupling versus Output Voltage Response for 3.2 V @10 A Table 2 shows an error budget calculation for a 0.25-µm processor. The example uses seven, 470-µF capacitors. Table 2. Voltage Error Budget for 0.25-Micron Processors (Models 7, 8, 9) Error Budget Component V (Set Point) V(ESR) V(ESL) Calculations* 1% 0.022 V 7.86 mΩ x 7.5A 0.059 V (7.86 mΩ = 55mΩ / 7) 0.2 nH x (7.5 A/10 nsec) {0.2 nH = (0.7 nH + 0.7 nH via) / 7} Total Budgeted Drop 0.015 V 0.096 V Note: * Calculations assume 7 capacitors Figure 8 on page 17 shows the voltage drop as a function of bulk decoupling for the 2.2V case. The graph was calculated using 55-mΩ ESR, 470-µF capacitors. 16 Decoupling and Layout Recommendations AMD-K6® Processor Power Supply Design 21103G/0—February 1999 Output Voltage vs. Capacitance 2.2 2.15 2.1 2.05 Voltage 2 1.95 Voltage 1.9 1.85 1.8 1.75 1.7 0 470 940 1410 1880 2350 2820 3290 3760 4230 4700 Capacitance in micro Farads Figure 8. Bulk Decoupling versus Output Voltage Response for 2.2 V @7.5 A Example 5 Hypothetical 2.4 V @ 15A This example assumes a device with a maximum processor current transient of 15A, the voltage tolerance of the processor is less than 100 mV, and the voltage regulator response time is 10µs. Using twelve tantalum capacitors with 60-mΩ ESR (the parallel resistance is 5 mΩ) as bulk capacitors, the minimum bulk capacitance is: CO ≥ ((15A/(0.100V – [15A • 5mΩ])) 10µs = 6000µF Twelve 470-µF tantalum capacitors with 55-mΩ ESR meet this requirement. Figure 9 on page 18 shows the voltage curve for this case. Note: The denominator of the C0 equation cannot be a negative value, which implies a negative capacitor (such as a battery). Decoupling and Layout Recommendations 17 AMD-K6® Processor Power Supply Design 21103G/0—February 1999 Output Voltage vs Capacitance 2.4 2.35 2.3 2.25 2.2 Voltage 2.15 2.1 2.05 Voltage 2 1.95 1.9 1.85 1.8 1.75 1.7 470 940 1410 1880 2350 2820 3290 3760 4230 4700 5170 5640 6110 Capacitance in micro Farads Figure 9. Bulk Decoupling versus Output Voltage Response for 2.4 V @15 A Output Voltage Response Measurement Techniques To measure output voltage response, run a program such as DOS EDIT and toggle STPCLK# every 40 µsec or slower. (AMD has developed the Maxpwr99.exe utility. See “Output Voltage R e sp o n s e M e a su re m e n t Ut i li t y ” on p a g e 19 fo r m ore information.) Measure the voltage at the back of the board right under the processor. Use a scope probe with a ground connection next to the tip. The 3 inch to 6 inch ground leads that come off the side of a scope probe have too much inductance for this type of measurement. The scope bandwidth can be limited to 20MHz, giving a clear indication of the power supplied. While limiting the scope bandwidth for bulk decoupling verification gives a clear indication of the low-frequency issues, AMD recommends rechecking with at least a 250 MHz bandwidth for verifying the high-frequency decoupling. 18 Decoupling and Layout Recommendations AMD-K6® Processor Power Supply Design 21103G/0—February 1999 AMD used a Tektronix 684B scope with 6245 probes and an HP54720 with 54701 probes. (There was no significant difference between these two instruments.) The data was taken over a 40-second window with the scope set to infinite persistence. For a good starting point, use a horizontal sweep rate of 500 nsec per division and a vertical scale of 0.1 V per division. AMD made measurements while running Winstone® 96 under the Windows® 95 operating system, running DOS EDIT pull dow n, and running Maxpwr99 . exe w hile t oggling STPCLK#. The latter case created the worst-case current transient in the measurements conducted by AMD. In addition, this is the case that requires the maximum decoupling capacitance. Those regulators that AMD believes can meet the processor requirements (with proper decoupling) are marked as tested in the tables shown in “Voltage Regulator Vendor Information” on page 53. The other listed regulators are expected to work, but were not tested in time for the printing of this document. Output Voltage Response Measurement Utility AMD has developed a software utility to assist in designing systems that comply with the processor power and thermal requirements. This utility can verify that the supply voltage remains stable during a transition to a higher power/current consumption level. This utility is DOS based. For systems based on the Windows 95 or Windows 98 operating system, re-boot in DOS mode or boot from a bootable DOS floppy disk that contains the utility. For systems based on the Windows NT® and OS/2 operating systems, boot from a bootable DOS floppy disk that contains the utility. The command line for this utility is as follows: (Note: Do not execute the utility in a DOS window or with a memory manager loaded.) c:\>Maxpwr99.exe The Maxpwr99.exe utility is available under a nondisclosure a g re e m e n t . C o n t a c t yo u r l o c a l A M D s a l e s o f f i c e fo r information. Decoupling and Layout Recommendations 19 AMD-K6® Processor Power Supply Design 21103G/0—February 1999 Decoupling Capacitance and Component Placement The high-frequency decoupling capacitors (C5–C31 in Figure 11 on page 22) should be located as close to the processor power and ground pins as possible. To minimize resistance and inductance in the lead length, the use of surface mounted capacitors is recommended. When possible, use traces to connect capacitors directly to the processor’s power and ground pins. In most cases, the decoupling capacitors can be placed in the Socket 7 cavit y on the sam e side of t he processo r (component side) or the opposite side (bottom side). Figure 11 on page 22 shows a suggested component placement for the decoupling capacitors. The values of the capacitors are specified in Table 5 on page 22. The split voltage planes should be isolated if they are in the same layer of the circuit board. To separate the two power planes, an isolation region with a minimum width of 0.254 mm is recommended. The ground plane should never be split. These recommendations are based on single-sided component assembly and general space constraints. The designer should assume these are minimum requirements. If double-sided component assembly is used, it is preferable to use more capacitors of a smaller value, which reduces the total ESR and total ESL of the capacitors. For example, instead of four 470-µF capacitors, use ten 47-µF capacitors. (Check the device specifications shown in Table 3. Occasionally a lower value capacitance has a higher ESR.) As the effective ESR is lowered, the total required capacitance is reduced. The breakdown voltage and case size both affect the ESR value. Table 3. Representative ESR Values Capacitance Device 1 Device 2 470 µF 55 mΩ 100 mΩ 270 µF 70 mΩ 100 mΩ 100 µF 90 mΩ 100 mΩ 68 µF 95 mΩ 100 mΩ 47 µF 120 mΩ 250 mΩ Via inductance can be reduced when using double-sided component assembly. Components can share vias on the top side and bottom side. This technique reduces the effective via inductance. Because double-sided assembly is rarely used in 20 Decoupling and Layout Recommendations AMD-K6® Processor Power Supply Design 21103G/0—February 1999 desktop systems, the most likely use for this technique is in portable systems. Figure 10 on page 21 shows another way to reduce via inductance — parallel vias. This technique is usually used on bulk decoupling capacitors. The inductance contribution numbers shown in Table 4 indicate that a poor layout can negate a good component. Pad Capacitor Dual Vias No Trace between Via and Pad Figure 10. Via Layout For Low Inductance Table 4. Inductance Contributions of Components Component Induction Comment Capacitor 0.6nH (approximately) ESL Via 0.7nH (approximately) – 100 mil Trace 1.6nH (approximately) 10 mil wide trace Decoupling and Layout Recommendations 21 AMD-K6® Processor Power Supply Design 21103G/0—February 1999 CC12 CC11 C21 C2 C11 CC4 + + CC5 CC6 C12 C13 VCC3 (I/O) Plane C26 CC10 C22 C23 C24 CC8 C1 + + + C25 C29 C27 C30 C28 C31 CC7 CC3 C15 C7 C9 C20 C16 C6 C10 C17 C19 C18 C5 CC9 C8 C14 0.254mm (min.) for isolation region VCC2 (Core) Plane CC1 CC2 Figure 11. Suggested Component Placement Table 5 lists the recommended capacitor values. Table 5. Decoupling Capacitor Values Item Qty 1 2 C1, C2 47µF 2 12 CC1–CC12 470µF 3 27 C5–C31 0.1µF 22 Location Value Footprint Description AVX Surface tantalum capacitor, AVX part number Size V TPSV476*025R0300 or equivalent AVX Surface tantalum capacitor, AVX part number Size V TPSV477*006R0100 or equivalent 0805 – Note VCC3 Decoupling VCC2 Decoupling C5–C13 for VCC3 C14–C31 for VCC2 Decoupling and Layout Recommendations AMD-K6® Processor Power Supply Design 21103G/0—February 1999 Table 6 lists recommended capacitor types. Table 6. Capacitor Recommendations Manufacturer Type Comment Web AVX TPS exceptional /www.avxcorp.com Vishay Sprague 594D exceptional /vishay.com/vishay/sprague Kemet T510 excellent /www.kemet.com Sanyo SA/SG / OS-CON 4SP560M excellent /www.sanyovideo.com Vishay Sprague 593D good /vishay.com/vishay/sprague Mallory T495 good /www.nacc-mallory.com Nemco SLR series good /www.nemcocaps.com Panasonic FA good /www.panasonic.com/pic Elna RJH/RJJ good /www.elna-america.com The recommendations in Table 6 are not the only solutions. Based on the availability of parts and the choice of controller, many correct solutions are possible. The information in Table 6 is intended to give insight into the requirements, and not to s p e c if y a p a r t i c u l a r s o l u t i o n . I n a dd i t io n, a lu m inu m electrolytics can be used instead of tantulum capacitors. This approach is acceptable as long as good quality, low-ESR parts are used. The biggest problem with aluminum electrolytics is the large decrease in capacitance as they age. High-Frequency Decoupling Inductance is also a concern for the high-frequency decoupling capacitors. Case size can be a significant factor affecting c a p a c i t o r i n d u c t a n c e . Fo r e x a m p l e , a 0 6 0 3 c a s e h a s significant ly more inductance than a 0612 case. AMD recommends the 0612, 1206, 0805, and 0603 case in order of best to worst. Inductance can also be reduced by directly connecting the capacitor to the power pin of the processor. In order to minimize its inductance, this trace must be short and as wide as possible. This technique effectively removes two via inductances between the capacitor and the processor as shown in Figure 13 on page 26. The dotted line shows that connecting the capacitor directly to the processor eliminates two series inductances. However, this trace also has inductance—if it is too long or too narrow it can be worse than the vias. Figure 12 on page 25 shows the effect of inductance at higher frequencies. (The numbers outside the X and Y axis indicate the minimum and maximum values plotted). The inductance Decoupling and Layout Recommendations 23 AMD-K6® Processor Power Supply Design 21103G/0—February 1999 used is 1.8 nH (two 0.7 nH for the vias and 0.4 nH for the capacitor itself). The capacitor is a 0.1-µF X7R multilayer Ceramic MLC. The inductance of a capacitor is a function of the case type. An 0612 case is assumed here. The following steps show how the number of required capacitors is calculated: 1. Decide what to allow as a ripple voltage budget. In this example calculation the ripple-voltage budget = 30mV. 2. The measured AC transient current is 0.75A. This transient current has a typical duration of 2.5 nsec. The amount of capacitance required can now be determined using the following equation: I = C (dv/dt) C = I (dt/dv) = 0.75A (2.5nsec/30mV) = 0.625µF This equation indicates that if the capacitors didn’t have inductance, only six 0.1-µF capacitors would be needed. 3. Determine the number of capacitors required based on the inductance of the capacitor. Use the following formula: V = L (di/dt) = L • (0.75A/2.5nsec) = 30mV Solving for L, the allowed budget is 100pH 4. The inductance of the capacitor and via = 1.8nH (two 0.7nH for the vias and 0.4nH from the capacitor itself). Because each capacitor usually has two vias (one on each end), the effective via inductance must be: 2 • 0.7nH + 0.4nH = 1.8nH 5. Solving the following equations for N: 1.8nH/N = 100pH N = 1.8nH/100pH = 18 The number of capacitors required is 18. 24 Decoupling and Layout Recommendations AMD-K6® Processor Power Supply Design 21103G/0—February 1999 The following steps repeat the calculation for I/O decoupling: 1. Determine the amount of capacitance required using the following equation: I= C (dv/dt) C = I (dt/dv) = 0.5 (2.5nsec/145mV) = 0.0086 µF This equation indicates that if the capacitors didn’t have inductance, only one 0.1-µF capacitors would be needed. 2. Using 0.5A as a typical ICC3 value, repeat the calculations to account for inductance: Note: The ripple budget is 145mV because the I/O drivers are not as sensitive to supply variations as the core and the current transient is smaller. L = V (dt/di) = 0.145 (2.5nsec/.5A) = 725pH Solving for L, the allowed budget is 725pH. The number of capacitors = 1.8nH/725pH = 2.5. Therefore, only three capacitors are needed on the I/O. AMD recommends a minimum of six capacitors. 15.9312 100 10 Zo( c2 , L , r , w ) Zo( c1 , L , r , w ) 1 0.191752 0.1 1 10 6 1e+006 1 10 7 1 10 w 8 1 10 9 1e+009 Figure 12. 0.1 µF (c1) and 0.01 µF (c2) X7R Capacitor Impedance versus Frequency Decoupling and Layout Recommendations 25 AMD-K6® Processor Power Supply Design 21103G/0—February 1999 Pad A Capacitor C Via to VCC Via to GND B D Processor via to VCC Cc Via Processor via to GND Processor A C Via B D VCC Plane GND Plane Figure 13. Decoupling Inductance Power Sequencing Although the processor requires dual power supply voltages, there are no special power sequencing requirements. The best procedure is to minimize the time between which V CC2 and VCC3 are either both on or both off (See Figure 14). However, a good design practice ensures VCC3 is always greater than VCC2. Volt VCC3 VCC2 Minimize Time Figure 14. Power Sequencing 26 Decoupling and Layout Recommendations AMD-K6® Processor Power Supply Design 21103G/0—February 1999 Power Supply Solutions The solutions provided in this section are not all-inclusive. Obtain additional circuit diagrams and application assistance from the manufacturers. The manufacturers may customize designs to an OEM’s requirements. The schematics shown in this document have not been tested by AMD and are provided as examples. Digital-to-Analog Converter (DAC) Table 7. Voltage Identification (VID) codes provide a way to program the Digital-to-Analog Converter (DAC) to supply a reference for d if f e re n t o u t pu t vo l t a g e s . M a ny m a nu fa ct u re rs h ave DAC-controlled devices, however, some do not follow the defined VID codes (designated as DAC in the Remarks column of the vendor table in “Voltage Regulator Vendor Information” on page 53). Table 7 shows the codes and corresponding voltage. Voltage Output VID Codes D4 D3 D2 D1 D0 Output Voltage D4 D3 D2 D1 D0 Output Voltage 1 0 0 0 0 3.50V 0 0 0 0 0 2.05V 1 0 0 0 1 3.40V 0 0 0 0 1 2.00V 1 0 0 1 0 3.30V 0 0 0 1 0 1.95V 1 0 0 1 1 3.20V 0 0 0 1 1 1.90V 1 0 1 0 0 3.10V 0 0 1 0 0 1.85V 1 0 1 0 1 3.00V 0 0 1 0 1 1.80V 1 0 1 1 0 2.90V 0 0 1 1 0 1.75V 1 0 1 1 1 2.80V 0 0 1 1 1 1.70V 1 1 0 0 0 2.70V 0 1 0 0 0 1.65V 1 1 0 0 1 2.60V 0 1 0 0 1 1.60V 1 1 0 1 0 2.50V 0 1 0 1 0 1.55V 1 1 0 1 1 2.40V 0 1 0 1 1 1.50V 1 1 1 0 0 2.30V 0 1 1 0 0 1.45V 1 1 1 0 1 2.20V 0 1 1 0 1 1.40V 1 1 1 1 0 2.10V 0 1 1 1 0 1.35V 1 1 1 1 1 OFF 0 1 1 1 1 1.30V Power Supply Solutions 27 AMD-K6® Processor Power Supply Design Cherry CS5166 21103G/0—February 1999 The CS5166 shown in Figure 15 on page 29 is a synchronous dual NFET buck regulator controller. It is designed to power the core logic of the processors in the AMD-K6 family. It uses the V2 control method to achieve fast transient response and good overall regulation. This proprietary control architecture makes use of the ramp signal developed across the ESR of the output capacitors. This signal is fed back to the CS5166 through two feedback loops. The CS5166 features a 5-bit DAC with 1% tolerance, programmable hiccup mode current limiting, adaptive voltage positioning, and over-voltage protection. The CS5166 buck regulators can deliver 14 A at 88% efficiency. The CS5166 minimizes external component count, total solution size, and cost. It operates over a 4.05 V to 20 V range using either single or dual input voltage. Table 8 on page 28 shows the bill of materials for the CS5166. Contact Information Cherry Semiconductor Corporation 2000 South County Trail East Greenwich, RI 02818-1530 Tel: (401) 885-3600 Fax: (401) 885-5786 www.cherry-semi.com Table 8. Cherry CS5166 Bill of Materials Reference Description Part Number Manufacturer C1 1 µF 499-717 Farnell/Newark C3, C4, C5 0.1 µF 1206B104K500NT Novacap C2 330 pF 0805N391J500NT Novacap C7–C14 1200 µF/10 V 10MV1200GX Sanyo R1 3.3K, 5%, 1/8 W RM73B2AT332J KOA R2 510 W, 1/8 W P-510-ECT-ND Digi-Key C6 1000 pF 0805N102J500NT Novacap Q1, Q2 (10 amp) N-Channel FET IRL3103 Intern.Rectifier Q1, Q2 (19 amp) N-Channel FET FS70VSJ-03 Mitsubishi L1 (10 amps) 2 µH/10 A S26-10006 Xformers L1 (10 amps) 2 µH/10 A S26-10006 Xformers L1 (19 amps) 1.2 µH/19 A XF0016-V04 Xformers U1 CS5166 CS5166DW16 Cherry 28 Power Supply Solutions 21103G/0—February 1999 Power Supply Solutions 12V 5V C1 1µF C7-C9 1200µF/10V x3 C2 330pf COFF VCC GATEH SS C3 0.1µF COMP C4 0.1µF VFB CS-5166 R1 3.3K Ω Q1 FS70VSJ-03 1.2µH "Droop" resistor (Free Current Sensing Element) 4m Ω Vcc L1 C10-C14 C6 1000pF 1200µF/10V x5 Vss VID0 VID1 GATEL VID2 PGND FS70VSJ-03 Q2 VID4 ISENSE LGND PWRGD C5 0.1µF R2 510 Ω Figure 15. Cherry CS5166 Switching Power Supply Design 29 AMD-K6® Processor Power Supply Design PWRGD VID3 AMD-K6® Processor Power Supply Design Elantech EL7571 21103G/0—February 1999 The EL7571 switching regulator is a flexible, high-efficiency, PWM controller that includes a five bit DAC adjustable output. This regulator employs synchronous rectification to deliver up to 15 A at efficiencies greater than 85% over a wide range of supply voltages. (Efficiencies up to 92% can be achieved at 10 A.) Figure 16 shows an EL7571 reference design. The VID code allows the output to be set between 1.3 V and 2.05 V (in 50 mV increments) and 2.1V and 3.5V (in 100 mV increments) with a 1% accuracy. The VID code should be set to 00011 for 3.2V output. Table 9 on page 31 shows the bill of materials for the EL7571. ENABLE VHI 20 1 OTEN 330pF C3 C4 2 CSLOPE C6 12V 1.3uH HSD 19 Q1 LX 18 3 COSC L2 0.1uF C1 3mF 5V Si4420X2 330pF C5 0.1uF L1 R1 C7 5 PWRGD POWER GOOD Voltage I.D. (VID[0:4]) VIN 17 4 REF VINP 16 5.1uH 1uF IC1 7.5mΩ VOUT C2 6mF 6 VID0 LSD 15 7 VID1 GNDP 14 8 VID2 GND 13 9 VID3 CS 12 10 VID4 FB 11 Q2 Si4420X2 Figure 16. Elantec EL7571 Switching Power Supply Design 30 Power Supply Solutions AMD-K6® Processor Power Supply Design 21103G/0—February 1999 Contact Information Elantec Corporation 675 Trade Zone Blvd. Milpitas, CA 95035-1323 Tel: (408) 945-1323 Fax: (408) 945-9305 www.elantec.com Table 9. Elantec EL7571 Bill of Materials Reference Description Part Number Manufacturer C1, C2 680 µF LXF16VB681M10X20LL United Chem-Con C3,C4 330 pF 08055A331JAT2A AVX C5, C6 0.1 µF 08053C104MAT2A AVX C7 1 µF TAJA105K025R AVX D1 Diode BAV99 Motorola, Siemens, et-al D2 Diode 32CTQ030 International Rectifier IC1 Controller EL7571CM Elantec L1 5.1 µH PE-53700 Pulse Engineering L2 (optional) 1.5 µH T30-26 7T AWG #20 Micro Metals R1 15 mΩ WSL-2512 Dale R2 5Ω RK73H2ATE05RoF KOA 2xQ1, 2xQ2 MOSFET Si4420 Siliconix Q1, Q2 MOSFET Si4410 Siliconix Power Supply Solutions 31 AMD-K6® Processor Power Supply Design Harris Semiconductor HIP6004 and HIP6005 21103G/0—February 1999 The Harris HIP6004 and HIP6005 are voltage-mode controllers with many functions pertinent to the processors in the AMD-K6 family. The HIP6004 is the heart of a standard step-down, or buck converter. It contains a high-performance error amplifier, a high-resolution, 5-bit digital-to-analog converter (DAC), a programmable free-running oscillator, and a floating MOSFET driver. This regulator can deliver up to 15 A at efficiencies greater than 80%. The VID code allows the output to be set between 1.3 V and 2.05 V (in 50 mV increments) and between 2.1 V and 3.5 V (in 100 mV increments) with a 1% accuracy. The HIP6004 is very similar to the HIP6005, but is targeted for buck converters with a synchronous rectifier design. Figure 17 shows the reference design and Table 10 on page 33 shows the bill of materials for the HIP6004. L1 +5VIN 1µH + GND BOOT VCC +12VIN R2 10k R1 2 15 14 GND PWRGOOD PGOOD VID0 VID1 VID2 VID3 VID4 SS C14 0.1µF OVP OCSET 1.5k 18 C4 1µF 13 Q1 HUF76129 UGATE PHASE 11 L2 VOUT 3.5µH 17 4 U1 16 HIP6004A 5 1 6 10 12 C1-3 3x1500µF VSEN1 FB C12 10nF R6 180k 3 9 + R3 7.50k C13 10pF 19 C5-10 6x1500µF PGND 7 8 Q2 HUF76129 LGATE R4 C11 2.43k 0.47µF R5 499k COMP 20 RT Figure 17. Harris HIP6004 1.3V–3.5V Switching Power Supply Design 32 Power Supply Solutions AMD-K6® Processor Power Supply Design 21103G/0—February 1999 Contact Information Harris Semiconductor P.O. Box 883, MS 53-210 Melbourne, FL 32902 Tel: (407) 729-4984 Fax: (407) 729-5321 www.semi.harris.com Table 10. Harris HIP6004 Bill of Materials Reference Description Part Number Manufacturer C1–C3, C5–C10 Aluminum Capacitor, 6.3 V, 1500 µF 6MV1500GX Sanyo C4 1.0 µF Ceramic Capacitor, X7S, 16 V 1206YZ105MAT1A AVX C11 0.47 µF Ceramic Capacitor, X7R, 16 V 0805YC474JAT2A AVX C12 0.01 µF Ceramic Capacitor, X7R, 16 V Various C13 10 pF Ceramic Capacitor, X7R, 25 V Various C14 0.1 µF Ceramic Capacitor, X7R, 16 V Various L1 1 µH Inductor, 7T of 16AWG on T50-52 core PO720 Pulse L2 3.5µH Inductor, 7T of 17AWG on T68A-52 core PO718 Pulse Q1, Q2 UltraFET MOSFET, 30 V, 16 mΩ Harris R1 1.5 kΩ Resistor, 5%, 0.1 W Various R2 10 kΩ Resistor, 5%, 0.1 W Various R3 7.50 kΩ Resistor, 1%, 0.1 W Various R4 2.43 kΩ Resistor, 1%, 0.1 W Various R5 499 kΩ Resistor, 1%, 0.1 W Various R6 18 0kΩ Resistor, 5%, 0.1 W Various U1 Synchronous PWM Controller Power Supply Solutions HUF76139S3S HIP6004ACB Harris 33 AMD-K6® Processor Power Supply Design 21103G/0—February 1999 Linear Technology LT1553 The LTC1553 is a high-power, high-efficiency (over 95% is p o s s i b l e ) sw i t ch i n g re g u l a t o r fo r 1 . 8 V – 3 . 5 V o u t p u t applications. It features a 5-bit DAC controlled output voltage, a precision internal reference that provides output accuracy of ±1.5% at room temperature, load current, and line voltage shifts. The LTC1553 uses a synchronous switching architecture (that free-runs at 300 kHz) with two external N-channel output devices, providing high efficiency. It senses the output current across the on-resistance of the upper N-channel FET, providing an adjustable current limit up to 19 A without an external sense resistor. Fast transient response minimiz es the output decoupling required. The design shown in Figure 18 on page 35 provides 14 A at efficiencies greater than 90%. Table 11 shows the bill of materials. Contact Information Linear Technology Corporation 1630 McCarthy Blvd. Milpitas, CA 95035-7417 Tel: (408) 432-1900 Fax: (408) 434-0507 www.linear.com Table 11. Linear LT1553 Bill of Materials Reference Description Part Number Manufacturer Cin 1200 µF 6.3V 20% aluminum electrolytic capacitor 10MV1200GX Sanyo Cout 330 uF 6.3 Volt Tantalum TPSE337M006R0100 AVX C1 150 pF 50 V 10% NPO chip capacitor 08055A151KAT1A AVX Css, Cs, Cvcc, Cvcc 0.1 µF 50 V 10% Y5V chip capacitor 08055G104KAT1A AVX Ccc 0.01 µF 50 V 10% Y5V chip capacitor 08055G103KAT1A AVX Cvcc, Cvcc 10 µF 35 V 20% tantalum capacitor TPSE106M035 AVX L0 2 µH 18 A inductor CTX02-13198 12TS-2R5SP Coiltronics Panasonic Q1, Q2 N-Channel MOSFET SUD50N03-10 Siliconix Rpu 5.6k 1/10W 5% chip resistor CR21-562J-T AVX Rfb 20 Ω 1/10W 5% chip resistor CR21-200J-T AVX Rmax 2.7k 1/10W 1% chip resistor CR21-2701F-T AVX Rc 8.2k 1/10W 1% chip resistor CR21-8201F-T AVX U1 20-lead narrow small outline IC LTC1553CG LTC 34 Power Supply Solutions 21103G/0—February 1999 Power Supply Solutions AMD-K6® Processor System 35 AMD-K6® Processor Power Supply Design Figure 18. Linear LT1553 1.8V to 3.5V Switching Power Supply Design AMD-K6® Processor Power Supply Design 21103G/0—February 1999 LINFINTY LX1664 and LX1665 The LX1664 and LX1665 are dual-output controllers that combine a programmable switch-mode controller with a linear regulator driver. The linear section is adjustable and can supply 5 A–7 A. The switch mode section uses a modulated constant off-time architecture with adaptive voltage positioning to achieve optimal transient response. The circuit offers pulse-by-pulse current limiting, short-circuit protection, and the LX1665 offers a power-good output and a crowbar driver for over-voltage protection. An input inductor is recommended to reduce ripple on the 5 V input. The internal 5-bit DAC provides an adjustable output of 1.3 V to 3.5 V. The circuit shown in Figure 19 on page 37 can deliver more than 15 A, dependent on choice of FETs and current limit set-point. The efficiency of this circuit is around 85–90%, depending on the choic e of components. Table 12 shows the bill of materials for the LX1664. Contact Information LINFINITY Microelectronics 11861 Western Avenue Garden Grove, CA 92841 Tel: (714) 372-8383 Fax: (714) 372-3566 www.linfinity.com Table 12. LINFINITY LX1664 Bill Of Materials Reference Description Mechanical Part Number Manufacturer U1 Dual output PWM controller SO-18 [LX1664 is SO-16] LX1665 LINFINITY Q1, Q2 MOSFET, 26 mΩ, 24 A TO-263 or TO-220 IRL3303S International Rectifier Q4 MOSFET TO-220 IRLZ44 International Rectifier L1 Inductor, 5 µH Thru-hole - - C1, C2 Capacitor, Al-Elec, 1000 µF, 6.3 Radial, 8x20mm V, low ESR 6MV1000GX Sanyo C7 Capacitor, Al-Elec, 330 µF, 6.3 V, low ESR 6MV330GX Sanyo C3 Capacitor, ceramic, 0.1 µF, X7R 0805 - - C4, C6 Capacitor, ceramic, 390pF, X7R 0805 - - C8 Capacitor, ceramic, 680pF, X7R 0805 - - C5 Capacitor, ceramic, 1 µF, Y5V 1206 - - R3, R4 Resistor, 1k, 5% 0805 - - R6 Resistor, 10k, 1% 0805 - - R6 Resistor, 10k, 1% 0805 - - R1 Power Resistor, 5 mΩ 1 % OARS-1 - - 36 Radial, 8x20mm Power Supply Solutions AMD-K6® Processor Power Supply Design Figure 19. LINFINITY LX1664 Switch-Mode Power Supply Design 21103G/0—February 1999 Power Supply Solutions 37 AMD-K6® Processor Power Supply Design Maxim MAX1638 21103G/0—February 1999 The MAX1638 is an ultra-high-performance, step-down DC-to-DC controller for processor power in high-end computer systems. It delivers over 35 A from 1.3 V to 3.5 V with ±1% total accuracy from a +5V ±10% supply. Excellent dynamic response corrects for output transient. This controller achieves over 90% efficiency by using synchronous rectification. The switching frequency is pin-selectable for 300 kHz, 600 kHz, or 1 MHz. Fast recovery from load transients is ensured by a GlitchCatcher current-boost circuit that eliminates delays ca us e d by t h e b uck in d u ct o r. O t h e r f e a t u re s i n c l u de over-voltage protection, internal digital soft-start, a power-good output, and a 3.5 V ±1% reference output. Figure 20 on page 40 shows a 14A reference design and Table 13 on page 39 shows the bill of materials. By changing the components as listed in the BOM, the circuit can be designed to supply up to 19 A. Contact Information Maxim Integrated Products 120 San Gabriel Drive Sunnyvale, CA 92841 Tel: (408) 737-7600 Fax: (408) 737-7194 www.maxim-ic.com 38 AMD-K6® Processor Power Supply Design 21103G/0—February 1999 Table 13. Maxim MAX1638 Bill of Materials Reference 2.2V 12A Load 2.2V 14A Load 2.2V 19A Load 1.3 V 19 A Load C1 (x2) Sanyo OS-CON 10SA220M (220 µF) (x3) Sanyo OS-CON 10SA220M (220 µF) (x4) Sanyo/OS-CON 10SA220M (220 µF) (x4) Sanyo/OS-CON 10SA220M (220 µF) C2 (x3) Sanyo OS-CON 4SP220M (220 µF) (x4) Sanyo OS-CON 4SP220M (220 µF) (x6) Sanyo OS-CON 4SP220M (220 µF) (x7) Sanyo OS-CON 4SP220M (220 µF) C4 1 µF ceramic or 2.2 µF TDK C3216X7R1C225M, Taiyo Yuden EMK316BJ225ML C5, C8 0.1 µF C6 Sprague 595D106X0010A2B (10 uF) C7 Sprague 595D475X0016A2T (4.7 uF) CC1 1000 pF CC2 0.056 µF D1 (optional) Nihon NSQ03A02 Schottky diode or Motorola MBRS340 or Central Semi NSC03A02 D2 Nihon NSQ03A02 Schottky diode or Motorola MBRS340 Central Semiconductor CMPSH-3 Coiltronics UP4-R47 (0.47 µH, 19 A, SMD) or Panasonic ETQP1F0R7H (0.70 µH, 19 A, 1.6 mΩ, SMD) Coiltronics UP4-R47 (0.47µH, 19A, SMD) or Panasonic ETQP1F0R7H (0.70 µH, 19 A, 1.6 mΩ, SMD) Panasonic ETQP2F1R0S (0.70 µH, 23 A, 0.94 mΩ, SMD) N1, N2 Int’l Rectifier IRL3103S Fairchild FDB7030L (10 mΩ) or Int’l Rectifier IRL3803S (9 mΩ) (x2) Fairchild FDB7030L (10 mΩ) or (x2) Int’l Rectifier IRL3803S (9 mΩ) P1/N3 Int’l Rectifier IRF7107 Int’l Rectifier IRF7105 (0.4 W/0.16 W) Int’l Rectifier IRF7307 (0.09 W/0.05 W) R1 (x2) Dale WSL-2512-R009-F (10 mΩ) (x2) Dale WSL-2512-R009-F (10 mΩ) (x2) Dale WSR-20.007 ±1% (7 mΩ) R2 Dale WSL-2512-R120-J (120 mΩ) Dale WSL-2512-R120-J (120 mΩ) L1 R3, R4 (Optional) 1-5 Ohms RC! 1K 5% resistor 39 21103G/0—February 1999 Figure 20. Maxim MAX1638 Switching Power Supply AMD-K6® Processor Power Supply Design 40 AMD-K6® Processor Power Supply Design 21103G/0—February 1999 Micro Linear ML4902 The ML4902 is designed to be configured as a synchronous buck converter with a minimum of external components. The ML4902 can generate voltages between 1.8V and 3.5V from a 5V supply at currents up to 14 A. Figure 21 shows an ML4902 reference design capable of 12 A at 90% efficiency. Table 14 on page 42 shows the bill of materials. 5VIN 12VIN C12 220nF 16V C10 220nF 16V C20 22µF 25V R3 1MΩ R2 1kΩ VID0 1 D0 PROTECT 20 VID1 2 D1 VDD 19 VID2 3 D2 VCC 18 VID3 4 D3 N DRV H 17 VID4 5 RANGE OUTEN 6 SHDN 7 N/C 8 PWR GOOD COMP 13 9 VREF ISENSE 12 PWRGD C22 1nF N DRV L 16 PWR GND 10 GND C14 220nF 16V C2 C3 C11 220nF 16V Q1, Q2 2X IRF7413 15 N/C 14 C1 Q3, Q4 2X IRF7413 L1 1.4µH VCCP R1 33Ω C19 C4-C9 R5 100Ω C15 VFB 11 U1 ML4902 VSS R4 100kΩ C13 1nF C1 - C9 - 1500µF, 6.3V, Sanyo 6MV1500GX C15, C19 - 100nF ceramic Figure 21. Micro Linear ML4902 Switching Power Supply Design Contact Information Micro Linear Corporation 2092 Concourse Dr. San Jose, CA 95131 Tel: (408) 433-5200 Fax: (408) 432-0295 www.microlinear.com 41 AMD-K6® Processor Power Supply Design 21103G/0—February 1999 Table 14. Micro Linear ML4902 Bill of Materials Reference C1–C9 Description 1500 µF, 6.3 V C10–C12, C14 0.22 µF ceramic 42 Part Number Manufacturer 6MV1500GX Sanyo 1206Y224Z205NT Novacap C13, C22 0.001 µF ceramic 0805N102J500N Novacap C15– C19 0.1 µF ceramic 1206B104K500NT Novacap C20 22 µF 25 V TPSE226M025 AVX L1 1.4 µH on T44-52 core CTX09-13336 Coiltronics Q1,Q2 Transistor IRF7413 International Rectifier Q3,Q4 Transistor IRF7413 International Rectifier R1 33 Ω 1% R2 1 KΩ 5% R3 1 MΩ 5% R4 100 KΩ 5% R5 100 Ω 1% AMD-K6® Processor Power Supply Design 21103G/0—February 1999 Fairchild RC5051 The RC5051 shown in Figure 22 combines a switch-mode DC-to-DC controller with a reference DAC in a single package. The DAC provides a mechanism to adjust the DC-to-DC converter output between 1.3 V and 3.5 V, which allows one motherboard design to accommodate several different processors. Table 15 on page 44 shows the bill of materials for the RC5051. This design provides up to 15 A at an 80% efficiency. +12V L2 +5V C2 2.5uH C1 0.1uF CIN* 0.1uF R2 47 D1 1N4735A C3 1uF C5 .1uF 11 10 12 9 13 8 14 7 15 6 C4 5 1uF 16 VREF RC5051 C8 17 0.1uF 18 3 19 2 20 1 GND Q1 L1 RSENSE* 1.3uH 4 DS1 MBRS320 Vo COUT* Q2 CEXT 100pF VC C VID4 VID3 VID2 VID1 R1 10K ENABLE *Refer to BOM for values of PWRGD C7 0.1uF C6 0.1uF RSENSE, COUT and CIN VID0 Figure 22. Fairchild RC5051 Power Supply Design 43 AMD-K6® Processor Power Supply Design Contact Information 21103G/0—February 1999 Fairchild Semiconductor (formerly Raytheon Electronics) 350 Ellis Street Mountain View, CA 94043 Tel: (415) 962-7982 Fax: (415) 966-7742 www.fairchildsemi.com . Table 15. Fairchild RC5051 Bill of Materials Reference Part Number and Description for 10 Amp Load Part Number and Description for 13 Amp Load Part Number and Description for 15 Amp Load Panasonic ECU-V1H104ZFX Capacitor, ceramic, 0.1 µF, X7R C1, C2, C5–8 C3,C4 Panasonic ECSH1CY105R Capacitor, ceramic, 1 µF, X7R CEXT Panasonic ECU-V1H121JCG Capacitor, ceramic, 100 pF, COG (3x) Sanyo 10MV1200GX, Capacitor, 10 V al-electrolytic, 1200 µF Cin Cout (4x) Sanyo 6MV1500GX Capacitor, 6.3 V al-electrolytic, 1500 µF (6x) Sanyo 6MV1500GX Capacitor, 6.3 V al-electrolytic, 1500 µF D1 Motorola 1N1545A Zener Diode Motorola 1N4735A Zener Diode Motorola 1N4735A Zener Diode 6.2 V,1 W 6.2 V,1 W DS1 General Instruments 1N5817 Schottky Diode General Instruments 1N5817 Schottky Diode L1 Skynet 320-8107 1.3 µH inductor 1.3µH, Isat>15Amp DCR-2.5mΩ 1.3µH, Isat>17Amp DCR-2.5mΩ L2 (optional) Skynet 320 6110 Bead Inductor Input Inductor 2.5µH, toroid, 10 2.5µH, Isat>11Amp DCR-6mΩ turns 17AWG R1 Q1, Q2 U1 44 Fairchild MBRS320 4 A, 20 V Schottky Diode 47 ohm resistor, 1/8 W, 5% R2 Rsense (8x) Sanyo 6MV1500GX Capacitor, 6.3 V al-electrolytic, 1500 µF Copel AWG #18–6 mΩ CUNi Alloy wire Sense resistor, 1 W, 10% 6.3 mΩ CUNi Alloy wire Sense resistor, 1 W, 10% Fairchild RC10-32, 5.2 mΩ, Wire resistor Panasonic ERJ-6ENF10.0KV 10 KΩ resistor, 1/8 W, 5% IRL3103 N-Channel MOSFET IRL2203 N-channel Power FET Fairchild RC5051M PWM Controller Fairchild FDB6030L 30V, 10mΩ, MOSFET AMD-K6® Processor Power Supply Design 21103G/0—February 1999 Semtech SC1182 and SC1183 The SC1182 and SC1183 combine a synchronous voltage-mode controller with two low-dropout linear regulators providing most of the circuitry necessary to implement three DC-to-DC converters for powering advanced processors such as the AMD-K6 family of processors. The SC1182 and SC1183 feature an integrated 5-bit DAC, pulse-by-pulse current limiting, integrated power-good signaling, and logic-compatible shutdown. The SC1182/3 switching section operates at a fixed frequency of 200 kHz. The integrated DAC provides output voltage programmability from 2.0V to 3.5V in 100 mV increments and 1.30V to 2.05V in 50 mV increments with no external components. The SC1182/3 linear sections are low dropout regulators. The LDOs can provide 3.3 V for operation of the I/O, cache, memory etc. The current capability of each LDO is determined by the MOSFET chosen. The circuit shown in Figure 23 on page 46 provides a current of 15 A at 85% efficiency. Table 16 on page 47 shows the bill of materials for the SC1182. Contact Information Semtech Corp 652 Mitchell Road Newbury Park, CA 91320 Tel: (805) 498-2111 Fax: (805) 498-3804 www.semtech.com 45 46 5V + PWRGD + C22 330uF C5 0.1uF 19 20 21 22 6 5 R14 * SC1182/3CSW LDOS2 GATE1 GATE2 PGNDL AGND EN VID3 VID2 VID1 VID0 OVP VCC U1 R18 10K Q4 BUK556 * R15 LDOS1 LDOV BSTL DL PGNDH DH BSTH VID4 PWRGOOD VO SENSE CS- CS+ C9 330uF 3 23 14 13 10 11 15 18 7 17 8 9 + + C13 0.1uF 4uH L1 Q1 BUK556 R12 * C10 330uF Q2 BUK556 R16 10k C14 1500uF + + + C12 330uF + + C17 1500uF C16 1500uF C11 330uF + C15 1500uF R5 2.32k C18 0.1uF VLIN2 VLIN1 GND VCC_CORE * SEE "SETTING LDO OUTPUT VOLTAGE" TABLE NOTE: FOR SC1182, R12,R13,R14 AND R15 ARE NOT REQUIRED CONNECT LDOS1 (PIN3) AND LDOS2 (PIN4) DIRECTLY TO VLIN1 AND VLIN2 RESPECTIVELY TO GENERATE 2.5V AND 1.5V OUTPUTS. R17 10K Q3 BUK556 * R13 R4 5mOhm R6 1.00k Figure 23. Semtech SC1182 Voltage Power Supply Design 4 2 24 12 1 R1 10 OVP 5V 330uF C21 VID4 VID3 VID2 VID1 VID0 R2 10k C3 1500uF + 16 C2 1500uF + EN C1 0.1uF 5V 12V AMD-K6® Processor Power Supply Design 21103G/0—February 1999 AMD-K6® Processor Power Supply Design 21103G/0—February 1999 Table 16. Semtech SC1182 Bill Of Materials Reference Description Part Number Manufacturer C1, C5, C13, C18 0.1µF 50V capacitor ECU-V1H104ZFX Panasonic C2, C3, C14–C17 Low ESR 1500 µF/6.3 V capacitor 63MV1500GX Sanyo C9–C12, C21, C22 330µF/6.3V L1 8 Turns 16AWG on T50–52D T50–52D core 4 µH Notes Micro Metals Phillips BUK556 Phillips or DIODS Inc Note 1 or DIODS Inc MMBT3904 or others or others Q1, Q2, Q3, Q4 IRC OAR-1 Series 5mΩ IRC OAR-1 Series 5mΩ 2.32 kΩ 1% resistor ERJ-6ENF2.32KV Panasonic R6 1.00 kΩ 1% resistor ERJ-6ENF1.00KV Panasonic R1 10 Ω, 5%, 1/8W R12 1%, 1/8W Note 2 R13 1%, 1/8W Note 2 R14 1%, 1/8W Note 2 R15 1%, 1/8W Note 2 R4 IRC OAR-1 Series 5 mΩ R2, R17, R18 10 kΩ, 5%, 1/8W R5 U1 SC1182/3CSW Semtech Notes: 1) FET selection requires a trade-off between efficiency and cost. Absolute maximum RDS(ON) = 22 mΩ for Q1,Q2 2) See Table 17 (Not required for SC1182) Table 17. LDO Voltage Selection VOUT LDO1 (LDO2) R12 (R14) R13 (R15) 3.45V 105Ω 182Ω 3.30V 105Ω 169Ω 3.10V 102Ω 147Ω 2.90V 100Ω 130Ω 2.80V 100Ω 121Ω 2.50V 100Ω 97.6Ω 1.50V 100Ω 18.7Ω 47 AMD-K6® Processor Power Supply Design 21103G/0—February 1999 Unisem US3004 The US3004A controller shown in Figure 24 on page 50 is a high-efficiency synchronous pulse width modulated (PWM) controller that provides in excess of 16 A of output current. The output voltage is selected by the 5-bit internal DAC. In addition, the US3004A features two uncommitted linear controllers that can provide a second regulated voltage of 3.3 V. The switcher also employs current sensing by using the R DS(ON) of the high-side power MOSFET as the sensing resistor. Other features include a power-good signal, under-voltage lockout for both 5 V and 12 V supplies, an external programmable soft-start function, and use of an external capacitor for programming the oscillator frequency. Table 18 shows the bill of materials for the US2075. Contact Information Unisem Corp. 32C Mauchly Irvine, CA 92618 Tel: (949) 453-1008 Fax: (949) 453-8748 www.unisem.com Table 18. Unisem US3004 Bill of Materials Reference Description Part Number Manufacturer Q1, Q2 MOSFET IRL3103 IRL3103S (Note 1) IR Q3 MOSFET MTP3055VL Motorola Q4 MOSFET NDP603AL National D1 Diode, GP 1N4148 Motorola L1 Inductor L=1 µH L2 Inductor Core: L=4 µH R=2 mΩ Micro Metal C3, C10 Capacitor, Electrolytic 6 MV1500GX, 1500 uF, 6.3 V, Sanyo C11 Capacitor, Electrolytic 220 µF, 6.3 V, ECAOJFQ221 Panasonic C9, C12, C13 Capacitor, Electrolytic 680 µF, 10 V, EEUFA1A681L Panasonic C2 Capacitor, Ceramic 0805Z105P250NT 1 µF, 25V, Z5U, SMT 0805 Novacap C4, C6 Capacitor, Ceramic 0805Z104P250NT 1 µF, 25V, Z5U, SMT 0805 Novacap Notes: 1) For the applications where it is desirable not to use the Heatsink, the IRL3103S MOSFET in the TO263 SMT package with 1 inch square of pad area using top and bottom layers of the board as a minimum is required. 2) R13 sets the Vcore, approximately 1% higher to account for the trace resistance drop. 48 AMD-K6® Processor Power Supply Design 21103G/0—February 1999 Table 18. Unisem US3004 Bill of Materials (continued) Reference Description Part Number Manufacturer C8 Capacitor, Ceramic 0.1 µF, SMT 0805 size C1 Capacitor, Ceramic 150 pF, X7R, SMT 0805 size C5 Capacitor, Ceramic 220 pF, SMT 0805 size C7 Capacitor, Ceramic 470 pF, SMT 0805 size C14, C15 Capacitor, Ceramic 0.01 µF, SMT 0805 size R1 Resistor 2.21 kΩ,1%, SMT 0805 size R2, R4 Resistor 10 Ω, 5%, SMT 1206 size R3 Resistor Short or 5 Ω, 5%, SMT 1206 size R5 Resistor 10 kΩ, 5%, SMT 0805 size R7 Resistor 267 Ω, 1%, SMT 0805 size R14 Resistor 180 Ω, 1%, SMT 0805 size R8, R15 Resistor 150 Ω, 1%, SMT 0805 size R6, R10 Resistor 1 kΩ, 5%, SMT 0808 size R9, R11 Resistor 100 Ω, 5%, SMT 0805 size R12 Resistor 100 Ω, 1%, SMT 0805 size R13 Resistor (Note 2) 10 kΩ, 1%, SMT 0805 size HS1 Q1 Heatsink 6270 Thermalloy HS2 Q2 Heatsink 6270 Thermalloy Notes: 1) For the applications where it is desirable not to use the Heatsink, the IRL3103S MOSFET in the TO263 SMT package with 1 inch square of pad area using top and bottom layers of the board as a minimum is required. 2) R13 sets the Vcore, approximately 1% higher to account for the trace resistance drop. 49 AMD-K6® Processor Power Supply Design 21103G/0—February 1999 L1 L2 Q1 5V C5 Vout 3 R1 C7 C3 C13 Q2 R2 C4 R3 C10 R13 C6 Q3 C9 12V 12 V12 5 V5 8 CS+ 9 HDrv 7 CS- 11 LDrv 10 Gnd 1 Ct C11 14 Vfb3 R7 R11 Lin1 2 US3004A C1 D4 15 C15 R10 13 SS C2 R8 Vfb1 3 D3 16 D2 17 D1 18 D0 19 VID4 VID3 VID2 VID1 VID0 PGd 6 Figure 24. Unisem US3004 Dual Supply Design Vfb2 4 Lin2 20 Q4 C12 R9 R14 3.3V R6 C14 3004Aapp3-1.2 R15 R5 Power Good C8 50 R12 R4 AMD-K6® Processor Power Supply Design 21103G/0—February 1999 Unitrode UCC3880 The UCC3880 PWM controller shown in Figure 25 combines a switch-mode DC-to-DC controller with a reference DAC, and a precision reference in a single package. The accuracy of the DAC/reference combination is 1.0%. Typical efficiency is greater than 83% at 11.2A. The DAC provides a mechanism to adjust the DC-to-DC converter output between 2.1V and 3.5V in 100 mV steps. Over-voltage and under-voltage monitors are also included. Table 19 on page 52 shows the bill of materials for the UCC3880 capable of up to 16 A. Figure 25. Unitrode UCC3880 Switching Power Supply 51 AMD-K6® Processor Power Supply Design Contact Information 21103G/0—February 1999 Unitrode Integrated Circuits 7 Continental Blvd. Merrimack, NH 03054 Tel: (603) 424-2410 Fax: (603) 424-3460 www.unitrode.com Table 19. Unitrode UCC3880 Bill of Materials Reference Description Part Number Manufacturer C12, C17, C21, C22 0.1µF 50V capacitor ECU-V1H104ZFX Panasonic C5 4.7µF 16V capacitor 595D475X0016A2B Sprague C11 100µF capacitor, 6.3 V tantalum 593D107X9010D2 Sprague C1- C4, C6 - C10 1500µF 6.3V electrolytic capacitor 6MV1500GX Sanyo C13, C14, C15 0.01µF 50V capacitor any any C16 1000pF ceramic any any C18 33pF NPO ceramic any any C19 1500pF ceramic any any C20 82pF NPO ceramic any any CR1 30V, 30A, Schottky Diode 32CTQ030 International Rectifier L1* 2 Turns #16 AWG, 1µH any (optional) any (optional) L2 10 Turns #16 AWG, 4.5µH T50-52B Micrometals Q1 N-Channel Logic Level Enhancement Mode MOSFET 30V, 56 A RL3103 International Rectifier R1 0.005Ω 1% power resistor WSR-2 Dale/Vishay R2 10Ω 5% 1/16 watt resistor any any R3 8.2kΩ 5% 1/16W resistor any any R4 6.81kΩ 1% 1/16W resistor any any R5, R8 3.92kΩ 1% 1/16W resistor any any R6 261 kΩ 5% 1/16W resistor any any R7 100 kΩ 5% 1/16W resistor any any R9 10.5 kΩ 5% 1/16W resistor any any Q1-HS TO-220 heat sink 576802 AAVID CR1–HS TO-220 heat sink 577002 AAVID Note: * 52 The L1 inductor is recommended for isolating the 5V input supply from current surges caused by MOSFET switching. L1 is not required for normal operation and may be omitted. AMD-K6® Processor Power Supply Design 21103G/0—February 1999 Voltage Regulator Vendor Information Company Name and Contact Part Number Type Remarks Cherry Contact: George Shuline (401)886-3821 CS5150 CS5151 CS5166 Switching Regulator Switching Regulator Switching Regulator a) Synchronous 4-bit DAC, 2.14 V min. b) Asynchronous 4-bit DAC, 2.14 V min. c) 5-bit VID, 1.3 V min. Corsair Microsystem Contact: John Beekley (408) 559-1777 SP52P6TS SP520P6CS SPX525P6TS Switching VRM Switching VRM Switching VRM a) 4-bit VID, 2.1 V min. b) 4-bit VID, 2.1 V min. c) 4-bit VID, 2.1 V min. Elantech Contact: Steve Sacarisen (408) 945-1323 EL7571 EL7556 Switching Regulator Switching Regulator a) Tested/ 5-bit VID DAC, 1.3 V min. b) 5-bit DAC, 1.3 V min. VRM designs are available Harris Contact: Steve River (407) 729-5949 HIP6002/3 Switching Regulator HIP6004/5/14 Switching Regulator HIP6019 2 Switchers + 2 Linear a) 4-bit DAC, 2.0 V min. b) 5-bit VID, 1.3 V min. VRM designs are available Linear Technology Corporation LT1430/35 Contact: Mike Gillespie LT1552/53 (408) 428-2060 Switching Regulator Switching Regulator a) Voltage set by resistor b) 5-bit VID, 1.8 V min. Linfinity Microelectronics Inc. Contact: Andrew Stewart (714) 372-8383 LX1660/61 LX1662/63 LX1664/65 Switching Regulator Switching Regulator Linear and Switcher a) External DAC or resistors b) 5-bit VID, 1.3 V min. c) 5-bit VID, 1.3 V min. Maxim Integrated Products Contact: Nancie George-Adeh (408) 737-7600 MAX1624 MAX1638 MAX1710 Switching Regulator Switching Regulator Switching Regulator a) 5-bit DAC, 1.1 V min. b) 5-bit VID, 1.3 V min c) 5-bit DAC, 1.1 V min. Micro Linear Contact: Doyle Slack (408) 433-5200 ML4900 ML4902 Switching Regulator Switching Regulator a) 4-bit DAC, 2.1 V min. b) 5-bit DAC, 1.8 V min. National Semiconductor http://www.national.com/pf/L M/LM2635.html LM2635 Switching Regulator a) 5-bit VID, 1.8 V min. (1.3V available) Fairchild Semiconductor Contact: David McIntyre (415) 966-7734 RC5041/42 Switching Regulator RC5051/53/54 Switching Regulator a) 4-bit VID 2.1 V min. b) 5-bit VID, 1.3 V min. Semtech Corporation Contact: Alan Moore (805) 498-2111 SC1172/73 SC1151/52 SC1182/83 a) 5-bit VID, 1.3 V min b) 5-bit VID, 1.3 V min. c) 5-bit VID, 1.3 V min. Switching Regulator Switching Regulator Switching Regulator Notes: 1) The lower value of the output voltage setting can vary between the parts listed in this table. 2) Parts with a DAC designation in the Remarks column do not follow the defined VID codes. For more information, see “Digital-to-Analog Converter (DAC)” on page 27. Voltage Regulator Vendor Information 53 AMD-K6® Processor Power Supply Design Company Name and Contact Part Number 21103G/0—February 1999 Type Remarks Texas Instruments http://wwws.ti.com/sc/psheets TPS5210 /slvs171/slvs171.pdf Switching Regulator a) 5-bit VID, 1.3 V min. Unisem Contact:Reza Amiranir (949) 453-1008 US2050 US3004 Switching Regulator Switching Regulator a) Voltage set by resistor b) 5-bit VID Unitrode Contact: John O’Connor (603) 429-8504 VXI –503-652-7300 UCC3882 UCC3881 VID073-207101 Switching Regulator Switching Regulator VXI VRM’s a) 5-bit VID, 1.8 V min. b) Voltage set by resistor VXI Contact: Joseph Chang Notes: 1) The lower value of the output voltage setting can vary between the parts listed in this table. 2) Parts with a DAC designation in the Remarks column do not follow the defined VID codes. For more information, see “Digital-to-Analog Converter (DAC)” on page 27. 54 Voltage Regulator Vendor Information