ETC 21103

®
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