U.S. patent application number 09/905187 was filed with the patent office on 2002-08-08 for power semiconductor switching devices, power converters, integrated circuit assemblies, integrated circuitry, power current switching methods, methods of forming a power semiconductor switching device, power conversion methods, power semiconductor switching device packaging methods, and methods of f.
Invention is credited to Eden, Richard C., Smetana, Bruce A..
Application Number | 20020105009 09/905187 |
Document ID | / |
Family ID | 22812786 |
Filed Date | 2002-08-08 |
United States Patent
Application |
20020105009 |
Kind Code |
A1 |
Eden, Richard C. ; et
al. |
August 8, 2002 |
Power semiconductor switching devices, power converters, integrated
circuit assemblies, integrated circuitry, power current switching
methods, methods of forming a power semiconductor switching device,
power conversion methods, power semiconductor switching device
packaging methods, and methods of forming a power transistor
Abstract
Power semiconductor switching devices, power converters,
integrated circuit assemblies, integrated circuitry, power current
switching methods, methods of forming a power semiconductor
switching device, power conversion methods, power semiconductor
switching device packaging methods, and methods of forming a power
transistor are described. One exemplary aspect provides integrated
circuitry including a monolithic semiconductive substrate; a power
semiconductor switching device comprising a plurality of field
effect transistors formed using the monolithic semiconductive
substrate and having a plurality of electrical contacts including a
plurality of gate contacts, a plurality of source contacts coupled
in parallel and a plurality of drain contacts coupled in parallel;
and auxiliary circuitry formed using the monolithic semiconductive
substrate and configured to couple with at least one of the
electrical contacts of the power field effect transistors
Inventors: |
Eden, Richard C.;
(Briarcliff, TX) ; Smetana, Bruce A.; (Colton,
WA) |
Correspondence
Address: |
WELLS ST. JOHN P.S.
601 W. FIRST
SUITE 1300
SPOKANE
WA
99201-3828
US
|
Family ID: |
22812786 |
Appl. No.: |
09/905187 |
Filed: |
July 12, 2001 |
Related U.S. Patent Documents
|
|
|
|
|
|
Application
Number |
Filing Date |
Patent Number |
|
|
60217860 |
Jul 13, 2000 |
|
|
|
Current U.S.
Class: |
257/138 ;
257/E23.015; 257/E23.062; 257/E23.067; 257/E23.079; 257/E29.116;
257/E29.136 |
Current CPC
Class: |
H01L 2924/01061
20130101; H01L 2924/10253 20130101; H01L 23/4824 20130101; H01L
2924/01005 20130101; H01L 2924/01019 20130101; H01L 2924/01077
20130101; H01L 2924/15747 20130101; H01L 2924/19041 20130101; H01L
2924/01011 20130101; H01L 2924/01074 20130101; H01L 2924/01076
20130101; H01L 2924/01082 20130101; H02M 3/33592 20130101; H01L
2924/01006 20130101; H01L 2924/01025 20130101; H01L 2924/01052
20130101; H01L 2924/01056 20130101; H01L 2924/14 20130101; H01L
24/17 20130101; H01L 2924/0101 20130101; H01L 2924/01029 20130101;
H01L 2924/1306 20130101; H01L 2924/01012 20130101; H01L 23/49822
20130101; H01L 2924/01023 20130101; H01L 2924/01046 20130101; H01L
2924/01068 20130101; H01L 2924/01043 20130101; H01L 2924/01078
20130101; H01L 2224/16225 20130101; H01L 2924/01051 20130101; H01L
2924/30107 20130101; Y02B 70/1475 20130101; H01L 2924/01015
20130101; H01L 2924/01039 20130101; H01L 2924/014 20130101; H01L
2924/01045 20130101; H01L 2924/15311 20130101; H02M 3/33507
20130101; H01L 2924/01057 20130101; H01L 2924/00014 20130101; H01L
2924/01037 20130101; H01L 2924/01073 20130101; H01L 2924/12032
20130101; H01L 2924/30105 20130101; H01L 23/642 20130101; H01L
29/4238 20130101; H02M 3/33576 20130101; H01L 27/088 20130101; H01L
29/41725 20130101; H01L 2924/15151 20130101; H01L 2224/13099
20130101; H01L 2224/16235 20130101; H01L 2924/1433 20130101; H01L
2924/13091 20130101; Y02B 70/10 20130101; H01L 23/50 20130101; H01L
2924/3011 20130101; H01L 2924/01013 20130101; H01L 2924/01027
20130101; H01L 2924/01033 20130101; H01L 23/49827 20130101; H01L
2924/0102 20130101; H01L 2924/01041 20130101; H01L 2924/01087
20130101; H01L 2924/19042 20130101; H01L 29/7802 20130101; H01L
2924/01094 20130101; H01L 2924/01018 20130101; H01L 2924/01042
20130101; H01L 29/78 20130101; H01L 23/49838 20130101; H01L
2924/01038 20130101; H01L 2924/01079 20130101; H01L 2224/16225
20130101; H01L 2924/13091 20130101; H01L 2924/10253 20130101; H01L
2924/00 20130101; H01L 2924/1306 20130101; H01L 2924/00 20130101;
H01L 2924/12032 20130101; H01L 2924/00 20130101; H01L 2924/15747
20130101; H01L 2924/00 20130101; H01L 2924/00014 20130101; H01L
2224/0401 20130101 |
Class at
Publication: |
257/138 |
International
Class: |
H01L 029/74 |
Goverment Interests
[0002] This invention was made with Government support under
Contract No. MDA-904-99-C-2644/0000 awarded by the Maryland
Procurement Office of the National Security Agency (NSA). The
Government has certain rights in this invention.
Claims
1. Integrated circuitry comprising: a monolithic semiconductive
substrate; a power semiconductor switching device comprising a
plurality of field effect transistors formed using the monolithic
semiconductive substrate and having a plurality of electrical
contacts including a plurality of gate contacts, a plurality of
source contacts coupled in parallel and a plurality of drain
contacts coupled in parallel; and auxiliary circuitry formed using
the monolithic semiconductive substrate and configured to couple
with at least one of the electrical contacts of the power field
effect transistors.
2. The circuitry of claim 1 wherein the field effect transistors
comprise planar field effect transistors.
3. The circuitry of claim 1 wherein the auxiliary circuitry
comprises a gate driver amplifier configured to provide a control
signal to the electrical contacts of the field effect transistors
comprising the gate contacts.
4. The circuitry of claim 1 wherein the auxiliary circuitry
comprises a power converter controller configured to provide a
control signal to the electrical contacts of the field effect
transistors comprising the gate contacts.
5. The circuitry of claim 1 wherein the gate contacts are coupled
in parallel.
6. The circuitry of claim 1 wherein the auxiliary circuitry
comprises an application specific integrated circuit.
7. The circuitry of claim 1 wherein the auxiliary circuitry
comprises a zero-current switching/timing circuit.
8. The circuitry of claim 1 wherein the auxiliary circuitry
comprises a load protection circuit.
9. The circuitry of claim 1 wherein the auxiliary circuitry
comprises an active snubber circuit.
10. The circuitry of claim 1 wherein the power semiconductor
switching device and the auxiliary circuitry are formed upon a
die.
11. The circuitry of claim 1 wherein the field effect transistors
comprise MOSFET devices.
12. A method of forming a power transistor comprising: providing a
monolithic semiconductive substrate having a surface; forming a
power field effect transistor using the monolithic substrate and
having a source contact and a drain contact adjacent to the
surface; and forming auxiliary circuitry using the monolithic
semiconductive substrate, the forming comprising coupling the
auxiliary circuitry with at least one contact of the power field
effect transistor.
13. The method of claim 12 wherein providing comprises providing
the substrate comprising a semiconductor die.
14. The method of claim 12 wherein the forming the power field
effect transistor comprises forming a plurality of planar field
effect transistors electrically coupled in parallel.
15. The method of claim 12 wherein the forming auxiliary circuitry
comprises forming a gate driver amplifier configured to provide a
control signal to a gate contact of the power field effect
transistor.
16. The method of claim 12 wherein the forming auxiliary circuitry
comprises forming a power converter controller configured to
provide a control signal to a gate contact of the power field
effect transistor.
17. The method of claim 12 wherein the forming auxiliary circuitry
comprises forming the auxiliary circuitry comprising application
specific integrated circuitry.
18. The method of claim 12 wherein the formings individually
comprise forming the power field effect transistor and the
auxiliary circuitry comprising CMOS devices.
19. The method of claim 12 wherein the forming auxiliary circuitry
comprises forming the auxiliary circuitry comprising zero-current
switching timing circuitry.
20. The method of claim 12 wherein the forming auxiliary circuitry
comprises forming the auxiliary circuitry comprising active snubber
circuitry.
21. The method of claim 12 wherein the forming auxiliary circuitry
comprises forming the auxiliary circuitry comprising load
protection circuitry.
22. The method of claim 12 wherein the forming the power field
effect transistor comprises forming a plurality of MOSFET devices.
Description
CROSS REFERENCE TO RELATED APPLICATION
[0001] This application claims priority from U.S. Provisional
Application Serial No. 60/217,860, which was filed on Jul. 13,
2000, titled "Low Cost Ultra-Low On-Resistance High-Current
Switching MOSFET for Low Voltage Power Conversion", naming Richard
C. Eden and Bruce A. Smetana as inventors, and which is
incorporated by reference herein.
TECHNICAL FIELD
[0003] This invention relates to power semiconductor switching
devices, power converters, integrated circuit assemblies,
integrated circuitry, current switching methods, methods of forming
a power semiconductor switching device, power conversion methods,
power semiconductor switching device packaging methods, and methods
of forming a power transistor.
BACKGROUND OF THE INVENTION
[0004] Computational power of digital processing circuitry is
related to the conversion of input DC power to waste heat. As
digital computational powers increase, the associated power
consumption and heat generated by processing devices also increase.
Power supply voltages of logic circuits have been reduced from 5
Volts to 1.2 Volts or less to alleviate excessive generation of
heat and power consumption. However, reduction of power supply
voltages has complicated other issues of power supply and
distribution to logic circuits. For example, electrical resistance
between power supplies and logic circuits has a more significant
impact upon efficiency as supply voltages continued to be
reduced.
[0005] Some designs have provided power to PC boards at high
voltages (e.g., 48 Volts) and then utilize on-board converters to
convert the received high voltage energy to 1.2 Volt or other low
voltage supply energy for application to logic circuits. To
minimize the size of such converters, the stored energy
requirements in the magnetics and capacitors can be reduced by
increasing the switching frequency of the converter. However,
conventional power semiconductor configurations utilized in
converters and capable of handling relatively large currents can
not typically switch efficiently at the desired switching
speeds.
[0006] FIG. 1 depicts a conventional vertical geometry power MOSFET
device having a plurality of n+ source contact regions 3 which lie
within p (body) regions 3P (typically formed as hexagonal islands),
where both the p (body) regions 3P and the n+ source contact
regions 3 are electrically connected to the upper source contact
metal 3M. The gate conductors 2 are insulated from source contact
metal 3M under which they lie by the insulator 31 and from the
silicon substrate by the thin gate insulator 21. The gate
conductors 2 cover the regions between the p (body) regions 3P,
extending across the edge (surface channel) portion of the
perimeter of the p (body) regions 3P to the n+ source regions 3.
When the gate conductors 2 are biased more positively than the
threshold voltage of this conventional n-channel power MOSFET,
electron flow through these surface channel regions is induced
which results in electron flow along indicated paths 4. Electron
paths 4 are formed from the adjacent n+ source regions 3
horizontally through the surface channel, vertically through the n-
drain drift region 5N to the n+ drain region 5 to the bottom drain
metallization contact 6 shown in FIG. 1. This current flow path
leads to values of source-drain ON resistance that are higher than
desired for efficient low voltage power conversion
applications.
[0007] The equivalent circuit of a conventional power MOSFET
illustrated in FIG. 1 is depicted in FIG. 2. A p-n body diode 7 is
provided from the source 3 to the drain 6 and comprises the p body
region 3P and the n- and n+ drain regions 5N and 5 shown in FIG. 1.
The body diode 7 is a relatively large p-n.sup.--n.sup.+ diode with
a very large diffusion charge storage capacity Q.sub.d.
Accordingly, when the body diode 7 is first reversed biased after
heavy forward conduction, a large transient reverse current i.sub.r
flows for a substantial period of time t.sub.r=Q.sub.d/i.sub.r
which can limit usable switching frequencies.
[0008] There exists needs for improved semiconductor devices and
methodologies which overcome problems associated with conventional
arrangements.
BRIEF DESCRIPTION OF THE DRAWINGS
[0009] The patent or application file contains at least one drawing
executed in color. Copies of this patent or patent application
publication with color drawing(s) will be provided by the Office
upon request and payment of necessary fees.
[0010] Preferred embodiments of the invention are described below
with reference to the following accompanying drawings.
[0011] FIG. 1 is a cross-sectional view of a conventional vertical
power MOSFET device.
[0012] FIG. 2 is a schematic representation of the conventional
power MOSFET equivalent circuit.
[0013] FIG. 3 is schematic illustration of an exemplary synchronous
rectification power converter.
[0014] FIG. 4 is an illustrative representation of exemplary
components formed upon a monolithic semiconductor die.
[0015] FIG. 5 is a side cross-sectional view of an exemplary
high-current, low resistance area interconnect die package
assembly.
[0016] FIG. 6 is another (top) cross-sectional view of the
high-current, low resistance area interconnect die package assembly
of FIG. 5.
[0017] FIG. 7 is a cross-sectional view of an alternative
high-current, low resistance area interconnect die package
assembly.
[0018] FIG. 8 is a detailed cross-sectional view of the package and
a planar high-current silicon switch die of FIG. 5 or FIG. 7.
[0019] FIG. 8A is another cross-sectional view of a portion of the
integrated circuit assembly of FIG. 8.
[0020] FIG. 9 is a cross-sectional view of another alternative
exemplary configuration of an integrated circuit assembly.
[0021] FIG. 10 is a plan view of an exemplary semiconductor die
provided in a flip-chip arrangement.
[0022] FIG. 11 is a side view of an exemplary integrated circuit
assembly including the integrated circuit die of FIG. 10.
[0023] FIG. 12 is another side view of the integrated circuit
assembly shown in FIG. 11.
[0024] FIG. 13 is an illustrative representation depicting a
plurality of exemplary planar MOSFET devices of a power
semiconductor switching device.
[0025] FIG. 14 is a plan view of an exemplary arrangement of the
planar MOSFET transistors of FIG. 13 fabricated within a
semiconductor die.
[0026] FIG. 15 is a schematic representation of an exemplary gate
driver amplifier coupled with the power semiconductor switching
device.
[0027] FIG. 16 is a schematic representation of an exemplary active
diode p-channel MOSFET coupled with the power semiconductor
switching device.
[0028] FIG. 17 is a graphical representation depicting drain
currents versus a plurality of drain-source voltages of an
exemplary power semiconductor switching device.
[0029] FIG. 18 is another graphical representation depicting drain
currents versus additional drain-source voltages of the exemplary
power semiconductor switching device graphed in FIG. 17.
[0030] FIG. 19 is yet another graphical representation depicting
drain currents versus drain-source voltages of the exemplary power
semiconductor switching device graphed in FIG. 17 and FIG. 18.
DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0031] This disclosure of the invention is submitted in furtherance
of the constitutional purposes of the U.S. Patent Laws "to promote
the progress of science and useful arts" (Article 1, Section
8).
[0032] Like reference numerals represent like components with
differences therebetween being represented by a convenient suffix,
such as "a".
[0033] Referring to FIG. 3, a switching power converter 10 is
depicted according to exemplary aspects of the present invention.
The switching power converter 10 illustrated in FIG. 3 is
configured as a synchronous rectification power converter. Other
configurations of power converter 10 are possible. The illustrated
embodiment of power converter 10 includes a transformer 12
comprising a primary side 14 and a secondary side 16. Power
converter 10 is configured to implement DC to DC power conversion
operations. High-current, low "on" resistance planar MOSFET power
semiconductor switching devices illustrated as reference 50
provided according to aspects of the present invention are included
on the secondary side 16 of power converter 10 in the depicted
embodiment. In a conventional diode rectified power converter, the
switches 50 would be replaced by a pair of Schottky diode
rectifiers with their anodes grounded.
[0034] Primary side 14 comprises a center-tapped primary
transformer winding 20 coupled with plural primary switching
devices 22, 24 and primary terminals 26. Switching devices 22, 24
are implemented intermediate one of primary terminals 26 and
primary transformer winding 20.
[0035] Secondary side 16 includes a center-tapped secondary
transformer winding 30 coupled with a plurality of secondary
switches 32, 34 and secondary terminals 36. Primary side 14
receives electricity via terminals 26 at a first voltage and a
first current in one implementation of power converter 10.
Secondary transformer winding 30 is coupled with primary
transformer winding 20 and is configured to provide electricity at
a second voltage different than the first voltage and a second
current different than the first current to terminals 36. For
example, in the depicted exemplary embodiment, primary side 14 is
configured to receive electricity at a first voltage greater than
the voltage provided upon secondary side 16. In addition, primary
side 14 receives current having a magnitude less than the magnitude
of current provided at terminals 36.
[0036] In one exemplary application, power converter 10 is
implemented to provide electricity to an associated microprocessor
and/or other processing device(s) in a low voltage application
(e.g., 1.2 Volts, 2.5 Volts) wherein utilization of diodes, for
example in a diode rectification power converter, is rendered
inefficient by the excessive voltage drops across the diode
rectifier switching devices. For example, power converter 10 may be
utilized in a personal computer, server, work station or other
processing and logic circuitry applications, including low voltage,
high current applications.
[0037] As shown, a controller 38 is provided coupled with primary
switches 22, 24, secondary switches 32, 34 and secondary terminals
36. Controller 38 is configured to monitor output voltage and
current via secondary terminals 36 to control operations of power
converter 10. For example, controller 38 controls the timing of
switching operations of primary switches 22, 24 and secondary
switches 32, 34 to implement appropriate power conversion
operations and to maintain the voltage and current of electricity
within secondary side 16 in a desired range.
[0038] As described in detail below, secondary switches 32, 34 are
configured to withstand power currents. Exemplary power currents
include currents experienced within power devices which are greater
than typical signaling currents which are usually a few tens of
milliamps. Exemplary power currents are greater than one Ampere and
reach magnitudes of 100 Amperes-200 Amperes or greater in exemplary
configurations. Secondary switches 32, 34 are implemented as power
semiconductor switching devices 50 described herein according to
additional aspects of the invention. Such devices 50 may be
configured to accommodate currents approaching or exceeding 1000
Amperes in exemplary configurations.
[0039] Power semiconductor switching devices 50 comprise high
current devices having ultra low ON resistance values (R.sub.on) in
exemplary configurations. Utilizing power semiconductor switching
devices 50 described herein, R.sub.on values less than 0.00015 Ohms
are possible for configurations capable of handling 200 Amperes.
Accordingly, it is favorable to use the low voltage, high current
power devices 50 as secondary switches 32, 34 having low R.sub.on
values for improved efficiency.
[0040] In the depicted power converter applications, controller 38
is configured to sense output voltages and currents at terminals
36. Responsive to such monitoring, controller 38 applies control
signals to primary switches 22, 24 and secondary switches 32, 34 to
control the operation of the switches to maintain the voltages and
currents at terminals 36 within a desired range. As illustrated,
controller 38 applies control signals to the gates of secondary
switches 32, 34 implemented as power semiconductor switching
devices 50 according to exemplary aspects of the invention.
[0041] Referring to FIG. 4, an exemplary configuration of secondary
switches 32, 34 configured as power semiconductor switching devices
50 according to aspects of the present invention is illustrated.
Power semiconductor switching devices 50 individually comprise a
power MOSFET transistor which may be formed by a plurality of
MOSFET transistors coupled in parallel according to additional
aspects of the invention. One common control signal is utilized to
control the individual MOSFET transistors of the power device 50.
Exemplary MOSFET transistors utilized to form an individual one of
the power semiconductor switching devices 50 are described below as
reference 174 in FIG. 13 and FIG. 14. Such transistors are
implemented as planar, horizontally configured n-channel MOSFET
devices in one exemplary aspect.
[0042] FIG. 4 depicts a monolithic semiconductor die 52 containing
plural power semiconductor switching devices 50 according to one
embodiment. The semiconductor die 52 is fabricated from a
monolithic semiconductive substrate 56 in the described exemplary
embodiment. For example, semiconductor die 52 is formed from a
semiconductive wafer (not shown) such as silicon, silicon carbide,
gallium arsenide or other appropriate semiconductive substrate.
[0043] In the described embodiment, portions of semiconductive
substrate 56 are doped with a p-type dopant providing a p-substrate
or p-well. In addition, an n-well may also be formed within
semiconductive substrate 56. As described further below, n-channel
MOSFET devices are formed within portions of substrate 56
comprising p-type portions or wells of substrate 56 and p-channel
MOSFET devices are formed within n-type portions or wells of
substrate 56. P-type and n-type portions of substrate 56 are shown
in FIG. 13 for example.
[0044] The depicted monolithic semiconductor die 52 also comprises
auxiliary circuitry 54 according to aspects of the present
invention. Auxiliary circuitry 54 is circuitry apart from circuitry
comprising power semiconductor switching devices 50. In some
applications, auxiliary circuitry 54 is coupled with and favorably
utilized in conjunction with power devices 50. For example,
auxiliary circuitry 54 comprises controller circuitry or driver
circuitry to control power semiconductor switching devices 50. In
other applications, auxiliary circuitry 54 is not coupled with and
is unrelated to operations of power semiconductor switching devices
50. In some configurations, auxiliary circuitry 54 is implemented
as application specific integrated circuitry (ASIC).
[0045] The depicted semiconductor die 52 is illustrated with
respect to the power converter 10 application of FIG. 3. Other
configurations of semiconductor die 52 including other arrangements
of power semiconductor switching devices 50 and auxiliary circuitry
54 configured for other applications are possible. In the
illustrated exemplary configuration, auxiliary circuitry 54 formed
upon monolithic semiconductor die 52 and comprising controller 38
and gate amplifiers 60 are configured to couple with at least one
of the electrical contacts (i.e., gate contact in the arrangement
of FIG. 4) of the respective power semiconductor switching devices
50.
[0046] Regardless of the application of die 52 or power devices 50,
aspects of the present invention provide auxiliary circuitry 54
upon monolithic semiconductor die 52 including power semiconductor
switching devices 50. Fabrication of power semiconductor switching
devices 50 using common CMOS processing methodologies according to
aspects of the present invention or other processing techniques
facilitates formation of auxiliary circuitry 54 using similar
processing methodologies upon monolithic semiconductor die 52. For
example, if CMOS processes are used to form devices 50, such CMOS
processes can also be utilized to form auxiliary circuitry 54, if
desired. Power semiconductor switching devices 50 may be fabricated
simultaneously with auxiliary circuitry 54 in such arrangements.
The capability to provide auxiliary circuitry 54 upon the
semiconductor die 52 is enabled by the fabrication of power
semiconductor switching devices 50 comprising planar MOSFET devices
according to aspects of the present invention which may be
fabricated within a standard CMOS foundry ordinarily used to
fabricate small-signal digital or analog circuits.
[0047] Apart from the illustrated fabrication efficiencies of power
semiconductor switching devices 50 and auxiliary circuitry 54,
there may be other reasons to provide auxiliary circuitry 54 upon
the same semiconductor die 52 as devices 50. For example, and as
described further below, exemplary configurations of power
semiconductor switching devices 50 individually comprise a
plurality of MOSFETs, including for example, thousands of
parallel-coupled MOSFETs (a plurality of MOSFET switching devices
of individual power devices 50 are depicted in exemplary
configurations in FIG. 13 and FIG. 14). Provision of auxiliary
circuitry 54 including control circuitry and driver circuitry upon
the same die 52 as devices 50 advantageously facilitates driving
the gate capacitances of the individual MOSFETs comprising power
devices 50 and minimizes the capacitance which must be driven by
external control circuitry such as the outputs from the converter
controller 38. Further details regarding driving gate capacitances
of devices 50 are described below and other advantages may be
gained by providing auxiliary circuitry 54 upon monolithic
substrate 56.
[0048] In particular, it may be desired to provide converter
controller 38 adjacent to the secondary side 16 in synchronous
rectifier switching applications. According to embodiments
described herein, auxiliary circuitry 54 is configured as converter
controller 38 provided upon die 52 including power devices 50.
Providing controller 38 adjacent to devices 50 is advantageous to
significantly reduce parts count in providing power converter
products. Controller circuitry 38 may be implemented using analog
or digital control circuit configurations.
[0049] As mentioned above, power converter controller 38 provides
respective control signals to control the operation of secondary
switches 32, 34 implemented as power semiconductor switching
devices 50 upon die 52 according to aspects of the present
invention. Controller 38 is also configured to apply control
signals externally of semiconductor die 52. For example, controller
38 is arranged in the exemplary configuration to apply control
signals (Phi.sub.p1, Phi.sub.p2) to primary switches 22, 24 to
control switches 22, 24, generally through some type of isolation
device to enable isolation between the input side 14 and output
side 16 grounds of power converter 10. Power controller 38 is also
coupled with output terminals 36 to sense voltages and currents
within the secondary side 16 of power converter 10. Bond pads (not
shown) are provided upon semiconductor die 52 to provide coupling
of die 52 with external circuitry including terminals 36 and
isolation devices coupling to primary switches 22, 24.
[0050] As discussed above, auxiliary circuitry 54 further includes
gate driver amplifier circuits 60 coupled intermediate controller
38 and respective power semiconductor switching devices 50. Gate
driver amplifier circuits 60 operate to improve power conversion
operations or other operations requiring controlled switching of
power semiconductor devices 50. Further details regarding an
exemplary configuration of amplifiers 60 are discussed below in
FIG. 15 as reference 180.
[0051] As mentioned above and according to aspects of the
invention, power semiconductor switching devices 50 may be
individually implemented as a plurality of parallel-coupled MOSFETs
including parallel-coupled gates, parallel-coupled sources and
parallel-coupled drains. Gate driver amplifier circuits 60 are
configured to provide respective control signals to the
parallel-coupled gates of the MOSFETs of power semiconductor
switching devices 50. Depending upon the application of power
devices 50 (e.g., application within power converter 10)
significant input currents may be required to drive the MOSFET
gates of individual power semiconductor switching devices 50.
Utilization of gate driver amplifiers 60 according to aspects of
the invention facilitates driving these input currents and
minimizes the capacitance which must be driven by the control
circuitry controlling the gates of power semiconductor switching
devices 50 providing improved switching speeds.
[0052] Further, large current spikes involved in charging and
discharging gate capacitance of power semiconductor switching
devices 50 comprising numerous parallel-coupled MOSFETs places
demands on available current and provides a serious limitation to
timing precision and practical switching speeds. In some
configurations, (e.g., some configurations of power semiconductor
switching devices 50 comprise 500,000 MOSFET devices or more
arranged in parallel as described below) 20 nF gate capacitances
are driven. The utilization of gate driver amplifiers 60 upon
semiconductor die 52 reduces the capacitance at the gate control
input to the gate drive amplifiers 60 to 4.5 pF of a 500,000 MOSFET
device 50 which may be easily controlled directly from digital
timing circuits such as controller 38 to precisions of a few
nanoseconds.
[0053] Alternative auxiliary circuitry 54 includes zero-current
switching/timing circuitry to detect the absence of currents within
secondary inductive device 30. Zero-current switching/timing
circuitry may be utilized to determine proper moments in time for
controlling switching of secondary switches 32, 34.
[0054] According to additional aspects, auxiliary circuitry 54
includes load protection circuitry configured to detect voltage
overage conditions and current overage conditions. Controller 38
coupled with appropriate load protection circuitry controls
operation of power semiconductor switching devices 50 responsive
thereto, including opening or closing appropriate switching devices
22, 24, 32, 34.
[0055] According to additional aspects, auxiliary circuitry 54
includes protection circuitry configured to detect drain-source or
gate-source voltage overage conditions within devices 50 which
could potentially be damaging to the devices and to institute
corrective action such as to mitigate the over voltage conditions.
Implementation of this function requires coupling from the drains
of the power semiconductor switching devices 50 through some type
of internal or external voltage discrimination elements (such as a
Zener diodes or transient voltage suppression (TVS) circuitry) back
to either auxiliary inputs on the respective gate drive amplifiers
60 or to the controller 38 (FIG. 3). When the voltage
discrimination element provides feedback that a dangerous
overvoltage condition exists at the drain of one of power MOSFET
devices 32 or 34, the protection circuitry functions to generate a
gate voltage on that device such as to cause a momentary drain
current pulse capable of suppressing the overvoltage condition in
the manner of an active snubber circuit.
[0056] Exemplary configurations of zero-current switching/timing
circuitry, load protection circuitry, active snubber circuitry or
other configurations of auxiliary circuitry 54 are implemented as
application specific integrated circuitry (ASIC) as mentioned
above.
[0057] The depicted exemplary configuration of device 50 within
semiconductor die 52 is provided for discussion purposes with
respect to the exemplary power converter 10 application. Other
configurations are possible including provision of a single power
semiconductor switching device 50, additional power semiconductor
switching devices 50 and/or other associated auxiliary circuitry 54
upon an appropriate semiconductor die 52. Other power conversion
configurations and other applications of power semiconductor
switching devices 50 described herein are possible. Power
semiconductor switching devices 50 are usable in other low voltage,
high current applications apart from exemplary applications
described herein.
[0058] Aspects of the present invention also provide devices and
methods for packaging of power semiconductor switching devices 50
and semiconductor die 52 including devices 50. As set forth above,
devices 50 according to exemplary configurations comprise plural
MOSFET devices coupled in parallel. Examples of such MOSFETs are
depicted in FIG. 13 and FIG. 14 as planar, horizontally configured
MOSFETs having high-current power electrodes (e.g., source and
drain) provided adjacent on a common surface as opposed to
conventional power MOSFETs wherein the source is provided upon an
upper surface and the drain is provided upon an opposing lower
surface, as illustrated in FIG. 1.
[0059] Various exemplary packaging configurations and integrated
circuit assemblies according to aspects of the present invention
are described hereafter with reference to FIG. 5-FIG. 12. The
illustrated configurations are exemplary and other packaging or
assembly arrangements are possible for power semiconductor
switching devices 50, including other combinations of the various
layers and electrical connections depicted in FIG. 5-FIG. 12.
[0060] Referring to FIG. 5 and FIG. 6, a first exemplary integrated
circuit assembly 70 is shown. Integrated circuit assembly 70
includes semiconductor die 52 including one or more power
semiconductor switching device 50 and auxiliary circuitry 54 (if
circuitry 54 is provided within die 52). In the illustrated
example, semiconductor die 52 is implemented in a flip chip
configuration. Other packaging designs of semiconductor die 52 are
possible.
[0061] Integrated circuit assembly 70 further includes a package 73
coupled with semiconductor die 52. Package 73 comprises one or more
intermediate layer 74, a source plane 76, and a drain plane 78
(plural layers 74, 74a are illustrated in the exemplary package
configuration of FIG. 5 and one intermediate layer is shown in the
exemplary arrangement of FIG. 3 in Appendix A).
[0062] Intermediate layer(s) 74 comprise "fineline" layer(s) in the
illustrated exemplary embodiment. For example, intermediate
layer(s) 74 are implemented as plated copper planes (5-10 microns
typical). Intermediate layer(s) 74 are capable of being patterned
to horizontal feature sizes sufficiently small as to allow area
array contact to the semiconductor die 52. Intermediate layer(s) 74
which provide electrical conduction in a horizontal direction may
be referred to as horizontal interconnect layers.
[0063] In one exemplary embodiment, drain plane 78 is implemented
as a plated copper plane (0.010 inches or 250 microns typical) and
plane 76 is implemented as a source plane comprising
copper-invar-copper (0.030 inches or 750 microns typical). Other
configurations of layer(s) 74 and planes 76, 78 are possible.
[0064] For example, the drain plane 78 may be divided into two or
more portions in an exemplary alternative configuration wherein
individual portions service respective portions (e.g., halves) of
the area of die 52 and coupled with two separate drain contact
areas. In such an arrangement, a die containing two power devices
50 as illustrated in FIG. 4 could be accommodated (whereas the
illustration of FIG. 5 and FIG. 6 shows only a single high-current
drain contact 82, along with the single high-current source contact
80, suitable for a die 52 having a single power device 50 in the
depicted exemplary configuration).
[0065] In addition, the integrated circuit assembly 70 includes a
single terminal source contact 80 and a single terminal drain
contact 82 in the depicted exemplary configuration. Opposing ends
of planes 76, 78 define terminal contacts 80, 82 in the exemplary
configuration. Terminal contacts 80, 82 are configured to couple
with devices external of assembly 70 upon installation of assembly
70 into final products (e.g., terminal contacts 80, 82 couple with
a motherboard in an exemplary computer application).
[0066] Referring to FIG. 6, one or more gate lead replaces a drain
solder ball in the array to provide connectivity to the gate of
device 50. The package lead which attaches to the gate may be
fabricated in layer 74 if the gate ball is at the perimeter. FIG. 6
depicts further details of the package 73 of FIG. 5 looking
downward through a cross-sectional line passing through
intermediate layer 74a. Die 52 and gate contact 89 fabricated using
layer 74 are shown in FIG. 6 for illustrative purposes.
[0067] A plurality of electrical interconnects 84 are depicted
which couple intermediate semiconductor die 52 and intermediate
layer 74 of package 73. Exemplary electrical interconnects 84
include respective gate, source and drain solder balls 87 coupled
with respective mating pads fabricated in intermediate layer 74 in
one exemplary configuration. Intermediate layer 74 also provides
contact to the gate (as mentioned above) and other lower current
connections to the semiconductor die 52 as well as providing
external package connection pads for coupling these to a circuit
board through suitable connections and can form, in conjunction
with solder balls 87 and intermediate layer 74a if desired, the
electrical connection between lower current connections of die 52
and their respective external package pads. Electrical
interconnects 84 coupled with respective source, drain, gate and
other bond pads of semiconductor die 52 (not shown) provide
connectivity of such bond pads to intermediate layer 74.
[0068] As shown in the depicted embodiment, intermediate layer 74a
is spaced from semiconductor die 52 including power semiconductor
switching devices 50 therein. Intermediate layer 74a provides
proper coupling of source bond pads of semiconductor die 52 with
source plane 76 and permits coupling of drain bond pads of
semiconductor die 52 with drain plane 78.
[0069] Intermediate layer 74a is generally implemented as a source
plane in the described embodiment. Layer 74a is coupled with source
plane 76 and source terminal contact 80 using a plurality of via
conductors 90 within vias 88 passing through, but insulated from,
drain plane 78. Via conductors 90 provide connectivity of source
plane 76 with the proper associated source electrical interconnects
84 and source portions of intermediate layer 74a.
[0070] As shown, layer 74a includes a plurality of vias 86.
Portions 77 of layer 74a within vias 86 and electrically insulated
from the remainder of layer 74a provide connectivity of respective
drain bond pads of die 52 to drain plane 78. Portions of layer 74a
within vias 86 provide lateral, horizontal matching of drain bond
pads and respective drain interconnects 84 to drain plane 78 as
shown illustrating layer 74a as a horizontal interconnect
layer.
[0071] Electrical interconnects 84 couple source bond and drain
pads of semiconductor die 52 with respective portions of layer 74
in the depicted embodiment. Although not illustrated, an
appropriate insulative dielectric material may be provided within
vias 86, 88, intermediate source plane 76 and drain plane 78, and
intermediate source and drain portions of layer 74a. Further,
dielectric material may provided at other desired locations to
provide appropriate electrical insulation, such as between planes
76 and 78, between plane 78 and layer 74a, and between layers 74
and 74a. Insulative underfill material may also be provided beneath
semiconductor die 52 to protect solder balls 87 and semiconductor
die 52.
[0072] As illustrated, FIG. 5 and FIG. 6 depict an exemplary
integrated circuitry assembly 70 comprising an area array bumped
flip-chip configuration of die 52 mounted to a fineline augmented
package 73. The high density solder bump array may be provided
directly upon die pads of semiconductor die 52 and contact a mating
high density pad array patterned in layer 74 over layer 74a of
package 73.
[0073] Package metallization layers thick enough to handle some
high source and drain currents can not be patterned fine enough to
directly mate to some chip solder bump pitches. Accordingly, in
some embodiments, one or more intermediate layer (comprising
fineline additive copper/polymer interconnect layers in one
embodiment as mentioned above) is implemented to provide an
exemplary solution to feature size mismatch of semiconductor die 52
and package 73.
[0074] The intermediate layers are implemented to include pattern
plated copper conductor layers with patterned benzocyclobutene
(BCB) sold under the trademark CYCLOTENE available from the Dow
Chemical Company, and comprising appropriate dielectric layers in
one exemplary embodiment. Other patternable dielectric materials
such as polyimide could also be used as an alternative to BCB. The
described exemplary intermediate layer 74 is capable of being
patterned with requisite area bump array feature sizes which are
thicker than on-chip metal layers (e.g., typically 5-20 microns of
copper compared to 0.4-0.8 microns of aluminum upon a CMOS die).
However, such are not as thick as 10 mils (254 microns) or more of
copper upon main packaging layers implemented as planes 76, 78 in
the described embodiment where undivided source and drain currents
pass from the package contacts. Further details regarding similar
or alternative configurations of package 73 are illustrated below
and in FIG. 6A and FIG. 6B of the Appendix and described in the
associated text thereof.
[0075] Referring to FIG. 7, an alternative configuration of
integrated circuit assembly is illustrated as reference 70a.
Integrated circuit assembly 70a may be utilized in applications
having finer solder ball pitch compared to assembly 70. Integrated
circuit assembly 70a includes package 73a including a plurality of
intermediate layers depicted as references 74b, 74c. Plural
intermediate layers 74b, 74pc comprising fineline layers for
example and configured as horizontal interconnect layers as shown
provide additional flexibility in accommodating feature size
mismatch of semiconductor die 52 and package 73a. Additional
horizontal interconnect layers for example implemented as
additional intermediate layers are utilized in other embodiments if
desired to couple additional source and drain bond pads of the
semiconductor die 52 with an associated package. Alternatively, in
other configurations, the packages are provided with no horizontal
interconnect layers and source and drain pads of semiconductor die
52 are coupled with respective source and drain planes 76, 78 using
appropriate via conductors 90 or other appropriate
configurations.
[0076] Layers 74b, 74c configured as horizontal interconnect layers
provide proper connectivity of source and drain electrical
interconnects 84a implemented as solder balls 87 with the
respective appropriate planes 76, 78a. Layer 74b provides lateral,
horizontal alignment for drain connections while layer 74c provides
lateral, horizontal alignment for source connections.
[0077] Via conductors 90 provide electrical coupling of appropriate
portions of layer 74c with source plane 76. The number of via
conductors 90 and vias 88 may be varied according to magnitude of
currents to be conducted. Although not shown in FIG. 7, appropriate
insulative material is provided in assembly 70a to effectively
insulate source and drain conductors of die 52 and package 73a.
[0078] The integrated circuit assemblies depicted herein minimize
resistance intermediate source and drain terminal contacts 80, 82.
In the exemplary configuration of FIGS. 5 and 6, for example, a
total (source+drain) package resistance of approximately 0.0001
Ohms is observed between source and drain terminals 80, 82 in a
configuration having a 4 mm.times.4 mm square semiconductor die 52
having a total of 256 solder balls on a 250 micron pitch (in
accordance with the geometry shown in FIGS. 6A and 6B of Appendix
A) of a device 50 implemented as a planar MOSFET and comprised of
500,000 MOSFET devices coupled in parallel capable of conducting
currents of approximately 200 Amperes. If a finer solder bump
pitch, such as 1600 solder balls on a 100 micron pitch were used
with the same sized semiconductor die 52 using a finer package
contact array pitch as illustrated in FIG. 7 but with the same
thicknesses of intermediate layer(s) 74 and planes 76, 78 cited in
conjunction with FIGS. 5 and 6, the calculated total (source+drain)
package resistance is approximately 0.00008 Ohms. (As shown in FIG.
10 of Appendix A, the modest 23% calculated reduction in package
resistance from 103 microOhms to 80 microOhms by going from 256 to
1600 solder balls is accompanied by a calculated reduction in
on-chip metallization [the resistance of the metallization layers
on the semiconductor die 52] resistance by a factor of 3.7 from 111
microOhms to 30 microOhms).
[0079] FIG. 8 and FIG. 9 depict cross-sectional views of exemplary
assemblies for power semiconductor switching devices 50 capable of
conducting currents in excess of 200 Amperes and comprising
approximately 500,000 planar switching MOSFET devices 174 coupled
in parallel (only three n-channel devices 174 are shown in each of
FIG. 8 and FIG. 9, while FIG. 8A shows an end view of 8 of these
n-channel devices). The assembly 70 of FIG. 8 generally corresponds
to the assembly depicted in FIG. 5 and FIG. 6. Further details of
similar or alternative constructions of FIG. 8 are illustrated as
FIG. 3 in the Appendix and described in the associated text
therein. Further details of assembly 70b of FIG. 9 are illustrated
in FIG. 7 of the Appendix and described in the associated text
therein.
[0080] Referring to FIG. 8, details of an exemplary 5-layer IC
metallization system 100 are depicted upon a CMOS semiconductor die
52. The illustrated metallization 100 may be utilized within the
assemblies 70, 70a, 70b, 70c and in conjunction with packages 73,
73a, 73b, 73c (or other assembly and package configurations)
although the metallization 100 of FIG. 8 is depicted with reference
to assembly 70 and package 73. FIG. 8 depicts a cross-sectional
view through planar MOSFET channel stripes and source and drain
buss bar stripes of metallization 100 looking in an "X" direction.
The depicted semiconductor die 52 includes a plurality of
source.backslash.drain regions 101 and gate regions 102
therebetween to form plural MOSFETs 174. Source.backslash.drain
regions 101 and gate regions 102 are preferably implemented as
silicide regions, comprising polysilicide for example, formed
adjacent to a surface 57 of substrate 56.
[0081] Semiconductor die 52 is coupled with an exemplary package
73. Electrical interconnects 84 comprise solder balls 87 having a
bump pitch of 50-250 microns utilizing normal or fine bump pitch
technology in exemplary configurations providing connectivity of
die 52 and package 73. Further, an intermediate layer 74a is also
depicted within package 73.
[0082] Metallization 100 includes plural metal layers including a
first metal layer 110, a second metal layer 111, a third metal
layer 112, a fourth metal layer 113 and a fifth metal layer 114
elevationally above surface 57 of semiconductive substrate 56.
First metal layer 110 depicts source and drain "Y" stripes having
dimensions of approximately 25 microns by 0.75 microns. Second
metal layer 111 depicts source and drain "X" buss bars having
dimensions of approximately 12 microns while portions corresponding
to gate 102 have a dimension of 2.5 microns. Third metal layer 112
depicts a source plane with drain holes having dimensions of
approximately 3 microns by 3 microns. Fourth metal layer 113
depicts a drain plane having source holes having dimensions of
approximately 3 microns by 3 microns. Fifth metal layer 114 depicts
a source and drain checkerboard configuration with ball pads
comprising source pads 116 and drain pads 118.
[0083] Further details of metallization 100 of a 200 Ampere NMOS
switching power transistor are discussed in the Appendix. Details
of an exemplary metal layer 110 are discussed in the Appendix with
reference to FIG. 4B and the associated text thereof. Further
details of an exemplary metal layer 111 are discussed in the
Appendix with reference to FIG. 4C and the associated text thereof.
Further details of an exemplary metal layer 112 are discussed in
the Appendix with reference to FIG. 4D and FIG. 5A and the
associated text thereof. Further details of an exemplary metal
layer 113 are discussed in the Appendix with reference to FIG. 5B
and the associated text thereof. Further details of an exemplary
metal layer 114 are discussed in the Appendix with reference to
FIG. 5C and FIG. 5D and the associated text thereof.
[0084] The illustrated intermediate layer 74 comprising a fineline
plated copper layer provides package solder bump drain and source
contact pads 122. In the illustrated embodiment, layer 74 is spaced
from substrate 56 comprising one or more power semiconductor
switching device 50. As described previously, this space may be
filled with insulative underfill material.
[0085] FIG. 8A depicts an illustrative representation of first and
second metal layers 110, 111 looking in "Y" direction while FIG. 8
and FIG. 9 look in an "X" direction. Portion 117 of layer 111
comprises a drain buss bar while portion 119 of layer 111 comprises
a source buss bar.
[0086] In an alternative embodiment to that illustrated in FIGS. 5,
6, 7 and 8, the intermediate layers may be fabricated on the
semiconductor die 52a instead of within the package 73. Referring
to FIG. 9, an alternative assembly 70b and package 73b are shown.
Plural intermediate layers 74d, 74e are depicted formed upon
semiconductor die 52a and metallization 100 including layers 110,
111, 112, 113 and 114a. Layer 74d comprises portions electrically
coupled with respective source bond pads 116 and drain bond pads
118 of metallization 114a. In FIG. 9 the area density or pitch of
electrical couplings between the semiconductor die metallization
114a and the layer 74d need not be limited to the solder bump
density, which allows much higher area array pad 116, 118 densities
to be utilized (e.g., pad pitches of 40 microns or smaller, or
10,000 or more on a 4 mm.times.4 mm semiconductor die 52a), which
substantially reduces the contribution of on-chip metallization 100
resistance to the overall chip plus package R.sub.on. The tight
mechanical coupling in FIG. 9 between the layers 74d, 74e and/or
their associated dielectric layers 124a, 124 with the chip
metallization 100 over substantially the complete area of the
semiconductor die 52a in one possible embodiment is anticipated to
offer improved robustness and potentially improved reliability
overthe fine-pitch solder bump approach of FIG. 7 or FIGS. 5, 6 and
8. BCB inter-layer dielectric 124 provides electrical insulation of
source and drain electrical connections. Other patternable
dielectric materials such as polyimide could be used here in place
of BCB. Layer 74e comprises integrated circuit source bond pads 130
and drain bond pads 132 coupled with respective portions of layer
74d. Solder balls 87 are coupled with respective source bond pads
130 and drain bond pads 132. Solder balls 87 are additionally
coupled with source solder pads 134 and drain solder pads 136 of an
external package to provide electrical connectivity to the external
package (only portions of one pair of pads 130, 132 and one pair of
pads 134, 136 are shown in FIG. 9). Exemplary external packages
include source and drain planes 76, 78 coupled with source and
drain terminal contacts 80, 82 as described previously, or the
package configuration 73c shown in FIG. 11. Other package
configurations are possible which will, in their structure,
incorporate pads 134, 136.
[0087] In an exemplary embodiment, fineline copper/polymer
intermediate layers 74d, 74e are added to a completed or
semi-fabricated semiconductor die 52a in a full-wafer process
according to exemplary aspects of the present invention. Because of
the very low sheet resistance of the fineline copper planes 74d and
74e, relatively coarse pitches may be utilized in joining the
semiconductor die 52 to relatively heavy package metallization
features without compromising ON resistance values (R.sub.on). Such
permits mating of semiconductor die 52 to relatively simple and
commercially available packages using coarse-pitch solder bump,
solder patch or other joining technologies.
[0088] Referring to FIG. 10, a lower surface of another arrangement
of semiconductor die 52b is depicted in a flip-chip configuration
comprising a plurality of solder balls 87 arranged in an array. As
depicted, source pads of semiconductor die 52b and solder balls
coupled therewith and drain pads of semiconductor die 52b and
solder balls coupled therewith are depicted in alternating columns
140, 142, respectively.
[0089] A column 144 comprises both source solder balls (S), kelvin
source and drain solder balls (K.sub.S, K.sub.D), temperature
sensing diode solder balls (D.sub.P, D.sub.N), gate drive amplifier
solder ball connections including ground (V.sub.SS), input
(G.sub.in), and 2.5 Volts (V.sub.DD) in one exemplary
embodiment.
[0090] In an alternative embodiment, one or more columns of drain
142 or source 140 solder balls can be assigned to V.sub.dd in order
to reduce the inductance and resistance of the V.sub.dd connection.
This can be accommodated at the external contact level of FIG. 12
by running the metal layers contacting these V.sub.dd columns out
the bottom direction in FIG. 12 (in which the source and drain are
on the right and left sides), or the V.sub.dd contact might be
extended to the right beyond the source contact region.
[0091] Referring to FIG. 11, an alternative package 73c of assembly
70c is depicted for coupling with electrical interconnects 84
comprising solder balls 87. Package 73c is implemented as a
vertically laminate package comprising a plurality of conductive
layers including source conductive layers 150 and drain conductive
layers 152 in an alternating arrangement to couple with solder
balls 87 of the flip-chip configuration of semiconductor die 52b
shown in FIG. 10. Layer 154 corresponds to the column 144 of
miscellaneous solder ball connections described above. Package 73c
may also be utilized in conjunction with other die configurations,
including the arrangements of semiconductor die 52 described above,
and die 52a having intermediate layers 74 illustrated in FIG. 9.
Other assembly configurations of dies and packages are
possible.
[0092] Still referring again to FIG. 11, a plurality of alternating
dielectric layers 153, 155 are provided intermediate appropriate
conductive layers 150, 152, 154 as shown. In one convenient
exemplary fabrication approach, dielectric layers 153 comprise PC
board layers to which the conductive layers 150, 152 are bonded,
while dielectric layers 155 comprise B-stage adhesive layers which
are used to bond the various PC board layers together using, for
example, the same type of lamination processes used to fabricate
multi-layer printed circuit boards. In FIG. 11 the metal layers are
shown extending a substantial distance 92 above the extent of the
circuit board and inter-board adhesive layers in the area where
contact to the solder balls is made. Typically this "pullback"
region 92 from which the PCB and inter-board adhesive layers are
absent can be fabricated by means of an etching or other removal
process after the vertical laminate package is fabricated. This
form of embodiment of the invention is anticipated to offer
potential benefits in reducing stress on the solder balls due to
differential thermal expansion between the semiconductor die 52b
and the package 73c by allowing the metal layers 150, 152 to bend
in the lateral direction in FIG. 11. This beneficial mechanical
compliance can be achieved in the opposite direction (that is, in
the lateral direction in FIG. 12) by patterning, for example,
suitably shaped vertical slots 93 in the source and drain metal
planes 150, 152 between the solder ball contact areas, such that
the solder ball contacts are made at the top of metal "towers"
which have substantial freedom to bend.
[0093] Flip-chip mounting large silicon die directly to copper or
thick PCB materials may cause severe reliability problems because
of the large differences in CTE between the materials (Si=3
ppm/.degree. C., Cu=16.6 ppm/.degree. C., PCB=19 to 32 ppm/.degree.
C.) which fatigues and breaks solder balls on thermal cycling. In a
typical PCB process, the metals are backed by a PCB dielectric. In
one fabrication method, the structure as shown is originally
fabricated with the PCB dielectrics going all of the way to the
top. The PCB dielectrics are etched away or otherwise removed to
provide flexible metal towers 95.
[0094] Referring to FIG. 12, vertical laminate package 73c is
illustrated in a side view coupled with semiconductor die 52b and
electrical interconnects 84. A source layer 150, shown in solid
outline, is coupled with electrical interconnects 84 comprising
source solder balls and extends to the right to provide source
terminal 80a. A drain conductive layer 152 is depicted extending in
an opposite direction from source conductive layer 150 to provide
drain terminal 82a. Both source 150 and drain 152 layers may be
patterned with vertical slots 93 or other suitable compliance
patterning features between solder balls to achieve lateral
compliance, particularly when used in conjunction with the pullback
92 of the PCB and inter-layer dielectrics from the ends of the
conductor towers 95 on which solder ball contact is made. Suitable
electrically insulative underfill material, not shown in FIG. 12,
may be provided intermediate die 52b and package 73c.
[0095] A vertical laminate package 73c provides ultra-low ON
resistance (R.sub.on) performance. As shown in FIG. 12, orientation
of alternating conductive metal layers 150, 152 in a laminate stack
perpendicular to a surface of semiconductor die 52b permits very
large reductions in package metal resistance by extending the
package structure vertically. Package 73c depicted in FIG. 12 may
be utilized by providing electrical interconnects 84 directly upon
semiconductor die 52 (e.g., flip chip arrangement) or with the
utilization of additive fineline metallization layers 74d, 74e upon
semiconductor die 52a as described above in the exemplary
configuration of FIG. 9.
[0096] Packaging concepts described herein provide high current
conduction while also taking advantage of low R.sub.on capabilities
of deep submicron lateral MOSFET devices realized not only at the
semiconductor device level but also at the packaged device level.
Currents applied to individual power semiconductor switching
devices 50 are divided into a large number of parallel paths with
increasing metallization or other conductor thicknesses as the
number of paths decreases. For example, it has been demonstrated
that 250,000 individual MOSFET source and drain electrodes are
coupled with 256 solder ball contacts through the utilization of
layers of metallization 100 described above upon semiconductor die
52. The packaging coupled with the semiconductor die further
reduces the number of contacts from 256 in the given example to a
single source terminal contact and a single drain terminal contact
comprising high current package leads.
[0097] Referring to FIG. 13, details regarding exemplary
transistors 174 utilized to form an exemplary power semiconductor
switching device 50 are illustrated. As described above, aspects of
the present invention provide power semiconductor switching device
50 comprising a plurality of transistors 174 coupled in parallel to
conduct the large currents (1-1000 Amperes) typically experienced
in power applications. The number of transistors 174 provided to
form a single device 50 is varied depending upon the particular
application of power semiconductor switching device 50 and the
magnitude of currents to be switched. Six n-channel transistors 174
are depicted in FIG. 13 for discussion purposes.
[0098] Semiconductor die 52 includes transistors 174 fabricated
using deep submicron CMOS integrated circuit processes according to
exemplary aspects of the present invention. CMOS integrated circuit
metallization layers (FIG. 8 and FIG. 9) are provided upon die 52
and are optimally patterned for distributing high currents with low
resistance and maximum current handling capability as described
previously in exemplary embodiments.
[0099] The present invention provides planar high-current
(i.sub.max=1 to 1000 Amperes) switching MOSFET devices having very
low ON resistance (R.sub.on=10 micro ohms to 1 milliohm typical)
and relatively low gate drive power requirements for very high
efficiency in low voltage power conversion applications. As
described above, the planar device structure of transistors 174
comprising power semiconductor switching device 50 provides
structures wherein current flows between closely spaced (e.g., 0.1
to 0.5 microns) source and drain electrodes on the same (top)
surface of the semiconductor die 52. Accordingly, aspects of the
present invention provide devices 50 comprising high current, low
R.sub.on parallel-coupled MOSFETs fabricated upon a relatively
small semiconductor die 52 using submicron to deep submicron CMOS
integrated circuit foundry processes.
[0100] FIG. 13 depicts a portion of the exemplary power
semiconductor switching device 50 comprising plural planar,
horizontal geometry high current switching MOSFET devices
implemented in a CMOS process. Power semiconductor switching device
50 is fabricated within a monolithic semiconductive substrate 56,
such as silicon, comprising die 52.
[0101] Portions of substrate 56 are formed to comprise p-type
substrate material 168 or p-wells in which to form n-channel
devices 174. In addition, other portions of substrate 56 may be
n-type doped to form n wells 170 to enable the formation of
p-channel devices 175 if desired. An inter-layer dielectric-filled
trench region 172 is typically provided for lateral electrical
isolation of n wells 170 from p-type substrate material or p wells
168.
[0102] First metal layer 110 of metallization 100 is depicted in
FIG. 13 comprising source electrodes 160 and drain electrodes 162
coupled with respective source diffusion regions 161, 163 which
correspond to diffusion regions 101. Gate electrodes 164 are
provided intermediate opposing source and drain electrodes 160, 162
and insulated from the semiconductor by a thin gate oxide to form
n-channel transistors 174.
[0103] A plurality of source regions 161 and drain regions 163 are
formed in p-type substrate 168 for the formation of n-channel
devices 174. Regions 161, 163 are doped with an n-type dopant to
form n+ source and drain regions in the exemplary embodiment. A
polysilicide layer 165 may be provided intermediate electrodes 160,
162 and respective diffusion regions 161, 163 in one embodiment to
minimize resistances therebetween. In FIG. 13, the via conductors
173 between the first level metal 110 and the polysilicide layer
165 are shown as part of the source electrodes 160 and drain
electrodes 162. Gate electrodes 164 individually comprise
polysilicide in one embodiment which are configured to couple with
layers of metallization 100. As utilized herein, the term "source"
refers to structures including electrically conductive structures
proximately coupled with a source of the power transistor and
including source contact 160 and/or source region 161 for example
and the term "drain" refers to structures including electrically
conductive structures proximately coupled with a drain of the power
transistor and including drain contact 162 and/or drain region 163
for example.
[0104] Source and drain regions 161, 163 are individually utilized
to form a plurality of adjacent transistors 174 in the depicted
exemplary embodiment. For example, a given source electrode 160 and
source region 161 are utilized to form a transistor 174 with the
drain electrode 162 and drain region 163 to the right as well as
being utilized in combination with the drain electrode 162 and
drain region 163 to the left of the given source electrode 160 to
form another transistor device 174. Accordingly, in a power device
50 configured according to this exemplary aspect and having x
number of transistors 174 coupled in parallel, x gates 164, x/2
source electrodes 160 and x/2 drain electrodes 162 are
utilized.
[0105] As depicted, semiconductive substrate 56 has a surface 57.
Source electrode 160, source region 161, drain electrode 162, drain
region 163 and gate electrode 164 are formed adjacent to surface 57
in the depicted embodiment according to the horizontal planar
configuration of CMOS-implemented transistors 174.
[0106] Source regions 161 may be connected with the p-wells in
order to avoid or minimize potential problems with floating p-wells
at excessive dV/dt occurrences. Other configurations of transistors
174 are possible.
[0107] Power semiconductor switching devices 50 individually
comprise a plurality of planar horizontally configured submicron
MOSFET transistors 174 individually including a source electrode
160, drain electrode 162, and gate electrode 164. According to one
exemplary embodiment, a single power semiconductor switching device
50 comprises 500,000 or more transistors 174 coupled in parallel to
provide a low voltage, high current power device 50. In such an
embodiment, source electrodes 160 of transistors 174 are coupled in
parallel, drain electrodes 162 are coupled in parallel and gate
electrodes 164 are coupled in parallel. Provision of parallel
coupled n-channel devices 174 enables power device 50 to conduct
currents in excess of one Ampere. An exemplary device 50 comprising
500,000 transistors 174 coupled in parallel on a 4 mm.times.4 mm
silicon die 52 enables conduction of currents up to approximately
200 Amperes.
[0108] The number of transistors 174 implemented within a given
device 50 varies depending upon the application or implementation
of device 50, as well as the width selected for the individual
transistors to be paralleled (taken as 25 microns for the examples
cited herein). The current handling capability, die size and
R.sub.on values for a device 50 vary corresponding to the numbers
of transistors 174 utilized. For example, a very small
semiconductor die 52 having an approximate area of 0.16 mm.sup.2
provides approximately 5,000 parallel-coupled transistors 174 which
conduct currents of approximately 2 Amperes with an R.sub.on of
approximately 0.01-0.02 Ohms (inclusive of n-channel MOSFET ON
resistance and on-chip metallization resistance, but not including
package resistance), while a die of 1.6 mm.sup.2 provides
approximately 50,000 parallel-coupled transistors 174 enabling
conduction of currents of approximately 20 Amperes with an R.sub.on
of approximately 0.001-0.002 Ohms, and a die area of 16 mm.sup.2
provides approximately 500,000 parallel-coupled transistors 174
which enables conduction of currents of approximate 200 Amperes
with an R.sub.on of approximately 0.0001-0.0002 Ohms, and a die
area of 80 mm.sup.2 provides approximately 2,500,000
parallel-coupled transistors 174 which conduct currents of
approximately 1000 Amperes with an R.sub.on of approximately
0.00004-0.00008 Ohms. Further details of FIG. 13 are discussed in
the Appendix with reference to FIG. 1B and the associated text
thereof.
[0109] FIG. 14 depicts an elevational plan view of a region of an
exemplary semiconductor die 50 illustrating n-channel transistors
174 forming power semiconductor switching device 50. The
illustrated region includes a plurality of rows 176 individually
including a plurality of transistors 174. The number of rows 176 is
varied and the number of transistors 174 within a row 176 is varied
depending upon the implementation of device 50, the magnitude of
currents to be conducted and desired R.sub.on values. In one
exemplary implementation of power device 50 having a row 176 height
of 25 microns (corresponding to the width of each of the individual
transistors), 500,000 transistors 174, 250,000 source regions 161,
250,000 drain regions 163 and 500,000 gates 164 are provided in a 4
mm.times.4 mm (16 mm.sup.2) die size as implemented in a nominal
0.24 micron feature size CMOS process. The R.sub.on and number of
transistors in a given die size are typically closely tied to the
IC feature size. The number of transistors also depends on the
selection of row height.
[0110] FIG. 14 depicts transistors 174 upon surface 57 of substrate
56 (FIG. 13). Source regions 161, drain regions 163 and gates 164
include polysilicide 165 (FIG. 13). A plurality of via conductors
173 are provided upon respective source regions 161 and drain
regions 163 and via conductors 177 are provided upon gates 164 to
provide vertical connectivity to first metal layer 110
elevationally over substrate 56 as shown in FIG. 8 and FIG. 9, for
example.
[0111] Individual rows 176 provide transistors 174 individually
having a channel length of approximately 25 microns. In an
exemplary nominal 0.24 micron feature size commercial CMOS process
using aluminum metallization, individual rows 176 provide W=250
microns of NFET width in a 10 micron horizontal distance.
Polysilicide 165 (FIG. 13) comprising source regions 161 and drain
regions 163 provides ohmic contacts of 4 Ohms/Sq. Via conductors
173, 177 provide 7.5 Ohms/cut to reduce current path resistance.
Polysilicide comprising gate 164 provides an ohmic contact of 7
Ohms/Sq. In the described exemplary configuration, metal layers
110, 111, 112 and 113 (FIG. 8 or FIG. 9) have sheet resistances of
0.08 Ohms/Sq., while the top metal layer 114 has a sheet resistance
of 0.04 Ohms/Sq., with a via resistance of 5 Ohms/cut between all
metal layers. Other constructions of transistors 174 and
connections to transistors 174 are possible. In particular, the Ron
and current carrying capacity of the devices could be improved if a
CMOS or other IC process using copper, rather than aluminum,
metallization is used for fabrication. Further details regarding
substantially similar or alternative constructions of FIG. 14 are
discussed in the Appendix with reference to FIG. 4A and the
associated text thereof.
[0112] Referring to FIG. 15, power semiconductor switching device
50 is illustrated as a power transistor having source terminal
contact 80, drain terminal contact 82 and a gate electrode 83.
Power semiconductor switching device 50 is coupled with an
exemplary gate driver amplifier 180 utilized to drive the gate
electrode of power device 50 coupled with a plurality of
parallel-coupled gates of transistors 174. Amplifier 180 is one
exemplary configuration of amplifiers 60 described above. More
specifically, gate driver amplifier 180 may be implemented as
auxiliary circuitry 54 upon semiconductor die 52 in one
configuration as described above and corresponding to amplifiers 60
of FIG. 4.
[0113] The illustrated exemplary gate driver amplifier 180 includes
a first stage 182 and a second stage 184. First driver stage 182
includes a p-channel device 186 wherein W.sub.d1p=2 mm and an
n-channel device 188 wherein W.sub.d1n=1 mm. Second driver stage
184 includes a p-channel device 190 wherein W.sub.d2p=100 mm and an
n-channel device 192 having W.sub.d2n=50 mm.
[0114] An input node 193 is configured to receive control signals
from an appropriate source, such as power converter controller 38,
in one embodiment. In the configuration wherein power semiconductor
switching device 50 comprises 540,000 n-channel devices an input
capacitance at node 193 is approximately 4.5 pF. At a node 194
intermediate first stage 182 and second stage 184, a capacitance of
approximately 225 pF is present. Node 195 of second driver stage
184 is coupled with gate terminal 83 of power semiconductor
switching device 50 where a capacitance of approximately 20 nF is
present.
[0115] Power semiconductor switching device 50 comprising 540,000
transistors 174 coupled in parallel provides We.sub.n=13,500 mm.
Power semiconductor switching device 50 conducts currents of 200
Amperes in the depicted embodiment with a 5 mm.times.5 mm die size
and can accommodate currents of 1000 Amperes if the switching
device, amplifier and die size are scaled up to a 10 mm.times.10 mm
die size. In the exemplary 0.24 micron CMOS process, V.sub.OD is
approximately 2.5 Volts and V.sub.SS ground in the illustrated
arrangement, assuming the source terminal 80 to be near ground
potential as used in FIGS. 3 and 4.
[0116] According to certain aspects of the invention, a bypass
capacitor 200 is coupled with the source of power semiconductor
switching device 50 and V.sub.DD. Bypass capacitor 200 is greater
than or equal to 20 nF in the depicted exemplary embodiment. Bypass
capacitor 200 is configured to provide adequate pulse current to
charge a capacitance of the gates of paralleled coupled transistors
174 of power semiconductor switching device 50 responsive to
control signals received via input 193.
[0117] In one embodiment of this invention, bypass capacitor 200 is
implemented monolithically on a CMOS chip using gate to channel
capacitance of a large number of large gatelength (e.g., L.sub.g=10
microns) n-channel MOSFETs in parallel with their common gate
electrode connected to V.sub.dd 184 and their source and drain
electrodes connected to the output source electrode 80.
[0118] In addition, connecting the source of n-channel device 192
with the source of power semiconductor switching device 50 obviates
a need for a separate body diode inasmuch as power semiconductor
switching device 50 turns on if the drain 82 thereof becomes
substantially more negative than the source 80. Alternative body
diode configurations are described below. Further details of the
exemplary circuitry of FIG. 15 are described below with reference
to FIG. 8 of the Appendix and the associated text of the
Appendix.
[0119] Referring to FIG. 16, power semiconductor switching device
50 is depicted coupled with an active diode connected p-channel
MOSFET 202. The drain 206 and gate 208 of the p-channel MOSFET 202
are connected as illustrated to provide a body diode circuit
turning on when the drain 206 becomes more negative than the source
204 by an amount greater than the threshold voltage V.sub.t of
transistor 202. As opposed to a conventional prior art vertical
geometry power MOSFET of FIG. 1 and 2 in which the diffusion stored
charge, Q.sub.d in the body diode 7 can be very large, and a
serious limitation to switching speed and efficiency, the stored
charge in exemplary devices 50 according to aspects of this
invention (e.g., FIG. 16) is very small, principally that stored in
the gate capacitance of the switching device 50.
[0120] If the gate 83 of power semiconductor switching device 50 is
constrained to go no more negative than the source 80, it
inherently acts as a body diode and transistor 202 may be omitted
if desired. As a consequence, with the gate driver amplifier
configuration of FIG. 15, in which the voltage at the gate terminal
83 of the switching device 50 is constrained to go no more negative
than that at its source 80, this n-channel MOSFET active body diode
is obtained automatically in the switching device 50. For this
exemplary device 50, the stored charge that is removed in switching
from a 200 Ampere active body diode current (typically at
V.sub.ds=-0.75 Volts) to V.sub.ds=0 Volts is less than about 15
nanoCouloumbs, which is far less than for conventional prior art
vertical geometry power MOSFET devices (FIGS. 1 and 2).
Alternatively, device 202 is implemented as an n-channel MOSFET in
another embodiment with the gate 208 thereof connected to the
source 204 thereof to also serve as an active body diode circuit.
Body diode circuit implementations are coupled with the source 80
and drain 82 of the power semiconductor switching device 50 in the
depicted embodiment to conduct free wheeling currents which may be
present within power converter 10 during switching operations or
present during operations in other applications. Further details of
FIG. 16 are discussed in the Appendix with reference to FIG. 2B and
the associated text thereof. Other circuit configurations to
conduct free wheeling currents are possible.
[0121] Aspects of the present invention provide a plurality of
planar, horizontally configured MOSFET devices 174 configured to
form a power semiconductor switching device 50 having contacts
including a high-current source terminal contact 80 and
high-current drain terminal contact 82 on a common surface 57 of a
semiconductor die 52. Additional aspects enable electrical
connectivity of terminal contacts 80, 82 to devices provided upon
the common surface 57 of semiconductor die 52 using convenient
package configurations. Other aspects of the invention are
contemplated and provided, some of which are described above and in
the attached Appendix, the contents of which are incorporated
herein by reference.
[0122] For comparison purposes, power semiconductor switching
devices 50 configured according to exemplary aspects of the present
invention including an exemplary 0.24 micron feature size (0.19
micron L.sub.eff) CMOS process are discussed below with respect to
a conventional vertical power MOSFET (FIG. 1) having equivalent ON
resistance Ron values. Power semiconductor switching devices 50 of
some aspects of the invention have twenty four times smaller die
area than conventional vertical arrangements, approximately thirty
times less gate capacitance, 478 times lower gate drive power at a
given frequency (P.sub.gate/f.sub.clock) and the exemplary
semiconductor die configurations described herein may be fabricated
by a standard CMOS integrated circuit foundry if desired as opposed
to processes to form vertical conventional configurations.
[0123] FIG. 17-FIG. 19 depict respective graphical representations
240, 250, 260 of electrical performance characteristics of an
exemplary power semiconductor switching device 50 which embodies
aspects of the present invention. The graphed power device 50
comprises a 200 Ampere power device 50 having 514,000 n-channel
MOSFETs 174 providing W=12,850,000 microns and L.sub.eff=0.19
microns.
[0124] FIG. 17 depicts drain currents of the exemplary power device
50 for values of V.sub.GS=0.5-2.5 Volts in 0.5 Volt increments over
its nominal 0 to 2.5 Volt drain voltage range and over an I.sub.d=0
to 8000 Ampere range high enough to include its I.sub.dss=7700
Ampere saturated drain current at V.sub.gs=+2.5 Volt value. Note
that sustained operation at high values of drain currents (e.g.,
above 200 Amperes) may not be possible because of metal migration
reliability issues, and operation at combinations of high drain
voltages and drain currents should be kept of short duration
because of thermal power dissipation and energy absorption
limitations.
[0125] FIG. 18 depicts drain currents of the exemplary power device
50 for values of V.sub.GS=0.45-0.7 Volts in 0.05 Volt increments
over its nominal 0 to 2.5 Volt drain voltage range and over the
I.sub.d=0 to 200 Ampere range within which sustained operation is
specified (subject to power dissipation and energy absorption
limitations).
[0126] FIG. 19 depicts drain currents for a V.sub.GS value of 2.5
Volts over the I.sub.d=0 to 200 Ampere range within which sustained
operation is specified and over the 0 to 0.05 Volt drain voltage
range within which sustained operation is normally seen for a
device of this size when the device is ON (Vgs=+2.5 Volts) in the
exemplary power conversion applications of the type illustrated in
FIGS. 3 and 4.
[0127] Reference 262 of FIG. 19 depicts, at any given drain current
within transistors 174 of an exemplary power device 50, the voltage
drop, V.sub.ds, measured from the transistor drain contact 163 to
source contact 161 in FIG. 13 assuming approximately equal sharing
of total current I.sub.d between all transistors 174. The slope of
this line is the R.sub.on of the transistor devices 174 themselves,
exclusive of on-chip or package metal resistance.
[0128] Reference 264 depicts, at any given drain current, the
voltage drop, V.sub.ds, measured from the chip drain pads 118 to
source pads 116 in FIG. 8, assuming approximately equal sharing of
total current I.sub.d between all source 116 and drain 117 pads.
The slope of this line is the R.sub.on of the transistor devices
174 themselves plus that of the on-chip metallization 100 for the
case of a 4 mm.times.4 mm die having a total of 256 solder balls,
exclusive of package metal resistance.
[0129] Reference 266 depicts at any given drain current, the
voltage drop, V.sub.ds, measured from the package drain contact 82
to package source contact 80 in FIGS. 5 and 6, assuming uniform
distribution of current across the width of these package contacts
80, 82, for the case of a 4 mm.times.4 mm die having a total of 256
solder balls and the package metallization thicknesses discussed in
conjunction with FIGS. 5 and 6, and in Appendix A. The slope of
this line is the total R.sub.on of the packaged transistor
inclusive of the transistor devices 174 themselves plus that of the
on-chip metallization 100 and the package metal resistance.
[0130] In compliance with the statute, the invention has been
described in language more or less specific as to structural and
methodical features. It is to be understood, however, that the
invention is not limited to the specific features shown and
described, since the means herein disclosed comprise preferred
forms of putting the invention into effect. The invention is,
therefore, claimed in any of its forms or modifications within the
proper scope of the appended claims appropriately interpreted in
accordance with the doctrine of equivalents.
APPENDIX A
SUMMARY OF THE INVENTION
[0131] The use of diodes for output current rectification in low
output voltage switching power converters or power supplies is
precluded by the severe loss in efficiency due to the forward
voltage drop of diodes becoming comparable to the desired output
voltage, V.sub.out,of the converter. The use of MOSFETs or other
similar high-current switching devices in a synchronous rectifier
approach can overcome this efficiency problem if the "on"
resistance, R.sub.on, of the current switching MOSFETs is
sufficiently low to make the I.sub.outR.sub.on voltage drop of
across the switches much lower than V.sub.out. This can be achieved
with conventional MOS power FETs by making the FET periphery (total
channel width, W) very large, either by making one very large
device, or by connecting many smaller devices in parallel. In
either approach, to achieve very low R.sub.on values, the device
cost becomes quite large due to the large size of a single die, or
the large number of paralleled devices. Also, the total gate
capacitance of all of these devices becomes very large, which
limits the attainable switching speed and causes significant losses
in converter efficiency due to the large ac gate drive power
requirements.
[0132] Aspects of the present invention take advantage of the fact
that in a switching converter application, the breakdown voltage,
V.sub.b, required for the synchronous rectifier switching devices
is, for an output voltage of V.sub.out, only about
V.sub.b=2V.sub.out. For example, for an output voltage of
V.sub.out=1.2V, a breakdown voltage of about
V.sub.b=2V.sub.out=2.4V is adequate for the synchronous rectifier
MOSFET switches. In aspects of this invention, this low breakdown
voltage requirement is exploited to enable implementation of the
synchronous rectifier switching MOSFETs with an inexpensive
commercial deep-submicron CMOS integrated circuit foundry process
(instead of the special higher voltage process normally used for
fabricating power MOSFET devices). For a given R.sub.on, compared
to standard MOS power FET processes, the use of a very short
channel length (e.g., L.sub.eff=0.19 .mu.m)/very small feature size
(e.g., 2.lambda.=0.24 .mu.m) commercial IC process for fabricating
the current switches dramatically reduces the size and cost of the
devices, and also greatly reduces the gate capacitance and
increases switching speeds for higher efficiency and smaller
converter size.
[0133] The dramatic size/cost reduction comes through the combined
effect of a reduction in the MOSFET periphery, W, required to
achieve a given value of R.sub.on (because of the very small
L.sub.eff), with the fact that the very small feature size allows
packing a given MOSFET periphery into a much smaller chip area. For
example, at the silicon device level, an extremely low "on"
resistance of less than R.sub.on=75 .mu..OMEGA. can be achieved for
a 200 amp switching MOSFET with a total L.sub.eff=0.19 .mu.m FET
periphery of less than W=13.5 meters, and with a 2.lambda.=0.24
.mu.m feature size, this W=13.5 meter MOSFET can be packed into a
chip only 4 mm.times.4 mm in dimensions. At an estimated foundry
CMOS foundry wafer cost of $1500 per 8" (200 mm) wafer, the
unpackaged die cost for such a 200 amp switching MOSFET chip should
be less than $1.00. This deep-submicron approach gives not only a
much smaller solution to achieving an extremely low R.sub.on 200
amp current switching MOSFET, but promises to be 10.times. to
100.times. cheaper than implementations using standard power MOSFET
processes. In addition, because of the reduced FET periphery and
very short (L.sub.g=0.24 .mu.m drawn) gate length, the gate
capacitance is far lower, and the switching speed much higher than
in the power MOSFET implementation, and the ac gate drive power at
a given switching frequency can be reduced by nearly a factor of
500 over conventional vertical geometry power MOSFETs.
[0134] Also important to some aspects of this invention is the
solution of the A problem of getting such large (e.g., 200 amp)
currents into and out of the small (e.g., 4 mm.times.4 mm) FET chip
without encountering metal migration reliability problems or unduly
compromising R.sub.on from the on-chip metal resistance or package
resistance. Conventional power MOSFETs take the source and drain
currents out of opposite faces of the die (e.g., all of the area
devoted to source contacts, plus the gate leads, are on the top
surface of the die, while the drain contact is the entire back
surface of the die. Having the high-current source and drain
contacts on opposing surfaces of the die makes the die
packaging/interconnect quite easy. A disadvantage of using a CMOS
integrated circuit process instead of a power MOSFET process is the
unavailability of this backside drain contact configuration. The
MOSFET geometries in commercial CMOS IC processes are purely
lateral; that is a source-gate-drain electrode structure on the top
surface of the semiconductor die. Hence, in order to implement
aspects of this invention and use a commercial deep-submicron CMOS
process to make the low R.sub.on current switching MOSFETs, we have
to be able to get both the high current source and drain contacts
(plus one or more gate leads) off of the same (top) surface of the
die. The key to the solution of this high-current chip
packaging/interconnect interface problem is to cover at least
substantially the entire area of the die with a fine pitch area
array of solder bump contacts which flip-chip mate with a matching
array on a very low resistance multi-layer metal package. It is
desirable that the pitch of these solder bump contacts be as low as
possible to minimize the metal resistance. While a standard
commercial practice area array bump pitch of 250 .mu.m is adequate
to give a total (i.e., silicon FET+on-chip metal+solder
bump+package metal) resistance of the order of 285.mu..OMEGA., with
a smaller, 100 .mu.m bump pitch (which is also commercially
available), the total R.sub.on can be reduced to less than
180.mu..OMEGA. for the MOSFET chip example cited above. This very
low resistance is a result of the sharing of the source and drain
current by a large number of solder bump contacts (e.g., 126 bumps
each for the 250 .mu.m pitch and about 800 bumps each for the 100
.mu.m pitch, which accounts for its better R.sub.on
performance).
[0135] In summary and in accordance with exemplary aspects of the
invention, the use of a commercial deep-submicron CMOS integrated
circuit process in conjunction with an area array solder bump
flip-chip packaging approach allows the achievement of high-current
(e.g. 200 amp), very low on resistance low voltage MOSFETs having
much lower cost and much lower gate capacitance, not to mention
much smaller size, than devices implemented using conventional
power MOSFET processes.
BACKGROUND OF THE INVENTION
[0136] A key focus of modern digital electronics is to achieve
extremely high computational densities; that is, to cram as much
computational throughput as possible into a limited space. The
relationship between something as seemingly tenuous or intangible
as computational power and a simple concrete physical quantity like
dc power input or waste heat generated may not be obvious. In fact,
computational throughput ("power") is, as a matter of physics, tied
to the conversion of dc input supply power to heat. Moreover, in
the most efficient forms of logic circuitry, the level power
consumption/heat generation is essentially proportional to the
computational throughput, albeit the exact value of the constant of
proportionality is dependent on details of the integrated circuit
(IC) technology used to implement the computer. Specifically, it
can be shown that for CMOS logic (in which the dynamic power
dissipation dominates), a chip with N.sub.g gates, each driving an
average load capacitance, C.sub.gl, operating at a supply voltage,
V.sub.dd, and at a clock frequency, F.sub.c, with a fraction,
f.sub.d, of those gate switching on each clock cycle, the chip
power, P.sub.d, will be given by
Pd=1/2f.sub.dN.sub.gC.sub.glV.sub.dd.sup.2 F.sub.c Eq. 1
[0137] Since a reasonable measure of computational activity is the
number of such logic gates that have switched, the measure of
computational throughput (rate) is simply the number of such logic
gates switching per unit time, f.sub.dN.sub.gF.sub.c, or
Computing Rate
.varies.f.sub.dN.sub.gF.sub.c=Pd/[1/2C.sub.glV.sub.dd.sup.2-
].varies.Pd Eq. 2
[0138] which shows the (technology dependent) proportionality
between computational throughput and the power consumed
(dissipated). What this means is that the quest for high
computational density forces designers to cope with both high power
supply densities, as well as the ability to handle high reject heat
densities.
[0139] Advances in integrated circuit technologies, principally
through the reduction in lithographic feature sizes and FET gate
lengths into the deep submicron region, have made it possible to
put millions of logic gates on a single monolithic IC chip, as well
as to increase clock rates, F.sub.c, into the 500 MHz to 1 GHz
region. This means, from Eq. 2, that since both n.sub.g and F.sub.c
have been increased, remarkably high computational throughputs can
be achieved even on a single chip microprocessor. While feature
size reductions also serve to reduce CMOS FET gate capacitances
(C.sub.gs and C.sub.gd), much of the gate loading capacitance,
C.sub.gl in Eqs. 1 and 2, is due to interconnect wire capacitance,
which is principally a function of wire length, and hence does not
scale strongly with feature size. This means that if the logic
power supply voltage were held constant, the chip dissipations
would become astronomical (e.g., many kilowatts) for these very
dense, high throughput chips. This problem has been recognized by
the semiconductor industry, and as a consequence the power supply
voltages have been gradually reduced over the last decade from
their original V.sub.dd=5.0 V level down to V.sub.dd=1.2 V for the
latest 2.lambda.=0.24 .mu.m feature size, L.sub.eff=0.19 .mu.m
gatelength CMOS logic.
[0140] While this reduction on Vdd has helped keep chip power
dissipations to a manageable level (typically under 100 W to 200 W
per chip), which helps the thermal management problem, in fact it
has in some ways made the power supply/distribution problem even
more difficult. The problem is that while the chip power
requirements have been maintained at tolerable levels, the power
supply current requirements have increased sharply, while at the
same time the tolerance for IR voltage drops in the power
distribution conductors has become more severe. For example,
consider a logic power (in one or more chips) requirement of
P.sub.d=240 W. In the "old days" of V.sub.dd=5 V logic, from
P.sub.d=V.sub.ddI.sub.s, the required supply current would be
I.sub.s=240/5=48 amps, and for a 10% dc noise margin allowable
V.sub.ndc=I.sub.sR.sub.s=0.5 V drop, the allowable wiring
resistance between the power supply and the logic is
R.sub.s=0.5/48=10.4 m.OMEGA.. With the reduced V.sub.dd=1.2 V
logic, the required supply current would be I.sub.s=240/1.2=200
amps, and for a 10% dc noise margin allowable
V.sub.ndc=I.sub.sR.sub.s=120 mV drop, the allowable power
distribution resistance from supply to logic is only
R.sub.s=0.12/200=600.mu..OMEGA.. It is probably obvious that it is
very difficult to achieve such low power distribution resistances.
While it is correct that the use of an on-board regulator near the
logic chips could make it much easier to meet the dc noise margin
requirements of the logic, note that at I.sub.s=200 amps, even a
R.sub.s=1.0 m.OMEGA. wiring resistance (including the regulator
resistance) between the power supply and the logic chips will
reduce the power efficiency by over 14%, so sub-millivolt
resistances are indeed required. Conventional circuit board
connectors, etc., are not capable of handling 200 amp currents with
sub-millivolt resistances, so another approach is needed.
[0141] The most attractive approach to solving this problem is to
distribute the power to the boards at a high voltage (e.g.,
V.sub.in=48 volts), using very efficient on-board power converters
to convert this to the V.sub.dd=1.2 volt logic supply level. In
order to insure that the on-board power conversion does not
unacceptably compromise the overall density of the electronics, it
is desirable to make the on-board power converter as small as
possible. In a switching converter, the stored energy requirements
in the magnetics (e.g., transformers and inductors) and in bypass
or filter capacitors can be reduced by increasing the switching
frequency, f.sub.s, of the converter. This is difficult, however,
because of the limited switching speeds of most semiconductor
devices capable of handling large currents. Addressing this
deficiency is one focus of aspects of this invention; that is to
provide low cost MOSFET switches capable of handling high currents
at high switching speeds.
[0142] While the use of on-board power conversion to efficiently
supply low voltage logic requirements in high-performance systems
is extremely attractive, there is an inherent difficulty with
achieving high efficiency, very small low voltage power converters.
The ordinary method of building a switching converter involves a
transformer to convert down to low voltage ac, with diodes to
rectify this to dc. If the forward voltage drop of these rectifier
diodes is V.sub.f and the output voltage is V.sub.out, then the
voltage loss in the diodes alone will limit the attainable
rectifier efficiency, .eta..sub.r, to a value less than
.eta..sub.r.ltoreq.V.sub.out/(V.sub.f+V.sub.out) Eq. 3
[0143] Typical Schottky diode forward voltage drops are of the
order of V.sub.f=0.55 V or more. In the "old days" of 5 volt
supplies, this diode efficiency limit, about .eta..sub.r=90%, was
quite acceptable. With the new low supply voltages, this is no
longer the case, as Eq. 3, with V.sub.f=0.55 V, gives a limiting
efficiency of about .eta..sub.r=69% at V.sub.out=1.2 volts, or
about .eta..sub.r=62% at V.sub.out=9 volts. These are unacceptably
low values of conversion efficiency.
[0144] The solution to this problem is to employ MOSFET current
switches in place of the diodes (this approach is called
synchronous rectification). If the MOSFET "on" resistance,
R.sub.on, is sufficiently low, this rectification efficiency loss
may be kept very low. Unfortunately, using commercial devices
fabricated with the standard vertical "power MOSFET" device
structure, to get a low enough R.sub.on value, it is often
necessary to parallel large numbers of power MOSFET devices in each
of the two current switching positions in the circuit. Also, the
enormous gate capacitance of all of these power MOSFETs makes gate
drive very difficult and causes significant efficiency loss due to
1/2 C.sub.gs.DELTA.V.sub.g.sup.2 F.sub.c gate drive power. This,
plus the limited switching speeds of the standard power MOSFET
devices, effectively limits the usable switching frequency of the
converters. Since the size of the magnetics (and capacitors) goes
inversely with the frequency, this makes it very difficult to
miniaturize the converters for on-board use (of course, having
large numbers of power MOSFETs gives a size problem itself).
DESCRIPTION OF EXEMPLARY ASPECTS OF THE INVENTION
[0145] Aspects of the invention achieve extremely low R.sub.on
values and extremely low gate capacitances/high switching speeds in
a MOSFET current switching device, including a packaging approach
enabling it to be used in circuits. Some concepts of the invention
use a completely different device structure from the vertical
geometry of power MOSFETs; in fact, to make current switching
MOSFETs in a horizontal (planar) geometry using the same 0.25 .mu.m
to 0.18 .mu.m CMOS IC processes that are used to fabricate the
high-speed logic chips themselves. While the transistors in these
deep-submicron short-channel CMOS processes have only about 2.5-3
volt breakdown voltages, for a synchronous rectifier having
V.sub.out=1.2 volt, a V.sub.br=2.4 volt breakdown voltage is
adequate for the transistors. With this approach, in a chip only 4
mm.times.4 mm in size, it is possible to switch currents of 200
amperes and to achieve a value of R.sub.on superior to that of the
parallel combination of dozens of power MOSFETs, and at a very
small fraction of the gate capacitance. This makes it possible to
reduce the ac gate drive power required to switch to a given Ron
value at a given switching frequency by nearly a factor of 500 over
conventional vertical power MOSFETs.
[0146] Other aspects of the invention provide a packaging approach
capable of getting 200 amperes into and out of such a small chip
without substantially degrading the R.sub.on value. Whereas the
vertical geometry of a conventional power MOSFET has the source
lead on the top surface of the die and the drain lead on the
bottom, in the horizontal or planar MOSFET structure, the source
and drain leads (as well as the gate connection) are on the top
surface of the device. On-chip metallization layers on
semiconductor die generally have substantial sheet resistances
(e.g., 40-80 m.OMEGA./square) and limited current carrying
capacity. This makes combining the distributed MOSFET currents from
the myriad of elemental MOSFET devices spread over the area of the
chip into single source and drain MOSFET chip contacts on the top
surface of the die without unacceptably degrading both R.sub.on and
the current switching capacity of the chip. In aspects of this
invention, a solution to this packaging/interconnection problem for
current switching MOSFETs fabricated in the horizontal or planar
structure is to utilize an area array of a large number (many
hundreds to thousands) of chip source and drain contact pads (plus
gate pads) covering the top surface of the die. With the very large
source and drain currents divided into a large number of parallel
paths, the aggregate current carrying capacity can be very large
and the metal contribution to the chip on resistance can be small.
The area pad array may be flip-chip mated to a mirror-image area
pad array in a very low resistance package. A multi-layer package
approach utilizing thick copper interconnect layers has been
designed to meet the requirements of contacting such an area array
planar current-switching MOSFET chip with capability for handling
200 ampere currents with package contribution to total on
resistance of less than 100.mu..OMEGA..
[0147] The achievement of an inexpensive, ultra-low on resistance
planar current switching MOSFET is anticipated to have a wide range
of applications in addition to efficient low-voltage on-board power
conversion. Very low R.sub.on MOSFETs could have application in a
wide variety of low-voltage high-current applications, ranging from
linear applications such as high-current regulators and drivers to
switching applications such as efficient high-density power
supplies and converters, including solar cell power conversion, and
switch-mode drivers for motors and actuators.
[0148] Comparison of Device Structure of Aspects of the Invention
to Conventional Power MOSFET
[0149] The dramatic difference between the horizontal or planar
geometry current switching MOSFET structure of aspects of this
invention and the vertical geometry of a conventional power MOSFET
is illustrated in FIG. 1a. FIG. 1a shows the device structure of a
conventional power MOSFET. In an N-channel device of this
structure, electrons flow from the n.sup.+ diffused source regions
horizontally across the p body region (which is electrically
shorted to the source), through a surface channel. The magnitude of
this electron current is under control of the insulated gate. Past
the p body region, the path of the electrons turns downward,
vertically crossing the n.sup.- epitaxial drift region and going
into the n+ substrate drain region, the contact for which is on the
bottom side of the power MOSFET die. From a MOSFET current control
standpoint, the FET channel length, L in FIG. 1a, is determined by
the difference in the lateral extent of the p body diffusion and
the n.sup.+ source diffusion, with values of the order of L=1 .mu.m
typical for power MOSFETs. It is important to note that the actual
on resistance, R.sub.on, of the standard power MOSFET device is
substantially higher than the channel resistance (specifically, the
channel resistance plus twice the source resistance, as for a
symmetrical planar MOSFET), due to the electron current path
passing vertically through the n.sup.- drift layer and the n.sup.+
substrate thickness to reach the drain contact.
[0150] Comparison of Die Area Required for Deep-Submicron Planar
Current Switching MOSFET to Achieve Given Value of R.sub.on to
Conventional Power MOSFET
[0151] The principal advantage of this conventional power MOSFET
device structure, in addition to its ability to handle
comparatively large drain voltages, is its convenient contact
arrangement. The high-current drain contact covers the entire
bottom of the die, while the high-current source contact covers
most of the top of the die, sharing that surface with a smaller
gate contact. This convenient die contact arrangement makes it easy
to package the chip in a simple high current package. The principal
disadvantage of the conventional power MOSFET structure is that
even though the structure of source islands with channel borders
shown in FIG. 1a is packed into a dense 2-dimensional array
covering the die surface, even with an aggressive 10.sup.6
islands/cm.sup.2 (10,000/mm.sup.2) array packing density, the
potential amount of MOSFET periphery (width, W, of the surface
channel per unit area, assuming W=20 .mu.m/island), is only about
W/A.sub.chip=200 mm/mm.sup.2 (reduced by .about.10% for gate
contact area). As a comparison, the planar horizontal geometry
current switching MOSFET structure according to aspects of this
invention gives W/A.sub.chip=926 mm/mm.sup.2 in a nominal
2.lambda.=0.24 .mu.m feature size CMOS process (or about
W/A.sub.chip=850 mm/mm.sup.2 including gate routing area), 4.63
times higher. Since for a given channel sheet conductance, the
channel resistance is proportional to L/W, the other critical
dimension for the power MOSFET is the channel length, L. As noted
in FIG. 1a, L=1.0 .mu.m is typical for the channel length of a
conventional power MOSFET. In comparison, the effective channel
length in a 2.lambda.=0.24 .mu.m feature size CMOS process is
typically L.sub.eff=0.19 .mu.m or less. Hence for the same MOSFET
periphery, the deep-submicron short-channel planar MOSFET devices
used in aspects of this invention offer nominally 5.26 times lower
channel resistance than a conventional power MOSFET. When this is
combined with the fact that the periphery available in a given chip
area is also 4.63 times higher, for a given chip size, the
2.lambda.=0.24 .mu.m feature size CMOS process planar current
switching MOSFETs of aspects of this invention will have a channel
resistance over 24 times less than that of a conventional power
MOSFET.
[0152] The structure of the horizontal or planar geometry short
channel current switching MOSFET is illustrated in Figure ib. The
basic device structure consists of a series of parallel n.sup.+
stripes with poly silicide contacts, which represent the
alternating source and drain electrodes, separated by poly silicide
gate electrodes insulated from the silicon surface by a very thin
(e.g., t.sub.ox=5.8 nanometers) insulating layer which allows
channel conduction to be controlled fully with very small gate
voltage changes (.DELTA.V.sub.gs.about.2.5V). The electron path is
essentially horizontal through the surface channel from the n.sup.+
source to the n.sup.+ drain electrodes. Typical dimensions in a
2.lambda.=0.24 .mu.m feature size CMOS process are a nominal gate
length of 2.lambda.=0.24 .mu.m (L.sub.eff=0.19 mm effective channel
length), with source and drain contact widths of 7.lambda.=0.84
.mu.m, for a total MOSFET pitch of 9.lambda.=1.08 .mu.m (the
reciprocal of which gives the W/A.sub.chip32 926 mm/mm.sup.2
periphery to area ratio cited above). In fact, it is possible to
fabricate substantially higher density MOSFETs in a such a
.lambda.=0.12 .mu.m CMOS process, with as low as
.about.6.lambda.=0.72 .mu.m pitch, but in aspects of this invention
it is desired to keep the total device resistance, including
contacts and metal interconnects, as small as possible. Because the
sheet resistance of the poly silicide source and drain contacts is
relatively high (.about.4 .OMEGA./square), it is important to
minimize the distance the current passes laterally through the poly
silicide contact before it reaches metal. As shown in FIG. 1b, a
7.lambda.=0.84 .mu.m source/drain (S/D) stripe width allows for a
continuous chain of poly silicide to M.sub.1 (first level
interconnect metal) via contacts to be placed down the center of
these S/D stripes, making the poly silicide degradation of R.sub.on
negligible.
[0153] Comparison of Gate Drive Power Required for Planar Current
Switching MOSFET to Achieve Given Value of R.sub.on to that of
Conventional Power MOSFET
[0154] To reiterate, the enormous (24.times. nominal) advantage in
required silicon die area to achieve a given on resistance of the
planar or horizontal geometry current switching MOSFET structure of
aspects of this invention as compared to a conventional vertical
power MOSFET structure comes about through the combination of the
4.63.times. greater FET periphery, W, that can be packed into a
given area (due to the smaller feature size), and the 5.26.times.
lower channel resistance for a given periphery (due to the shorter
L.sub.eff). This greatly reduced die size offers both a greatly
reduced die cost and the opportunity for miniaturizing power
conversion products. An additional benefit of this approach is the
enormous reduction in gate drive power it affords. Ignoring
parasitic contributions to gate capacitance, the charge, Q.sub.g,
furnished to the gate electrode will nominally equal the charge
induced in the surface channel of the MOSFET. For an n-channel
MOSFET with channel length, L, having a channel electron mobility
of .mu..sub.e, the channel resistance, R.sub.ch, at low drain
voltages will be given as
R.sub.ch=L.sup.2/(.mu..sub.eQ.sub.g) (for very low V.sub.ds) Eq.
4
[0155] Comparing the L=1.0 .mu.m channel length of conventional
vertical power MOSFETs with the L.sub.eff=0.19 .mu.m channel length
of the planar devices of aspects of this invention means that, for
the same ate, the gate charge to reach a given R.sub.ch value will
be 27.7 times smaller for the new planar current switching MOSFETs.
Further, with a gate voltage swing, .DELTA.V.sub.gs, the ac gate
drive power, P.sub.gd, required to furnish this gate charge (i.e.,
to charge and discharge the gate capacitance, C.sub.gs+C.sub.gd) at
a switching frequency, F.sub.c, will be given by
P.sub.gd=1/2(C.sub.gs+C.sub.gd).DELTA.V.sub.gs.sup.2F.sub.c=1/2Q.sub.g.DEL-
TA.V.sub.gsF.sub.c Eq. 5
[0156] The magnitude of the gate voltage swing, .DELTA.V.sub.gs,
required to transfer the gate charge,
Q.sub.g,=(C.sub.gs+C.sub.gd).DELTA.V.sub.gs is determined by the
gate oxide thickness, t.sub.ox. The gate capacitance of a MOSFET
with channel length, L, and periphery, W, is given by
C.sub.gs+C.sub.gd=.epsilon..sub.o.epsilon..sub.rL W/t.sub.ox Eq.
6
[0157] where .epsilon..sub.0=8.85.times.10.sup.-14 F/cm and
.epsilon..sub.r=3.9 for an SiO.sub.2 gate dielectric. From Eqs.
4-6, the channel resistance of the MOSFET will be given by
R.sub.ch=L.sup.2/(.mu..sub.eQ.sub.g)=L
t.sub.ox/(.mu..sub.e.epsilon..sub.0-
.epsilon..sub.rW.DELTA.V.sub.gs) Eq. 7
[0158] This means that the gate voltage above threshold,
.DELTA.V.sub.gs, required to reduce the channel resistance to a
given R.sub.ch value, as given by
.DELTA.V.sub.gs=L
t.sub.ox/(.mu..sub.e.epsilon..sub.0.epsilon..sub.rWR.sub- .ch) Eq.
8
[0159] is proportional to the oxide thickness, t.sub.ox.
Substituting Eq. 8 into Eq. 5 shows that the amount of ac gate
drive power, P.sub.gd, required to switch the MOSFET to a channel
resistance, R.sub.ch, will be given by
P.sub.gd=1/2F.sub.cL.sup.3t.sub.ox/(.mu..sub.e.sup.2R.sub.ch.sup.2.epsilon-
..sub.0.epsilon..sub.rW) Eq. 9
[0160] This is an important result which indicates that a given
switching frequency, F.sub.c, for MOSFETs with the same W/L ratio
and channel mobility, in addition to the factor of L.sup.2 that
comes from the required gate charge (Eq. 5), the gate drive power
expression has an additional factor of t.sub.ox in it. Hence, if we
compare the planar geometry current switching MOSFET of aspects of
this invention with L=L.sub.eff=0.19 .mu.m and t.sub.ox=5.8 nm with
a typical vertical geometry power MOSFET having L=1.0 .mu.m and
t.sub.ox=100 nm, we see that in addition to the L.sup.2 advantage
of 27.7.times., we have a t.sub.ox advantage of 17.24.times., which
means that the ac gate drive power is reduced by an factor of
478.times. by going from a conventional vertical power MOSFET to
the deep-submicron planar MOSFET structure of aspects of this
invention. This amazing (nearly 500.times.) reduction in gate drive
power lets this new planar short channel current switching MOSFET
be operated at much higher switching rates without seriously
compromising power efficiency, which in turn allows for greatly
reducing the size of switching power converters, etc., because of
the reduced sizes of the capacitors and magnetic elements.
[0161] It should be noted that while Eq. 5 is exact, due to higher
vertical electric fields at the gate oxide interface, the channel
mobilities, .mu..sub.e3, in deep-submicron MOSFETs will be somewhat
smaller than those for power MOSFETs. On the other hand, while the
total MOSFET device "on" resistance, R.sub.on(f), of the planar
geometry MOSFETs of aspects of this invention (FIG. 1b) are given
from the channel resistance, R.sub.ch, and the source and drain
contact resistances, R.sub.s and R.sub.d, (normally
R.sub.s.congruent.R.sub.d due to symmetry) by
R.sub.on(f)=R.sub.ch+R.sub.s+R.sub.d.congruent.R.sub.ch+2R.sub.S(for
planar MOSFETs) Eq. 10
[0162] for the unsymmetrical vertical geometry power MOSFET
structure of FIG. 1a, there is a substantial additional drain
resistance component due to the n.sup.- drift layer in FIG. 1a that
makes R.sub.on(f) substantially greater than R.sub.ch+2R.sub.s.
Hence a substantially lower R.sub.ch value must be achieved in the
vertical power MOSFET in order to achieve the same total device
resistance, R.sub.on(f). Additionally, while Eq. 6 is a good
approximation for the gate capacitance in the planar MOSFET
structure, the large gate area over the n.sup.- drain region (see
FIG. 1a) makes the actual gate capacitance at low V.sub.ds for the
conventional vertical power MOSFET structure much larger than
indicated by Eq. 6. Hence the nearly 500.times. gate drive power
advantage of the planar geometry short channel MOSFETs of aspects
of this invention over conventional vertical power MOSFETs should
at least be reasonably accurate, and probably understates the
advantage.
[0163] Capability of Some Embodiments of Invention to Realize Very
High-Speed Ultra-Low Q.sub.d Body Diode
[0164] The capability of a CMOS integrated circuit fabrication
process to make, in addition to the n-channel MOSFETs discussed
above, isolated p-channel MOSFETs (as shown in FIG. 1b), can also
be exploited to advantage in designing high current switching
planar MOSFETs as well. FIG. 2a shows the equivalent circuit of a
conventional vertical power MOSFET. Note that, in parallel to the
n-channel MOSFET device, there is a robust p-n junction body diode
which is capable of passing high levels of current if the drain is
made more negative than the source by V.sub.fp-n=1.0V or so. While
in some switching circuit applications where the synchronous
rectifier switch timing is not precisely matched to the secondary
current timing, the "freewheeling" currents can be handled by the
p-n body diode. Unfortunately, the relatively high forward voltage
drop of the body diode will degrade the power efficiency of low
voltage converters substantially if the duration of the
freewheeling currents is a significant fraction of the total
switching cycle. In addition to the high p-n body diode forward
voltage drop, they also have serious charge storage problems (large
Q.sub.d values) that limit usable switching frequencies. In the
short-channel CMOS-implemented planar current switching MOSFET
device of aspects of this invention, it is possible to use the
p-channel device to implement a relatively low forward voltage drop
"diode" having virtually no charge storage (near-zero Q.sub.d). The
optional circuit for adding this "body diode" capability to an
n-channel device is illustrate in the equivalent circuit of FIG.
2b. This "body diode" function is emulated by an active
diode-connected p-channel MOSFET (i.e., the gate is connected to
the drain). When the drain electrode becomes more negative than the
source by a voltage equal to the gate threshold voltage, V.sub.tp,
the p-channel device will begin strong conduction. Since a typical
value for this gate threshold voltage is V.sub.tp32 -0.28V, the
V.sub.fpch forward conduction voltage drop can be much lower than
for a p-n junction body diode. When the drain terminal is more
positive than the source, as when the n-channel device is
conducting normally, the p-channel device conduction is cut off.
There is negligible charge storage associated with the transient
transition of this diode-connected p-channel MOSFET "body diode"
from strong "forward" conduction to "reverse bias" cutoff
(principally the small C.sub.gs+C.sub.gd of the p-channel MOSFET
plus some n-well to p-substrate capacitance).
[0165] As noted in FIG. 2b, this "body diode" function can equally
well be implemented with an n-channel MOSFET whose gate is
connected to its source. When the drain electrode becomes more
negative than the source by a voltage equal to the gate threshold
voltage, V.sub.tn, the n-channel device will begin strong
conduction. This is because when the drain of the symmetrical
n-channel MOSFET is taken more negative than the source, the source
behaves like the drain and visa versa, so the gate is now connected
to the (effective) drain, which is the usual configuration for
connecting a MOSFET as an active diode. Note that this also means
that if we constrain the gate voltage of the "OFF" large n-channel
current switching MOSFET to go no more negative than its source
[(V.sub.gs).sub.off=0], then this n-channel current switching
MOSFET will act, by itself, as a "body diode" with no additional
structures necessary.
[0166] Interconnect/Packaging Approach for Ultra-Low Resistance
Connection of Deep-Submicron Planar Current Switching MOSFET Die to
Circuits
[0167] While the electrical characteristics of the planar geometry
deep-submicron current switching MOSFET structure of aspects of
this invention represent an enormous improvement over conventional
power MOSFET devices in terms of much smaller die sizes, higher
switching speeds and remarkably lower gate drive requirements to
achieve a given value of on resistance, R.sub.on, it requires a
carefully designed interconnect and packaging approach to realize
the die-level performance in actual circuit application. The basic
approach to achieving this is illustrated in FIG. 3, which shows a
cross-sectional view of the planar (horizontal) geometry MOSFET
structure with its on-chip interconnects, solder ball array die to
package mating, and high-current low-resistance package. Note that
in order to maintain visibility of the on-chip interconnect
structure, the vertical dimensions of the thicker elements such as
silicon die thickness, die to package (and solder ball) thickness,
and thickness of package copper layers have been reduced in FIG.
3.
[0168] The basic interconnect/packaging approach is to make use of
an area array of a large number of source (S) and drain (D) pads,
in addition to a few gate (G) pads, covering the full area of the
planar geometry current switching MOSFET die. In a preferred
embodiment of this invention, illustrated in FIG. 3, solder balls
are formed on each of these pads for flip-chip mating to the
mirror-image pad array on the high-current package. As the density
of the pad array is increased, or pitch between the solder balls
decreased, the height of the solder balls is generally
proportionally decreased. Hence as the center-to-center pitch,
P.sub.sb, of the solder balls is reduced, the number of balls,
N.sub.sb, on a chip of area, A.sub.chip,
N.sub.sb=A.sub.chip/P.sub.sb.sup.2 Eq. 11
[0169] increases as 1/P.sub.sb.sup.2, while the resistance per
ball, R.sub.psb, increases as 1/P.sub.sb (since the ball height and
diameter go as P.sub.sb, the height to area ratio goes as
P.sub.sb/P.sub.sb.sup.2=1/P- .sub.sb), so the combined solder ball
resistance varies as
R.sub.sb=4R.sub.psb/N.sub.sb.varies.1/P.sub.sb Eq. 12
[0170] (The factor of 4 in Eq. 12 comes from assuming that if there
are N.sub.sb total solder balls on the die, approximately 50% will
be used for source contacts and the other 50% for drain contacts,
accounting for a factor of 2, while the R.sub.on will be degraded
as the sum of the source and drain contributions, accounting for
the other factor of 2.) What Eq. 12 illustrates is that the total
pad contact resistance can be reduced by making the solder ball
pitch as small as possible (the ball array density as large as
possible). The pitch effect is even stronger in the reduction of
the metal spreading resistance, which tends to fall nearly as
1/N.sub.sb or 1/P.sub.sb.sup.2. Hence the use of a high-density
array of as many solder bumps as possible covering the surface of
the die reduces both the solder bump and the metallization (on-chip
and package) contributions to the "on" resistance of the packaged
planar current switching MOSFET devices. While typical commercial
flip-chip solder bump array pitches of P.sub.sb=250 .mu.m are
adequate for useful performance, the total metallization plus ball
resistance can be several times the intrinsic MOSFET device
resistance (Eq. 10) (or somewhat less if a copper interconnect
metallization technology is used on the chip). By using the highest
commercially available solder bump array density, P.sub.sb=100
.mu.m, the total (chip plus package) metallization plus solder ball
resistance may be reduced by about a factor of three, to make it
comparable with the intrinsic MOSFET device resistance (e.g.
.about.75.mu..OMEGA. for a 4 mm.times.4 mm die with a typical
aluminum metallization technology).
[0171] On-Chip Interconnects
[0172] In the cross-section drawing of FIG. 3, there are 5 levels
of on-chip metallization illustrated, typical for a current
2.lambda.=0.24 .mu.m feature size commercial CMOS IC foundry
process. In the cross section drawing of FIG. 3, and in the
associated mask layer sequence layout drawings in FIGS. 4A-4B and
5A -5B, the use of the various chip metallization layers is shown.
The chip layout actually begins with the shallow trench isolation
of the p-well in which the n-channel MOSFETs are defined. While the
shallow trench isolation is not illustrated in FIGS. 4A-4B, it
would occupy the 3 .mu.m space between the 25 .mu.m high rows of
MOSFET channels, somewhat wider than the M.sub.1 and M.sub.2 gate
(G) busses. The upper left layout (FIG. 4A) shows the
7.lambda.=0.84 .mu.m wide by 25 .mu.m long poly silicide
source/drain (S/D) stripes, plus the row of 2.lambda.=0.24 .mu.m
square via cuts which connect them to the M.sub.1 stripes above
them. These M.sub.1 stripes, plus their vias up to M.sub.2 are
shown at the upper right (FIG. 4B). The structure up through
M.sub.1 is also illustrated in the cross sections of FIGS. 1b and
3. As seen at the lower left (FIG. 4C), M.sub.2 is dedicated mainly
to horizontal S and D busses, plus the smaller horizontal gate bus
(which is in parallel to its counterpart on M.sub.1). M.sub.3, as
seen at the lower right (FIG. 4D) (and reproduced at the upper left
FIG. 5A) is a dedicated S (source) plane, with a substantial number
of drain (D) feedthrus (and a much smaller number of gate (G)
feedthrus) penetrating it. Similarly, as seen at the upper right in
FIG. 5B, M.sub.4 is a dedicated D plane, with S and G feedthru
patches to reach up to the S and D contact "checkerboard" (and gate
contact "grid" in between) that makes up M.sub.5. The lower left
drawing in FIG. 5C shows a portion a source and a drain contact pad
and the vertical M.sub.5 gate "grid" stripe between them (these tie
all of the horizontal M.sub.1/M.sub.2 G busses together into a
single gate connection, to which are assigned four M.sub.5 gate
pads). The total chip layout at the lower right illustrates a 4
mm.times.4 mm chip carrying 256 solder ball pads on a P.sub.sb=250
.mu.m pitch. As illustrated in the cross-sectional drawing of FIG.
3, the actual solder ball base is a 125 .mu.m diameter circle,
centered on each of the square M.sub.5 contact pads, that is
defined by a hole in the oxide protection coating ("scratch
mask").
[0173] Package Interconnects
[0174] Because of the concentration of current, and requirement for
carrying the current laterally over substantial (many mm)
distances, the metallization thicknesses in the package must be
quite substantial (e.g., hundreds of microns of copper in the
thicker layers). FIG. 6A and FIG. 6B show respective
cross-sectional and plan views of an example of the layout of a
package with planar bottom source and drain contacts (e.g.,
suitable for surface mounting on a circuit board). In the example
illustrated, the base substrate is a 500 .mu.m (20 mil) thick
copper-invar-copper (to provide a coefficient of thermal expansion
[CTE] reasonably closely matched to that of the silicon die). While
initially this base substrate is continuous, during processing it
is sawn or etched through for electrical isolation, with the major
area chosen as the source (S) package contact, while a smaller end
region (the right end in FIGS. 6A and 6B) is the drain (D) package
contact. The lateral current flow from the drain solder balls from
the die to the drain contact end is through a thick copper layer
(e.g., 250 mm [20 mils] thick plated copper in FIG. 6B). The
lateral current from the source solder balls to the opposite,
source contact, end of the package is through the 20 mil thick
copper-invar-copper substrate, but the current reaches this
substrate by passing through the 10 mil thick Cu drain plane via an
array of isolated copper "studs" or "plugs" penetrating the
plane.
[0175] While it would be advantageous to make the center-to-center
pitch of these studs as fine as possible (e.g., the same pitch as
the solder bumps on the die would be ideal), due to the 10 mil
thickness of the copper layer through which they pass this is
impractically difficult. In typical circuit plating processes, the
aspect ratio (metal layer thickness divided by minimum horizontal
feature size) is typically kept to 1:1 or less. (Higher aspect
ratios can be fabricated using more exotic processes, but their
higher costs would make them economically unsuitable for most
applications at present.) In the package structure illustrated in
FIGS. 6A and 6B, the smallest feature to be defined in the lo mil
thick plated copper layer would be the narrow isolating gap
("moat") surrounding each source plug (via) penetrating the layer.
The plug itself is of reasonable dimension to carry S current
vertically with low resistance, while the outer rims of the
isolation moats do not get too close together or they will impede
the lateral flow of drain current through this 10 mil thick plated
copper drain plane. Hence, keeping the aspect ratio near 1:1, with
a 10 mil thick drain plane, the nearest center-to-center stud pitch
(diagonal in FIG. 6A) cannot be much less than 25-30 mils. It is
advantageous in the layout to make the horizontal stud (via plug)
pitch an even integer multiple of the die solder ball pitch, so for
the P.sub.sb=250 .mu.m (.about.10 mil) ball pitch, a 1.000 mm
(.about.40 mil) horizontal stud pitch (0.7071 mm [27.84 mils]
minimum [diagonal] pitch) is chosen in FIG. 6A.
[0176] While the drain plane lies directly under the drain solder
balls (unless the via stud isolating moats are too large, causing
some of the drain solder ball contacts to fall "off the edge" of
the drain plane metal onto the moat), this is not the case for the
source solder ball contacts, only a small fraction of which fall on
studs. The connection of the source solder balls to the studs (via
plugs) utilizes an additional layer or two of copper that is
capable of being more finely patterned than the drain plane (and
which hence are much thinner). This source plane metallization
(seen more easily in the cross-section of FIG. 3 than in that of
FIG. 6B) could have one or two layers of fineline copper about 10
.mu.m in thickness (about {fraction (1/25)}.sup.th of the thickness
of the 10 mil thick drain plane). Creating these higher density
(smaller feature size) multi-layer interconnects requires a
reasonably flat surface to start with, which means that the moats
are filled around the plugs with a suitable dielectric material and
planarize the dielectric and copper surface before the copper/BCB
(or other suitable multilayer dielectric material) interconnect
layer(s) are applied. While the moat diameter shown in FIG. 6A
would allow drain contacting and the source plane to be created in
one metal layer, in most cases it would be advantageous to have at
least two fineline copper interconnect layers, both to facilitate
getting the drain solder ball contacts down to the 10 mil copper
drain plane and to reduce the spreading resistance from the source
studs out to the majority of the source solder balls that do not
sit on studs, as well as to facilitate making contact to the gate
pads without significantly degrading this source plane
resistance.
[0177] Alternative Wafer-Level Additive Copper/Polymer Packaging
Approach
[0178] As described in the following section (Table 1), the
contribution of metal resistance, both on-chip and package, to
R.sub.on can be reduced by increasing the density of solder ball
contacts between the die and the package. In the packaging approach
illustrated in FIGS. 3 and 6A and 6B, the solder balls ("bumps")
are fabricated directly on the completed IC wafer using standard
commercial bumping processes (it is also possible to bump
individual die, but wafer bumping is cheaper). While 100 .mu.m
solder bump pitches in area arrays are commercially available, the
technology is not as well developed as the more common 250 .mu.m
pitch, and the reliability is problematical where significant
package to silicon wafer CTE differences may exist. Also, with a
100 .mu.m bump pitch, the need for one or two layers of "fineline"
additive interconnect metal (as illustrated in FIGS. 3 and 6A and
6B) to the high-current package structure is desired in some
aspects. This adds to package complexity and makes finding suitable
commercial sources for the package more difficult.
[0179] The alternative packaging approach illustrated in FIG. 7
achieves the very low R.sub.on of the 100 .mu.m solder ball
approach (or exceeds its performance) while reducing the demands on
both the solder bump mating technology and the package. The concept
is to fabricate the one or two (as illustrated in FIG. 7)
"fineline" metal layers (e.g., 5 .mu.m to 20 .mu.m of copper) with
polymer (e.g., BCB or polyimide) interlayer dielectric directly on
the completed IC wafer in a full-wafer process. Applying the
additive copper/polymer interconnects directly to the "completed"
IC wafer (i.e., the wafer has passed through all of the normal
fabrication steps of the CMOS IC foundry) will in general be less
expensive and more practical than adding them on the package side.
It is more practical because this copper/polymer technology is not
generally available at package vendors. It is less expensive
because of the fact that die area is smaller than that of the
package; hence a smaller area of material is subjected to the cost
of the copper/polymer processing when it is placed directly on the
die. (In the case of very low yields of "good" die on the IC
wafers, the cost of putting the "fineline" copper/polymer
interconnects directly on the IC wafers could exceed that of
applying them to the packages, but such low yields of good die
would not be expected for mature CMOS foundry processes.)
[0180] In addition to being more practical and lower cost, applying
the additive copper/polymer interconnects directly to the
"completed" IC wafer offers the advantage that the density of
interconnections between the normal CMOS on-chip metallization and
the much thicker "fineline" copper layers is not limited by any
solder bump or other joining technology, but only by the via
density and metal linewidths available in the first layer of
copper/polymer interconnects. The interconnect densities practical
between the IC chips and an additive copper/polymer are not only
much higher than solder bump densities, but there is no reliability
penalty associated with going to very high via densities on a chip
like there is in joining technologies such as solder bumps. In the
approach of FIG. 7, there is very little R.sub.on penalty
associated with using relatively coarse solder bump (or solder
stripe, etc.) pitches because the of the low sheet resistance of
the additive copper layers. Hence, the solder contacts may go
directly from the die (with the copper/polymer "fineline" layers on
it) to very thick metal package structures having very coarse
feature sizes (such as going to an interdigitated thick (e.g.,
>20 mils) copper-invar-copper package or some form of laminate
package).
[0181] Alternative Vertical Laminate Package Approach
[0182] The package geometries illustrated in FIGS. 3 and 6A and 6B
are horizontal, in that layers of metal conductors carrying the
high MOSFET source and drain currents lay parallel to the surface
of the MOSFET die. In this geometry, via structures carry current
between the different patterned layers of package metallization.
These vias (more specifically, the spreading resistance of getting
currents into and out of the inter-layer via areas) represent a
significant contribution to the overall package metal resistance,
as well as a source of concern for package reliability. Also, the
fact that the vertical thickness of the metal layers is limited by
practical fabrication considerations that limit the layer
thicknesses to be of the order of the horizontal feature sizes
provides a further limitation on package resistance for the
horizontal package geometry.
[0183] These limitations can be overcome and extremely low package
resistances achieved by going to a vertical laminate package
geometry, as illustrated for the case of the
monolithically-integrated gate drive amplifier (FIG. 8) in FIG. 9A
and 9B. In this structure, the high source and drain currents are
carried in vertical metal layers (that is, perpendicular to the
surface of the current switch MOSFET die). Their (horizontal)
thickness is limited by the feature size of the die contacts (e.g.,
solder bump pitch), but their vertical extent is unlimited, so that
extremely low package metal resistance levels can be achieved. Note
that ordinarily, the separation of the high current paths is
achieved in the vertical laminate package by sweeping, for example,
the layers contacting the source rows of solder balls in one
direction and those contacting the drain layers in the opposite
direction (e.g., in the illustration at the bottom (FIG. 9B), the
source layers might be swept in the direction into the page, and
the drain layers out of the page). Alternatively, rows may be
dedicated to other purposes such as low inductance V.sub.dd
connections and they can be taken out of the package bottom, in
addition to the opposing sides (or either the source or drain
connection could be taken out of the package bottom if desired).
Note that while the attachment of this vertical laminate package
directly to solder balls on die metallization pads is shown in FIG.
9B, this package configuration can also be very advantageously used
with a die with the additive "fineline" thick copper interconnects
added as illustrated in FIG. 7. When the additional copper/polymer
interconnect layers are added to the die, the pitch of the solder
contacts to the package may be greatly increased without
significantly degrading R.sub.on. This means that fewer layers of
laminate are needed for the package, making it easier to match the
laminate layer pitch to the solder contact pitch, and potentially
making practical the use of materials like copper-invar-copper
(CIC) for the metal layers in the laminate stack for improved CTE
match to the silicon die, which would offer the potential for
improved reliability. Note that while the use of CIC in the
laminate package stack would alleviate CTE mismatch in the plane of
the metal layers, there could still be CTE mismatch in the
perpendicular direction due to the CTE of the organic dielectric
layers in the stack. If, however, the organic dielectric
interlayers are pulled back from the solder interface to the die
(further than the pullback illustrated at the bottom FIG. 9B), then
this perpendicular CTE mismatch may be compensated for by a small
amount of bending in the laminate metal layers.
[0184] Calculated Performance for Packaged Planar Current Switching
MOSFET Devices
[0185] It is of value to calculate the intrinsic device, on-chip
metal, and package metal contributions to the on resistance,
R.sub.on, as well as other critical performance parameters that
would be expected for a planar current switching M.NOSFET device of
the type described in aspects of this invention. Since, as noted in
Eq. 12, the metal interconnect resistance decreases as the density
of the solder ball interconnects is increased, two cases will be
presented; one assuming 125 .mu.m balls on a 250 .mu.m pitch
(standard commercial area array solder ball practice), and the
other assuming 50 .mu.m balls on a 100 .mu.m pitch (highest area
array solder ball density that is commercially available at
present). A silicon die area of 4 mm.times.4 mm was assumed for the
planar current switching MOSFET device, so that the total number of
solder balls was 256 for the 250 .mu.m pitch or 1600 for the 100
.mu.m pitch case. Using design rules for a commercial
2.lambda.=0.24 .mu.m/L.sub.eff=0.19 .mu.m CMOS process, with a
layout of the type illustrated in FIGS. 3-5D, a total FET width of
13.5 meters is achievable in the 4 mm.times.4 mm chip size. This
process has a t.sub.ox=5.8 nm gate oxide thickness which gives an
N-channel resistance, R.sub.ch (Eq. 7), at V.sub.gs=+2.5V of
43.mu..OMEGA. for the L.sub.eff=0.19 mm/W=13.5 m MOSFET of only
43.mu..OMEGA.. When the R.sub.s=R.sub.d=14.1.mu..OMEGA. source and
drain resistances are added to R.sub.ch (as per Eq. 10), the
intrinsic device on resistance is calculated to be only
(R.sub.on).sub.dev=71.1 mW.
[0186] The CMOS process analyzed has a 5-layer aluminum on-chip
metallization system, with metal layers M.sub.1-M.sub.4 having a
resistance of 0.088.OMEGA./sq and M.sub.5 having 0.044.OMEGA./sq,
while the S/D silicide has 4.0.OMEGA./square. The via cut
resistances are 7.5.OMEGA./cut between the S/D silicide and
M.sub.1, and 5.OMEGA./cut between M.sub.1 and M.sub.2 or between
other metal layers. While a 7.5.OMEGA./cut S/D silicide to M.sub.1
metal layer resistance may sound serious, for a W=13.5 meter MOSFET
width, there are 13.5 million of these via cuts in parallel in both
the source and drain paths to share the current, so the addition to
R.sub.on is only 1.11.mu..OMEGA.. This parallelism applies to the
other metal resistance contributions, as a consequence of a chip
interconnect layout (FIGS. 4 and 5) that is well thought out to
minimize metal resistance. A resistance build-up analysis of the
various spreading resistances in the on-chip metal planes and the
inter-layer via cut resistances gives a calculated on-chip metal
resistance of 54.6.mu..OMEGA. in the source lead and
55.9.mu..OMEGA. in the drain lead (110.5.mu..OMEGA. total S+D) for
the 250 .mu.m solder bump pitch, which drops to 14.1.mu..OMEGA. in
the source lead and 15.5.mu..OMEGA. in the drain lead
(29.6.mu..OMEGA. total S+D) for the finer 100 .mu.m solder bump
pitch.
[0187] Similarly, the package metal resistance is also a function,
though slightly less dramatic, of the solder bump pitch in the
flip-chip area array interconnects between the planar current
switching MOSFET die and its package. A resistance build-up
analysis of the various spreading resistances in the package metal
planes and through-layer "plug" resistances gives a calculated
package metal resistance of 54.6.mu..OMEGA. in the source path and
55.9.mu..OMEGA. in the drain path (110.5.mu..OMEGA. total S+D) for
the 250 .mu.m solder bump pitch, which drops to 33.6.mu..OMEGA. in
the source path and 46.0.mu..OMEGA. in the drain path
(79.7.mu..OMEGA. total S+D) for the finer 100 .mu.m solder bump
pitch. Table 1 summarizes the contributions to the packaged device
Ron (total source+drain) of the MOSFET device, the on-chip
metallization, and the package metallization (including solder
balls) for the "standard" (250 .mu.m) and "high-density" (100
.mu.m) chip-to-package solder ball array pitches.
1TABLE 1 Calculated contributions to R.sub.on for planar L.sub.eff
= 0.19 .mu.m current switching MOSFET (C.sub.gs + C.sub.gd = 20 nF
for W = 13.5 m, 4 mm .times. 4 mm die). Solder Ball Pitch Cases
Standard High-Density P.sub.sb = 250 .mu.m P.sub.sb = 100 .mu.m
Silicon/Transistor 71.1 .mu..OMEGA. 71.1 .mu..OMEGA. (V.sub.gs =
+2.5 V) On-Chip Metallization: 110.5 .mu..OMEGA. 29.6 .mu..OMEGA.
Package Metal: 103.2 .mu..OMEGA. 79.7 .mu..OMEGA. Total Resistance
284.8 .mu..OMEGA. 180.4 .mu..OMEGA. (Si + Metal + Package):
[0188] Note in Table 1 that even with the high-density solder ball
pitch, the largest contribution to R.sub.on is the metallization,
not the MOSFET device itself, and that the largest contribution to
the R.sub.on resistance is from the package metallization. This is
a result of the fact that we have assumed a convenient planar,
surface-mountable package configuration (FIGS. 6A and 6B). It would
be possible to substantially reduce the package metal resistance by
reducing the horizontal distance through which the source/drain
current must pass between the die and the package S/D contacts. One
package configuration which would accomplish this would be a
vertical package geometry with the source lead on the bottom (as it
is under the chip in FIG. 6B), but with the drain lead on the top,
surrounding (and probably covering) the planar MOSFET die. This
could be accomplished, for example, by bringing the 10 .mu.m thick
plated copper drain plane up to the surface around the die, and
then using either a plated copper wall or preform to bring the
contact surface above the height of the die. It would be possible
to use a cover lid on this drain contact ring, or leave the back of
the die exposed (with suitable underfill passivation). This
configuration would reduce both the total package area and its
contribution to R.sub.on substantially, and could represent a
convenient form factor for some applications.
[0189] It should be noted that the flip-chip die underfill noted
above would probably be utilized in either the flat, surface-mount
version of the package shown in FIGS. 6A and 6B, or in a vertical
package geometry of the type just discussed. One reason for this is
that underfill is an effective means of enhancing solder ball
reliability in the face of differential thermal expansion between
package and silicon die. Another is that the constraint provided by
the underfill could be of substantial value in reducing lifetime
degradation due to electromigration in the solder bump metal. The
planar MOSFET die in this example should be able to switch currents
of 200 amperes, even though the die is only 4 mm.times.4 mm in
size. This pushes, however, the conventional operational current
density of solder ball contacts to die for high reliability. This
is not viewed as a serious concern, however, both because the
constraint of the underfill should minimize the problem, and
because of the way the solder balls are used in this application.
In usual flip-chip reliability analyses, it is assumed that the
failure of any single solder ball represents a functional failure
of the part. In this planar current switching MOSFET structure,
there is massive parallelism between the current-carrying (S/D)
solder balls. For example, for the high-density (P.sub.sb=100 .mu.m
ball pitch) case, there will be a total of 1600 solder balls, or
800 balls in parallel in the source path and 796 balls in parallel
in the drain path (assuming 4 balls for gate leads). The complete
failure of even 10% or 20% of the solder balls mating the die to
the package would make a barely-perceptible change in the R.sub.on
of the packaged device. (This is because the total S+D solder ball
resistance contribution to the 180.4.mu..OMEGA. R.sub.on value is
only 8.4.mu..OMEGA., and while failed balls would also locally
increase the metal plane spreading resistances, the effect of even
20% randomly-distributed ball failures is very small.) The
favorable statistics in this application where 10% to 20% ball
failures have no significant effect, as compared to the usual "one
ball failure is death" condition allows us to push to higher than
normal current densities in the solder balls. Of course, if solder
ball electromigration remains an issue in some applications, solder
ball metal compositions more resistant to electromigration effects
at the desired operating temperature can be utilized. Note that as
discussed in the previous section, the alternative packaging
approach of fabricating copper/polymer "fineline" interconnect
layer(s) directly on the completed CMOS wafers has the advantage of
achieving Ron performance at least as good as shown for 100 .mu.m
solder balls in Table 1, while using a relatively coarse feature
size joining technology which should offer reliability advantages
over fine-pitch solder balls.
[0190] It should be noted that whether the planar package
configuration of FIGS. 6A and 6B (R.sub.on=180.4.mu..OMEGA.) or the
vertical package geometry (R.sub.on=125 to 150.mu..OMEGA.
estimated) is used, the packaged device resistances are almost
unbelievably small for such a small device (4 mm.times.4 mm die, 12
mm.times.7 mm planar package, or about 6 mm.times.8 mm for a
vertical package with the same die). Even at a 200 amp drain
current, R.sub.on=150.mu..OMEGA. represents only a V.sub.ds=30 mV
voltage drop.
[0191] Gate Drive Requirements and Optional Device Features
[0192] A dramatic difference between the horizontal or planar
geometry deep-submicron channel length current switching MOSFET
structure of aspects of this invention and conventional vertical
geometry power MOSFETs is in their gate drive requirements. In the
example cited above, the total calculated gate capacitance is only
C.sub.gs+C.sub.gd=20 nF (Eq. 6, with .epsilon..sub.r=3.90, L=0.24
.mu.m, W=13.5 m and t.sub.ox=5.8 nm gives C.sub.gs+C.sub.gd=19.28
nF). This is far (1 to 2 orders of magnitude) less than the gate
capacitance of conventional vertical power MOSFETs having a
comparable R.sub.on values. (Actually, many conventional power
MOSFETs connected in parallel would be required to reach this Ron
level; it is the total C.sub.gs+C.sub.gd value for the paralleled
power MOSFETs that is compared to the C.sub.gs+C.sub.gd=20 nF value
of the planar switching MOSFET of aspects of this invention.) Also,
while conventional power MOSFETs require V.sub.gs=10 V or more gate
voltages to reach low R.sub.on, with aspects of this invention the
values in Table 1 are reached with only a V.sub.gs=2.5 V gate drive
voltage. As discussed previously, the combination of the lower
value of C.sub.gs+C.sub.gd and the lower V.sub.gs gate drive
voltage requirement drastically lowers the (Eq. 5)
P.sub.gd=1/2(C.sub.gs+C.sub.gd).DELTA.V.su- b.gs.sup.2 Fc gate
drive power requirements for the planar short-channel current
switching MOSFET of aspects of this invention (by nearly a factor
of 500 over conventional power MOSFETs).
[0193] One of the implications of this enormous reduction in ac
gate drive power is that it becomes practical to sharply increase
the switching frequency, F.sub.c, without serious efficiency loss
due. This makes it possible to reduce the size and weight of the
capacitive and magnetic energy storage devices in a power converter
circuit, to allow, for example, very small on-board switching power
converters to be realized. For example, an V.sub.out=1.2V switching
converter having a maximum output current of 200 amperes (240 watts
peak output) could be fabricated with two of these planar geometry
current-switching MOSFETs, which would have, even at a very high
F.sub.c=1 MHz switching rate a total gate drive power of only
P.sub.gd=0.125 watts for the two synchronous rectifier switches, or
0.052% of the peak output power. This means that it would be quite
practical to increase switching frequencies up into the many tens
of megahertz range before gate drive power became a serious
efficiency concern, which would allow for remarkable
miniaturization of power converters.
[0194] The low gate drive power requirements, along with the fact
that the planar geometry deep-submicron current switching MOSFET
devices are conveniently (and inexpensively) fabricated using a
standard deep-submicron CMOS foundry process, suggests a very
attractive option: the inclusion of the gate driver amplifier on
the same chip with the switching MOSFET. With conventional power
MOSFETs, furnishing clean ("square-wave") gate drive voltages is a
major problem, even at switching frequencies of the order of
F.sub.c=100 KHz. While the planar geometry deep-submicron current
switching MOSFET devices have low C.sub.gs+C.sub.gd=20 nF gate
capacitances and are capable of operation at >10 MHz switching
frequencies at low gate drive powers, for clean, efficient
square-wave gate voltages the pulsed gate currents will be
substantial. For example, if it is desired to keep the gate voltage
risetimes and falltimes to less than 2.5% of the switching period
at F.sub.c=10 MHz, the voltage risetimes must be under
t.sub.r=t.sub.f=2.5 ns. With .DELTA.V.sub.gs=2.5V and
C.sub.gs+C.sub.gd=20 nF, the gate drive current will pulse to
i.sub.g=(C.sub.gs+C.sub.gd)(.DELTA.V.sub.gs/t.sub.r- )=.+-.20
ampere levels in each MOSFET during the rise and fall intervals.
The distributed inductance/characteristic impedance of usual
on-board interconnects virtually precludes furnishing gate drive
signals of this quality in a practical multi-package on board
environment.
[0195] These gate drive signal limitations of conventional
packaging can be overcome in the case of the planar geometry
deep-submicron current switching MOSFET devices of aspects of this
invention by placing the gate driver amplifier on the same silicon
chip as the MOSFET, as illustrated in FIG. 8. The gate driver
amplifier would consist of a short chain of CMOS inverters (e.g., 2
to 4 stages typically; FIG. 8 shows a 2-stage gate driver) between
the gate input to the package/die and the actual MOSFET gates. In
order to minimize ac gate current distribution problems, it would
probably prove useful to distribute the last (largest) stage of the
gate amplifier chain in sections around the chip. The sizing of the
p- and n-channel MOSFET devices in the CMOS driver inverters can be
estimated from the device characteristics and the desired gate
drive currents. Consider a 2-stage CMOS inverter driver for the 200
ampere current switching MOSFET example, as shown in FIG. 8. The
saturated drain current for the L.sub.eff=0.19 .mu.m n-channel
MOSFETs in the 2.lambda.=0.24 .mu.m feature size commercial CMOS
foundry process is I.sub.dss/W=0.6 ma/.mu.m with an R.sub.onW=960
.mu.m "on" resistance. The p-channel devices are about half as good
(half the I.sub.dss/W and roughly twice the R.sub.onW), so for
balanced current drive the p-channel MOSFETs are scaled about twice
as wide as the n-channel devices. For convenience, the sizing
(width, W) of only the n-channel devices will be cited, with the
understanding that, as shown in FIG. 8, the matching p-channel
pull-up devices will be twice as wide. For a desired gate drive
current pulse of 20 amperes into the W=13.5 meter, 200 amp current
switching FET, a second-stage driver MOSFET width of Wd.sub.2=50 mm
(I.sub.dss=30 amps) would more than suffice. Its
R.sub.on=0.0192.OMEGA. on resistance would give an
t.sub.rc=R.sub.on(C.sub.gs+C.sub.gd)=0.384 ns time constant with
the (C.sub.gs+C.sub.gd)=20 nF gate capacitance, which would give
excellent gate voltage settling. The input gate capacitance,
C.sub.in2, of this W.sub.d2=50 mm driver output stage would be
about 75 pF for the n-channel device and 150 pf for the (2.times.
wider) p-channel MOSFET, or C.sub.in2=225 pF total. To drive this
225 pF capacitance over its .DELTA.V.sub.gs=2.5V range in
.DELTA.t.sub.r=1 ns would require
I.sub.in2=C.sub.in2.DELTA.V.sub.gs/.DELTA.t.sub.r=563 ma gate drive
current pulses, which require that the prior first-stage driver
have an n-channel MOSFET width of about W.sub.d1=1.0 mm
(I.sub.dss=600 ma). The input capacitance for this first driver
stage will be about 1.5 pF for the n-channel, and 3 pF for the
p-channel MOSFETs, or 4.5 pF total gate input capacitance for the
chip. Since a 4.5 pF gate driver input capacitance is very easy to
drive with very fast edge rates with typical Z.sub.o=50.OMEGA.
characteristic impedance circuit board interconnects, the two-stage
driver is more than adequate.
[0196] The inclusion of the on-chip gate driver amplifier allows
the gate input of this 200 amp planar current switching MOSFET chip
to be driven with remarkable timing precision with nothing more
than conventional digital logic signals and circuit board
interconnects. From a terminal count standpoint, it adds only two
package connections; the V.sub.dd=+2.5Vdc and V.sub.ss=ground
driver amplifier power inputs. While it would be possible, in
principle, in some applications to tie the input stage amplifier
V.sub.SS power lead to the source lead of the current switching
MOSFET, in many cases this could lead to feedback coupling
problems/potential for oscillations on transitions, so a separate
input stage V.sub.SS ground lead is preferred. On the other hand,
the high peak gate current spikes into the switching demand a very
low impedance (particularly inductance) connection between the
driver output stage and the switching MOSFET gate, most easily
accomplished if we tie the driver V.sub.SS power lead to the source
lead of the current switching MOSFET. A side benefit to tying the
V.sub.SS lead of the amplifier driver (output) stage to the source
of the switching MOSFET is that it automatically achieves the "body
diode" of FIG. 2B. When the input in FIG. 8 is taken "LOW"
(.apprxeq.V.sub.SS) to turn off the current switching MOSFET, its
gate will essentially tied its source through the "ON" W.sub.d2n=50
.mu.m driver MOSFET. In this state, as noted in FIG. 2B, the
W.sub.n=13,500 .mu.m current switching MOSFET will begin to conduct
any time its drain voltage becomes more negative than its source by
more than its gate threshold voltage, V.sub.tn. While connecting
the source of the W.sub.d2n=50 .mu.m n-channel driver MOSFET
directly to the source of the output current switch MOSFET solves
the problem for handling the large (.apprxeq.20 ampere) negative
gate current peaks, the equally large positive gate current spikes
out of the W.sub.d2p=100 .mu.m p-channel driver MOSFET must also be
dealt with. In the gate driver amplifier design in FIG. 8, a large
(>20 nF) bypass capacitor is included between V.sub.dd and the
source of the current switching MOSFET. In the absence if this
bypass capacitor, the full magnitude of the positive gate current
spikes would have to be furnished through the V.sub.dd=+2.5V gate
driver amplifier supply, which would require an extremely low ac
impedance (particularly inductance) to the V.sub.dd lead in order
to avoid "voltage starvation" during positive gate current
transients which would compromise the switching process. The
inclusion of the on-chip bypass capacitor as an extremely low
inductance source of the charge required to charge the gate
capacitance of the large current switching MOSFET makes an
ultra-low inductance Vdd connection to the chip unnecessary. It
has, however, some chip area/cost penalty. The increase in chip
area due to the gate driver amplifier itself is minimal, adding a
total of only W.sub.d(n+p)=153 mm (W.sub.d2=50 mm+W.sub.d1=1.0 mm,
or 51 mm of n-channel MOSFET width, plus twice that, or
W.sub.d1p+W.sub.d2p=100 mm, of p-channel MOSFET width). Hence, this
on-chip gate driver would require only about 1.1% of the die area.
The capability for dispatching the ordinarily-odious gate drive
problem at an increase of die cost of only 1% to 2% is a remarkable
side-benefit of implementing the planar current switching MOSFET of
aspects of this invention in a standard deep-submicron CMOS
integrated circuit foundry process. The inclusion of the >20 nF
bypass capacitor is more costly. Assuming the IC process has no
special very high capacitance per unit area (C/A) capability (such
as vertical geometry capacitors or ferroelectric dielectrics like
many DRAM processes have, then the highest available C/A is the
gate to channel capacitance. Implementing, using this gate oxide
capacitance, a bypass capacitor whose capacitance equals the gate
capacitance of the W.sub.n=13,500 .mu.m current switching MOSFET
requires an area equal to approximately 20% of that of the
W.sub.n=13,500 .mu.m current switching MOSFET. Again, this is a
small price to pay for a virtually effortless way of solving the
normally odious gate drive problem in high-speed switchers.
[0197] The capability for integrating other functionality on the
current-switching MOSFET die by including CMOS circuitry of various
types is also enabled by the use of a CMOS IC process for
fabricating the devices. These could include various types of
analog or digital circuitry that would otherwise require separate
IC chips in the various applications. One example of this would be
the inclusion of the switcher control circuitry for a power
converter (i.e., the analog circuitry that compares the actual VOUt
dc output voltage of the converter to the voltage setpoint value
and generates the various primary switching and secondary output
synchronous rectifier switching timing signals needed for the
operation of the power converter. While this would require
additional package pins, the value added in reduced parts count and
potential for miniaturization would be substantial. Another
example, shown in FIG. 2b, would be the inclusion of the "optional
ultra-low Q.sub.d diode" to replace the "body diode" (FIG. 2a) of a
normal vertical-geometry power MOSFET. This "body diode" function
is emulated by an active diode-connected p-channel MOSFET (i.e.,
the gate is connected to the drain). This device structure and its
applications were discussed in a previous section. Note that it is
possible to include both the gate driver amplifier and active
diode-connected p-channel MOSFET "body diode" options at the same
time (i.e., as additions to the same planar current switching
MOSFET die), as could be other application-specific circuitry.
Other applications for the planar current switching MOSFET devices
of aspects of this invention could benefit from the addition of
specialized circuitry designed to serve the needs of each
particular application.
FIGURE CAPTIONS
[0198] FIG. 1a. [Top] Device structure of a conventional vertical
geometry power MOSFET showing lateral electron path through surface
n-channel transitioning to vertical path down to drain contact.
[0199] FIG. 1b. [Bottom] Device structure of n-channel
planar/horizontal geometry high-current switching MOSFET of aspects
of this invention implemented in a deep-submicron CMOS IC process,
showing very short lateral electron path for very low R.sub.on.
Also illustrated at right is the capability to include p-channel
MOSFETs on same die.
[0200] FIG. 2a. [Left] Equivalent circuit for conventional vertical
geometry n-channel MOS power FET device, showing body diode. Note
that very large stored diffusion charge, Q.sub.d, in body diode
severely limits switching speeds at which it can be used.
[0201] FIG. 2b. [Right] Equivalent circuit for planar/horizontal
geometry power high-current switching MOSFET of aspects of this
invention, including optional ultra-low stored charge diode
implemented with an active diode-connected p-channel MOSFET for use
when "body diode" functionality is required.
[0202] FIG. 3. Cross-section drawing of planar deep-submicron
current switching MOSFET of aspects of this invention including
high-current die and package interconnects from source and drain
electrodes. Die features are drawn to scale for a 2.lambda.=0.24
.mu.m feature size CMOS process, but some package/solder ball
dimensions reduced to maintain visibility of die features.
[0203] FIGS. 4A-4B. Plan view of layout example for planar
deep-submicron current switching MOSFET of aspects of this
invention drawn to scale for implementation in a 2.lambda.=0.24
.mu.m feature size CMOS process. The source/drain and gate poly
silicide layers plus the via cuts (small squares) up to the first
level metal (Ml) are illustrated at the upper left. At the upper
right, M.sub.1, the first level of interconnect metal, is added,
and the via cuts to M.sub.2 are shown. At the lower left, M.sub.2,
the second level of interconnect metal, is added, and the via cuts
to M.sub.3 are shown. At the lower right, M.sub.3, the third level
of interconnect metal, dedicated principally to the source plane,
is added, and the via cuts to M.sub.4 are shown.
[0204] FIGS. 5A-5B. (Continued from FIG. 4). Plan view of layout
example for planar deep-submicron current switching MOSFET of
aspects of this invention drawn to scale for implementation in a
2.lambda.=0.24 .mu.m feature size CMOS process. At the upper left,
M.sub.3 and the via cuts to M.sub.4 are shown, as in FIG. 4D. At
the upper right, M.sub.4, the fourth level of interconnect metal,
dedicated principally to the drain plane, is added, and the via
cuts to M.sub.5 are shown. At the lower left, M.sub.5, the fifth
and final level of interconnect metal, used principally for the
source, drain and gate solder ball pads and gate interconnect
lines, is added (the scale does not permit showing the "pad mask"
dielectric opening that defines the circular solder ball contact
areas). At the lower right, the scale has been changed to show the
entire 4 mm.times.4 mm die, showing the M.sub.5 (top metal) source,
drain and gate solder ball pads with gate interconnect lines
running in between them, and the "pad mask" dielectric opening that
defines the circular solder ball contact areas.
[0205] FIGS. 6A and 6B. Plan view [top] and cross-sectional view
[bottom] of example of a planar package for the planar
deep-submicron current switching MOSFET of aspects of this
invention. Because the high current (e.g., 200 amp) source and
drain leads are coplanar on the package bottom (as well as the ends
of the top surface), it would be suitable for surface mount
applications (with an appropriate gate lead extension from the gate
pad on the top surface).
[0206] FIG. 7. Example of including optional two CMOS inverter
stage gate driver amplifier on the same die as a 200 amp planar
deep-submicron current switching MOSFET of aspects of this
invention. The driver-amplifier would be capable of supplying 20 to
30 amp gate pulses to the W=13,500 m [(C.sub.gs+C.sub.gd)=20 nF]
planar switching MOSFET for 1.5 to 2 ns risetimes of the 0 to 2.5V
gate voltage, while presenting only about a 4.5 pF capacitive load
at the package gate input and requiring less than 1.5% to 2% of the
die area.
[0207] FIG. 8. Example of a gate driver amplifier.
[0208] FIGS. 9A and 9B. Example plan and side views of a vertically
laminated package.
[0209] FIG. 10. 250 .mu.m Bump Pitch Calculation of 0.19 .mu.m
Power Switching FET Resistance with On-Chip Metal and Package
R.
* * * * *