Miscellaneous active electrical nonlinear devices – circuits – and – Signal converting – shaping – or generating – Phase shift by less than period of input
Reexamination Certificate
1997-06-27
2001-03-06
Callahan, Timothy P. (Department: 2816)
Miscellaneous active electrical nonlinear devices, circuits, and
Signal converting, shaping, or generating
Phase shift by less than period of input
C327S112000, C327S295000, C330S264000
Reexamination Certificate
active
06198325
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention is related generally to local power supplies for integrated circuits with a high clock frequency, and in particular to circuits for use in limiting inductance-induced ripple voltage in integrated circuits with high dynamic power consumption.
2. Description of Related Art
Over the past few years clock speeds of central processing units (CPUs) have increased from a few MHz to 500 MHz or more. This increase in clock speed requires that components within the CPU, i.e., the CPU core, as well as components that communicate with the CPU operate at ever increasing clock speeds.
The increase in CPU clock speeds has been accompanied by an increase in the number of transistors on the CPU die, i.e., in the integration on the chip. The greater integration results in a larger die size, which in turn means that some circuits on the die are separated by greater distances, and in addition, more pins are required to take information to and from the die. Both of these effects complicate electromagnetic interference (EMI) shielding at the higher clock speeds.
FIG. 1
is a first order lumped model of a typical CPU configuration as seen by the power supply pins. A first inductor
145
represents the inductance of the board plane and via layer, and is connected to power supply line
110
. Connected in series with inductor
145
is a second inductor
140
that represents the inductance of the socket and packaging of the CPU. Connected in series with inductors
140
and
145
is a series combination of a third inductor
137
and a resistor
135
. The series combination of third inductor
137
and resistor
135
represents the inductance and resistance, respectively of either a ball grid array, or a lead frame with bond wires, and the power grid of CPU
100
. The CPU die has an intrinsic RC characteristic that is represented by a series combination of resistor
151
and capacitor
152
connected to resistor
135
by local power supply line
110
A and to ground line
111
by local power supply line
111
A.
Typically, at the start of each clock cycle, many elements in CPU
100
change state, which in turn causes a momentary increase in current draw, i.e., the current draw changes as a function of time. The voltage generated by inductors
145
,
140
and
137
is directly proportional to the changes in current with respect to time. Specifically, as the time derivative of the current increases, inductors
145
,
140
,
137
create a positive voltage drop which in turn reduces the voltage across local power supply lines
110
A and
111
A. Consequently, sufficient power cannot be provided instantaneously to the elements in the CPU core changing state, and so the voltage difference between local power supply lines
110
A and
111
A decreases, i.e., collapses.
However, once the various elements have switched state, the change in current draw with respect to time diminishes and the voltage difference recovers. In addition, inductors
145
,
140
, and
137
supplement the voltage across local power supply lines
110
A and
111
A because the time derivative of the current is negative.
These swings in voltages caused by the inherent inductances, including parasitic inductances, in response to changes in the current draw over time, are called inductance-induced ripple voltages or sometimes simply bounce. The inductance-induced ripple voltages have many undesirable features. For example, if the voltage collapse is too great, operation of CPU
100
can become unreliable. Also, the inductance-induced ripple voltages radiate from at least the pins of the circuit, and can radiate from power supply lines in CPU
100
that function as antennas. This requires additional shielding or other design changes to assure that CPU
100
complies with all relevant EMI standards. Additionally, if the frequency of the inductance-induced ripple voltage approaches the resonance frequency of the package, the voltage collapse and EMI noise is effectively amplified which in turn further exacerbates the problems associated with the parasitic inductances.
Various techniques have been used to minimize the effects of the inductance-induced ripple voltages in attempting to provide a stable power supply voltage across a CPU core. Specifically, power supply decoupling was normally achieved by placing banks of capacitors on the die between the local power supply lines carrying power supply and ground potentials so as to minimize the effects of parasitic inductances and resistances. However, each bank of capacitors has a limited band in the frequency spectrum where the bank is effective in smoothing out the inductance-induced ripple voltage, and has a limited capacity to decouple.
Moreover, as both the power consumption and the clocking speeds increased, the switching current at local power nodes within the integrated circuit required a relatively large capacitance to offset the power losses associated with the parasitic board and package inductances near the resonance frequency of the package. This, in turn, meant that a larger number of high frequency capacitors were required.
However, as the high frequency circuits become more highly integrated, the real estate on the die available for capacitors diminished as the requirement for the number of capacitors increased due to the increased power consumption. This limitation forced consideration of alternative physical and manufacturing configurations to provide the required amount of passive capacitance.
A first approach was to connect passive capacitance
160
on the board between the CPU socket power supply connectors and ground, i.e., between the connection of inductors
145
and
140
and ground as illustrated in FIG.
2
A. It should be understood that passive capacitance
160
includes parasitic inductance and resistance that are connected in series with passive capacitance
160
. However, this approach was not completely successful because this configuration did not effectively offset the effects of inductors
140
and
137
.
Consequently, some manufacturers use discrete capacitors that are placed on top or below the package. In this configuration, passive capacitance
165
is connected between inductors
140
and
137
and ground as illustrated in FIG.
2
B. While this configuration is better than the configuration of
FIG. 2A
, passive capacitance
165
does not directly affect the voltage across local power supply lines
110
A and
111
A.
Hence, other manufacturers include a separate chip of low inductance capacitors in the package containing the high frequency die and attempt to connect capacitance
170
on the separate chip between local power supply line
110
A and
111
A as shown in FIG.
2
C.
While the current manufacturing techniques and physical configurations may be adequate for current microprocessor clock speeds and power consumption, as the power consumption and clock speeds continue to increase other solutions will be required. Unfortunately, the next generation of high-speed circuits will draw even more power, because typically the power consumption is proportional to the clock speed. With the higher power consumption and the fast clock speeds, the changes in the time derivative of the current will be more extreme which in turn indicates problems associated with inductance-induced ripple voltages will be further exacerbated.
The solution to power supply inductance-induced ripple voltages appears to be limited to the use of passive capacitors. Other techniques for offsetting the effects of parasitic inductance and resistance are not of use considering the feature sizes of the next generation of integrated circuits. Consequently, the dynamic power consumption of future generations of circuits may be limited by the inability to effectively offset the effects of parasitic inductances and resistances on power supply voltages.
SUMMARY OF THE INVENTION
In accordance with the principles of this invention, an active digital voltage regulator circuit overcomes the problems with stabilizing on-board power supply v
Ang Michael Anthony
Taylor Alexander Dougald
Callahan Timothy P.
Gunnison Forrest
Gunnison, McKay & Hodgson LLP
Nguyen Minh
Sun Microsystems Inc.
LandOfFree
Differencing non-overlapped dual-output amplifier circuit does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Differencing non-overlapped dual-output amplifier circuit, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Differencing non-overlapped dual-output amplifier circuit will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2513978