Miscellaneous active electrical nonlinear devices – circuits – and – Signal converting – shaping – or generating – Synchronizing
Reexamination Certificate
2000-03-31
2004-01-20
Nguyen, Minh (Department: 2816)
Miscellaneous active electrical nonlinear devices, circuits, and
Signal converting, shaping, or generating
Synchronizing
C327S237000, C327S269000, C370S508000
Reexamination Certificate
active
06680636
ABSTRACT:
FIELD OF THE INVENTION
The field of the present invention pertains to data communications between digital systems. More particularly, the present invention relates to a method and system for high performance source synchronous data communication between integrated circuit devices.
BACKGROUND OF THE INVENTION
The field of data communications represents one of the most rapidly evolving technologies in wide spread use today. Data communications and data processing has become important to virtually every segment of the nation's economy. Whole new industries and companies have organized around the need for, and the provision of, data communications. Through the use of specialized semiconductors for signal processing and data compression, various multimedia applications are evolving which orient data communications to the transport of voice, data, and video information, the types of information desired by the everyday consumer.
Recently, the computer and data processing industries are seeing a large expansion in the requirements for high performance, high speed data communications between multiple integrated circuits on, for example, printed circuit boards. For example, it is becoming increasingly common to implement high-performance digital systems using multiple integrated circuit modules, or chips, interconnected on a high-speed printed circuit board. The multiple chips are typically highly integrated, having several million transistors per chip, and operating at very high speeds (e.g., 500 MHz or above). With such technology, the speed and integrity of the data communications between the chips becomes very critical.
Data is commonly transferred between computer systems and terminals by changes in the current or voltage on a metal wire, or channel, between the systems. These interconnections are typically etched into the material of the printed circuit board itself. A data transmission in which a group of bits moves over several channels simultaneously is referred to as a parallel transmission. A data transmission where the bits move over a single channel, one after the other, is referred to as a serial transmission. Computers and other data processing systems which are located on a single printed circuit board normally use parallel transmission because it is much faster.
As the level of integration and the operating speeds of the multiple chips on printed circuit boards increases, the transmission of data between and among the multiple chips via the channels of the printed circuit board suffer a number of limitations. One such limitation is due to the fact that most digital systems are designed to operate synchronously with respect to the individual integrated circuits which comprise the system. For example, the multiple chips coupled to a printed circuit board are typically designed to operate synchronously with respect to one another, using well-defined clock frequency and phase relationships. However, as the operating speeds of the multiple chips increases, the tolerance of the system for “timing skew” among the multiple data channels decreases. The timing relationship between, for example, the clock signal shared among the multiple chips and the corresponding data signals conveyed across the channels becomes increasingly critical. Prior Art
FIG. 1
below illustrates this problem.
Prior Art
FIG. 1
shows a typical high-speed multichip device
10
. Device
10
includes a first chip
11
(e.g., chip
1
) a second chip
12
(e.g., chip
2
). Chips
11
and
12
are communicatively coupled via data channels of a printed circuit board. One such data channel
14
is shown. Chips
11
and
12
share a common clock signal
16
and operate synchronously with respect to clock signal
16
.
As described above, as the operating speeds of multichip device (e.g. system
10
) increases, the tolerance of the system for timing skew among the data channels and with respect to the clock signal decreases.
FIG. 1
depicts this problem. The high level of integration of chips
11
and
12
causes a clock insertion delay, depicted as clock insertion delay
15
, as the clock signal
16
propagates among the millions of transistors comprising chips
11
and
12
. As the clock signal reaches logic elements
17
and
18
deeply within chips
11
and
12
, the phase relationship of the output of logic element
17
and the input of logic element
18
can vary significantly. For the device
10
to remain synchronous, the output of logic element
17
needs to be received at the input of logic element
18
prior to the next cycle of clock
16
. The propagation delay, Tpd
13
, from the output pin of chip
11
to the input pin of chip
12
constitutes a significant portion of this delay.
As system
10
is designed, engineers account for the various delay factors in designing system
10
to operate at its maximum speed. For example, the delay of clock
16
propagates to each chip is accounted for by, for example, precisely defining the length of the channels transmitting clock
16
to each chip. Similarly, the length of the data channels, such as data channel
14
, between the chips is precisely defined. However, the clock insertion delay incurred in each chip as clock
16
propagates among the millions of transistors comprising the chips cannot be as precisely controlled. Numerous variables (e.g., fabrication process variation, temperature, voltage fluctuation, etc.) affect the propagation delay, and unfortunately, many of these variables cannot be precisely ascertained or controlled. The variables affect the “setup-and-hold” timing tolerances of the overall device.
Prior Art
FIGS. 2A-2C
illustrate the setup-and-hold timing tolerance problem.
FIG. 2A
shows a typical logic element
21
as contained in chips
11
and
12
. Element
21
depicts an edge triggered flip-flop having a data input, a data output, and a clock input as shown.
FIG. 2B
shows a diagram of the proper timing relationship between data
22
and the clock signal
23
. As depicted in
FIG. 2B
, ideally, the rising edge of the clock signal
23
is placed such that perfectly corresponds to the setup time
24
of the data input
22
and the hold time
25
. This provides the maximum likelihood that the correct value of the data input is clocked into logic element
21
.
FIG. 2C
shows a diagram of an improper timing relationship between data
22
and the clock signal
23
. In this case, the phase relationship between the clock signal
23
in the data
22
has deteriorated such that the setup and hold times
24
and
25
are not properly placed with respect to the phase of the data signal
22
. In this case, the rising edge of clock signal
23
does not correspond to the correct value of the data input
22
, leading to “indeterminate” operation of the logic element
21
. This deterioration is typically due to the uncontrollable variables described above (e.g., fabrication process variation, temperature, voltage fluctuation, etc.).
Hence, a significant amount of uncertainty exists regarding the maximum possible speed of the multichip device, which leads to extensive testing to determine “safe” operating margins, device malfunctions, and/or less than optimal device configurations. Device
10
must be engineered such that it retains enough margin to ensure proper operation taking into account performance variables such as process variation, temperature, and the like.
One attempted solution creates individual serial data bit streams out of each channel. This scheme encodes the clock signal directly into the bit stream, recovering the clock signal at the receiver and reconstructing the data word through signal processing techniques. This system requires complex (e.g., expensive) signal processing at the transmitting chip and the receiving chip and is thus generally impractical for printed circuit board type devices.
Another attempted solution performs a complex set of analyses on test signal patterns on each of the channels between the multiple chips. The results of the analysis are used to reconfigure compensation or filter circuits between
Collins Hansel
Everhardt Paul
Parry David
Nguyen Minh
Schwegman Lundberg Woessner & Kluth P.A.
Silicon Graphics Inc.
LandOfFree
Method and system for clock cycle measurement and delay offset does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method and system for clock cycle measurement and delay offset, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method and system for clock cycle measurement and delay offset will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3231202