Architecture for large capacity high-speed random access memory

Static information storage and retrieval – Addressing – Plural blocks or banks

Reexamination Certificate

Rate now

  [ 0.00 ] – not rated yet Voters 0   Comments 0

Details

C365S230010, C365S189011, C365S063000

Reexamination Certificate

active

06215718

ABSTRACT:

TECHNICAL FIELD
The present invention relates generally to semiconductor devices, and more particularly to random access memories (RAMs) with rapid access speeds and high capacities.
BACKGROUND OF THE INVENTION
The increasing operating speed and computing power of electronic systems has given a rise to the need for memory devices having rapid access times and large capacities. In order to ensure that computing cycles are not wasted by a system, large amounts of data must be provided by a semiconductor memory device at as high a rate as possible.
In a typical read operation, a number of memory cells are selected by the application of an address. The data stored within the memory cells are then accessed according to one or more timing signals. For example, in the case of an asynchronous dynamic random access memory (DRAM), a row address is applied in combination with a row address strobe (RAS) timing signal to select a row of memory cells. A column address is then applied in combination with a column address strobe (CAS) timing signal to access selected cells within the row. In the case of a synchronous DRAM (SDRAM), memory cells are selected according to a system clock.
As semiconductor memory devices increase in capacity, they also typically increase in physical size. Larger physical sizes can impact the overall speed of a device, as data and timing signals must propagate across larger distances, adding to the response time of the device. The placement of the conductive lines (the routing) that carry timing and data signals can thus play an important role in the speed of a memory device.
The millions of memory cells within a high-capacity memory device are typically arranged into a number of arrays that are further divided into a number of array banks. Data is output from each array bank by input/output (I/O) lines. A problem with high-capacity memory devices is that all of the I/O lines must be routed to the same I/O circuit. The I/O circuit includes the input buffers, output buffers, and latches necessary to store incoming data in a write operation, or drive outgoing data in a read operation. In the event the I/O circuits are located in a central portion of the semiconductor memory device, a routing “bottleneck” can occur in the central portion of the device. The bottleneck results in large numbers of I/O lines overlapping one another and limiting the available space in the central portion of the device.
One example of a memory device having a routing bottleneck in a central portion of the device is set forth in FIG.
1
.
FIG. 1
is a top plan view of a SDRAM, illustrating the placement of memory cell banks and various other circuit blocks. The SDRAM is designated by the general reference character
100
, and is shown to include four array banks
102
a
-
102
d
. In the particular example of
FIG. 1
, the SDRAM has a storage capacity of 256 megabits (Mb); thus each array bank (
102
a
-
102
d
) includes 64 Mb. Each array bank (
102
a
-
102
d
) is further divided into a first and second sub-banks. The first sub-banks are shown as
104
a
-
104
d
and the second sub-banks are shown as
106
a
-
106
d
. The memory cells within each sub-bank are accessed by activating associated row address circuitry
108
and column select circuitry
110
.
Data within the SDRAM
100
are accessed by way of a first I/O circuit
112
, situated between array banks
102
a
and
102
c
, and a second I/O circuit
114
, situated between array banks
102
b
and
102
d
. The I/O circuits (
112
and
114
) possess the structures necessary to input data for write operations, and to output data in read operations, including I/O pads. When the SDRAM
100
is active in a read or write cycle, the memory cells within one of the array banks (
102
a
-
102
d
) are accessed. In order to make this possible, each array bank (
102
a
-
102
d
) has a data I/O bus that couples data from the array bank to the I/O circuits (
112
and
114
). Because there are two I/O circuits (
112
and
114
), each data I/O bus is further divided into two sub-bank buses. One sub-bank bus couples the data from a first sub-bank (
104
a
for example) to the first I/O circuit
112
, while the other sub-bank bus couples the data from a second sub-bank (
106
a
for example) to the second I/O circuit
114
.
In the particular example of
FIG. 1
, the general path of only selected I/O lines is illustrated to not unduly clutter the view of the figure. In particular, the first I/O lines (
116
a
,
116
b
,
116
c
and
116
d
) and last I/O lines (
118
a
,
118
b
,
118
c
and
1186
d
) of the sub-bank buses for array banks
102
a
and
102
b
are illustrated. Thus, first I/O line
116
a
and last I/O line
118
a
are used to represent a first sub-bank bus
120
a
that connects first sub-bank
104
a
with the first I/O circuit
112
. A second sub-bank bus
120
b
connects the second sub-bank
106
a
to the second I/O circuit
114
, and is defined by first I/O line
116
b
and last I/O line
118
b
. In the same general fashion, sub-bank bus
120
c
, defined by first I/O line
116
c
and last I/O line
118
c
, connects first sub-bank
104
b
to first I/O circuit
112
, and sub-bank bus
120
d
, defined by first I/O line
116
d
and last I/O line
118
d
, connects second sub-bank
106
b
to the second I/O circuit
114
. It is understood that sub-banks
104
c
,
106
c
,
104
d
and
106
d
are connected to the first and second I/O circuits (
112
and
114
) in mirror image fashion.
It is noted that the SDRAM
100
further includes a timing circuit
122
located in the central portion of the device. The timing circuit
122
receives timing signals, such as the system clock signal, and in response thereto, activates circuits within the SDRAM that are necessary to access data within the memory cells. The timing path of a memory cell access operation is shown in
FIG. 1
by dashed line
124
. In response to a clock signal applied to the timing circuit
122
, a signal is activated which runs to the sub-bank
106
c
, and places data on an I/O line. The I/O line is coupled to an I/O bus line, which connects the sub-bank
106
c
to the second I/O circuit
114
. The central location of the timing circuit
122
allows for shorter timing path distances with respect to all of the array banks. Also set forth in
FIG. 1
, are the word lines
126
that are activated in order to provide the access operation illustrated by line
124
.
Referring now to
FIGS. 2A and 2B
, a portion of
FIG. 1
is set forth in a top plan view to provide one representation of the sub-bank bus lines.
FIG. 2A
provides a representation of sub-bank buses
120
a
and
120
b
. It is understood that each of the bus lines set forth in
FIG. 2A
represents four actual bus lines, giving a total of
32
bus lines in each sub-bank bus. Sub-bank buses
120
c
and
120
d
are omitted in FIG.
2
A.
FIG. 2B
sets forth the same view as
FIG. 2A
, but omits sub-bank buses
120
a
and
120
b
, and includes sub-bank buses
120
c
and
120
d
. A comparison between
FIGS. 2A and 2B
illustrates that sub-bank buses
120
b
and
120
c
must both travel over the same location, and so overlap one another in the central portion of the SDRAM. This results in an undesirable routing bottleneck in the center of the device.
Referring now to
FIG. 3
, a top plan view of an alternate SDRAM architecture is set forth. The SDRAM is designated by the general reference character
300
, and is shown to include many of the same general elements as FIG.
1
. To this extent, like elements will be referred to by the same reference characters, but with the first number being a “3” instead of a “1.” Accordingly, the SDRAM
300
of
FIG. 3
includes four array banks (
302
a
-
302
d
), each of which includes a first sub-bank (
304
a
-
304
d
) and a second sub-bank (
306
a
-
306
d
). Unlike the SDRAM
100
of
FIG. 1
, the array banks (
306
a
-
306
d
) extend in the horizontal direction across the entire SDRAM
300
. Similarly, the first and second sub-banks (
304
a
-
304
d
and
306
a
-
306
d
) extend roughly halfway across the SDRAM in

LandOfFree

Say what you really think

Search LandOfFree.com for the USA inventors and patents. Rate them and share your experience with other people.

Rating

Architecture for large capacity high-speed random access memory does not yet have a rating. At this time, there are no reviews or comments for this patent.

If you have personal experience with Architecture for large capacity high-speed random access memory, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Architecture for large capacity high-speed random access memory will most certainly appreciate the feedback.

Rate now

     

Profile ID: LFUS-PAI-O-2512008

  Search
All data on this website is collected from public sources. Our data reflects the most accurate information available at the time of publication.