Electrical computers and digital processing systems: memory – Storage accessing and control – Specific memory composition
Reexamination Certificate
2001-03-22
2003-09-16
Nguyen, Hiep T. (Department: 2187)
Electrical computers and digital processing systems: memory
Storage accessing and control
Specific memory composition
C711S110000, C711S167000
Reexamination Certificate
active
06622198
ABSTRACT:
BACKGROUND OF THE INVENTION
The present invention relates, in general, to the field of integrated circuit (“IC”) devices. More particularly, the present invention relates to a look-ahead, wrap-around first-in, first-out (“FIFO”) integrated circuit device architecture.
U.S. Pat. No. 5,996,052 issued Nov. 30, 1999 describes a conventional method and circuit for enabling a clock-synchronized read-modify-write operation on a memory array. The design illustrated includes a FIFO memory, configured with a predetermined number of memory stages, which is incorporated into the input side of the device address decoder and is used only for writing. The write decoder circuitry of the design shown constitutes a FIFO memory and a write only address decoder.
Conventional data capturing techniques using non-look-ahead FIFOs, as represented by the '052 patent, require extremely tight timing constraints between the data-in and clocking signals. Moreover, in order to ensure overall high speed data input/output (“I/O”) operations, read-modify-write operations must be able to complete in the same time frame as read only cycles.
SUMMARY OF THE INVENTION
In order to implement such high speed data I/O operations, a write address FIFO is disclosed herein which, in an exemplary embodiment is organized as an eight stage by 6 bits configuration, that may be operated in a wrap-around fashion and wherein conventional status flags (i.e. full, half-full, empty, etc.) are not required. The particular embodiment of the FIFO of the present invention is able to operate at 714 MHz (1.4 ns cycle time) or faster and data loaded into the FIFO is available for read accessing from the output of the FIFO one cycle after it is loaded. In essence, the design disclosed serves to effectively minimize device power requirements while simultaneously enabling high speed operation utilizing separate read data and write address data paths to the memory array.
Functionally, a unique latching technique is utilized such that data is transparently captured into a FIFO stage implemented in the form of a single stage latch which functions as the storage element. More specifically, the technique disclosed herein serves to initialize the latency level of the first stage of the FIFO in the “enabled” (or “transparent”) state whereby it becomes “inactive” when a load command is executed. When the load command is executed to the first stage, its latch is “disabled”, but the following stage's latch becomes “active” in anticipation of the next stage being loaded by a subsequent load command. Latching in this fashion results in a single-cycle, look-ahead function.
In an exemplary embodiment, since the latching sequence operates in a single cycle look-ahead manner, it is desirable to utilize N+1 FIFO stages, where N is the stage depth of the FIFO. Using this as an example, an eight-stage FIFO would then require nine FIFO stages to implement single-cycle look-ahead functionality. This wrap-around operation can continue as long as the following condition is satisfied:
NL−NR≦T
where NL is the number of loads; NR is the number of reads and T is the logical depth of the FIFO. In other words, a nine-stage FIFO using single-cycle look-ahead provides an eight-stage logical depth and
NL−NR≦
8
By extending the transparent single-cycle latching technique disclosed herein, the frequency performance of a DRAM memory macro can be greatly improved at only a slight increase in power requirements through the use of two-cycle, look-ahead operation. In this manner, the timing constraints otherwise associated with enabling the latch on the same cycle as a load command occurs are removed even further increasing the ease in handling back-to-back loads and enabling even higher frequency operation.
Utilizing this technique, when a FIFO load occurs implemented with two-cycle look-ahead operation, the latching signal for the targeted stage goes “inactive” (i.e. capturing data or “disabled load”) while the next stage remains “active” (i.e. “transparent loading”) and the latch on the following stage (two stages after the loaded stage) goes “active” in such a way that there are always two stages with their latches “active” (or “enabled”). When this two-stage transparency operation is implemented, N+2 stages are required for wrap-around operation for a FIFO stage equivalency of N stages.
Wrap-around operation in this manner then also requires:
NL−NR≦T
as for the previously described single-cycle look-ahead function. With respect to the exemplary embodiments disclosed herein, a primary difference between one-stage and two-stage look-ahead operation is the generation of the LWORDB and WWORDB signal and the number of stages in the FIFO (or CWRBIT circuits).
Particularly disclosed herein is a DRAM macro which employs only a single common set of column address inputs for reading and writing data. A write address FIFO and independent internal read and write address and data paths allow for simultaneous read/write operation which effectively doubles the data rate for read-modify-write or read-write cycles. For a read cycle, the column address is loaded into a write address FIFO for use in a later write cycle. In each clock cycle where the write address strobe (“WASB”) is active “low”, the in-coming column address is loaded into an eight stage FIFO. Although the FIFO stores eight addresses, there are nine stages, in the case of single cycle look-ahead loading, so that the load pointer can “look-ahead” for improved FIFO speed. In each write cycle (WRITEB “low”), a write address is read from the FIFO and sent to the write column decoders and a column address can be loaded into the FIFO during a read cycle. This means that the read address can be stored in the FIFO and retrieved in a subsequent write cycle without having to supply a write address to the macro. This feature provides read-modify-write operation with simultaneous read and write cycles and only the read addresses need be supplied to the macro. It should also be noted that write only cycles may also be implemented through the conjunctive use of the WASB and WRITE commands.
REFERENCES:
patent: 4734850 (1988-03-01), Torii et al.
patent: 4914652 (1990-04-01), Nguyen
patent: 4951280 (1990-08-01), McCool et al.
patent: 5027326 (1991-06-01), Jones
patent: 5420997 (1995-05-01), Browning et al.
patent: 5646700 (1997-07-01), Park
patent: 5838933 (1998-11-01), Molnar et al.
patent: 5901103 (1999-05-01), Harris, II et al.
patent: 5978935 (1999-11-01), Kim et al.
patent: 5996052 (1999-11-01), Taniguchi et al.
patent: 6026473 (2000-02-01), Cross et al.
patent: 6172927 (2001-01-01), Taylor
patent: 6269413 (2001-07-01), Sherlock
patent: 6366979 (2002-04-01), Narayana et al.
Hogan & Hartson LLP
Kubida William J.
Meza Peter J.
Nguyen Hiep T.
United Memories Inc.
LandOfFree
Look-ahead, wrap-around first-in, first-out integrated... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Look-ahead, wrap-around first-in, first-out integrated..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Look-ahead, wrap-around first-in, first-out integrated... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3047080