Self-timed strobe generator and method for use with multi-strobe random access memories to increase memory bandwidth

This application claims priority under 35 U.S.C. §119(e) to U.S. Provisional Patent Application Ser. No. 60/410,977 entitled SELF-TIMED STROBE GENERATOR AND A METHOD TO USE IT WITH MULTI-STROBE RANDOM ACCESS MEMORIES TO DOUBLE BANDWIDTH filed in the name of Heragu et al. on Sep. 16, 2002, the entirety of which is hereby incorporated by reference.

This invention generally relates to electronic digital logic circuitry, and in particular it relates to the clocking of logic stages.

A chip that includes a memory, such as a static random access memory (SRAM), a dynamic random access memory (DRAM), or a synchronous dynamic random access memory (SDRAM), typically uses a system clock to generate signals for allowing memory accesses, including read and write functions. Rising and/or falling edges of the system clock signal are used to trigger the strobe inputs.

In such memories with multiple strobe inputs, read or write memory operations may occur at a rising edge of any strobe signal. One way to generate two strobes is to use the system clock to generate the first strobe signal and an inversion of the system clock to generate the second strobe signal. This introduces a requirement to maintain a duty cycle for memory functions that does not substantially deviate from 50% of the clock cycle, in order to optimize memory access bandwidth.

A precise 50% duty cycle is often difficult to achieve or maintain, particularly as the complexity of an integrated circuit (IC) increases. With the increasing functionality of modern electronic devices, there are corresponding increases in the demands being placed on memory bandwidth. An economical solution for increasing memory bandwidth would therefore be beneficial, particularly where such a solution could be adapted for existing memory devices.

Accordingly, the present disclosure introduces a self-timed strobe generating circuit and a method of using such a circuit cooperatively with, for example, a multi-strobe random access memory device to increase memory bandwidth. In particular, one aspect of the disclosure includes providing a clock pulse to a first strobe input of the memory for initiating a first memory access. The data input may be determined based on a state of a SELECT signal generated by the self-timed strobe generating circuit. The memory generates a strobe ready signal after the first memory access, which results in the self-timed circuit generating a second strobe signal, after inherent timing delays in the circuit. This delayed strobe signal, in turn, is provided to a second strobe input of the memory during the clock pulse for initiating a second memory access in the same clock cycle, thus doubling the stand-alone access bandwidth of a like memory operating without the self-timed circuit.

In certain embodiments, the self-timed circuit includes first and second standard flip-flops, or other sequential elements of similar functionality, each having a clock input, a data input and a data output. The clock input of the first flip-flop receives the clock pulse from a system clock that is used to initiate a first memory access while a SELECT signal is low. The data output of the first flip-flop transmits a high SELECT signal in response to the rising edge of the clock pulse. The data output of the first flip-flop is further connected to the data input of the second flip-flop for transmitting the high SELECT signal thereto. The clock input of the second flip-flop receives a strobe ready signal from the memory after the first memory access. In response to the rising edge of the strobe ready signal and the high SELECT signal, the data output of the second flip-flop transmits a second strobe signal to initiate a second data access for the memory during the same clock pulse.

Further aspects of the present disclosure will be more readily appreciated upon review of the detailed description of its various embodiments, described below, when taken in conjunction with the accompanying drawings, of which:

FIG. 1 is a circuit diagram of an exemplary circuit for generating a self-timed strobe signal according to the present disclosure;

FIG. 2 is a timing diagram for the circuit of FIG. 1 ;

FIG. 3 is a circuit diagram for a first exemplary embodiment of a random access memory controlled by the circuit of FIG. 1 ; and

FIG. 4 is a circuit diagram for a second exemplary embodiment of a random access memory controlled by the circuit of FIG. 1 .

Referring now to FIGS. 1-4 , wherein similar components of the present disclosure are referenced in like manner, various embodiments of a self-timed circuit and methods of using the same with a memory to increase memory bandwidth are disclosed.

Turning now to FIG. 1 , there is depicted one embodiment of a self-timed circuit 100 for allowing increased memory bandwidth access. In this embodiment, the self-timed circuit 100 includes first and second flip-flops 102 , 103 , that may both be well-known timing components commonly referred to as a D flip-flop. Although the following descriptions reference the use of D flip-flops, it should be readily apparent to one of skill in the art that various combinations of devices may be used to accomplish the functionality of the self-timed circuit 100 described herein For example, either or both flip-flops 102 , 103 may be known J-K or R flip-flops with appropriate connections, and perhaps additional circuit components, to achieve the memory access timing depicted in FIG. 2 . Components other than flip-flops may also be used.

A standard D flip-flop has a data input (D), a clock input (CLK) and a data output (Q). The output Q always takes on the state of the input D at the moment of a rising clock edge on CLK, for positive-edge triggered versions of the flip-flop device. When the input to CLK falls to logic 0, the output Q can change states. When the input to CLK is at either the high or low level, the D input signal has no effect at the output Q. The D flip-flop may also have a direct clear input (CLRZ) for clearing the output Q upon receipt of a particular signal thereto.

The self-timed circuit 100 accepts an input of system clock (also referred to herein as a first strobe signal or STRB 0 ) at the clock input CLK of a first flip-flop 102 . The CLK signal may be generated from an external system clock that puts out a periodic series of clock pulses at a frequency (F), commonly measured in megahertz (MHz).

In order to perform two memory accesses per clock cycle on a two strobe RAM, the STRB 0 signal may be communicated to a first strobe input of the RAM, for performing a first memory access, as well as to the clock input CLK of the first flip-flop. The output Q of the first flip-flop 102 is referred to as a SELECT signal that is used in conjunction with STRB 1 to perform a second memory access in a single clock pulse. In particular, the state of the SELECT signal may be used to select appropriate data and address inputs for the memory from more than one data and address input lines of the memory. A second strobe signal is generated at the data output Q of the second flip-flop 103 in response to a strobe ready signal generated after the first memory access. The second strobe signal, and additional associated control signals generated by the first and second flip-flops 102 , 103 , together enable the second memory access after the first memory access and within the time of the same clock pulse.

In the particular configuration shown in FIG. 1 , the two flip-flops 102 , 103 may be configured as follows. The data input D of the first flip-flop 102 receives a constant logic high signal (VDD) so that the data output Q will always be logic 1 on the positive edge of a clock pulse STRB 0 . The clock input CLK of the first flip-flop 102 receives STRB 0 from a system clock, used in conjunction with a low SELECT signal to perform a first memory access. The data output Q of the first flip-flop 102 generates a high SELECT signal in response the positive edge of STRB 0 , and after a brief propagation delay, the high SELECT signal is used by a memory to select a data input for a second memory access in conjunction with the STRB 1 signal. The data output Q of the first flip-flop 102 is also connected to the data input D of the second flip-flop 103 for transmitting the high SELECT signal thereto. The clock input CLK of the second flip-flop 103 receives a strobe ready (SR) signal from the memory after the first memory access. In response to the rising edge of the SR signal, in conjunction with the high SELECT signal, the data output Q of the second flip-flop 103 then generates a second strobe signal (STRB 1 ) used by the memory conjunction with the high SELECT signal to initiate a second data access during the same clock pulse of STRB 0 . Particular interactions of the inputs and outputs of the self-timed circuit 100 in conjunction with a memory are described further below with respect to FIGS. 3-4 .

The self-timed circuit 100 may also include an inverter 104 and a buffer 106 for providing clear signals to the first and second flip-flops 102 , 103 . The input of the inverter 104 is connected to the data output Q of the second flip-flop 103 for receiving and inverting the strobe signal STRB 1 . The output of the inverter 104 is connected to the clear input CLRZ of the first flip-flop 102 for transmitting a clear signal (STRB 1 _N) that is low when STRB 1 is high. The output of the inverter 104 is also connected to an input of the buffer 106 for transmitting the clear signal (STRB 1 _N) thereto. The output of the buffer 106 , in turn, is connected to the clear input (CLRZ) of the second flip-flop 103 for transmitting a second delayed clear signal (STRB 1 _N_DEL) to clear the output Q of the second flip-flop 103 after an appropriate delay. Components to be used as the buffer 106 may be selected based on a desirable propagation delay of the outputs of such component.

FIG. 2 shows a timing diagram of the inputs to and outputs from the self-timed circuit 100 of FIG. 1 , including associated exemplary waveforms. Although the waveforms are displayed as substantially square waveforms with positive clock pulses, any suitable signaling may be used with a variety of known components in various equivalent configurations to achieve the timing of the signals displayed in FIG. 2 .

As shown therein, the rising edge of the clock pulse STRB 0 , in conjunction with constant input VDD, causes the output Q (or SELECT) of the first-flip 102 to go high after a propagation delay associated with the first flip-flop 102 . The memory 110 also responds to the rising edge of STRB 0 and generates SR after a strobe cycle time associated with the memory 110 , during which time a first memory access may be performed. The rising edge of SR, in turn, is provided to the CLK input of the second flip-flop 103 , which responds by generating a STRB 1 signal in conjunction with the SELECT input that is now high. The rising edge of STRB 1 serves not only to clear the first flip-flop 102 via the inverter 104 using STRB 1 _N, but also to clear itself after a certain delay associated with the second buffer 106 using STRB 1 _N_DEL. The rising edge of STRB 1 will also cause the memory 110 to generate a second rising edge on SR after a second memory access cycle time, but this has no effect on STRB 1 since both SELECT and STRB 1 are low by that point, as shown in FIG. 2 .

FIG. 3 shows a first embodiment of the self-timed circuit 100 shown in FIG. 1 in communication with a memory 110 . The memory 110 is depicted as a simplified version of a RAM, but the techniques described herein are applicable and can be extended to work with any multi-strobe or multi-clock RAM device, including SRAMs, DRAM, SDRAMs and the like which may have additional numbers of the data inputs and data outputs than as shown. The following descriptions of memory 110 in conjunction with FIG. 3 are applicable to the embodiment in FIG. 4 as well, and will not be repeated in the description of FIG. 4 , where only the differences in the two embodiments will be referenced.

In the embodiment shown in FIG. 3 , the inputs of the memory 110 are two strobe inputs (STRB 0 and STRB 1 ) each for selecting an associated internal memory register (not shown) that receive address information through input A and/or data information through input D to be stored or retrieved from the address.

The memory 110 may further include a write enable (WE) input (not shown), the function and operation of which is well known in the art. A logic high on WE indicates a write operation, during which data on D is written into a memory location, whose address is specified by A on the rising edge of STRB 0 or STRB 1 . A logic low on WE indicates a read operation, during which stored data in the memory 110 whose address is specified by A, is then output on the memory output Q.

The data corresponding to DIN 0 , DIN 1 , ADDR 0 and ADDR 1 used for memory read/write functions may be handled by one or more multiplexers 108 that are responsive to the SELECT signal to select the data and addressing information received from external computing components (not shown) to pass to inputs D and A of the memory 110 in the following manner. A rising edge on STRB 0 triggers a logic high on SELECT that is used to choose registered versions of DIN 1 and ADDR 1 as inputs into the memory corresponding to STRB 1 , identified as input “1” on the input multiplexers 108 . A rising edge on STRB 1 triggers a logic low on SELECT that is used to choose DIN 0 and ADDR 0 as inputs into the memory corresponding to STRB 0 , identified as input “0” on the input multiplexers 108 . STRB 1 is also used to register the memory output Q substantially halfway through a clock cycle.

The memory 110 further has two memory select inputs (MS 0 and MS 1 ) for recognizing a signal to the first or second strobe inputs (STRB 0 and STRB 1 , respectively) of the memory 110 . A logic one on the MS 0 input indicates that any signal to the strobe input STRB 0 is to be recognized and a logic zero on MS 0 indicates that any input to STRB 0 is to be ignored. Signaling on MS 1 operates likewise with respect to the strobe input STRB 1 .

Another output from the memory 110 is denoted as strobe-ready (SR). SR will output a logic high pulse when the strobe cycle time of the memory is met, thus indicating that a next memory access may be performed. Such memory accesses or read/write operations are thus enabled at each rising edge of either strobe STRB 0 or STRB 1 . The SR signal is used by the self-timed circuit 100 in the manner described with respect to FIG. 1 .

In the embodiment of FIG. 3 , the STRB 0 input receive the system clock signal STRB 0 . The STRB 1 input is tied to the data output Q of the second flip-flop 103 of the self-timed circuit 100 . The MS0 and MS1 inputs both receive a constant input signal VDD. The SR signal output by the memory 110 is communicated to the clock input of the second flip-flop 103 of the self-timed circuit 110 .

FIG. 4 shows an alternate embodiment of the self-timed circuit 100 of FIG. 1 in communication with the memory 100 . In this embodiment, only the MS0 input is tied to a constant signal VDD and the MS1 input is tied to the SELECT signal output by the self-timed circuit 100 of FIG. 1 . The SR signal output by the memory 110 is communicated to the clock input of the second flip-flop 103 of the self-timed circuit 110 , and is further communicated to the STRB 1 input of the memory 110 . This configuration may be used if the second memory access in the cycle needs to happen earlier than with the configuration shown in FIG. 3 .

In either of the manners depicted in FIG. 3 and FIG. 4 , as well in other equivalent embodiments that are contemplated but not particularly described herein, the self-timed circuit 100 achieves an effective doubling of memory access bandwidth of the memory 110 by allowing two read/write operations in a single clock cycle. This further is achieved without any resulting burst mode uncontrolled memory accesses that could happen, for example, if SR was simply directly tied to a MS input to provide a second strobe input STRB 1 . Such burst mode accesses can occur particularly as the speed of the memory 110 used increases.

In all embodiments described herein, it is readily contemplated that negative (or falling) edge triggering could be employed in place of positive (or rising) edge triggering, with appropriate inverters and the like placed in the system. Other equivalent configurations for any of the functions described herein are readily contemplated as being part of the present disclosure.

Although the best methodologies of the invention have been particularly described in the foregoing disclosure, it is to be understood that such descriptions have been provided for purposes of illustration only, and that other variations both in form and in detail can be made thereupon by those skilled in the art without departing from the spirit and scope of the present invention, which is defined first and foremost by the appended claims.