Complete refresh scheme for 3T dynamic random access memory cells

[0002] The invention relates to dynamic random access memory (DRAM), particularly to three-transistor (3T) DRAM.

[0003] Various forms of static and dynamic semiconductor storage cells are known in the art. Static cells (usually 6T-SRAM) continue to store data for as long as power is applied to them. In contrast, a dynamic storage cell (e.g., 1T-DRAM, 3T-DRAM or 4T-DRAM) must be periodically refreshed or it loses the stored data. Static cells are generally faster, consume less power and have lower error rates, but have the disadvantage of requiring more space on a semiconductor chip. Generally speaking, refreshing scheme on the dynamic storage cells only creates the pseudo static storage cells because the external access command is unpredictable and can't be executed when the heavy external access occurs and interferes with the internal refresh operation. One way to solve the access/refresh conflict problem is to insert the refresh operation after the external access operation in the same clock cycle but it causes more cycle time or poorer performance.

[0004] Various conventional circuitries use the dynamic storage cells but provide the static storage effect to reduce the space on the semiconductor chip. 4-T SRAM cell is given a higher leaky current from the pre-charged bit line to the storage node via the pass transistor to retain the data. 1T-DRAM is the smallest in area but the capacitor included in the memory cell has a three-dimensional configuration that increases the process numbers and the production cost. Moreover, because of the required destructive read and write-back, access time is increased when compared with a case in which 6T-SRAM is employed. As such, these are not suitable for system-on-chip (SOC) applications since most of these SOC applications use the generic process provided by the majority of the foundries.

[0005] A three-transistor DRAM (3T-DRAM) cell (see FIG. 1) is not required to conduct the process to form a capacitor having a three-dimensional structure and can be fabricated in the same transistor processes as for the memory including 6T-SRAM. The access speed can be similar to the 6T-SRAM since the 3T-DRAM read operation is non-destructive. However, 3T-DRAM has the drawback of needing more frequent refresh operations to retain the data due to a small storage capacitance in the memory cell. This increases the possibility of conflict between an access to the 3T-DRAM cell for external access and an internal access for the refresh operation. Thus, for 3-T DRAM, a need exists to have internal refresh operation running independently in the memory block no matter what the external access is, such that the speed advantage of 6T-SRAM and the area advantage of DRAM can be obtained at the same time.

[0006] The accompanying drawings which are incorporated in and form a part of this specification, illustrates embodiments of the invention and together with the description, serve to explain the principles of the invention:

[0007] FIG. 1 shows a three-transitor dynamic random access memory (3T-DRAM) cell that is deployed in a DRAM refreshing scheme in accordance with one embodiment of the invention.

[0008] FIG. 2 shows a sub-array of a 3T-DRAM that is deployed in a DRAM refreshing scheme in accordance with one embodiment of the invention.

[0009] FIG. 3 shows a timing diagram of a DRAM refreshing scheme for a 3T-DRAM in accordance with one embodiment of the invention.

[0010] FIG. 4, shows a clock generator for a DRAM refreshing scheme of a 3T-DRAM in accordance with one embodiment of the invention.

[0011] Reference is made in detail to the preferred embodiments of the invention. While the invention is described in conjunction with the preferred embodiments, the invention is not intended to be limited by these preferred embodiments. On the contrary, the invention is intended to cover alternatives, modifications and equivalents, which may be included within the spirit and scope of the invention as defined by the appended claims. Furthermore, in the following detailed description of the invention, numerous specific details are set forth in order to provide a thorough understanding of the invention. However, as is obvious to one ordinarily skilled in the art, the invention may be practiced without these specific details. In other instances, well-known methods, procedures, components, and circuits have not been described in detail so that aspects of the invention will not be obscured.

[0012] Referring now to FIG. 1, a three-transistor dynamic random access memory (3T-DRAM) cell 100 is shown for facilitating description of a 3T-DRAM refreshing scheme in accordance with various embodiments of the invention. In accordance with the present invention, an SRAM compatible device is designed using cells such as 3T-DRAM cell 100, whose transistors can be the same type of devices or any combination.

[0013] As shown, transistors 101, 102 are pass transistors that couple respectively with a write bit line (WBL) 150 and a read bit lines (RBL) 160. Activation of transistors 101 and 102 are respectively controlled by a write word line (WL) 170 and a read word line (RL) 180. A third transistor 103 provides the gate capacitance to retain the data and the cell current for the read operation.

[0014] As will be described below, the design of 3T-DRAM cell 100 is incorporated into a DRAM refreshing scheme in accordance with various embodiments of the invention.

[0015] Referring now to FIG. 2 in view of FIG. 1, a 3-T DRAM sub-array 200 is shown in accordance with one embodiment of the invention. Sub-array 200 comprises a sense amplifier (SA) 290, 3T-DRAM cells 210-230, WBL 250, RBL 260, WL's 271-272, and RL's 281-282. Although not explicitly shown here in FIG. 2, many other cells beside cells 210-230 are understood to be parts of sub-array 200.

[0016] Coupled to both WL 271 and RL 281, cell 210 is coupled to SA 290 via WBL 250 and RBL 260. Similar to cell 100 (as shown in FIG. 1), cell 210 as shown in FIG. 2 comprises three transistors 211-213. Transistors 211-212 are pass transistors that couple with WBL 250 and RBL 260. Activations of transistors 211 and 212 are respectively controlled by WL 271 and RL 281. Transistor 213 provides the gate capacitance to retain the data and the cell current for the read operation.

[0017] Coupled to both WL 272 and RL 282, cell 220 is coupled to SA 290 via WBL 250 and RBL 260. Similar to cell 100 (as shown in FIG. 1), cell 220 as shown in FIG. 2 comprises three-transistors 221-223. Transistors 221-222 are pass transistors that couple with WBL 250 and RBL 260. Activations of transistors 221 and 222 are respectively controlled by WL 272 and RL 282. Transistor 223 provides the gate capacitance to retain the data and the cell current for the read operation.

[0018] In the present embodiment, an external memory access to a cell in sub-array 200 comprises a read followed by a write.

[0019] As an example, for a write operation to cell 210, WL 271 is asserted to couple WBL 250 to the gate capacitance of transistor 213. Thus, the voltage on WBL 250 will be written into the gate capacitance of transistor 213 and the charge will be retained for a certain time after de-asserting WL 271.

[0020] For a read operation to cell 210, RL 281 is asserted to couple RBL 260 to the drain node of transistor 213. Transistor 213 provides one-way current to RBL 260 for the data sensing when transistor 213 is on. Otherwise, no current is provided to RBL 260 when transistor 213 is off.

[0021] As shown, RBL 260 coupled to the numerous memory cells is coupled to sense amplifier 290. Sense amplifier 290 can sense, amplify the voltage/current change on RBL 260. In turn, RBL 260 is decoupled from sense amplifier 290 such that RBL 260 may be pre-charged after sense amplifier 290 is activated. In the mean time, sensed data will be latched in sense amplifier 290 and transferred to WBL 250 or to an external data bus.

[0022] On the other hand, the sensed data may be updated by the external data bus for the external write operation. After writing the data to the memory cells of sub-array 200, WBL 250 may be pre-charged and sense amplifier 290 may be pre-charged and equalized.

[0023] As understood herein, since the sensed data has been latched and isolated from RBL 260, for the memory cells of sub-array 200 (i.e., memory cells that share RBL 260 and WBL 250), a memory cell with the asserted read word line RL does not interfere with the other cell with the asserted write word line WL.

[0024] As mentioned above, the possibility of the conflict between an external memory access and an internal memory refresh is increased due to the small storage capacitance. Thus, sub-array 200 is likely to require an internal refresh operation every clock cycle. For the memory array operation, a memory cell is only accessed in a portion of the clock cycle. The rest of the clock cycle can be non-memory array operations such as address decoding, data output etc.

[0025] For high-density memory block, the memory array access time may occupy only a small portion of the clock cycle. In the present embodiment of the invention, the memory array that includes sub-array 200 is preferably arranged in a way such that the memory array operation only takes half or less of the clock cycle. That is, the memory array can be accessed twice in one clock cycle. In other words, both the external access and the internal refresh combined can be executed in one clock cycle period. This can be done by partitioning the memory array into the small grains of the sub-arrays (e.g., sub-array 200) such that the asserted sub-array can be freed up for the next internal refresh operation after the external access while the data registered at the local sense amplifier transfer to the output buffer.

[0026] On the other hand, the memory array is preferably arranged in a way such that when the write word line is asserted, the numerous memory cells will be written from the write bit lines. Since the write data are less than the number of the memory cells accessed, some of them are required to do the read-then-write-back operation similar to the internal refresh operation. Therefore, whether read or write external access is performed, the memory operation can be partitioned in two steps (read and write) and executed in a read-then-write order.

[0027] For the external read access, the write portion is similar to the internal refresh operation. Or, the external read access can be skipped without losing the data.

[0028] For the external write access to those memory cells not written by the external data bus, the write portion is similar to the internal refresh operation. Since the refresh operation is a read-then-write-back operation, the external access operation and the refresh operation can be interleaved. That is, the memory access can be operated in an order of the external read, external write/internal refresh read and internal refresh write, wherein external write and internal refresh read overlap. The internal refresh write can be extended to the next clock cycle before the external read occurs.

[0029] For 3T-DRAM memory cells, separate word line and bit line are used respectively for the read and write. As such, for example, an external write to cell 210 and an internal refresh read to cell 220 can be executed at the same time without interfering each other. Thus, memory speed is not degraded and the refresh operation can be executed in every clock cycle period no matter what the external operation is.

[0030] Referring now to FIG. 3 in view of FIG. 2, a timing diagram 300 is shown for 3-T DRAM sub-array 200 in accordance with one embodiment of the invention. To interleave the external access and internal refresh, a clock cycle is first partitioned into time slots. In turn, operations to be performed are assigned to their respective pre-defined time slots. As understood herein, a clock cycle partition need not be limited to the clock cycle partition as shown here. For example, in another embodiment, a clock cycle is partitioned into a different number of time slots than the one shown in timing diagram 300.

[0031] MCLK 310 is the external clock. RASCK 320, PRCCK 330 and SENCK 340 are the internally generated clock signals derived from MCLK 310 or from each other. RASCK 320, PRCCK 330 and SENCK 340 have two pulses in one MCLK 310clock cycle. RASCK 320 can be generated from the rising edge of MCLK 310 or the rising edge of SENCK 340 when MCLK 310 is high.

[0032] The first pulse of PRCCK 330 can be obtained from RASCK 320 after the predetermined delay line. The second pulse of PRCCK 330 can be obtained either from RASCK 320 after the pre-determined delay line or after the MCLK 310 high goes low depending on which is later. SENCK 340 can be obtained from PRCCK 330 after the pre-determined delay line. Each delay line is associated with the timing of the memory operation, and is furthermore adjusted to guarantee the proper memory operation.

[0033] Specifically, by gating with MCLK 310, RASCK 320, PRCCK 330 and SENCK, 7 independent time slots can be created in the present embodiment.

[0034] As shown, time slot 1 is from the rising edge of MCLK 310 to the rising edge of RASCK 320.

[0035] Time slot 2 is from the rising edge of RASCK 320 to the rising edge of PRCCK 330.

[0036] Time slot 3 is from the rising edge of PRCCK 330 to the rising edge of SENCK.

[0037] Time slot 4 is from the rising of SENCK 340to the rising edge of the second pulse of RASCK 320.

[0038] Time slot 5 is either from the rising edge of the second pulse of RASCK 320 to the rising edge of the second pulse of PRCCK 330 or the falling edge of MCLK 310 to the rising edge of the second pulse of PRCCK 330.

[0039] Time slot 6 is from the rising edge of the second pulse of PRCCK 330 to the rising edge of the second pulse of SENCK.

[0040] Time slot 7 is from the rising edge of the second pulse of SENCK 340 to the next rising edge of MCLK 310.

[0041] In terms of memory cell operations, time slot 2, 3, 4 and 5 are dedicated for external access operation. Time slot 5, 6, 7, 1 (next cycle), 2 (next cycle) are dedicated for internal refresh operation. Both start from the rising edge of RASCK 320 to the rising edge of the next following PRCCK 330. As understood herein, the memory array can be arranged so that time slot 6 and 7 as well as 3 and 4 can be reduced. Thus, inserting the refresh operation after the external operation will not degrade the speed performance.

[0042] More specifically, at time slot 1, the external addresses are registered and decoded.

[0043] At time slots 2 and 3, read word line (RL) 281 is asserted associated with the external addresses no matter the external operation is read or write.

[0044] At time slot 4, read sense amplifier 290 latches voltage differential at read bit line (RBL) 260. In turn, RBL 260 may be pre-charged to a pre-defined voltage level.

[0045] At time slots 4 and 5, the write word line WL associated with a cell being accessed externally can be asserted for the external write operation. For the external read operation, this operation can be skipped. For the refresh operation, the refresh addresses can be registered and decoded before time slot 5. For example, in the scenario shown in FIG. 3, when performing an external access to cell 210, WL 271 can be asserted for the external write operation. For the external read operation, this operation can be skipped. For the refresh operation to be performed in the same clock cycle in cell 220, the refresh addresses can be registered and decoded before time slot 5.

[0046] At time slots 5 and 6, RL 282 associated with the refresh addresses can be asserted after RBL 260 is pre-charged at time slot 4. As shown here, the external write operation is overlapped with the internal refresh read operation at time slot 5. Since the read data have been latched at sense amplifier 290, any change on RBL 260 won't affect the data being written to the memory cell. In the case where row addresses are the same for both the external write and refresh operation, the refresh operation on this row can be skipped without losing any data.

[0047] At time slot 6, WL 271 associated with the external addresses is de-asserted. The write bit line WBL may be pre-charged and the sense amplifier may be pre-charged and equalized.

[0048] At time slot 7, read sense amplifier 290 latches/amplifies the data on RBL 260 and RBL 260 may be pre-charged at time slot 7 and the next cycle's time slot 1. Meanwhile, the WL 272 associated with the refresh addresses will be asserted for the write-back operation for time slot 7 and the next cycle's time slots 1 and 2.

[0049] At time slot 3 of the next cycle, WL 272 associated with the previous refresh addresses will be de-asserted. WBL 250 may be pre-charged and sense amplifier 290 may be pre-charged and equalized. Thus, the external access operation and the internal refresh operation have been executed in one clock cycle period though part of the refresh operation has been extended to the next cycle. As understood herein, although part of the refresh operation extends to the next cycle, both the external access operation and refresh operation can be initiated within one clock cycle period. As such, both operations can be considered as being executed in one clock cycle period.

[0050] Referring now to FIG. 4 in view of FIGS. 2 and 3, a block diagram of a clock generator 400 for a 3-T DRAM is shown in accordance with one embodiment of the invention.

[0051] For clock generator 400, flip-flops 401, 402, 403 and 404 are D-type flip-flops with the RESET input pin. A flip-flop 405 is D-type flip-flop with the SET and RESET input pins. Input D receives the data and transfers to the output Q when CK changes from low to high. Input RESET forces output Q to logic low when it is asserted regardless with CK. On the contrary, input SET pin forces output Q to logic high regardless with CK. Delay lines 410, 411, 412, 413 and 414 are the non-inverting delay lines. Additionally, an OR gate 420 and a switch 430 are also shown here.

[0052] For simplicity, the logics initializing the states of each flip-flop is not drawn in FIG. 4. However, after the initialization or the power-on reset, RASCK 320, PRCCK 330 and PRCCKEN 441 are reset to logic low. SENCK 340 is set to logic high. When MCLK 310 changes from logic low to logic high and propagates through DELAY A 410, RASCK 320 will be set to logic high via D flip-flop 401. In turn, RASCK 320 will reset SENCK 340 to logic low. The delay time from DELAY A (time slot 1 in FIG. 3) is understood to be greater than the time to latch and decode the external addresses. The output Q of D flip-flop 401 also sets PRCCKEN 441 to logic high after the delay from DELAY E when SENCK 340 is logic low. DELAY E is used to guarantee signal PRCCKEN 441 can be set to logic high after the first pulse of RASCK 320 is asserted.

[0053] Continuing with FIG. 4, once the signal PRCCKEN 441 is asserted, switch 430 is closed so the first pulse of RASCK 320 can propagate to D flip-flop 403 and cause PRCCK 330 to change from low to high. In turn, PRCCK 330 comes back to D flip-flop 401 and 402 to reset RASCK 320. PRCCK 330 will set SENCK 340 to logic high after propagating through DELAY D. In turn, SENCK 340 will come back to D flip-flop 403 to reset PRCCK 330 to logic low.

[0054] As understood herein, the combined delay time for DELAY C and DELAY D (see time slots 2 and 3 shown in FIG. 3) is understood to be greater than the time to turn on the read word line and build up enough voltage or current change on the read bit line before the sense amplifier is activated.

[0055] Once SENCK 340 is set to high, the signal PRCCKEN 441 will be reset to logic low and force the switch 430 to be open. At the same time, SENCK 340 propagates through DELAY B and asserts RASCK 320 via D flip-flop 402. The delay time from DELAY B (time slot 4 shown in FIG. 3) is understood to be greater than the time required to pre-charge the read bit line. The second pulse of RASCK 320 can not set PRCCK 330 high when the switch 430 is open. PRCCKEN 441 will be set to logic high again once MCLK 310 changes from high to low. The delay time from the rising edge of the second pulse of RASCK 320 to the second pulse of PRCCK 330 is either the delay from DELAY C or the elapsed time from the second rising edge of RASCK 320 to the falling edge of MCLK 310 and depends on which comes later.

[0056] As understood herein, the above delay and the delay time for DELAY B together are understood to be greater than the time required to write the data from the write bit line to the memory cell. Again, SENCK 340 is set high by the PRCCK 330 after the delay of DELAY D and remains the logic high until RASCK 320 is asserted in the next clock cycle.

[0057] Also as understood herein, when the external clock speed is lower than expected (This usually happens in system-on-chip application), the time (time slots 5 and 7 in FIG. 3) allowed for the memory cell's read and write can be extended automatically by the external clock cycle time except for the first memory read time (i.e., time slots 2 and 3 shown in FIG. 3). This approach can also be adjusted to give more margin for the memory read by tuning the delay time of DELAY C and DELAY D. Thus, each time slot in one clock cycle can be optimized to give the best margin for the memory read and write for both external access and internal refresh.

[0058] The foregoing description of specific embodiments of the invention have been presented for purposes of illustration and description. They are not intended to be exhaustive or to limit the invention to the precise forms disclosed. Obviously, many modifications and variations are possible in light of the above teaching. The embodiments were chosen described in order to explain the principles and the application of the invention, thereby enabling others skilled in the art to utilize the invention in its various embodiments and modifications according to the particular purpose contemplated. The scope of the invention is intended to be defined by the claims appended hereto and their equivalents.