Patent Grant
9153298
The present invention generally relates to the field of semiconductor memory devices, and more particularly, to semiconductor memory devices configured for stacked arrangements.
In many multichip arrangements, devices (e.g., integrated circuits, dice, or chips) may be placed in side to side configurations. Examples of these types of arrangements include single in-line memory module (SIMM) and dual in-line memory module (DIMM) configurations. Other multichip arrangements include multichip module (MCM) structures of chips that may share a common package. However, such conventional multichip approaches may not be suitable for relatively large I/O and density structures. For these applications, vertical stacking using through-silicon vias (TSVs) and die-to-die vias represents a more promising technology. However, these technologies may be limited by various drawbacks and challenges related to on-chip circuitry for the stacked devices.
Reference will now be made in detail to particular embodiments of the invention, examples of which are illustrated in the accompanying drawings. While the invention will be described in conjunction with particular embodiments, it will be understood that they are not intended to limit the invention to these 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 present invention, numerous specific details are set forth in order to provide a thorough understanding of the present invention. However, it will be readily apparent to one skilled in the art that the present invention may be practiced without these specific details. In other instances, well-known methods, procedures, components, etc. have not been described in detail so as not to unnecessarily obscure aspects of the present invention.
Some portions of the detailed descriptions which follow are presented in terms of processes, procedures, block arrangements, logic blocks, functional blocks, processing, and/or other symbolic representations of operations or device arrangements. These descriptions and representations are generally used by those skilled in the relevant arts to effectively convey the substance of their work to others skilled in the art.
Particular embodiments can advantageously provide circuits, methods, and structures, to facilitate circuit operation for stacked memory devices. For example, particular embodiments can advantageously provide approaches that allow for devices to be manufactured using a same mask set and layout patterns regardless of a device's ultimate position in a stack of memory devices.
Various embodiments relate to stacked memory devices, such as dynamic random access memories (DRAMs), static RAMs (SRAMs), electrically erasable programmable read-only memories (EEPROMs), resistive RAMs (ReRAMs), field-programmable gate arrays (FPGAs), and content addressable memories (CAMs). For example, identifiers that indicate a stack position of a device may be assigned or otherwise determined, and used by each memory device to make a number of adjustments on that device. In one embodiment, a self-refresh rate of a DRAM can be adjusted based on the SPID of that device. In another embodiment, a latency of a memory device can be adjusted based on the SPID. In another embodiment, internal regulation signals may be connected with other devices by way of through-silicon vias (TSVs) and control circuitry, such as when operating in a test mode. In another embodiment, adjustments to internally regulated signals can be made based on the SPID of a particular device. In another embodiment, serially connected signals can be controlled based on a chip SPID (e.g., an even or odd stack position), and whether the signal is an upstream or a downstream type of signal. These and various other embodiments will be discussed in more detail below.
Referring now to
Interconnections between one or more of the DRAMs and memory controller 102 can be accommodated by use of through-silicon vias (TSVs) 110 and/or die-to-die vias 112. TSVs 110 may also include thermal TSVs (TTSVs) for use in managing heat dissipation, and reducing temperature differences, within structure 100. In addition, interlayer dielectric (ILD) and signal layers may also be included in the regions between the DRAMs. However, pad arrangements and associated control circuitry on each chip in particular embodiments can allow for simplified signal layer routing, or in some cases the complete removal of such a signal layer.
While memory controller 102 is shown, any suitable interface chip or device is contemplated in certain embodiments. Any type of memory interface chip, processor, type of controller, switch fabric, other memory chip, optical processing layer, or package substrate, can be utilized in place of memory controller 102 in some applications. Further, while pads described herein may include metal pads or other connections to TSVs, particular embodiments may also not include actual metal pads, but rather other types of connections to TSVs. For example, pads can include connections below a surface of a semiconductor device, such as when a connection layer (e.g., layers of metallization for signal connections) is used between a TSV and the semiconductor device surface. However, metal pads on a semiconductor surface may be desirable so that the probe testing as part of a test flow (e.g., including final test, etc.) can be performed to determine known good die (KGD) prior to placing the die in a stack to form a stack the memory devices. In any event, “pads” as described herein can represent any interconnect position on a semiconductor device. As will be discussed in more detail below, testing may also be performed once the memory devices are secured in a stack.
TSVs can be formed using any suitable fabrication process. For example, a front-end-of-line (FEOL) process that creates active devices can proceed or follow TSV formation. In this case, wafer thinning to expose the TSVs on a wafer backside can occur with a back-end-of-line (BEOL) process that creates conductive lines for power and signals between devices and off-chip connections. In another example, TSV formation can occur from the wafer backside after device wafer fabrication and thinning. TSVs and/or die-to-die vias can be filled with any suitable conductive material (e.g., doped polysilicon, tungsten, copper, etc.). In any event, particular embodiments are suitable for stacked memory devices utilizing TSV, die-to-die via, as well as other chip interconnections.
TSV processing steps can include via etch or drill, an insulator and seed layer deposition, and via fill. The via etch step can include a deep reactive ion etching (DRIE) process, or a laser etching process using ultraviolet lasers. The insulating layer may include SiO2, and may be deposited using chemical vapor deposition (CVD) techniques. Organic layers (e.g., parylene) may also be utilized as an insulating layer in some embodiments. Also, a diffusion barrier layer (e.g., TiN) may follow deposition of the insulating layer. A seed layer (e.g., TiN and Cu) may then be deposited (e.g., via CVD techniques) to facilitate metallization in the TSV. The TSV may then be filled, such as by depositing tungsten using a CVD process, or electrochemical plating for copper metallization. Of course, other methods and materials as to TSV, die-to-die via, and/or other chip interconnections can also be employed in particular embodiments.
In addition, while TSVs and die-to-die vias are discussed in various examples herein, other types of connections suitable for interconnections among stacked devices can also be employed in particular embodiments. For example, one or more signal layers, interposers, or other substrates can be utilized between memory devices in the stack. Also, TSVs may be mated to microbumps that may then be connected to other TSVs and/or a signal layer or substrate. Thus, particular embodiments are suitable to a variety of types of interconnections between chips that are stacked on top of one another.
Particular embodiments are suited to chips or devices that may not utilize different mask patterns based on an ultimate position in a stack of memory devices. In this way, high-volume manufacturing of individual memory devices can occur without regard to possible applications for the memory devices that might include stacked arrangements. This is accomplished by design techniques that allow each memory device to adapt dynamically to any position in the stack of memory devices.
Referring now to
The particular example of
In particular embodiments, multiple pads (e.g., different pads or external interconnect positions) that are allocated to a same signal can be configured on a memory device based on a stack position. In one embodiment, a semiconductor memory device can include: (i) a plurality of pads allocated to a signal on the semiconductor memory device; (ii) a stack position identifier for identifying a position of the semiconductor memory device in an aligned vertical stack of a plurality of semiconductor memory devices; and (iii) a pad selector configured to select one of the plurality of pads to be connected to the signal on the semiconductor memory device in response to the stack position identifier, where the pad selector is configured to disconnect each of a remaining of the plurality of pads from the signal based on the stack position identifier.
Referring now to
When viewed in conjunction with
Similarly, as shown in 300-1, a next in stack DRAM-1 can have a next in line pad (e.g., second pad from the left) enabled, while others in the mapping are disabled. Note also that the leftmost pad in DRAM-1 may not have a die-to-die via or connection to the TSV, while other pads in the mapping do. Note also that the DQ signal described here may represent an internal data input and/or output signal, and may be a signal that would otherwise connect to the pad absent electrostatic discharge (ESD) and/or driver or other input path and/or output path control circuitry. As such, the DQ signal may represent any signal internal to the chip that may be mapped to one of eight (or any number of) pads or interconnect positions on a given memory device. As described herein, this arrangement represents one example of mapping an internal signal for connection on different external paths as a result of a stack position of the device.
As shown in 300-2, DRAM-2 may have a next in line pad enabled connection, and so on up the stack for DRAM-3 as shown in 300-3, DRAM-4 as shown in 300-4, DRAM-5 as shown in 300-5, DRAM-6 as shown in 300-6, and DRAM-7 as shown in 300-7. As shown in 300-7, each of eight pads or external connection positions 302-0, 302-1, . . . 302-7 can be mapped to a single DQ signal. In the case of DRAM-7, a rightmost pad connection may be enabled for a path between DRAM-7 and memory controller 102. As shown in 300-7, each of the remaining possible pad connections in the leftmost seven positions can be disabled as a result of control based on the stack position. In this way, each DRAM or other memory device in the stack of devices can be customized to external (i.e., inter-chip) connections by enabling different possible interconnect positions in a mapping to a common signal (e.g., a DQ signal) based on stack position.
Of course, while eight DRAMs and eight DQ pad mappings are shown in this particular example, any number of memory devices and mappings to a common on-chip signal can be accommodated in particular embodiments. In addition, the memory devices in the stack of memory devices can be oriented either right side up or upside down. Further, references to “top,” “bottom,” “up,” “down,” “upstream,” “downstream,” “on top of,” “above,” “below,” “vertical,” “adjacent,” etc., are made with reference to other devices in a stack of devices, and do not limit the invention to a particular orientation of the entire stack of devices, but are rather directed to relative positioning of the devices in a stack of memory devices, as well as any associated substrates, controllers, and the like.
Referring now to
Based on the SPID (e.g., bit values as shown in Table 1), pad selectors can be used to isolate various pads or interconnect positions from the DQ signal. This selection may effectively remove excess capacitance due to the inter-chip connections on alternative pad positions (e.g., pad positions 302-0, 302-1, . . . 302-6 in 300-7 of
In this example, pad position 302-0 may be disabled for this DQ signal unless SPID=000, representing the lowest memory device in the stack (see, e.g., 300-0 in
In addition, pad selection circuitry may be replicated for different signals. For example, such pad selection circuitry may be shared among different DQ signals on a same memory device. This is because a given selection (e.g., 1 of 8) of pad positions may be the same for each such signal (e.g., DQ0, DQ1, DQ2, . . . DQ15). Alternatively, pad selection circuitry 400 may be replicated for each mapping of common signals and pads. Further, pad selection circuitry 400 may also be used to steer a redundant pad path to replace that of a de-allocated memory device, such as where a top memory device in the stack may replace a de-allocated memory device in the stack. Also, as one skilled in the art will recognize, other digital logic or circuitry can be employed in order to achieve a similar type of decoding. In addition, various ESD considerations may also be accounted for in the design of pad selection circuitry in particular embodiments.
Referring now to
Referring now to
In particular embodiments, stack position identifiers can be automatically determined based on a sensing of the connections on pads that are mapped to a common signal. In one embodiment, a method of determining a stack position of a semiconductor memory device, can include: (i) comparing, for each of a plurality of pads allocated to a same signal on the semiconductor memory device, a first capacitance on a first pad against a second capacitance on a second pad; (ii) setting a latch if the second capacitance is greater than the first capacitance by at least a predetermined amount; (iii) counting a number of the latches that remain reset to determine a stack position identifier for the semiconductor memory device; and (iv) storing the stack position identifier in a register on the semiconductor memory device.
In this way, stack position identifiers can automatically be determined by each memory device in the stack of memory devices based on comparing capacitances between associated pads. This allows for determination of stack positions by using circuitry found on each memory device, without having to program any stack position identification in advance. Certain embodiments provide for determination of a device's position in a stack of devices after the devices have been placed in the stack. Thus, stack positions can be determined based on connections made external to the memory device, where those connections may be in furtherance of a multiple pad to same signal mapping as part of a stack memory device structure.
Referring now to
In the example circuit of
An example circuit to drive a high or low value (from CTDA) onto pad 302-A while CTEN is high, and to tri-state when CTEN is low, may include NAND gate 702-A, PMOS transistor 708-A, NOR gate 704-A, inverter 706-A, and NMOS transistor 710-A. A similar circuit that may be used to drive a high or low value (from CTDB) onto pad 302-B while CTEN is high, and to tri-state when CTEN is low, may include NAND gate 702-B, PMOS transistor 708-B, NOR gate 704-B, inverter 706-B, and NMOS transistor 710-B. During operation, CTDA and CTDB may be set to opposite values such that substantially equally differing levels are applied to pads 302-A and 302-B during the CTEN pulse.
Referring also to
At the falling edge of CTEN, an edge-triggered pulse CSEN indicating that capacitance sensing is enabled may be generated as shown in example waveform 808. For example, CSEN may be generated using delay circuit 712, inverter 714, and NOR gate 716 to generate a positive going edge-triggered pulse having a width of about that of delay circuit 712. For example, the delay of delay circuit 712 may be long enough to ensure adequate time to complete the charge sharing process between pads 302-A and 302-B. When CSEN is high (e.g., during the pulse) and its compliment is low via inverter 722, the pass gate formed by NMOS transistor 718 and PMOS transistor 720 can be enabled to allow charge sharing to occur between pads 302-A and 302-B. Because CSEN may not be triggered until a falling edge of CTEN, pass gate 718/720 may only be enabled when CTEN is low. Thus, pads 302-A and 302-B may be coupled together for charge sharing purposes only when the circuits that drive data onto the pads are tri-stated. For example, when CTEN is low, transistors 708-A and 710-A for pad 302-A, and transistors 708-B and 710-B for pad 302-B are off to place each path in tri-state.
An example of this charge sharing mechanism on pads 302-A and 302-B is shown respectively as waveforms 810-A and 810-B. In this particular example, both of pads 302-A and 302-B start out low, but pad 302-B may be charged high via transistor 708-B when CTEN is high. Thus in this example, CTDA may be low and CTDB may be high, as shown in respective waveforms 804 and 806. Once CTEN goes low and CSEN goes high to enable charge sharing between pads 302-A and 302-B, some intermediate voltage level can be established at the positive input of differential amplifier 724. As shown in example waveform 810, this intermediate voltage level after charge sharing may be higher than a middle value between the high and low levels applied, thus indicating that pad 302-B has a higher capacitance than pad 302-A.
Differential amplifier 724 may be used to compare the charge shared value input to a positive terminal against a predetermined reference level, VREF. For example, if the charge shared value is greater than VREF, then the differential amplifier output (DAOUT) may be high, but if the charge shared value is less than VREF, then DAOUT may be low. In order to ensure that differential amplifier 724 performs this comparison after the charge sharing is complete, a differential amplifier enable signal (DAEN) can be generated after a falling edge of CSEN. Thus, the falling edge of CSEN may trigger a high going pulse on DAEN by, for example, using the edge triggered pulse generation circuit that includes delay circuit 728, inverter 730, and NOR gate 732. Waveform 812 shows an example operation of DAEN, which can pulse high upon a falling edge of CSEN.
Once DAEN is high, NMOS transistor 726 may allow current to flow through and/or otherwise enable differential amplifier 724 such that DAOUT may provide an accurate output. DAEN may also be used with inverter 738 to open a pass gate formed by NMOS transistor 740 and PMOS transistor 742 to capture the value of DAOUT. When pass gate 740/742 is open by DAEN being high, which also enables differential amplifier 724, a value of DAOUT (see, e.g., waveform 814) can override the value stored in the latch formed by inverters 734 and 736. The output of latch 734/736 can be buffered by inverters 744 and 746 to provide output signal DBCH (see, e.g., waveform 816). For example, DBCH may go high to indicate that the capacitance at pad 302-B is greater than the capacitance at pad 302-A. Conversely, DBCH may be low the capacitance at pad 302-A were detected as greater than the capacitance at pad 302-B.
In this way, a capacitance comparison between two pads, or other external connection interfaces, can be performed in particular embodiments. Further, such capacitance comparison circuits may be employed as part of a test mode and/or as part of a standard power-up sequence. In certain embodiments, test modes and/or power-up sequences can include a sequenced activation for each memory device in the stack so that capacitance comparisons on a given chip are not interfered with by capacitance comparisons ongoing on a different chip. Further, such capacitance comparisons may be employed for other pad pairs, such as those that may not be mapped to a same signal. This may be useful as part of a test mode to detect external (e.g., TSV connections, etc.) on different pad or external connection pairs.
A test mode for capacitance comparisons may also be performed to determine whether a given pair of pads (e.g., pads 302-A and 302-B) have a same or matching capacitance. For example, matching capacitances can be determined by adjusting the level of VREF and/or by performing multiple tests with different values on CTDA and CTDB (e.g., first ‘1’ and ‘0’, then ‘0’ and ‘1’). In this way, external connections to TSVs on a given device can be tested by using a test mode on that memory device. Further, other types of comparisons, such as resistance or current-based, as opposed to capacitance-based, may also be employed in particular embodiments.
Referring now to
Example capacitance comparison circuits 700, as discussed with reference to
Once the count of zeros is completed, these values may be stored in SPID registers 904 to supply identifier bits SPID[2], SPID[1], and SPID[0]. Once the SPID register values are updated, SPID_Ready may be activated to indicate that the values stored in SPID registers 904 are ready for use. For example, SPID_Ready may be low to disable various circuits from using SPID values, and then may be brought high to indicate that SPID values are ready for use. An example of such zero value counting corresponding to SPID values is shown below in Table 2. Alternatively, “1” values may also be counted, or other correlations between adjacent pad capacitance determinations and ultimate SPID values can be performed. This particular example applies to the example external connection arrangement as shown in
SPID registers 904 may also be part of, or otherwise associated with, a mode register, such as a mode register on a double data rate (DDR) DRAM. In one embodiment, default values may be stored in SPID registers 904 and/or the associated mode register positions for SPID. For example, a mode register may contain designated bits for SPID, or any other unique ID for each of a group of memory devices in a stack. The mode/SPID register may have a default setting upon power-up whereby each memory device in the stack has a same initial SPID value (e.g., SPID=000). Then, each command may be applied and recognized by each memory device of the stack. SPIDs for each memory device can subsequently be assigned or determined, such as by using the approach described above.
In one example, SPID registers 904 may contain resettable values to allow for commands to be accepted by each device in the stack of devices. For example, a mode register may be programmed to indicate a test mode or global command configuration, whereby SPID registers 904 are subsequently reset to a given value in response thereto. This approach provides an override mechanism for SPID registers 904 that can then be undone upon completion of the particular test mode or global command operation. For example, the determined SPID values could be stored in one latch and preserved throughout the test mode or global command configuration that overrides the SPID values, such that the determined SPID values could be restored thereafter. A control signal can be used to control a multiplexer to select either a stored SPID value from SPID registers 904, or an override value supplied via a mode register. As another example, a separate control signal (e.g., a global chip select signal) can be used when asserted to override SPID values to allow a same command to be accepted by each memory device in the stack.
Other example approaches for automatically determining SPID values can also be employed in certain embodiments. For example, a sequential SPID assignment approach may include use of a serially connected signal, as will be discussed below. In such connections, each memory device can receive a signal from an adjacent memory device, and perform operations (e.g., data modification, combinatorial logic, clocking adjustments, etc.) thereon. For example, given default SPID values of zero, an SPID value may be sent from one memory device to a next memory device located above in the stack of memory devices, and the receiving device may effectively add one to the received value, and then store that value as its SPID. Thus, each device can sequentially assign itself an SPID value. Certain embodiments may also support such a serially assigned SPID comparison mode against an SPID determined through another approach (e.g., capacitance comparison), as a way to double check SPID determination. For example, an error indication may be generated if such values do not match.
Also in particular embodiments, a test mode can connect one or more of various internal supplies together for one or more devices in a stack of devices. In one embodiment, a semiconductor memory device can include: (i) a stack position identifier for identifying a position of the semiconductor memory device in an aligned vertical stack of a plurality of semiconductor memory devices, where the plurality of semiconductor devices are coupled together via a common vertical connection; (ii) a test mode detector configured to determine a test mode in response to a received command and a match of the stack position identifier against the received command; (iii) a supply generator configured to generate an internal supply for the semiconductor memory device; and (iv) a supply coupler configured to couple the internal supply to the common vertical connection in response to the determined test mode.
Referring now to
One or more of the memory devices, such as DRAM-5 (see, e.g., 1002) and DRAM-3 (see, e.g., 1004) may have their internal supplies connected to an appropriate common vertical connection to allow for testing and/or monitoring via memory controller 102. In this way, internal supply generation circuits may be tested once the memory devices are placed in a stacked configuration. In addition, this configuration may also allow for an override of the internal supply of one or more memory devices in order to test the effects of different supply values. In this case, the internal supply circuitry may be disabled by this test mode for a given chip in order to allow for the override. Furthermore, if it is determined that a particular internal supply regulator has failed after the devices have been secured in the stacked configuration, this failed regulator may be supplemented or effectively replaced by another functioning regulator in a different device in the stack. For example, if the VPP charge pump is determined to have failed on DRAM-3, the VPP charge pump for DRAM-5 or an adjacent DRAM, may be connected to the VPP supply for DRAM-3 via the appropriate common vertical connection. Further, control for regulators of different chips in the stack may utilise SPIDs. In this way, a level of redundancy may be provided for internal supplies across memory devices.
In certain embodiments, test modes can allow for one, two, or any number of memory devices in a stack to connect internal supplies to a common vertical connection. For example, one test mode may provide for a sequential connection of a designated internal supply to a common vertical connection such that memory controller 102 or a test system may test a regulated supply from each memory device in the stack. The sequencing, which may be in any order of memory devices, such as a pre-programmed or random ordering of memory devices, can be controlled by use of SPIDs. For example, the test mode can be generated by a test device or memory controller 102, and associated test commands may be issued with an SPID designating that an internal supply be connected to a common vertical connection. In this way, internal supplies from each memory device in a stack of memory devices can be measured via a test mode after the memory devices are placed into the stack.
Referring now to
In 1102, VPP generator 1108 may normally be used to generate word line high supply VPP. In this example, test mode signal TMVPEXT may be used to indicate connection of VPP test pad 1110 to internal supply VPP by PMOS transistor 1112. Level shifter 1116 may be used to shift from a VDD-based signal (e.g., TMVPEXT) to a VPP or VPP+ (e.g., a level at least as high as VPP or the level of VPP test pad 1110, so as not to forward bias a P-N junction of transistor 1112), and inverter 1114 may also be supplied at this VPP+ level. In a mode whereby the signal at VPP test pad 1110 is to override the internally generated VPP supply, TMVPEXT may be used to disable VPP generator 1108. However, TMVPEXT or other suitable test mode and/or control signals may also be used to indicate to VPP generator 1108 that it needs to enable additional or redundant charge pump capabilities therein, such as when VPP generator 1108 is to act in a redundant capacity for a different DRAM device, as discussed above.
In 1104, internal voltage regulator 1118 may normally be used to generate regulated supply VREG. In this example, test mode signal TMVREXT may be used to indicate connection of VREG test pad 1120 to internal supply VREG by the pass gate formed by NMOS transistor 1122 and PMOS transistor 1124. Inverter 1126 may be used to control PMOS transistor 1124 and in complementary fashion to control NMOS transistor 1122. In a mode whereby the signal at VREG test pad 1120 is to override the internally generated VREG supply, TMVREXT may be used to disable internal voltage regulator 1118. However, TMVREXT or other suitable test mode and/or control signals may also be used to indicate to internal voltage regulator 1118 that it needs to enable additional or redundant supply capabilities therein, such as when internal voltage regulator 1118 is to act in a redundant capacity for a different DRAM device, as discussed above.
In 1106, VBB generator 1126 may normally be used to generate bulk low supply level VBB, to reverse bias P-WELLS under the memory cells for cell leakage reduction. In this example, test mode signal TMVBEXT may be used to indicate connection of VBB test pad 1130 to internal supply VBB by NMOS transistor 1132. Negative level shifter 1134 may be used to shift from a VDD-based signal (e.g., TMVBEXT) to a VBB or VBB− (e.g., a level at least as low or negative as VBB or the level of VBB test pad 1130. In a mode whereby the signal at VBB test pad 1130 is to override the internally generated VBB supply, TMVBEXT may be used to disable VBB generator 1126. However, TMVBEXT or other suitable test mode and/or control signals may also be used to indicate to VBB generator 1126 that it needs to enable additional or redundant negative charge pump capabilities therein, such as when VBB generator 1126 is to act in a redundant capacity for a different DRAM device, as discussed above.
Thus, each of VPP generator 1108, internal voltage regulator 1118, and VBB generator 1126 may be designed such that additional regulation or pumping capabilities may be enabled to support other memory devices in the stack via an appropriate common vertical connection. Further, each of VPP generator 1108, internal voltage regulator 1118, and VBB generator 1126 may be designed to be disabled under certain situations, such as when it is to be replaced by a corresponding regulator/generator (e.g., one with enhanced regulating or pumping capabilities enabled) from another memory device and the stack.
In particular embodiments, SPID bits may be used in order to isolate which memory devices are to be configured in a test mode, such as in a supply override test mode, or a redundant supply test mode, etc., as discussed above. For example, circuitry to generate various test mode signals (e.g., TMVPEXT, TMVREXT, TMVBEXT, etc.) may be generated by use of chip select signals sent with the test mode command (e.g., from memory controller 102) matching with the SPID of a particular device. As another example, bit settings to provide a mapping as to the types of supply accesses via a common vertical connection may also be utilized. For example, a table that includes SPIDs and corresponding settings, such as VPP override, VBB override, providing redundancy for a device identified by different SPID, etc., may be utilized in order to manage various supply related test modes. In some applications, such a table may be stored on memory controller 102 such that test mode or other such regulator setting commands as discussed herein may be sent to the proper memory devices for configuration of the supplies by use of and SPIDs corresponding common vertical connections.
In various embodiments, a self-refresh timing adjustment can be made based on a position of a device in a stack of such devices. In one embodiment, a semiconductor memory device can include: (i) a stack position identifier for identifying a position of the semiconductor memory device in an aligned vertical stack of a plurality of semiconductor memory devices; (ii) a self-refresh oscillator configured to generate an oscillating signal to control a rate of self-refresh operations for the semiconductor memory device; and (iii) an oscillator adjustor configured to increase the self-refresh oscillator from a first frequency to a second frequency in response to the stack position identifier when the stack position identifier indicates that the semiconductor memory device is higher in the aligned vertical stack than at least one of the plurality of semiconductor memory devices.
Referring now to
A standard refresh command (e.g., from address or other control signals) may be detected using refresh command detector 1202. A separate self-refresh counter 1204 (e.g., a wraparound counter) can be used to determine when a self-refresh cycle should occur. Thus, a refresh operation can effectively occur when either a refresh command is detected or a self-refresh counter indicates that it is time for a self-refresh operation. For example, NOR gate 1210 and inverter 1212 can be used to trigger a refresh operation from either refresh command detector 1202 or self-refresh counter 1204.
A self-refresh enable signal (SREN) may be used to enable isolation of self-refresh oscillator 1206. Thus, if self-refresh is not enabled, SREN may be low to disable operation of self-refresh oscillator 1206, resulting in a fixed, as opposed to an oscillating, signal at SROSC. If SROSC is fixed instead of oscillating or clocking, self-refresh counter 1204 may not update to get closer to another self-refresh operation. However, if SROSC is oscillating, a rate or frequency of oscillation therefrom may determine how often self-refresh counter 1204 reaches a value (e.g., all ‘1’ or logic high levels) that can result in initiation of a refresh operation.
In certain embodiments, self-refresh oscillator adjustor 1208 can receive SPID bits, and provide an adjustment to self-refresh oscillator 1206 in response thereto. For example, the higher the memory device is in the stack of memory devices, as indicated by its SPID, the higher the frequency of SROSC. This represents one example of how self-refresh frequency may be increased to help compensate for increased storage cell leakage due to increased temperatures. Of course, one skilled in the art will recognize that adjustments may also, or alternatively, be made to self-refresh counter 1204 in order to increase a self-refresh frequency as one goes higher up in a stack of memory devices. Alternatively, as discussed above, a device can be configured to change from a default refresh rate to a higher refresh rate based on SPID matching, where the device does not necessarily have to be higher in the stack than a given device in order to make the adjustment.
Referring now to
The second stage of self-refresh oscillator 1206 may be inverter 1314, and the third stage may include PMOS transistors 1316 and 1320, and NMOS transistors 1322 and 1324. The fourth stage may include PMOS transistors 1330 and 1334, and NMOS transistors 1336 and 1338. The fifth stage may include inverter 1344 that drives SROSC and provides feedback to control the first stage at transistors 1306 and 1308. Various other transistors may be included in order to reduce resistance in a particular path when enabled, in order to speed up that particular stage. For example, NMOS transistor 1326 can be enabled by SPID[2] such that a pull down path through transistor 1322 can be sped up when SPID[2] is high because of the parallel resistance effect of transistors 1324 and 1326. SPID[2] being high can indicate that the memory device is in a top half of the stack of memory devices, and SPID[2] being low may indicate that the memory device is in a bottom half of the stack of memory devices. Thus, memory devices in a top half of the stack may have an increased self-refresh rate based on their SPIDs.
The third stage may also be adjusted via PMOS transistor 1318 which may be on (by way of inverter 1342) when SPID[1] is high. The fourth stage may include adjustment PMOS transistor 1332 which can be turned on by way of inverter 1328 when SPID[0] is high. Also, the fourth stage may include adjustment NMOS transistor 1340 that can be turned on when SPID[1] is high. Table 3 below shows various adjustments in oscillator stages that can be made based on SPID values.
In this way, a self-refresh oscillator can be adjusted according to SPID bits. Of course, this represents just one example, and certain embodiments are amenable to a variety of circuit design choices as to how refresh frequency adjustment can be made based on SPID values. For example, different stages can be adjusted in the self-refresh oscillator, and also different numbers of stages can be employed. Further, self-refresh counter 1204 may alternatively, or in addition to, be adjusted based on SPID values. Finally, SPID-based self-refresh frequency adjustment may supplement other forms (e.g., temperature sensor-based) of self-refresh adjustment. For example, SPID-based self-refresh adjustment may represent a coarse tuning, while temperature sensor-based self-refresh adjustment may represent a finer adjustment, or vice versa.
In particular embodiments, internal regulated supply adjustment can be performed based on stack position identification. In one embodiment, a semiconductor memory device can include: (i) a stack position identifier for identifying a position of the semiconductor memory device in an aligned vertical stack of a plurality of semiconductor memory devices; (ii) a regulator circuit configured to generate a supply for the semiconductor memory device; and (iii) a regulator adjustment circuit configured to adjust the supply from a first voltage level to a second voltage level in response to the stack position identifier when the stack position identifier indicates that the semiconductor memory device is higher in the aligned vertical stack than at least one of the plurality of semiconductor memory devices.
Referring now to
When OSC is low, PMOS transistors 1404 and 1408 may be off, while NMOS transistor 1410 may be on to discharge the source/drain side of transistor 1412, which is connected in a capacitor configuration. Diode connected NMOS transistors 1414 and 1416 (when PMOS transistor 1418 is off) may charge a gate side of transistor 1412 to a level of about VDD−2*VTN. When OSC is high, NMOS transistor 1410 may be off, while PMOS transistors 1404 and 1408 may be on in order to charge the source/drain side of transistor 1412. Through capacitive coupling, the gate of NMOS transistor 1420 may be raised to a level of about 2*(VDD−VTN).
Also when OSC is low, and the low level has passed through delay 1422 and inverter 1424, inverter 1440 may drive a high level at the gates of PMOS transistors 1426 and 1430, and NMOS transistor 1432. This may discharge the source/drain connection of capacitor-connected NMOS transistor 1434. When OSC goes high, after delay 1422, transistors 1426 and 1430 may charge the source/drain of 1434 to a VDD level. The previous precharge cycle on the gate side of 1434 may have precharged via NMOS transistor 1420 (with a gate that is capacitively coupled to greater than VDD+VTN) such that a charge pump action may reach 2*VDD at a gate side of 1434, pumping out of diode connected NMOS transistor 1436 to the VPP node a level of about 2*VDD−VTN.
Of course, this represents only one example of a possible charge pump circuit, and any suitable charge pump or regulating circuit may be adjusted based on a position of the memory device in a stack by using SPID. In the particular example of
Other example adjustments can include SPID[0] controlling PMOS transistor 1406, which can reduce the resistance through the pull-up path for the source/drain connection of 1412 when SPID[0] is low. Also, PMOS transistor 1428 can reduce the resistance through the pull-up path for the source/drain connection of main pumping capacitor 1434 when SPID[1] is low. Table 4 below shows example states of various adjustment transistors, which may generally be configured to increase a level of VPP as one goes up a stack of memory devices.
Of course, other internally regulated supplies (e.g., VBB, VBLR, VREG, etc.) can also be adjusted based on SPID in particular embodiments. For example, a word line low level (e.g., to a negative value to reduce leakage effects as temperature increases) may also be adjusted using SPID values. Further, SPID-based supply regulation adjustment may supplement other forms (e.g., temperature sensor-based) of regulation adjustment. For example, SPID-based supply adjustment may represent a coarse tuning, while temperature sensor-based supply adjustment may represent a finer adjustment, or vice versa.
In particular embodiments, a chip select valid command detector can be based on stack position identifier. In one embodiment, a semiconductor memory device can include: (i) a stack position identifier for identifying a position of the semiconductor memory device in an aligned vertical stack of a plurality of semiconductor memory devices, where the plurality of semiconductor devices are coupled together via a plurality of common vertical connections; (ii) a receiver configured to receive a plurality of chip select signals from the plurality of common vertical connections; and (iii) a command detector configured to detect whether a command is intended for the semiconductor memory device based on a comparison of the received plurality of chip select signals against the stack position identifier.
Referring now to
SPID_Ready (see, e.g.,
Thus when either the SPID signals are not enabled (e.g., SPID_Ready is low), or when a bit-by-bit matching of CS and SPID indicates that a command is intended for a given memory device, a valid command indicator may be sent to command detector 1516. In this way, command detector 1516 may apply the valid command indicator to a command received (e.g., via address and/or control signals) in order to determine a detector command. Thus, a command received may be ignored by particular device if any bit of the bit-by-bit matching of CS and SPID results in a mismatch.
Particular embodiments may utilize the common vertical connections (see, e.g.,
A chip-level redundancy or de-allocation of a memory device using SPIDs may also be supported in particular embodiments. If, for example, even though KGD testing may indicate that each memory device to be placed in the stack is good, some failure modes may occur as a result of chip interconnections or other factors after a device is secured in the stack. In certain embodiments, redundancy or chip disable may be employed to address failed chips after placement in the stack. For example, the chip enable signal may be used to disable further commands to a given device via command detector 1516 when low. This chip enable signal may be derived from a latch or other setting that indicates that the given chip is to be disabled. In this case, some other device operations (e.g., internal voltage regulation and supply, memory cell reading and writing, etc.) may be disabled, but other device operations (e.g., passing signals between adjacent devices for serially connected pads, as discussed below) may be maintained.
Further, chip-level redundancy can be employed, such as in case when 9 memory devices instead of 8 are placed in the stack, where the extra memory device can be substituted for the bad device by using SPIDs and other controls. For example, address and other control signals may be applied in a parallel fashion to each memory device in the stack, but DQ signals may be parallelized such that each device contributes a given number of DQs (e.g., 64) in order to form a wide I/O (e.g., 512-bits wide) memory structure. Thus, the memory device that provides redundancy can have its DQ signals multiplexed or selected in place of the de-allocated memory device. In this way, the redundant memory device DQ signals may be steered (e.g., using pad selection circuitry as discussed above with reference to
In particular embodiments, signal latency may also be adjusted based on a position of a device in a stack of devices by using a stack position identifier. In one embodiment, a semiconductor memory device can include: (i) a stack position identifier for identifying a position of the semiconductor memory device in an aligned vertical stack of a plurality of semiconductor memory device; (ii) a latency determiner configured to determine a programmed latency from a mode register on the semiconductor memory device; and (iii) a latency adjustor configured to adjust the programmed latency in response to the stack position identifier.
Referring now to
Referring now to
In particular embodiments, latency adjustment or alignment can be performed using a data strobe on a common vertical connection. In one embodiment, a semiconductor memory device can include: (i) a stack position identifier for identifying a position of the semiconductor memory device in an aligned vertical stack of a plurality of semiconductor memory devices, where the plurality of semiconductor devices are coupled together via a data strobe connection; (ii) a phase detector configured to detect a phase of the data strobe relative to a data signal; and (iii) a phase adjustor configured to delay the data signal to match the data strobe, wherein the data strobe is driven by one of the plurality of semiconductor memory devices based on the stack position identifier.
Referring now to
Referring now to
In this particular example, a top memory device and a stack of memory devices may be designated to drive pad 302-DS, and in particular for read or downstream operations. Here, if each of bits in SPID[2:0] are high, an output of NAND gate 1904 may be low. This can indicate a top device (e.g., DRAM-7 with SPID=111) is allowed to drive pad 302-DS because NAND gate 1908 may be enabled by a high input thereto, and NOR gate 1912 may be enabled via a low input thereto by way of inverter 1910. Then, if the data strobe input is high, an output of NAND gate 1908 may be low to drive pad 302-DS high via pull-up PMOS transistor 1914. If in this case the data strobe input is low, an output of NOR gate 1912 may be high to drive pad 302-DS low via pull-down NMOS transistor 1916.
Driving capability for a given memory device on pad 302-DS may also be enabled when a DS enable (e.g., for a non-stacked device configuration) or test mode signal is activated such that inverter 1902 presents a low at an input to NAND gate 1906. In this way, a test mode or other DS enable control may be utilised to allow any device, or a particular separately designated memory device and the stack, to drive pad 302-DS. For other memory devices that do not match SPID[2:0]=111, such as any of DRAM-0, DRAM-1, . . . DRAM-6, circuit 1900 may be used to tri-state pad 302-DS such that the present memory device may not drive this common vertical connection. In this case, NAND gate 1906 may output a low that turns off pull-up transistor 1914 via NAND gate 1908, and pull-down transistor 1916 via NOR gate 1912. In this way, only a predetermined one (e.g., a top memory device in the stack) memory device may drive the common connection data strobe signal (e.g., via pad 302-DS) for a given operation (e.g., a read operation). Of course, one skilled in the art will recognize that other types of circuits, as well as other predetermined SPID bit inputs can also be employed in certain embodiments.
Referring now to
Transition detector 2002 may generate a pulse in response to a transition on the common vertical DS signal (labeled “DS pad” in
An internally generated data strobe signal can be received at inverter 2018, and supplied to a flip-flop circuit. This internally generated data strobe signal may be a normal such data strobe signal as would be generated if the memory device were configured for non-stacked usage. The flip-flop circuit can include a first pass gate formed by NMOS transistor 2020 and PMOS transistor 2022, a first latch formed by inverters 2024 and 2026, a second pass gate formed by PMOS transistor 2028 and NMOS transistor 2030, and a second latch formed by inverters 2032 and 2034. Transistors 2022 and 2030 may be controlled by the output of NAND gate 2014, and transistors 2020 and 2028 may be controlled by an inverted output of NAND gate 2014 by way of inverter 2016.
In operation, the internally generated data strobe signal can be stored in the first latch when an output of NAND gate 2014 is low. This value from the first latch may be transferred to the second latch when the output of NAND gate 2014 is high, such as during the pulse indicating detection of a transition on the DS pad. The output of the second latch may then be inverted at 2036 and provided as an adjusted data strobe signal. In this way, an internally generated data strobe signal can be adjusted to align with a DS pad signal that is driven by another memory device (e.g., a memory device at the top of a stack of memory devices). For example, this adjusted data strobe may then be used to clock data out in a read operation (e.g., at DQ connections) such that data from each memory device in the stack essentially arrives at a memory controller or common interface chip at substantially a same time.
In this fashion, a data strobe signal on a common vertical connection may be used to align signal timing from multiple memory devices at a common device. The same, or a different, common vertical connection may be utilised in a write path whereby the signal is driven from memory controller 102. In this case, a bottom memory device (e.g., DRAM-0), or simply the memory controller or logic chip itself, may drive the DS signal in an upstream signal flow. Then, each memory device of the stack of memory devices may synchronize write data to the DS signal in the same fashion. In this way, both reads and writes may be effectively synchronized among multiple memory devices using a common vertical signal. Further, any such common vertical signal, or any signal even common to two adjacent devices in a stack, may be used in the same fashion to synchronize from one device to another.
In particular embodiments, serially connected signal interconnects among stacked memory devices may also be accommodated by control of the signal paths at each memory device in the stack. In one embodiment, a semiconductor memory device can include: (i) a stack position identifier for identifying a position of the semiconductor memory device in an aligned vertical stack of a plurality of semiconductor memory devices; (ii) a first pad coupled to a first signal and a first TSV; (iii) a second pad coupled to a second signal and a second TSV; and (iv) a pad signal path determiner configured to couple a signal path from the first pad to the second pad in response to a first state of the stack position identifier, and to couple a signal path from the second pad to the first pad in response to a second state of the stack position identifier.
Referring now to
Serially connected signals as described herein can be any types of signals, such as control and/or data signals, that may interface with a memory device. For example, data I/O signals can be serially connected, such as in a master-slave data arrangement. In another example, each or substantially all data and/or control signals can be serially connected between the memory devices so as to avoid long TSV and/or die-to-die via connections across several devices. In this way, clocking and control can be passed from one memory device to another in serial fashion. Further, pads as described herein can represent any suitable interconnect positions or interfaces on a semiconductor device.
Referring now to
In the example of
In certain embodiments, a state of SPID[0], which may indicate an even or odd numbered memory device position in the stack, can be used to determine a direction of the signal path. For example, an even-numbered memory device may receive a signal from pad 302-A and send the signal to pad 302-B, while an odd-numbered memory device may receive a signal from pad 302-B and send the signal to pad 302-A. A true value of SPID[0] can be used to enable the tri-statable driver for pad 302-A, while a compliment of value of SPID[0] can be used (e.g., via inverter 2202) to enable the tri-statable driver for pad 302-B.
Thus, SPID values (e.g., a state of SPID[0]) can be used to effectively steer a signal from one pad to another, particularly when that signal is part of a serially connected path, such as that shown in
Referring now to
As one example use of signal modifiers (e.g., via 2302 or 2304), phase differences can be determined between signals propagating via one or more memory devices in the stack. Then, latency may be added where appropriate in order to adjust and synchronize signaling across memory devices. In another example, a clock frequency may be set during test, with counting of clocks based on SPID values, and may be used to determine a latency difference between memory devices (e.g., adjacent devices) in a memory stack. The signal modifiers may also be used to provide results of a capacitance comparison test, such as discussed above with reference to
Test modes may also make use of, or otherwise be incorporated with, signal modifiers (e.g., via 2302 and/or 2304), such as for boundary scan or other design for test (DFT) applications. The signal modifiers can employ any number of logical combinations, test registers, and the like. For example, test registers may be read out from one of pad A or pad B, and passed to the other as part of a test mode. Oscillation modes may also be created in this fashion by using the signal modifiers to invert each signal as it passes through a given memory device. In this way, clocking can occur across multiple memory devices in a stack of memory devices. For example, the serially connected signal may enter a chip with a high value, and send a low value to an adjacent chip, which then send a high value to its adjacent chip, and so on, thus creating a cross-chip clock. Also, one serially connected signal path may be designated or used as such a cross-chip clock signal for an upstream path, while another such serially connected signal path may be designated or used as a cross-chip for a downstream path. A cross-chip clocking loop may also be formed by connecting the upstream cross-chip clock path to the downstream cross-chip clock path at a designated memory device (e.g., a top memory device in the stack). Such cross-chip clocking may be particularly useful for test modes that test across different memory devices, including those that test memory devices in a stack.
Referring now to
However, if the serially connected signal (e.g., 2102) is for a downstream signal path, the pass gate formed by NMOS transistor 2404 and PMOS transistor 2406 may be turned on by the upstream indicator signal being low. The upstream indicator being low can allow for the compliment of SPID[0] (e.g., via inverter 2402) to control the tri-statable driver for pad 302-A, and the true value of SPID[0] (via inverter 2414) to control the tri-statable driver for pad 302-B. Thus, when the upstream indicator is inactive or low, indicating that the signal associated with the serially connected signal path is a downstream signal, even-numbered memory devices may send the signal from pad B to pad A, while odd-numbered memory devices may send the signal from pad A to pad B.
In this way, serially connected signal paths that employ TSVs and two possible external connection positions or pads on each memory device, may be configured to flow through a given device in a direction that is controlled by SPID bits. Of course, depending on the particular patterns of TSVs and/or die-to-die vias, other bits of SPID values could be employed. For example, a signal may be in a common vertical connection position for a bottom half of the memory devices, and can switch over to another common vertical connection position for a top half of the memory devices, or any combination thereof. Based on the particular pattern and type of signaling (e.g., upstream, downstream, modifiable, etc.), different SPID bit values can be used to control the directional flow of the signal through a given memory device.
In particular embodiments, a content-addressable memory (CAM) device may be configured for serial searching in a stacked arrangement. In one embodiment, a method of controlling a search in a first CAM device can include: (i) receiving a search key and a search control signal from a second CAM device, where the second CAM device is a next CAM device below the first CAM device in an aligned vertical stack of a plurality of CAMs devices; (ii) performing a search of the first CAM device in response to the received search key and the received search control signal; (iii) sending the search key and the search control signal from the first CAM device to a third CAM device if the search of the first CAM device results in no match on the first CAM device, where the third CAM device is a next CAM device above the first CAM device in the aligned vertical stack of the plurality of CAMs devices; and (iv) sending a search result to the second CAM device if the search of the first CAM device results in a match on the first CAM device.
Referring now to
In order to save power, if a search of a given CAM results in a match condition, that CAM can send the search results downstream via serially connected signals 2504. Thus, if a search key is presented to CAM-0, and a search is performed on CAM-0 that results in no matches, CAM-0 can then forward the search key upstream to CAM-1 so that CAM-1 can perform a search. This can continue until a particular CAM obtains a match and search result from the search using the search key. The CAM with the match can then send the search results downstream, without notifying any further upstream CAM devices. This has the advantage of saving power by not having to search each CAM device in a stack of such CAM devices. In addition, a priority based on stack position is also provided because of the order of performing searches. For example, CAM-0 may be a higher priority than CAM-1, and CAM-1 may be a higher priority than CAM-2, and so on up the stack of CAM devices.
Referring now to
In 2650, an example flow for a CAM device that has already performed a search on a given search key, and resulted in no match, as shown. The flow begins (2652), and at 2654 a search result is received from an upstream chip. Because the given CAM device has already been searched using the given search key, thus the search key may be passed downstream on a serially connected signal path (e.g., search results 2504). Thus at 2656, the search result may be sent to a downstream chip, completing the flow at 2658.
Other CAM functions that can be performed in a sequential fashion by using serially connected signals as described herein, may also be supported in particular embodiments. For example, reads and writes of values to the CAMs may be performed in a sequential fashion, particularly when considering pipelining or interleaving various types of operations, such as reads/writes mixed in with search operations. In one example, in order to maintain timing consistency, search operations may be passed through to each device up the stack and returned from a top device in the stack. For example, even if a search of CAM-3 results in a match to the search key, the search key may still propagate up the stack of CAM devices to reach CAM-7, but with a flag set that indicates that no actual search is to be performed on CAM-4, CAM-5, CAM-6, or CAM-7. For example, such a flag may be applied on another common vertical connection (e.g., a command signal), or may be encoded within search key 2502.
Alternatively, a count of clock cycles may be made in the CAM device having a match of the search key (e.g., CAM-3), and search results 2504 can be driven by CAM-3 at an appropriate time or clock cycle consistent with the search result being driven by the top CAM in the stack (e.g., CAM-7). In either the clock cycle delay, or the actual propagation up to a top CAM and back, additional CAM device searches (e.g., those CAMs above CAM-3 in the stack) may still be suppressed in order to save power. Also, strobe signals can also be employed as discussed above in order to synchronize results from one camp to another CAM. Thus, a consistent number of clock cycles may be designated to a search function no matter which CAM in the stack provides the search result. While this increases a latency of a search result for each search except those that only find a match in the top CAM in the stack, a consistent latency can allow for deeper pipelining of CAM operations.
In another example of search functionality for stacked CAM devices, a search key may be applied to each of the CAM devices in parallel. Then, each CAM in the stack may hold this search key until receiving notice from a CAM position below that CAM in the stack of CAMs. For example, CAM-4 may hold the received search key until receiving an indication from CAM-3 that a search on this search key has resulted in no match, and thus CAM-4 may proceed with a search on that search key. In this approach, fewer signals may be of the serially connected variety, and search key signals may be direct or parallel connected common vertical signals, which may help to reduce CAM chip size.
In another variation, each CAM in the stack of CAMs may hold a supplied search key for possible searching on that CAM for a predetermined number of clock cycles based on the SPID of that CAM device. This predetermined number of clock cycles may include a number of clock cycles that it takes to perform a search operation on a CAM device multiplied by a number representing the stack position via the SPID. For example, if it takes five clock cycles to perform a CAM search, CAM-3 may perform a search of a given search key after 15 clock cycles, upon receiving an indication that none of CAM-0, CAM-1, nor CAM-2 have found a match to the given search key. In this way, CAM operations can be serialized and synchronized for multiple CAM devices in a stack of such devices.
In addition to stacks of memory devices that include a single type of memory device, hybrid structures that include at least two different types of memory devices are also contemplated. For example, one or more SRAMs may be coupled to one or more DRAMs in a stack of such devices. In one example, one or more DRAMs may be stacked on top of one SRAM, and this pattern may be replicated (e.g., one SRAM, followed by three DRAMs, followed by another SRAM, and then another three DRAMs). Such a memory device arrangement may be applicable to cache functionalities and/or applications. Further, SPIDs associated with such hybrid memory device stack structures may include one identifier per repeatable pattern (e.g., one SRAM and three DRAMs), or designated bits (e.g., the two least significant bits) that indicate a specific memory device within a repeatable pattern of memory devices.
As another example hybrid memory device stack structure, a CAM device may be followed by one or more DRAMs devices in a repeatable pattern. With this structure, the DRAMs may be configured to perform CAM operations, or extended search operations, such as when a given CAM does not find a match case. For example, the CAM devices in this repeatable pattern may operate much as discussed above, but a mismatch may first result in a search of associated DRAM devices, prior to being sent to a next CAM device up the stack of memory devices. Further, the DRAMs associated with a given CAM may also be searched in a sequential order. As can be appreciated by one skilled in the art, any suitable variety of memory devices may be stackable, and identifiable via stack position identifiers as discussed herein.
The foregoing descriptions of specific embodiments of the present 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, and obviously many modifications and variations are possible in light of the above teaching. The embodiments were chosen and described in order to best explain the principles of the invention and its practical application, to thereby enable others skilled in the art to best utilise the invention and various embodiments with various modifications as are suited to the particular use contemplated. It is intended that the scope of the invention be defined by the claims appended hereto and their equivalents.
This application claims priority from U.S. Provisional Patent Application Ser. No. 61/570,215, entitled “MEMORY DEVICES CONFIGURED FOR STACKED ARRANGEMENTS,”, filed on Dec. 13, 2011, which is hereby incorporated by reference as if set forth in full in this application for all purposes.
| Number | Name | Date | Kind |
|---|---|---|---|
| 5719445 | McClure | Feb 1998 | A |
| 6233157 | Yoon et al. | May 2001 | B1 |
| 6480931 | Buti et al. | Nov 2002 | B1 |
| 7446420 | Kim | Nov 2008 | B1 |
| 7480166 | Jeung et al. | Jan 2009 | B2 |
| 7791918 | Ruby | Sep 2010 | B2 |
| 7791919 | Shimizu | Sep 2010 | B2 |
| 7822910 | Kao | Oct 2010 | B2 |
| 7834440 | Ito et al. | Nov 2010 | B2 |
| 7916511 | Park | Mar 2011 | B2 |
| 7930492 | Chung et al. | Apr 2011 | B2 |
| 8054597 | Gebreselasie et al. | Nov 2011 | B2 |
| 8270228 | Mizukane et al. | Sep 2012 | B2 |
| 8378500 | Choi | Feb 2013 | B2 |
| 8397013 | Rosenband et al. | Mar 2013 | B1 |
| 8405221 | Imai | Mar 2013 | B2 |
| 8437163 | Nakanishi et al. | May 2013 | B2 |
| 8492905 | Suh | Jul 2013 | B2 |
| 8599595 | Stephens, Jr. | Dec 2013 | B1 |
| 8659928 | Stephens, Jr. | Feb 2014 | B1 |
| 8891278 | Stephens, Jr. | Nov 2014 | B1 |
| 20020052633 | Prutchi et al. | May 2002 | A1 |
| 20060052961 | Best | Mar 2006 | A1 |
| 20080283995 | Bucki et al. | Nov 2008 | A1 |
| 20090097393 | Diab et al. | Apr 2009 | A1 |
| 20090268539 | Ruckerbauer et al. | Oct 2009 | A1 |
| 20100034005 | Lee | Feb 2010 | A1 |
| 20110044084 | Chung et al. | Feb 2011 | A1 |
| 20110194326 | Nakanishi et al. | Aug 2011 | A1 |
| 20120122251 | Yamada et al. | May 2012 | A1 |
| 20120249219 | Shoemaker | Oct 2012 | A1 |
| Entry |
|---|
| Non-Final Office Action on U.S. Appl. No. 14/021,534, mailed Dec. 4, 2013. |
| Non-Final Office Action on U.S. Appl. No. 13/403,122, mailed Jun. 14, 2013. |
| Non-Final Office Action on U.S. Appl. No. 13/491,090, mailed Nov. 21, 2013. |
| Non-Final Office Action on U.S. Appl. No. 13/964,764, mailed May 23, 2014. |
| Non-Final Office Action on U.S. Appl. No. 14/010,864, mailed Jun. 20, 2014. |
| Non-Final Office Action on U.S. Appl. No. 14/071,443, mailed Jan. 22, 2014. |
| Notice of Allowance on U.S. Appl. No. 13/466,736, mailed Oct. 28, 2013. |
| Notice of Allowance on U.S. Appl. No. 13/349,833, mailed Oct. 18, 2013. |
| Notice of Allowance on U.S. Appl. No. 13/357,384, mailed Jul. 17, 2013. |
| Notice of Allowance on U.S. Appl. No. 13/425,636, mailed Sep. 3, 2013. |
| Notice of Allowance on U.S. Appl. No. 13/455,893, mailed Feb. 24, 2014. |
| Notice of Allowance on U.S. Appl. No. 13/491,090, mailed Dec. 6, 2013.. |
| Notice of Allowance on U.S. Appl. No. 13/534,402, mailed Jun. 9, 2014. |
| Notice of Allowance on U.S. Appl. No. 14/021,534, mailed Jan. 17, 2014. |
| Notice of Allowance on U.S. Appl. No. 13/403,122, mailed Aug. 23, 2013. |
| Notice of Allowance on U.S. Appl. No. 14/071,443, mailed Apr. 2, 2014. |
| Notice of Allowance on U.S. Appl. No. 13/365,413, mailed Jul. 29, 2014. |
| Final Office Action on U.S. Appl. No. 14/010,864, mailed Nov. 6, 2014. |
| Notice of Allowance on U.S. Appl. No. 13/964,764, mailed Oct. 24, 2014. |
| Non-Final Office Action on U.S. Appl. No. 14/105,547, mailed Feb. 20, 2015. |
| Notice of Allowance on U.S. Appl. No. 14/010,864, mailed Feb. 17, 2015. |
| Non-Final Office Action on U.S. Appl. No. 14/060,276, mailed Mar. 4, 2015. |
| Non-Final Office Action on U.S. Appl. No. 14/540,196, mailed Mar. 17, 2015. |
| Notice of Allowance on U.S. Appl. No. 14/105,457, mailed Jun. 10, 2015. |
| Notice of Allowance on U.S. Appl. No. 14/060,276, mailed Jul. 15, 2015. |
| Number | Date | Country | |
|---|---|---|---|
| 20150009740 A1 | Jan 2015 | US |
| Number | Date | Country | |
|---|---|---|---|
| 61570215 | Dec 2011 | US |
