Sense amplifier circuit, corresponding memory device and operating method

By designing a sense amplifier circuit without input capacitors in the phase change memory, using the combination of floating comparator and sense amplifier core, the problem of limited read speed and accuracy is solved, and more efficient memory operation is achieved and low power consumption.

CN119068927APending Publication Date: 2024-12-03STMICROELECTRONICS INT NV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410685388.7
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Priority Date
2024-05-29
Filing Date
2024-05-30
Publication Date
2024-12-03

AI Technical Summary

Technical Problem

Prior Art In the read operation of phase change memory, the sensing amplifier is difficult to operate effectively when the bit line voltage is higher than the minimum internal power supply voltage, resulting in limited read speed and accuracy, and the discharge window and function are present in the input capacitor. Consumption problem.

Method used

A sense amplifier circuit without input capacitor was designed. Through the combination of floating comparator and sensing amplifier core, differential current is generated by using the differential input node and output node of the comparator to realize effective sensing of bit line voltage. and amplify to reduce discharge windows and power consumption.

Benefits of technology

Improves memory access time and reduces power consumption, shortens the discharge window duration, maintains high sensing accuracy, and reduces DC current consumption.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119068927A_ABST
    Figure CN119068927A_ABST
Patent Text Reader

Abstract

The invention relates to a sense amplifier circuit, a corresponding memory device and an operating method. First and second input terminals of the sense amplifier are coupled to first and second memory sense nodes. The first input transistor has a channel arranged between the first comparator input and the first comparator output, and a control terminal at a bias node. The second input transistor has a channel disposed between the second comparator input and the second comparator output, and a control terminal at the bias node. The first and second comparator inputs may be selectively coupled to each other in response to a compensation signal assertion, or may be selectively coupled to the first and second input terminals in response to a compensation signal de-assertion. The bias node may be selectively coupled to a comparator bias node in response to a bias enable assertion, or floating in response to a bias enable de-assertion. The sense circuit generates a read signal as a function of a difference between the first and second currents at the comparator output.
Need to check novelty before this filing date? Find Prior Art

Description

[0001] Priority claim

[0002] This application claims the benefit of Italian Patent Application No. 102023000011031 filed on May 31, 2023, the contents of which are incorporated herein by reference in their entirety to the maximum extent permitted by law. Technical Field

[0003] The present description relates to circuits and methods for amplifying a differential voltage sensed at a bit line of a memory device via a sense amplifier circuit.

[0004] For example, such a sense amplifier may be used during a read operation from a non-volatile memory (NVM) such as an embedded phase change memory (embedded PCM or ePCM). Such a memory may be applied to, for example, general purpose microcontrollers (GPMCUs) and secure microcontrollers. Background Art

[0005] A read operation (eg, data extraction) from the phase change memory is used to output (eg, fetch, extract) the contents of a PCM data word on an output data bus (eg, 145-bit bus DATA_OUT[144:0]).

[0006] Figure 1 is an exemplary (simplified) circuit diagram of a read structure in a PCM memory, and Figure 2 Yes Figure 1 The timing diagram of the signal waveform during the read operation in the structure of FIG. Figure 1 As illustrated in , the memory 10 includes an array of memory cells 102 (e.g., phase change memory cells) arranged in bit lines 103B (illustrated here as vertical lines) and word lines 103W1, 103W2, 103W3 (illustrated here as horizontal lines, also collectively or individually referred to as word lines 103W) according to a memory architecture known per se. Specifically, each memory cell 102 may be arranged in series with a corresponding selection transistor 104 (e.g., a pnpBJT transistor) between a ground voltage terminal and the corresponding bit line 103B. The selection transistors 104 of the same word line 103W receive the same control signal at their control (e.g., base) terminals, such as Figure 1 and Figure 2 Each bit line 103B can be selectively coupled (eg, via a corresponding p-channel MOS transistor 106 connected in series with the bit line and receiving the control signal Y_SEL) to provide a regulated voltage V REG (For example, with a target value V BL_CHARGE) of the memory supply voltage rail 108. In addition, a respective clamp transistor 110 (e.g., a pnpBJT transistor) may be coupled between each bit line 103B (e.g., at the source terminal of transistor 106) and a ground terminal. The clamp transistors 110 may receive the same control signal WL at their control (e.g., base) terminals. CLAMP Due to the capacitance of the memory cell 102 and / or the transistors 106, 110, each bit line 103B has a respective capacitive load C BL .

[0007] As expected, it can be Figure 2 As shown in the example, Figure 1 A read operation in a phase change memory is illustrated to read (and provide as output) the contents of multiple PCM words on the output data bus DATA_OUT[144:0]. Generally speaking, in a phase change memory, information can be stored associating two cells with each bit (data): a SET cell (with low resistance and high current) and a RESET cell (with high resistance and low current). The memory sensing circuit can read information from the cells in a differential mode.

[0008] In order to provide fast reading of memory cells, the memory access time (T ACC ) may be particularly demanding (e.g., on the order of 10ns). Additionally, read parallelism (e.g., reading two words in parallel) may be used to increase the throughput of the output data. As stated previously, the sensing of the contents of the memory cells may be performed in differential mode: the sensing circuit reads from the bit line BL_SET and the bit line BL_RESET in differential mode, thereby selecting the SET memory cell and the RESET memory cell. The SET memory cell and the RESET memory cell are selected by appropriately driving the control signals (word select signals) WL1, WL2, WL3: for example, Figure 2 As illustrated in FIG. 1 , by maintaining signals WL1 and WL3 at a high logic value (e.g., equal to value V BL_CHARGE ) and by driving the signal WL2 to a low logic value (e.g., equal to the ground voltage 0V), the unit 102 S is selected as a SET unit and unit 102 R Selected as RESET unit.

[0009] A voltage regulator (eg, a low dropout LDO regulator) generates a regulated voltage V at the power rail 108. REG (For example, with a target voltage value V BL CHARGE ). Voltage V REG The value of V BL_CHARGEThe voltage VBL at the selected bit line is clamped to the value of the base-emitter voltage V of the clamp transistor 110 prior to sensing (e.g., differentially between BL_SET and BL_RESET). BE , and then by activating transistor 106 (see Figure 2 In the control signal WL CLAMP Switching from low value to high value V BL_CHARGE To turn off transistor 110, the control signal Y_SEL changes from the high value V BL_CHARGE switches to a low value to turn on transistor 106), through the regulated voltage V REG Precharge the bit line to the value V BL_CHARGE The control signals WL1 and WL3 of the unselected word lines are maintained at a high voltage value (eg, V BL_CHARGE ) in order to keep the corresponding selection transistor in a non-conducting state.

[0010] Thus, an embedded phase change memory with a BJT selector as illustrated herein may rely on the bit line being biased (e.g., precharged) to a target voltage level VBL_CHARGE that is higher than the minimum internal supply voltage (VDD) and the external supply voltage (VCC). However, the sense amplifier (SA) cannot be provided to operate above the bit line voltage VBL_CHARGE, otherwise it would require an elevated supply voltage (e.g., generated by a charge pump circuit), resulting in higher power consumption during the read operation. In addition, the read operation must obtain sensing information from a shifted voltage domain, where the sense amplifier operates at voltage VDD and the bit line is biased at a higher voltage VBL_CHARGE. Shifting methods known in the art obtain the bit line voltage VBL_CHARGE from an input capacitor, but this known method has certain limitations that limit the read speed of the sense amplifier to its accuracy. For example, a long discharge window (e.g., approximately 4ns) may be required to record differences in the input capacitor; the input capacitor is shorted and connected in parallel with the sense amplifier core during commutation, which may affect the switching speed; and the input capacitor may slow down the offset compensation mechanism (e.g., because a long equalization time is required).

[0011] Therefore, there is a need in the art to provide improved read structures for phase change memories, and in particular to provide improved sense amplifiers that allow differential voltage sensing without relying on input capacitors.

[0012] There is a need in the art to provide such an improved sense amplifier for phase change memories. Summary of the invention

[0013] One or more embodiments may be directed to a sense amplifier circuit.

[0014] One or more embodiments may be directed to a corresponding memory device.

[0015] One or more embodiments may be directed to corresponding operating methods (eg, memory reading methods).

[0016] In one embodiment, a sense amplifier circuit includes a first input terminal and a second input terminal configured to be coupled to a first memory sense node and a second memory sense node, respectively. The sense amplifier includes a comparator circuit, which in turn includes a first input node and a second input node, and a first output node and a second output node, which are configured to generate a first output current and a second output current, respectively. The comparator includes: a first input transistor having a conductive path disposed between a first comparator input node and a first comparator output node and a control terminal coupled to an internal bias node; and a second input transistor having a conductive path disposed between a second comparator input node and a second comparator output node and a control terminal coupled to the internal bias node. The first comparator input node and the second comparator input node are selectively coupled to each other in response to an asserted compensation signal, or selectively coupled to the first input terminal and the second input terminal, respectively, in response to a deasserted compensation signal. The internal bias node is selectively coupled to the comparator bias node in response to a bias enable signal being asserted, or floats in response to the bias enable signal being deasserted. The sense amplifier further includes a sensing circuit coupled to the first comparator output node and the second comparator output node and configured to generate a memory read signal as a function of a difference between the first output current and the second output current.

[0017] Thus, one or more embodiments may provide a sense amplifier for a memory device that does not include an input capacitor, which may improve memory access time and / or reduce power consumption.

[0018] According to another aspect, a memory device includes a plurality of memory groups. As described above, each memory group is coupled to a respective sense amplifier circuit. Each memory group includes an array of memory cells arranged in a plurality of bit lines and a plurality of word lines. Each memory cell is arranged in series to a corresponding selection transistor between a ground terminal and a corresponding bit line, and each bit line can be selectively coupled to a stable power supply node providing a stable power supply voltage or to one of a first and a second memory sense node.

[0019] According to another aspect, a method of operating a memory device includes: i) asserting the bias enable signal to couple the internal bias node of the comparator circuit to the comparator bias node, and asserting the compensation signal to couple the first comparator input node and the second comparator input node to each other; ii) coupling at least one pair of selected bit lines to a regulated power supply node to precharge the selected bit lines to a target voltage; iii) when the precharge of the selected pair of bit lines expires: decoupling the pair of selected bit lines from the regulated power supply node; coupling the selected pair of bit lines to the first and second memory sensing nodes, respectively; deasserting the bias enable signal to float the internal bias node of the comparator circuit (e.g., isolating it from the comparator bias node); enabling and equalizing the sensing circuit; v) when the equalization interval expires, deasserting the compensation signal to couple the first comparator input node and the second comparator input node to the first input terminal and the second input terminal, respectively, and activating the sensing circuit; vi) when the sensing interval expires, generating a memory read signal. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] One or more embodiments will now be described by way of example with reference to the accompanying drawings, in which:

[0021] Figure 1 is a simplified circuit diagram of an example of a read architecture in a PCM memory previously proposed;

[0022] Figure 2 It was previously proposed that Figure 1 A timing diagram of a signal waveform during a read operation in an architecture of FIG. 1 ;

[0023] Figure 3 is an exemplary circuit diagram of a read architecture in a PCM memory;

[0024] Figure 4 is an exemplary circuit block diagram of a read architecture;

[0025] Figure 5 Yes Figure 4 A timing diagram of a signal waveform during a read operation in an architecture of FIG. 1 ;

[0026] Figure 6 is an exemplary circuit block diagram of a sense amplifier circuit for use in a memory device;

[0027] Figure 7 is an exemplary circuit block diagram of a set of sense amplifier circuits and associated bias circuits;

[0028] Figure 8 is an exemplary circuit diagram of a bias buffer circuit for use with a sense amplifier circuit;

[0029] Fig. 9 is an exemplary circuit diagram of a floating comparator circuit for a sense amplifier circuit; and

[0030] Fig.10 is a circuit diagram of an exemplary sense amplifier circuit. DETAILED DESCRIPTION

[0031] In the following description, one or more specific details are shown for the purpose of providing a deeper understanding of the examples of the embodiments of the present description. The embodiments may be obtained without one or more specific details, or with other methods, components, materials, etc. In other cases, well-known structures, materials, or operations are not shown or described in detail so as not to obscure certain aspects of the embodiments.

[0032] References to "one embodiment" or "an embodiment" in the framework of this specification are intended to indicate that a particular configuration, structure, or characteristic described with respect to that embodiment is included in at least one embodiment. Therefore, phrases such as "in an embodiment" or "in an embodiment" that may appear in one or more points of this specification do not necessarily refer to the same embodiment. Furthermore, in one or more embodiments, particular configurations, structures, or characteristics may be combined in any appropriate manner.

[0033] The headings / reference signs used herein are provided for convenience only and do not limit the scope of protection or the scope of the embodiments.

[0034] In the drawings attached hereto, unless the context otherwise indicates, the same parts or elements are denoted by the same reference numerals / numbers, and the corresponding description will not be repeated for the sake of brevity.

[0035] By way of introduction to the detailed description of the exemplary embodiments, reference is first made to Figure 3 , Figure 3 is an exemplary circuit diagram of the read structure of the PCM memory group 10B. It should be noted that for ease of explanation, Figure 3 A single word line 103W and a plurality of bit lines 103B can be seen in the memory bank 10B. It will also be noted that for ease of illustration, certain elements of the array of memory bank 10B are also shown only once. The bit lines are arranged into two main groups (e.g., a left group and a right group), wherein the two main groups include the same number of bit lines (e.g., 16 bit lines) to allow differential reading. During a read operation, a differential voltage is sensed between two bit lines (a SET line and a RESET line), one bit line belonging to the left group and the other bit line belonging to the right group. The two main groups are each subdivided into n subgroups or "main bit lines" 103BM, e.g., 2 subgroups, each subgroup including m bit lines, e.g., 8 bit lines. Each main bit line has an associated parasitic capacitance C MAIN(For example, the master bit line in the left group is C MAIN,L And the main bit line in the right group is C MAIN,R The bit lines 103B of the same master bit line 103BM are connected at corresponding common nodes 300 (eg, node 300L for the master bit lines in the left group and node 300 for the master bit lines in the right group). R ) and ground, and can be selectively coupled to the common node 300 via corresponding switches arranged in series with the bit lines. These switches are controlled by corresponding selection signals NYO (e.g., NYO <j>Each common node 300 (eg, each master bit line) is controlled by a corresponding switch with a voltage V REG The NYN signal may be selectively coupled to the power supply voltage rail 108 or coupled to a corresponding sensing node 302 (eg, a left sensing node 302L for a master bit line of a left group and a right sensing node 302R for a master bit line of a right group). The first master bit line in the left group and the first master bit line in the right group can be coupled to rail 108 according to the value of signal NYN<i+1> The second master bit line in the left group and the second master bit line in the right group can be coupled to rail 108, and so on (i=1...n). In a similar manner, according to the corresponding signal YN ,YN<i+1> The value of , two master bit lines can be coupled to the corresponding sensing node 302 at a time L and 302 R The sense amplifier core 30 (eg, a latch-type sense amplifier) ​​includes a differential input node 304. L and 304 R . Node 304 L The switch SW YMPS,L selectively coupled to node 302 L , and node 304 R The switch SW YMPS,R selectively coupled to node 302 R In particular, the switch SW YMPS,L and SW YMPS,R is configured to connect node 304 L and 304 R With node 302 L and 302 R Isolation, because node 302 L and 302 R A high voltage value may be used during a write operation. Node 304 L and 304 L The switch SW CONNECT The first input capacitor C of the sense amplifier 30 is selectively coupled to each other. INL having a coupling to node 304 L The first terminal is coupled to the node 306 L The second terminal of the sense amplifier 30 and the second input capacitor C INR having a coupling to node 304 R The sense amplifier core 30 further includes a first terminal coupled to the node 306 and a second terminal coupled to the node 306R. L The input terminal is coupled to node 308 R The output terminal of the first inverter INV1 is coupled to the node 308 R and 306 R A first capacitor CR between the first and second switches SW1 of the bypass inverter INV1 (eg, coupled between the input and output terminals of the inverter INV1) has a first capacitor CR coupled to the node 306. R The input terminal is coupled to node 308 L The output terminal of the second inverter INV2 is coupled to the node 308 L and 306 L A second capacitor CL is provided between the inverters INV1 and INV2, and a second switch SW2 bypassing the inverter INV2 (eg, coupled between the input and output terminals of the inverter INV2). The inverters INV1 and INV2 are enabled by the enable signal EN_SA and are typically lower than the regulated voltage V REG The voltage V DD Provide (for ease of explanation, in Figure 3 Switches SW1 and SW2 can be controlled by the same equalization signal EQ (for ease of illustration, in Figure 3 (not visible in the image).

[0036] This can be performed according to the procedure described below. Figure 3 The read operation in the memory group 10B is illustrated in FIG.

[0037] In the first step, the master bit lines (eg, all nodes 300L and 300 R ) is connected to the power supply voltage rail 108 and is powered by the voltage V REG Charge to target value V BL_CHARGE (eg, about 1.7 V). The word line signal WL applied to word line 103W is also driven to a value of V BL_CHARGE In addition, the signal YN , YN<i+1> is set so that a node 300 of the left group L The corresponding (eg, homologous, symmetric) node 300 of the right group R are connected respectively, thereby selecting the main bit line for reading. In addition, the sensing node 302 L By closing the switch SW YMPS,L The input node 304 connected to the sense amplifier core 30 L , and the sensing node 302 is closed by the switch SW YMPS,R and connected to the sense amplifier core 30 input node 304 R The remaining switches are open during the first step, i.e., all nodes 300 not corresponding to the selected master bit line are decoupled from the corresponding sense nodes 304, all bit lines are decoupled from the corresponding nodes 300 (at this stage), and the switches SW of the sense amplifier core 30 are closed. CONNECT , SW1 and SW2 are also disconnected. The enable signal EN_SA is set to a value that turns off the inverters INV1 and INV2, that is, it is de-asserted (eg, set to a low logic value, a ground voltage 0V).

[0038] In the second step, switches SW1 and SW2 are closed (e.g., equalization signal EQ is asserted) and enable signal EN_SA is asserted (e.g., set to a high logic value) to activate inverters INV1 and INV2. In this step, the input offset of the sense amplifier core 30 is compensated and the input capacitor C is INL , C INR Pre-charge.

[0039] In the third step, by signaling NYO <j>Closing the corresponding switch will connect the selected master bit line (i.e., currently connected to node 302) belonging to the left group to the selected master bit line. L The selected bit line of the left main bit line of the node 300 is connected to the corresponding main bit line (eg, node 300 L ). Similarly, by passing the signal NYO <j>Close the corresponding switch, belonging to the right group and belonging to the selected master bit line (ie, currently connected to node 302 R The corresponding selected bit line of the right master bit line of the node 300 is connected to the corresponding master bit line (eg, node 300 R By doing so, the selected bit line is precharged, i.e., the corresponding capacitor C BL From the voltage V BE (e.g., about 0.6V) to the target voltage V BL_CHARGE .

[0040] In the fourth step, by disconnecting the signal NYN The corresponding switches are controlled to disconnect the selected master bit lines (left and right) from the power rail 108.

[0041] In the fifth step, the word line signal WL applied to a certain word line 103W is driven to the ground voltage, thereby activating the selection transistor 104 and performing the actual word line selection (ie, selecting two specific memory cells 102 L and 102 R As a result, the capacitive node 300 L and 300 R (For example, the bit line capacitance C BL and the main bit line capacitance C MAIN ) starts to discharge, and node 300 L and 300 R (connected to the sensing node 302 L and 302 R ) decreases during the so-called "discharge window". L and 102 R The resistance and capacitor C INL Charged to a certain voltage V LEFT , capacitor C INR Charged to a certain voltage V RIGHT .

[0042] In the sixth step, the switch SW YMPS,L and SW YMPS,R The sense amplifier core 30 (whose input terminals are charged) is disconnected from the memory array. Switches SW1 and SW2 are also disconnected, and switch SW CONNECT In this state, the differential voltage (for example, V LEFT With V RIGHT The sensing of the difference between φ and φ) is performed by the sense amplifier core 30.

[0043] Therefore, if Figure 3 The read structure shown can rely on providing an input capacitor C INL and C INR to shift the input voltage of the sense amplifier core 30. Moreover, the corresponding reading process described above may require a discharge window with a non-negligible duration. Therefore, it is desirable to provide an improved reading structure (particularly an improved sense amplifier) ​​that does not include an input capacitor and / or reduces the duration of the discharge phase.

[0044] Therefore, one or more embodiments rely on Figure 4 The architecture shown in Figure 4 FIG. 1 is an exemplary circuit block diagram of a memory read circuit including a sense amplifier 40, which includes a floating comparator 42 (e.g., a distributed input capacitor) and a sense amplifier core (SA core) 44. In essence, the floating comparator 42 has two input nodes 402 that receive respective input signals COMP_IN_L and COMP_IN_R. L and 402 R Or, node 402 L Can be coupled to node 304 L To receive the first sensing signal YMS_LL from the selected (eg, left) master bit line (see also Figure 3 ), or coupled to the power supply voltage rail 108 to receive the regulated voltage V REG Similarly, node 402 R Alternatively coupled to node 304 R To receive the second sensing signal YMS_RR from the selected (eg, right) master bit line (see also Figure 3 ), or coupled to the power supply voltage rail 108 to receive a regulated voltage V REG Specifically, if the offset compensation signal SA_CONNECT_BL is asserted (eg, asserted low), the node 402 L and 402 R The two terminals may be connected to each other (and specifically to the node 108) via switches, and if the signal SA_CONNECT_BL is deasserted, that is, if the complementary signal SA_CONNECT_BL_N is asserted (eg, asserted low), they may be connected to the node 304 via other switches, respectively. L and 304 R These features are used for input offset compensation, as discussed further below. In addition, the floating comparator 42 receives a bias voltage from the bias node BIASPCOMP and receives a common source and common gate enable signal SA_ENAB_CASC. The bias of the floating comparator 42 can be controlled by a switch SW disposed between the bias input COMPINT and the bias node BIASPCOMP and controlled by the bias enable signal SA_ENAB_BL. BIAS The floating comparator 42 generates an amplified differential signal between its output nodes YMS_LLL_N and YMS_RRR_N, which are coupled to the input node 306 of the sense amplifier core 44. L and 306 R The sense amplifier core 44 receives the differential signal between the nodes YMS_LLL_N and YMS_RRR_N, receives the enable signal EN_SA (similar to Figure 3 ), equalize the signal EQ (similar to Figure 3 ), latch enable signal SA_LATCH, and generate output (read) signal SA_OUT.

[0045] Floating comparator 42 takes the voltage from node 304 L , 304 R The amplification of the differential signal is provided to the differential input of the SA core 44, and thus facilitates the sensing of the expected (e.g., shortening or contraction of the discharge window), such as by referring to Figure 5 For further information, Figure 5 is a time diagram including waveforms of signals YMS_LL and YMS_RR during a read operation. As illustrated, due to the pre-charge phase, signals YMS_LL and YMS_RR are initially both high (e.g., forced to a value VBL_CHARGE of approximately 1.7V). During the discharge window DW, signals YMS_LL and YMS_RR decrease at different rates due to the different resistance values ​​of the corresponding memory cells, and generate an input differential signal for the sense amplifier 40. The closer the sensing moment is expected (i.e., closer to the start of the discharge window DW at time t=0), the smaller the amplitude of the differential signal (i.e., the difference between signals YMS_LL and YMS_RR) input to the sense amplifier 40. Therefore, this amplification of the differential signal allows the duration of the discharge window DW to be reduced without affecting the read capability.

[0046] in addition, Figure 4 The architecture helps to solve the common mode (CM) input voltage problem (for example, the common mode voltage is higher than the supply voltage V of the SA core 44). DD , the common-mode voltage is variable within the VBL_CHARGE range, and / or the common-mode voltage is a function of the cell current). Implement an offset compensation mechanism (see Figure 4 is coupled to node 402 L , 402 R switches) to allow the use of small and fast devices in the implementation of the read circuit, for example to avoid mismatches and obtain high accuracy. In this regard, Figure 6 is a circuit block diagram of a sense amplifier circuit 40' that can be used as a reference to understand possible problems that may occur during reading due to offset compensation of the input of the SA core 44', regardless of the type of comparator 42'. Figure 6 , the following currents and voltages can be defined: I d is the differential current output by comparator 42', I cm is the common mode current, I L is from the first output node of the comparator 42' to the first input node 306 of the SA core 44'. L The current, I R is from the second output node of the comparator 42' to the second input node 306 of the SA core 44'. R The current, V YMS_LLL_N It is node 306 L At (e.g., node 306 L and ground), and V YMS_RRR_N It is node 306 R At (e.g., node 306 R Therefore, the following definitions and equations (1) to (4) apply:

[0047]

[0048] Regardless of the offset compensation used by the comparator 42', the offset difference before and after the equalization of the SA core 44' may be critical. cm (0) (i.e., the current I at t = 0 cm ) and zero differential current I d (0) = 0 to perform offset compensation (i.e., signal EQ is asserted and switches SW1 and SW2 are closed). At a subsequent time t = t1 (where t1>0), when equalization is removed (i.e., signal EQ is deasserted and switches SW1 and SW2 are opened), if I cm (t1)≠I cm (0) (i.e., if the common-mode current changes simultaneously due to the change of the operating point), the offset compensation may be impaired, i.e., the following condition may apply: V YMS_LLL_N (0)≠V YMS_LLL_N (t1) and V YMS_RRR_N (0)≠V YMS_RRR_N (t1). Therefore, I cm (t1)≠I cm The change of the operating point of the SA core 44' at (0) may cause the SA core 44' to commutate at t=t1, even if I d (t1)=0 (ie, false commutation). As further described below, one or more embodiments based on the floating comparator 42 may advantageously result in compensation performed at t=0 being valid also for subsequent times.

[0049] Figure 7 1 is an exemplary circuit block diagram of a group of sense amplifier circuits 40 (e.g., one sense amplifier circuit for each memory group 10B of the memory 10) and an associated bias circuit 72 (e.g., a bias comparator buffer). In particular, the bias circuit 72 is configured to receive a regulated voltage V from a node 108. REG , and generates a bias voltage BIASPCOMP for the floating comparator 42 of the sense amplifier 40 .

[0050] Figure 8 is an exemplary circuit diagram of a possible implementation of bias circuit 72. Bias circuit 72 includes a first current flow line disposed between supply voltage node 108 and ground. The first current flow line includes a diode-connected p-channel MOS transistor MP1, whose conductive path is disposed between node 108 and node 81 (e.g., transistor MP1 has a source terminal coupled to node 108, a drain terminal coupled to node 81, and a gate terminal coupled to its drain terminal). The first current flow line further includes a p-channel MOS transistor MP2 and a diode-connected n-channel MOS transistor MN1, whose conductive channels are arranged in series between node 108 and node 81 (i.e., in parallel with the conductive channel of transistor MP1), wherein transistor MP2 is configured to receive the same gate control signal as transistor MP1 (e.g., transistor MP2 has a source terminal coupled to node 108, a drain terminal coupled to node 82, and a gate terminal coupled to the gate terminal of transistor MP1; transistor MN1 has a drain terminal coupled to node 82, a source terminal coupled to node 81, and a gate terminal coupled to its drain terminal). The first current flow line further includes a diode-connected p-channel MOS transistor MP4, whose conductive channel is arranged between node 81 and node 83, and current generator 84 is coupled to node 83 (e.g., transistor MP4 has a source terminal coupled to node 81, a drain terminal coupled to node 83, and a gate terminal coupled to its drain terminal). The first current flow path further includes a current generator 84 disposed between the node 83 and ground and configured to sink a current 2*I BIAS Transistors MP1 and MP2 have the same size and receive the same control signal, so that the current I BIAS is equally divided in the two upper branches of the first current flow line (ie, the current I BIAS flows through transistor MP1 and the same current I BIAS Flows through transistors MP2 and MN1). Bias circuit 72 includes a second current flow line arranged between power supply voltage node 108 and ground. The second current flow line includes an n-channel MOS transistor MN2, whose conductive channel is arranged between node 108 and output node 85, wherein transistor MN2 is configured to receive the same gate control signal as transistor MN1 (e.g., transistor MN2 has a drain terminal coupled to node 108, a source terminal coupled to node 85, and a gate terminal coupled to the gate terminal of transistor MN1). The second current flow line further includes a p-channel MOS transistor MP5, whose conductive channel is arranged between output node 85 and ground, wherein transistor MP5 is configured to receive the same gate control signal as transistor MP4 (e.g., transistor MP5 has a source terminal coupled to node 85, a drain terminal coupled to ground, and a gate terminal coupled to the gate terminal of transistor MP4). In addition, bias circuit 72 may include a capacitor C arranged between power supply node 108 and output node 85. BIAS .

[0051] Fig. 9 is an exemplary circuit diagram of a floating comparator circuit 42 coupled to a sense amplifier core 44 in a sense amplifier circuit 40. Note that the comparator 42 and the SA core 44 operate at different voltage levels. Specifically, the SA core 44 may operate at a low voltage (e.g., between ground voltage and V DD Between DD 8V to 0.9V), while comparator 42 operates at a higher voltage (e.g., between ground and V BL_CHARGE Between BL_CHARGE Therefore, Fig. 9 The sensing architecture of is intended to amplify the differential input signal (eg, VL(t)-VR(t)) of the comparator 42 to generate an output differential current signal for the SA core 44 while using a differential input architecture. Fig. 9 One or more of the embodiments illustrated in rely on a differential input pair with input signals COMP_IN_L and COMP_IN_R applied to the source terminals of the input transistors (rather than to their gate terminals).

[0052] Specifically, the floating comparator 42 includes a low voltage p-channel MOS transistor MP CL (eg, a first input transistor) having a source terminal coupled to the first input terminal COMP_IN_L to receive a voltage V L (t), the gate terminal is coupled to the internal bias node COMPINT, and the drain terminal is configured to be coupled to the first input node 306 of the SA core 44 L (For example, to generate a first output current I L (t)). The floating comparator 42 includes a low voltage p-channel MOS transistor MP CR (e.g., a second input transistor) having a source terminal coupled to the second input terminal COMP_IN_R to receive a voltage V R (t), the gate terminal is coupled to the internal bias node COMPINT, and the drain terminal is configured to be coupled to the second input node 306 of the SA core 44 R (For example, to generate a second output current I R (t)). Transistor MP CL and MP CL , and may be the same size as transistors MP1 and MP2 of bias circuit 72. As expected, internal bias node COMPINT may be switched via switch SW controlled by signal SA_ENAB_BL. BIAS Selectively coupled to bias voltage BIASPCOMP.

[0053] Using a differential input pair that receives input signals at the source terminals is advantageous because it allows comparator 42 to be biased without the use of a charge pump (which would be required if the input signals were received at the gate terminals of the input transistors). However, since the input signals at terminals COMP_IN_L and COMP_IN_R vary in time between the moment when equalization is applied (before reading) and the moment when the input signals are sensed (during reading), transistor MP CL and MP CR The gate bias of transistor MP should not be kept at a fixed value (e.g., stably connected to BIASPCOMP), otherwise common-mode compensation will not be maintained. CL and MP CR The gate terminal of MP is coupled to the bias node BIASPCOMP before reading (e.g., during the equalization phase before the discharge window DW), and is then decoupled from the bias node BIASPCOMP during reading (during the discharge window DW) so as to "float" during the reading phase. CL and MP CR The gate terminal of the comparator 42 is floating during operation, and considering that the transistor MP CL and MP CR Strong gate-source capacitance coupling (for example, considering the parasitic capacitance CP between node COMPINT and ground and the capacitance of transistor MP CL and MP CR The gate-source capacitance is negligible), and the voltage V COMPINT (t) follows the change of common-mode input voltage, even if the input signal V L (t) and V R (t)(in Figure 5 Also referred to as YMS_LL and YMS_RR in , changes during sensing, and the bias current of the differential input pair does not change.

[0054] Specifically, if the capacitive coupling between the node COMPINT and ground is negligible relative to the capacitive coupling between the node COMPINT and the input nodes COMP_IN_L and COMP_IN_R, then the voltage V COMPINT (t) can be described according to the following equations (5) and (6):

[0055]

[0056] Where V BL_CHARGE is the regulated voltage V REG The target value, and V BIASPCOMP is the bias voltage at node BIASPCOMP (eg, generated by bias circuit 72).

[0057] Define the following small signal:

[0058]

[0059] V can then be expressed using the aforementioned equations (1) and (2) and the following equation (7): COMPINT (t), I L (t) and I R (t):

[0060]

[0061] Using equations (6) and (7), the small signal voltage at node COMPINT can be expressed using the following equation (8):

[0062]

[0063] Assume that g m It is transistor MP CL and MP CR Assuming equal transconductance (current gain) and ignoring body bias effects, the output current I can be expressed using equations (9) and (10) L (t) and I R (t):

[0064]

[0065] Using equations (9), (10), (3) and (4), the differential output current I of comparator 42 is d (t) and common-mode output current I cm (t) can be expressed using the following equations (11) and (12):

[0066]

[0067] According to equations (3), (4), (11) and (12), the common-mode current I cm (t) is constant and is defined at t = 0 (during equilibrium, when node COMPINT is connected to bias voltage BIASPCOMP); therefore, the output current I L (t) and I R (t) depends on the difference V when t>0 L (t)-V R (t) (during reading) and the constant component at t = 0 (during equalization):

[0068]

[0069] Therefore, use the reference Fig. 9 For the floating comparator discussed, the offset compensation performed at t=0 (before sensing) remains valid for t>0 (during sensing), and the floating comparator 42 amplifies the differential input voltage V L (t)-V R (t).

[0070] In one or more embodiments, transistor MP CL and MP CR Strong gate-source capacitive coupling can be achieved by appropriate design of the transistor itself. Additionally or alternatively, the floating comparator 42 may include a transistor MP coupled to the floating comparator 42. CL The capacitor C between the gate terminal and the source terminal CL , and coupled to transistor MP CR The capacitor C between the gate terminal and the source terminal CR , to increase (eg, enhance) the capacitive coupling between them. Capacitor C CL and C CR can have the same capacitance.

[0071] Alternatively, the floating comparator 42 may include a low voltage p-channel MOS transistor MP EL (eg, a first cascode transistor), whose conductive channel is connected to the transistor MP CL The conductive channels of transistor MP are arranged in series and are configured to receive the common source common gate enable signal SA_ENAB_CASC at their gate terminals. Specifically, transistor MP EL may have a transistor MP coupled to it CL A source terminal of a drain terminal of the SA core 44, a gate terminal coupled to the cascode control node SA_ENAB_CASC, and a first input node 306 configured to be coupled to the SA core 44. L (For example, to provide a first output current I L The floating comparator 42 may include a low voltage p-channel MOS transistor MP ER (eg, a second cascode transistor) whose conductive channel is connected to the transistor MP CR The conductive channels of transistor MP are arranged in series and are configured to receive a common source and common gate enable signal SA_ENAB_CASC at its gate terminal. ER may have a transistor MP coupled to it CR A source terminal of the drain terminal of the SA core 44, a gate terminal coupled to the cascode control node SA_ENAB_CASC, and a second input node 306 configured to be coupled to the SA core 44 R (For example, to provide a second output current I R When the sense amplifier 40 is turned off, the common-source common-gate transistor MP CL and MP CL It can be used to cut off the current flowing through the comparator 42.

[0072] Fig.10 It is shown Figure 4 The circuit block diagram of the detailed implementation of the read structure of FIG. 4 is a circuit diagram of the read structure of FIG. 4 , wherein the sense amplifier circuit 40 includes Fig. 9 The floating comparator 42 and Figure 6 The sense amplifier core 44 is shown in FIG. Figure 6 The SA core, in Fig.10 The architecture of SA core 44, node 306 L and 306 R The power supply voltage V may be selectively coupled to the power supply voltage V, for example, via respective switches SW3 and SW4 controlled by the signal EN_SA. DD .in addition, Fig.10 The SA core 44 includes a latch circuit 46 controlled by a signal SA_LATCH, which provides as an output a detection value SA_OUT as a function of the signals YMS_LLL_N and YMS_RRR_N.

[0073] It should also be noted that Fig.10 In the SA core 44 of the architecture, the arrangement of capacitors is different from that of Figure 3 The SA core 30 of the architecture is arranged in FIG. In particular, the first capacitor C' L Set at node 306 L and the input terminal of the inverter INV1, the second capacitor C' R Set at node 306 R The output terminal of inverter INV1 is directly connected to node 306. R , the output terminal of inverter INV2 is directly connected to node 306 L This is consistent with the fact that the differential signal received by the SA core 44 is a differential current generated by the comparator 42, which converts the differential voltage into a differential current. The two components of the differential current are injected into the output terminals of the two inverters INV1 and INV2 (where balanced currents exist due to the previous compensation). This current injection causes the latch to shift from its equilibrium state.

[0074] use Fig.10 The read operation of the read structure can therefore be performed according to the process described below, wherein the precharge process is substantially the same as in reference Figure 3 The structure is implemented as discussed above.

[0075] In the first step, the master bit lines (eg, all nodes 300 L and 300 R - See Figure 3 ) is connected to the power supply voltage rail 108 and is powered by the voltage V REG Charge to target value V BL_CHARGE (eg, about 1.7 V). The word line signal WL applied to word line 103W is also driven to a value of V BL_CHARGE In addition, the signal YN , YN<i+1> ...is set to put the node 300 of the left group L and the corresponding (eg, homologous, symmetric) node 300 of the right group R One of them is respectively connected to the sensing node 302 L and 302R, thereby selecting the main bit line for reading. In addition, the sensing node 302 L By closing the switch SW YMPS,L The input node 304 connected to the sense amplifier 40 L , and the sensing node 302 is closed by the switch SW YMPS,R The output node 304 connected to the sense amplifier 40 R In addition, the signal SA_ENAB_BL is asserted (low), causing the bias node COMPINT of the comparator 42 to be initialized to the bias value BIASPCOMP provided by the bias circuit 72. The remaining switches are open during the first phase, i.e., all nodes 300 that do not correspond to the selected master bit line are decoupled from the corresponding sense nodes 302, all bit lines are decoupled from the corresponding nodes 300 (in this phase), and the switches SW1 and SW2 of the sense amplifier core 44 are also open. During this first phase, the sense amplifier 44 is off (inactive).

[0076] In the second step, switches SW1 and SW2 are closed (asserting the equalization signal EQ). In this phase, the input offset of the sense amplifier core 44 is compensated.

[0077] In the third step, by signaling NYO <j>The corresponding switch is closed to connect the selected bit line belonging to the left group and belonging to the selected master bit line (i.e., the left master bit line currently connected to the node 302L) to the corresponding master bit line (e.g., the node 300 L ). Similarly, by passing the signal NY0 <j>By closing the corresponding switch, the corresponding selected bit line belonging to the right group and belonging to the selected master bit line (ie, the right master bit line currently connected to the node 302R) is connected to the corresponding master bit line (eg, the node 300 R By doing so, the selected bit line is precharged, i.e., the corresponding capacitor C BL From the voltage V BE (e.g., about 0.6V) to the target voltage V BL_CHARGE .

[0078] During these three initial steps, signal SA_CONNECT_BL is asserted to close the corresponding switch, and its complement SA_CONNECT_BL_N is deasserted to open the corresponding switch, so that the internal nodes COMP_IN_L and COMP_IN_R of comparator 42 are shorted between themselves and to input node 304. L and 304 R In addition, the signal EN_SA is low, so that the switches SW3 and SW4 are closed and the inverters INV1 and INV2 are disabled. The signal SA_LATCH is low, so that the latch 46 is disabled.

[0079] In the fourth step, the selected master bit lines (left and right) are disconnected from the supply voltage node 108 by opening the respective switches controlled by the signal NYN. At the same time, the signal SA_ENAB_BL switches to deasserted (high, for example, equal to V BL_CHARGE ) state, so that the switch SW BIAS is disconnected and node COMPINT is isolated from bias BIASPCOMP (ie, it is floating).

[0080] In the fifth step, the signal EN_SA switches to a high value (V DD , for example 0.9V), in order to open switches SW3 and SW4 and enable inverters INV1 and INV2. At the same time, the signal SA_ENAB_CASCODE is changed from the value V BL_CHARGE (e.g., about 1.7 V) is switched to a value V BL_CHARGE -V DD (For example, about (1.7-0.9)V=0.8V). During this phase, node COMPINT is floating and equalization is performed. In particular, offset compensation of comparator 42 begins with current I L (0) and I R (0) respectively from node 306 L and 306 R flows to the input of the SA core 44, and the comparator input is shorted to V REG (At the value V BL_CHARGE ).

[0081] In the sixth step, the word line signal WL applied to a certain word line 103W is driven to the ground voltage, thereby activating the selection transistor 104 and performing the actual word line selection (ie, selecting two specific memory cells 102 L and 102 R As a result, the capacitive node 300 L and 300 R (For example, the bit line capacitance CBL and the main bit line capacitance C MAIN ) starts to discharge, and node 300 L and 300 R The voltage at V drops during the discharge window. At this stage, the signal SA_CONNECT_BL switches to the deasserted state (e.g., high, equal to V BL_CHARGE ) to open the corresponding switch, and its complementary SA_CONNECT_BL_N switches to the asserted state (e.g., low) to close the corresponding switch, so that the internal nodes COMP_IN_L and COMP_IN_R of the comparator 42 are from the voltage V REG disconnected and respectively coupled to the input node 304 L and 304 R to receive signals YMS_LL and YMS_RR respectively. At this stage, the current I L (t1) and I R (t1) respectively from node 306 L and 306 R flows to the input of the SA core 44. Subsequently, the equalization signal EQ switches to a de-asserted state (eg, low) so as to open the switches SW1 and SW2 and perform a sensing operation.

[0082] In the seventh step, after the equalization signal EQ switches to the deasserted state, the differential signal is established as the difference between the signals YMS_LLL_N and YMS_RRR_N, which is received by the latch 46. Once the signals YMS_LLL_N and YMS_RRR_N have switched, the signal SA_LATCH is asserted (e.g., high, equal to V DD ), and provides the read output value as signal SA_OUT.

[0083] therefore, Fig.10 The read architecture illustrated in FIG. 4 may rely on providing a floating comparator 42 without using an input capacitor (eg, Figure 3 C INL and C INR ).

[0084] Thus, one or more of the embodiments illustrated herein may provide higher sensing speeds compared to conventional solutions. This higher sensing speed may be due, at least in part, to an associated shortening of the discharge window (e.g., down to approximately 1 ns compared to approximately 4 ns for the previous solution). The shortening of the discharge window may be obtained because the comparator stage amplifies the differential signal and therefore allows sensing at a reduced differential input voltage, and because there are no input capacitors to be charged, which allows for an accelerated equalization step. In addition, the higher sensing speed may be due, at least in part, to an acceleration of the commutation of the sense amplifier core, which is also allowed by the removal of the input capacitors. For example, one or more embodiments may exhibit a memory access time (from the moment a bit line is connected to its corresponding master bit line to the moment a read output value SA_OUT is available at the output of latch 46) in the range of approximately 3.8 ns to 3.9 ns, where the supply voltage V DD In the range of about 0.75V to 1.05V, and the temperature is in the range of -40°C to 140°C.

[0085] Another advantage of one or more embodiments is high sensing accuracy despite the shortened discharge window, which is facilitated by the fact that the comparator offset is compensated together with the equalization of the sense amplifier core.

[0086] Another advantage of one or more embodiments is lower power consumption relative to conventional solutions due to reduced DC current consumption as the equalization window is shortened. For example, one or more embodiments may exhibit a current consumption of approximately 0.391 μA / MHz per sense amplifier, while previous solutions may exhibit a current consumption of approximately 0.52 μA / MHz per sense amplifier.

[0087] Without prejudice to the essential principles, the details and embodiments may vary even significantly with respect to what has been described by way of example, without departing from the scope of protection.

[0088] The claims are an integral part of the technical teaching provided here regarding the embodiments.

[0089] The scope of protection is determined by the appended claims.< / j> < / j> < / j> < / j> < / j>

Claims

1. A sense amplifier circuit for a memory device, comprising: a first input terminal and a second input terminal configured to be coupled to a first memory sensing node and a second memory sensing node, respectively; A comparator circuit comprising: a first input node and a second input node; A first output node and a second output node are configured to generate a first output current and a second output current, respectively; a first input transistor having a conductive path disposed between the first input node and the first output node of the comparator circuit and a first control terminal coupled to an internal bias node; a second input transistor having a conductive path disposed between the second input node and the second output node of the comparator circuit and a second control terminal coupled to the internal bias node; wherein the first input node and the second input node of the comparator circuit are selectively coupleable to each other in response to assertion of a compensation signal and selectively coupleable to the first input terminal and the second input terminal, respectively, in response to de-assertion of the compensation signal; wherein the internal bias node is selectively coupleable to a comparator bias node in response to assertion of a bias enable signal and selectively remains floating in response to de-assertion of the bias enable signal; A sensing circuit is coupled to the first output node and the second output node of the comparator circuit, the sensing circuit being configured to generate a memory read signal as a function of a difference between the first output current and the second output current.

2. The sense amplifier circuit of claim 1 , wherein the first input transistor comprises a p-channel MOS transistor having a source terminal coupled to the first input node and a drain terminal configured to be coupled to the first output node, and the second input transistor comprises a p-channel MOS transistor having a source terminal coupled to the second input node and a drain terminal configured to be coupled to the second output node. 3 . The sense amplifier circuit of claim 1 , wherein the first input transistor and the second input transistor have the same size.

4. The sense amplifier circuit of claim 1 , wherein capacitive coupling between the internal bias node and ground is negligible relative to capacitive coupling between the internal bias node and the first and second input nodes of the comparator circuit.

5. The sense amplifier circuit of claim 1, wherein a capacitance between the internal bias node and ground is at least ten times lower than a capacitance between the internal bias node and the first and second input nodes of the comparator circuit.

6. The sense amplifier circuit of claim 1, wherein a capacitance between the internal bias node and ground is at least twenty times lower than a capacitance between the internal bias node and the first and second input nodes of the comparator circuit.

7. The sense amplifier circuit of claim 1, wherein a capacitance between the internal bias node and ground is at least fifty times lower than a capacitance between the internal bias node and the first and second input nodes of the comparator circuit.

8. The sense amplifier circuit of claim 1, wherein a capacitance between the internal bias node and ground is at least one hundred times lower than a capacitance between the internal bias node and the first and second input nodes of the comparator circuit. 9 . The sense amplifier circuit of claim 1 , wherein the comparator circuit comprises a first capacitor disposed between the internal bias node and the first input node, and a second capacitor disposed between the internal bias node and the second input node.

10. The sense amplifier circuit of claim 1 , wherein the comparator circuit comprises: a first cascode transistor having a conductive channel arranged in series with the conductive channel of the first input transistor and a control terminal configured to receive a cascode enable signal; as well as A second cascode transistor has a conductive channel arranged in series with the conductive channel of the second input transistor and a control terminal configured to receive the cascode enable signal.

11. The sense amplifier circuit of claim 1 , wherein the first input node and the second input node of the comparator circuit are further selectively coupled to a regulated power supply node that provides a regulated supply voltage in response to assertion of the compensation signal.

12. The sense amplifier circuit of claim 1 , wherein the sense circuit comprises: a first capacitor having a first terminal coupled to the first output node; a first inverter having an input terminal coupled to the second terminal of the first capacitor and an output terminal coupled to the second output node; a second capacitor having a first terminal coupled to the second output node; a second inverter having an input terminal coupled to the second terminal of the second capacitor and an output terminal coupled to the first output node; as well as A latch circuit is configured to sense a differential voltage between the first output node and the second output node and, in response to assertion of a latch enable signal, generate a digital sense signal as a function of the latch enable signal.

13. A memory device comprising: Multiple memory groups; in: Each memory bank is coupled to a respective sense amplifier circuit according to claim 1; and Each memory group includes an array of memory cells arranged as a plurality of bit lines and a plurality of word lines, wherein each memory cell is arranged in series to a corresponding selection transistor between a ground terminal and a corresponding bit line of the plurality of bit lines, and wherein each of the plurality of bit lines can be coupled to: a regulated power supply node providing a regulated power supply voltage or one of the first memory sensing node and the second memory sensing node.

14. The memory device of claim 13, further comprising a bias circuit coupled to the regulated power supply node and configured to generate a comparator bias voltage at the comparator bias node.

15. The memory device of claim 14, wherein the bias circuit comprises a first current flow line arranged between the regulated power supply node and ground, the first current flow line comprising: a first diode-connected transistor having a conductive path disposed between the regulated power supply node and an internal node; a second transistor and a third diode-connected transistor having their conductive channels arranged in series between the regulated power supply node and the internal node, wherein the second transistor has the same size as the first diode-connected transistor and is configured to receive the same control signal as the first diode-connected transistor; a fourth diode-connected transistor having a conductive path disposed between the internal node and another internal node; as well as a current generator disposed between the other internal node and ground and configured to sink current from the other internal node; wherein the bias circuit comprises a second current flow line arranged between the regulated power supply node and ground, the second current flow line comprising: a fifth transistor having a conductive path disposed between the regulated power supply node and the comparator bias node, wherein the fifth transistor is configured to receive the same control signal as the third diode-connected transistor; and a sixth transistor having a conductive path disposed between the comparator bias node and ground, wherein the sixth transistor is configured to receive the same control signal as the fourth diode-connected transistor; Wherein the bias circuit preferably comprises a capacitor arranged between the regulated power supply node and the comparator bias node. 16 . The memory device of claim 15 , wherein the first diode-connected transistor and the second transistor have the same size as the first input transistor and the second input transistor of the comparator circuit.

17. The memory device of claim 15, further comprising an operation mode configured to: i) asserting the bias enable signal to couple the internal bias node of the comparator circuit to the comparator bias node, and asserting the compensation signal to couple the first input node and the second input node to each other; ii) coupling at least one pair of selected bit lines to the regulated power supply node to precharge the selected bit lines to a target voltage; iii) upon expiration of the precharge period of the selected pair of bit lines: decoupling the selected pair of bit lines from the regulated power supply node; coupling a selected pair of bit lines to the first memory sense node and the second memory sense node, respectively; deasserting the bias enable signal to float the internal bias node of the comparator circuit; and enabling and equalizing said sensing circuit; iv) upon expiration of an equalization interval, de-asserting the compensation signal to couple the first input node and the second input node to the first input terminal and the second input terminal, respectively, and activating the sensing circuit; as well as v) When the sensing interval expires, generating the memory read signal.