Improvements to and regarding flash memory
By generating a reference voltage from evenly distributed memory cells with the same size and capacity, the solution addresses variability in flash memory arrays, improving read accuracy and extending lifespan while reducing production costs and interference susceptibility.
Patent Information
- Application Number
- DE102019118117
- Authority / Receiving Office
- DE · DE
- Patent Type
- Patents
- Current Assignee / Owner
- Priority Date
- 2018-07-13
- Filing Date
- 2019-07-04
- Publication Date
- 2024-11-28
- Estimated Expiration
- 2039-07-04
AI Technical Summary
Flash memory arrays face challenges in maintaining operational robustness due to inherent variability in memory cell responses under varying process, voltage, and temperature conditions, leading to reduced data retention and increased susceptibility to interference, which affects the memory's lifespan and production costs.
Generate a reference voltage (VREF) using a plurality of memory cells in different states, evenly distributed across the array, with the same size and capacity as other cells, to create a centered and narrow distribution that compensates for manufacturing variations and uneven cell distributions, thereby improving read accuracy and extending the memory's lifespan.
The solution enhances memory robustness by maintaining a stable reference voltage distribution, reducing calibration time, and extending the memory's operational life while minimizing production costs and interference susceptibility.
Smart Images

Figure 00000000_0001_ABST 
Figure 00000000_0000_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present invention relates to the field of memory design. In particular, the invention relates to a non-volatile flash memory and the conversion of analog memory data to digital memory data. BACKGROUND
[0002] The demand for low-cost, yet high-performance embedded flash memory is constantly increasing. Such flash memory is expected to operate over an ever-expanding voltage and temperature range and achieve ever-larger data retention while also reducing susceptibility to noise. This is expected to be achieved at ever-decreasing costs.
[0003] One factor that influences the cost of flash memory is the time required to test and calibrate the memory. Reducing this time typically results in lowering the memory's production costs. The test and calibration time depends in part on optimizing the memory's read path architecture.
[0004] With reference to Fig. 1, a flash memory 101 is typically incorporated into an array structure with a number of rows and columns. Rows are referred to as word lines 102, while columns are referred to as bit lines 103. Each memory cell 104 of the array has an address or coordinate (X, Y) that designates its position in the array. Each memory cell can be selected by applying a combination of the addresses or coordinates to X and Y decoders 105, 106. The array is divided into N sticks 107a, 107b, 107n, each stick comprising a predetermined number of bit lines connected to a respective multiplexer 108a, 108b, 108n, which in turn is connected to an input of a respective comparator 109a, 109b, 109n. The purpose of the stick / multiplexer arrangement is to reduce the number of comparators required in the array, thereby reducing the complexity of the array architecture.Comparators 109a, 109b, and 109n are responsible for reading the memory cells. The output of each comparator is a digital bit (i.e., a 1 or a 0). Collectively, the N bit lines (i.e., the predetermined number of bit lines connected to multiplexers and comparators) are referred to as a digital bus. The comparators compare an input signal with a reference signal and output a signal indicating which is larger. If the input signal is larger than the reference signal, the output corresponds to a "1" (i.e., a written state). If the input signal is smaller than the reference signal, the output corresponds to a "0."
[0005] The main operating modes of flash memory are write, erase, and read. Write and erase modes are used to store data, while the read operation is used to convert the analog information of the state of the memory cells into digital values.
[0006] A basic feature of non-volatile memory transistor cells is the ability to shift their threshold voltage by applying a high voltage to their terminals. The threshold voltage is the minimum gate-source voltage required to create a conductive path between the source and drain terminals of a transistor.
[0007] During the erasing process, the threshold voltage is set to an erased state (V TH_E ), and during the write process the threshold voltage is set to a programmed state (V TH_P), which is also known as a written state. In both cases, during erase and write operations, the high voltage is applied to a group of selected memory cells. The smallest number of memory cells that can be erased simultaneously is all the memory cells within a word line. Another erase mode is "flash (or bulk) erase," where all memory cells of the entire array are erased at the same time. Once a memory cell has been erased, the write mode can be performed. Any memory cell on a selected word line can be changed to a programmed state or left in a previous state.
[0008] With reference to Fig. 2, during the read operation, a particular memory cell can be selected / addressed based on its (X, Y) coordinate by activating the gate of the selected transistor of a word line and activating the current flow of the selected bit line to a sense amplifier 201. A separate bit line, called the reference bit line, is used to provide a reference threshold voltage value (V REF ) that is located between the deletion (V TH_E ) and programmed (V TH_P ) voltage values. The purpose of the reference threshold (V REF ) is to provide a reference value against which the deleted (V TH_E ) and (V TH_P ) programmed values can be compared to determine the state of a given cell.
[0009] As mentioned above with reference to Fig. As described in Figure 2, in a commonly used readout approach, the sense amplifier(s) 201 are implemented as a comparator (e.g., a latched comparator) 202 that compares the voltage levels generated by integrating the input signal current 203 (i.e., from the memory cell being read) with the voltage level generated by integrating the reference current 204 of the reference bit line. Each of the N sticks has the same sense amplifier structure. As described above, the sticks are collectively referred to as a digital bus.
[0010] Each memory cell has a slightly different threshold voltage value (and thus different discrete erase (V TH_E ) and programmed (V TH_P) voltage values) resulting from, for example, manufacturing inconsistencies and defect locations. One of the reasons for the inherent variability in the operating response of each cell is that the minimum transistor sizes required for proper memory functioning result in different cell responses to different process, voltage, and temperature conditions. For example, the resistance of one cell might change more or less than the resistance of another cell under the same temperature change. Due to the large number of memory cells present in a given array, it is not possible to precisely determine or calibrate the array to operate at these unique discrete erase (V TH_E ) and programmed (V TH_P ) voltages for each memory cell.
[0011] Instead, the discrete erase and programmed voltages of each cell are considered to be distributed around the mean values of V TH_E and V TH_P in distributions in the form of a Gaussian curve, each of which has a standard deviation σ TH_E and σ TH_P (which may or may not be identical to each other). This is shown in distributions 300 and 400 in Fig. 3 and Fig. 4. The distribution 301 / 401 is a Gaussian distribution around the mean of V TH_E and the distribution 302 / 402 is a Gaussian distribution around the mean of V TH_P .
[0012] Likewise, the cells in the reference bit line, which are used to generate the reference voltage (V REF ) suffer from inherent variability and therefore do not always produce the same reference voltage. Therefore, the reference voltage can also be around an average value of V REFdistributed and a standard deviation of σ REF This is shown in the distribution 303 in Fig. 3. The respective standard deviations define the width of the respective distributions.
[0013] The separation between the maximum level (V H ) the V TH_E distribution and the minimum level (V L ) the V TH_P distribution, the effective memory cell window (MW EFF ) 403 (see Fig. 4).
[0014] If the reference voltage value (V REF ), which is output by the reference bit line, is always between V H and V L (ie the maximum or minimum level of the V REF -distribution never overlaps with V H or V L ), a read operation, if initiated, will be successful.
[0015] Due to memory cell transistor reliability effects, such as data retention, endurance cycling, and noise effects, the response of the memory cells under different process, voltage, and temperature conditions becomes less consistent over time, causing the standard deviation σ TH_E and σ TH_P the V TH_E - and V TH_P distributions become larger and / or cause the means of one or both distributions to shift in position over time. These effects cause the MW EFF deteriorates over time (i.e. the size of the separation between V H and V L , into the V REFdistribution without overlap decreases). If it is no longer possible to avoid overlap, some read operations will fail, and the memory is said to have reached its end of life (EOL). It is desirable to minimize the effects of MW degradation EFF to overcome in order to increase the lifespan of the memory.
[0016] In an ideal situation, the greatest operational robustness (i.e. consistency of response under a wide variety of process, voltage and temperature ranges) can be achieved, and thus the effect of the MW EFF Degradation can be reduced if each cell in an array has its own uniquely calibrated reference bit line, which is specifically calibrated to ensure that the output V REF between V H or V LHowever, this would not be practical given the very large number of additional cells that would be required. Furthermore, when the number of cells is large, it is difficult to manufacture all cells to be identical. As such, there is an inherently large variability in responses to different process, voltage, and temperature ranges in a single memory device. This makes it difficult to estimate the standard deviations σ TH_E and σ TH_P the V TH_E - and V TH_P distributions. In fact, the V TH_E - and V TH_P -Distributions usually broad relative to MW EFF and continue to increase at the end of the storage life.
[0017] One option, described in US2011 / 194330 A1, is to try to estimate the standard deviation σ REF the reference voltage V REF In US2011 / 194330 A1 this is achieved by V TH_E and V TH_Pof two individual cells (also described as memory elements), each in a different state (i.e., one memory element in an erased state and one in a written / programmed state). V REF to read is the numerical average of these two values.
[0018] Another technique, described in US 6 687 150 B1 (Joachim et al.), is to use a plurality of differently sized cells (i.e. cells with different channel lengths (e.g. higher channel lengths) or different channel widths or a combination of both, so that they have a different, higher cell capacitance and switching behavior compared to other memory cells in a memory array) within a word line in order to achieve as close as possible to a mean level of V REFThe size of the reference cells of US 6 687 150 is chosen to produce a higher cell capacitance compared to the regular memory cells of the array, to help distinguish between the different states (programmed or erased) of regular memory cells. We also note that while Fig. 1 of US 6 687 150 B1 (Joachim et al) shows a memory cell using a capacitor 140, the embodiments described herein do not require such a capacitor.
[0019] US 2013 / 0 148 429 A1 discloses a memory device comprising a memory cell coupled to a switch such that the switch controls the timing of read or write operations in the memory cell in response to a control signal. SUMMARY
[0020] This invention provides a flash memory and a method of operating a flash memory as set out in the appended claims.
[0021] Exemplary embodiments of the present invention will now be described, by way of example only, with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS Fig. 1 shows a known flash memory array structure. Fig. 2 shows a well-known digital bus architecture. Fig. Figure 3 shows a graph illustrating voltage distributions during a memory read operation. Fig. Figure 4 shows a graph illustrating the effective memory window during a memory read operation. Fig. Figure 5 shows a diagram illustrating the relationship between the end of memory lifetime and the width of a reference voltage distribution. Fig. Figure 6 shows a diagram illustrating the relationship between the end of memory lifetime and the width of a reference voltage distribution that has been narrowed. Fig. Figure 7(a) shows a digital bus architecture. Fig. Figure 7(b) shows a digital bus architecture according to an embodiment of the invention. Fig. Figure 8(a) shows a graph illustrating a yield loss where a broad reference voltage distribution overlaps with the voltage distributions of the erased and programmed states. Fig. Figure 8(b) shows a graph illustrating yield loss where a narrow reference voltage is not centered between the voltage distributions of the erased and programmed states. Fig. Figure 9 shows a graph illustrating a narrowed, centered reference voltage distribution obtained using an embodiment of the invention. DETAILED DESCRIPTION
[0022] We attempt to provide a method to achieve memory robustness by reducing the effects of MW EFF degradation caused by transistor reliability effects under different process, voltage, and temperature ranges. This can be achieved by using a reference voltage V REF provided which is formed from a plurality of memory cells in a plurality of reference bit lines, each cell being in different memory states (ie, erased or programmed) and each of the plurality of memory cells having the same size (eg, same channel length and width and construction and thus the same cell capacitance) as the rest of the memory cells in a memory array.
[0023] As described above, there is an inherent variability in the response that different memory cells have when operating under a range of process, voltage, and temperature regimes. As the memory ages, this variability increases, and this reduces the robustness of the memory as the width of MW EFF decreases.
[0024] The inventors have recognized that the technique used in US 6687150 B1 to obtain an average V REF with a narrow σ REF to receive, no V REF near the middle between V H and V L generated. When the center of the V REF -Distribution too far from the center and too close to V H or V L positioned, it does not matter how narrow it is, because the memory begins to fail as soon as there is an overlap between the V REF -Distribution and the V H - or V L -values. If V REFis very close to the V H - or V L -values when MW EFF begins to narrow, the end of its life is often reached very quickly.
[0025] One of the causes of an uncentered V REF -Distribution in US 6 687 150 B1 is that the reference cells used to determine V REF to generate, have a different size and therefore cell capacity compared to the cells of the rest of the storage. These differently sized reference cells have different response characteristics compared to the other cells of the array when operating under varying process, voltage, and temperature ranges. This results in the center of the V REF -Distribution generated by the different sized reference cells at a distance from the center of the V TH_E - and V TH_P distributions (ie the mathematical mean between V H and V L). In particular, depending on the operating conditions, it will be positioned closer to either V H or V L be.
[0026] An example of a process variation effect is one that arises as a result of poly-etching techniques used during manufacturing. In particular, tolerated manufacturing variations caused by the poly-etching technique lead to variations in the effective channel length of different cells (and thus how they respond under different conditions); the shorter the length, the greater the influence a manufacturing variation can have on the cell response. Because reference cells have a different channel length compared to the memory cells in US 6 687 150 B1, the influence that the process variation has on the response of the shorter-length cells is different from the influence the variations have on the longer-length cells. Similarly, other manufacturing techniques that allow tolerated manufacturing variations may suffer from similar types of process dependencies.
[0027] An example of a voltage variation effect is that the threshold voltage of a transistor depends on its channel length. Different channel lengths result in different voltage dependencies (i.e., different response characteristics under different voltages).
[0028] The inventors have recognized that this disadvantage can be overcome by using a V REF-distribution is not generated by separate reference cells of larger cell capacity, but by memory cells taken from the rest of the cells that make up the memory array and thus have the same size (e.g. channel length, width and construction and thus cell capacity) as the other cells in the array. This results in all cells having the same process, voltage and temperature dependence. This is in contrast to the differently sized reference cells of US 6 687 150 B1. As described above, it is achieved by using memory cells with the same cell size (e.g. channel length, width and construction and thus capacity) as the other array cells to generate the V REFDistribution means that both the reference cells and the read cells have a much closer dependence and operating response to changing process, voltage, and temperature ranges than reference cells with different cell sizes and thus capacities. Since all cells (i.e., reference cells and the other cells in the array) are the same size, they will inherently respond in the same way to different conditions, and the V REF -Distribution will be much closer to the actual center of MW EFF , ie the mathematical middle between V H and V L over a much wider set of process, voltage and temperature ranges.
[0029] The inventors have also recognized that the technique described above can be implemented by generating a V REFDistribution using a mean calculated from memory cells selected to be evenly distributed across the memory array and in a combination of erased and programmed (i.e., written) states. The total number of cells from which the mean is calculated is three or more. This is in contrast to the technique used in US 2011 / 194330 A1, which uses a mean V REF using only two individual cells positioned at individual points in the array (i.e., with individual point characteristics that depend on their position in the array), one in an erased state and one in a written / programmed state. Although the technique used in US 2011 / 194330 A1 has a narrow V REFdistribution, it cannot compensate for the scenario where there is an uneven distribution of cells in erased and programmed states around the memory array. The uneven distribution of cells in different states around a memory array leads to a shift in the positions and / or widths of the V TH_E - and V TH_P distributions. An example reason for this is that tolerated manufacturing process variations may have resulted in certain areas of the array having longer channel lengths compared to other areas and thus exhibiting different response characteristics. Thus, even if V REF narrow when the positions and widths of the V TH_E - and V TH_P -distributions have shifted due to an unequal distribution to such an extent that they are at odds with V REFoverlap, the memory may still start to fail and reach the end of its lifespan sooner. Creating a V REF Distribution using an average calculated from multiple reference bit lines constructed from a plurality of memory cells taken from positions evenly distributed throughout the array overcomes this disadvantage.
[0030] The inventors have also recognized that the techniques described above can be further improved by changing the ratio of cells in erased to programmed state in the plurality of memory cells used to store the V REFdistribution. This ratio can be defined during a memory array characterization step (e.g., during array fabrication and calibration). Once the ratio is defined, it can be used during all write operations. As described above, the voltages that determine the V TH_E distribution (ie deleted cells) lower than the voltages that the V TH_P distribution (ie written / programmed cells). By changing the ratio, the V REF -distribution (which is generated as the mean of the higher V TH_P -Voltages and lower V TH_E ) have a center position shifted up or down. By including several deleted cells, the center of the V REF -Distribution closer to V L In contrast, by including several programmed (ie written) cells, the center of the V REF-Distribution closer to V H delay.
[0031] The techniques described above, individually and together, contribute to allowing the center of the generated V REF -Distribution much closer to the mathematical middle between V H or V L positioned than is possible with the technique of US 6687150 B1, while at the same time the means for shifting the center of the V REF distribution, if necessary, to respond to unequal distributions of erased and written cells, which is not possible using the technique of US 2011 / 194330 A1.
[0032] The effect of the techniques described above is in the Fig. 5 and Fig. 6. In the graph 500, the reference distribution 501 is relatively broad, so that it reaches the highest voltage in the V TH_E -Distribution 502 and the lowest voltage in the V TH_Pdistribution 503, which means that the end of life is reached much earlier. In addition, although in Fig. 5 or Fig. 6 not shown, an overlap may also occur earlier if the reference distribution is not centered, resulting in the end of life being reached earlier.
[0033] In contrast, the graph 600 shows where the end of life could occur with a narrower reference distribution 601 that is more closely centered than in the graph 500. Even if the size of MW EFF decreases over time, the reference distribution 601 is narrow enough and centered enough to be between the highest voltage in the V TH_E -Distribution 602 and the lowest voltage in the V TH_P-Distribution 603 for a much longer period of time. In other words, the time between the start of the memory's lifetime and the end of the lifetime is increased, which helps enable the memory designer to more easily meet a selected design goal.
[0034] Fig. 7(a) and (b) illustrate the difference between a known digital bus 700 and a digital bus 701 according to an embodiment of the invention. Fig. 7(a) is identical to the known digital bus 201 shown in Fig. 2 is shown. Fig. Figure 7(b) shows a digital bus 701 with a reference block 702 configured to receive signals from at least three memory cells selected from evenly distributed positions around the array. A first number of the selected memory cells are in an erased state, and a second number of the memory cells are in a programmed state. The ratio of the first number to the second number is a factor in determining the center position of the V REF distribution. The ratio can be changed (e.g., during an array characterization step during manufacturing and calibration) to adjust the center position of the V REF distribution. To change the ratio, a write operation can be performed on the memory cells of the reference block. If the position of V REFhigher, write operations can be performed such that more cells in the reference memory block are set to their written state. Conversely, if the position of V REF lower, write operations can be performed such that more cells in the reference memory block are set to their erased state. The total number of selected cells is a factor in determining the width of the V REF Distribution, as described below. The total number of cells in the reference memory block and the ratio of written cells to erased cells are intended to be determined during an array characterization step during manufacturing and calibration. The output from the reference element block 702 is the input to each of the comparators of the digital bus 701.
[0035] Fig. 8(a) and Fig. 8(b) show graphs 800, 801 of stress distributions obtained using known techniques. The graph 800 in Fig. 8(a) shows V TH_E -, V TH_P - and V REF -Distributions 802, 803, 804, which correspond to Fig. 3 and show a yield loss 805 (ie, failed memory reads), with the distributions overlapping. The graph 801 in Fig. 8(b) shows V TH_E -, V TH_P - and V REF -Distributions 806, 807, 808, where the V REF -Distribution 808 is narrowed but not in the middle between V H or V L centered. Accordingly, it is too close to the edge of the V TH_E distribution 806, which leads to overlap and yield loss 809.
[0036] In contrast, Fig. 9 shows a graph 900 of V TH_E -, V TH_P -, and V REF-distributions 901, 902, 9034, obtained using the techniques described above. The V REF -Distribution 903 is narrowed and is between the V TH_E -, V TH_P -Distributions 901, 902 are centered without overlap, thus guaranteeing successful reads without loss of yield.
[0037] A flash memory according to the techniques described above comprises a first plurality of memory cells, each memory cell of the first plurality of memory cells being selectively connected to a first input of a comparator, where selectively connected can be understood to mean that each cell can be addressed and read accordingly using an X, Y decoder. A second plurality of memory cells is selectively connected to a second input of the comparator. In contrast to the first plurality of memory cells, the second plurality of memory cells are all connected to all second inputs of each of the comparators of the memory (where multiple comparators are present). A first number of the second plurality of memory cells are in an erased state. A second number of the second plurality of memory cells are in a written state.Each memory cell of the first plurality of memory cells and each memory cell of the second plurality of memory cells has the same cell size and thus capacity. The sum of the first number and the second number is at least three. Each of the memory cells of the first plurality of memory cells is capable of generating a first signal to be received at the first input of the comparator (i.e., the signal associated with the cell(s) being read). The second plurality of memory cells together generate a second signal to be received at the second input of the comparator (i.e., the reference voltage signal).
[0038] It is contemplated that all memory cells of the second plurality of memory cells are in at least one erased or written state. However, it is also possible to implement the invention where this applies only to a majority (but not all) of the memory cells of the second plurality of memory cells. This allows the techniques described above to be combined with existing techniques to create a hybrid system.
[0039] As described above, during a read operation, a digital bit is captured by comparing the reference voltage signal at the comparator with an input voltage signal received from the memory cell being read. Because there are multiple comparators, multiple digital bits (i.e., N digital bits) can be captured at the same time. As described above, due to the inherent variability in response under different process, temperature, and voltage operating conditions, both the reference signal and the N input signal voltages will be distributed around a respective mean and have a respective standard deviation.
[0040] Each of the N signal voltages is calculated by integrating the signal current (I H or I L, depending on the state of a given cell) over the capacity for a predetermined time, which generates a voltage at the end of the integration time associated with the cell read.
[0041] The reference voltage V REFis determined in the same way by taking the integrated current across the integrated capacitance for each of the memory cells selected as reference cells and taking an average of the output values. As described above, the memory cells selected as reference cells are selected from positions evenly distributed around the array, are in a combination of erased and programmed states, and are three or more in number. In the extreme case where there are only three, the average will be heavily skewed to one side because the number of reference memory cells in one state is twice or half the number of cells in the other state (e.g., two cells in the written state and one in the erased state will consume the V REF -Middle position clearly to the mean of V TH_PAs described below, although this will not be useful in all cases, the larger possible shift of the reference voltage means that it can still be used to shift the center position by a very large amount without having to provide a separate biasing device. Conversely, at the other extreme, a very large number of memory cells in different states selected as reference cells provides a higher granularity of control to make small, subtle shifts in the center position of V REF to produce.
[0042] If the input signal voltage is higher than the reference voltage, the comparator outputs that the cell is in its programmed state (i.e., a digital "1"). If the generated signal voltage is lower than the reference voltage, the comparator outputs that the cell is in its erased state (i.e., a digital "0"). It is assumed that due to the mismatch variation of the signal current and capacitance (i.e., the signal currents and cell capacitances also exhibit some variability and are thus also distributions with a respective mean and standard deviation), the integrated signal voltages would, in practice, exhibit mismatched distributions. In other words, the width of the V TH_E - and V TH_P -distributions are not necessarily the same in practice. However, for this explanation, it is assumed that both distributions have the same standard deviation value σ.
[0043] The ratio of cells in erased state to cells in programmed state can be adjusted during manufacturing with a write operation to adjust the width and position of the V REF distribution to adapt it to whatever variations in the V TH_E - and V TH_P distribution to minimize any overlap and thus provide a greater degree of control during a memory array characterization step during manufacturing.
[0044] As a backup option to act as an additional degree of freedom to assist in positioning the center of V REF , a suitable scaling factor (also known as a bias factor) can be set to V REF The purpose of the scaling factor can be to determine the center position of V REF to move so that they are between V H and V Lfalls. In practice, this can be done by scaling the reference capacitance or the current.
[0045] As described above, the reference block consists of several bit lines, which are described herein as reference bit lines, but which are otherwise the same as the bit lines of the rest of the array. These are connected in parallel M times to the N comparators (see Fig. 7(b)). In particular, the outputs of the reference bit lines are all connected to each other and to the reference input of each of the comparators, so that the reference bit lines are connected in parallel to each corresponding reference input of the plurality of comparators. The effect of this is to reduce the standard deviation σ REF the V REF distribution by a factor of √M, ie the V REF -Standard deviation is now σ REF / √M. The larger the value of M, the narrower the V REF -Distribution. In Fig.7(b) the reference bit lines are labeled to run from M-1 to 0.
[0046] If the width of the V REF distribution becomes narrower, the time until the end of life is reached is extended because the memory cells can continue to be read successfully for longer, as the narrower V REF -distribution does not agree with V H or V L overlapped until a longer period of time has passed and a greater degree of deterioration has occurred.
[0047] Furthermore, the more reference bit lines and thus memory cells there are in the reference block, the less effect parasitic capacitances and chip-to-chip variations (e.g., minor structural differences within manufacturing tolerances) will have on the operation of the memory, since these will be averaged out.
[0048] An additional advantage of the techniques described above is that the calibration time of a memory can be significantly reduced, as it is much easier and faster to determine the position and width of the V REF -Distribution than is possible using known techniques. This reduces overall production time and costs without sacrificing memory performance in terms of process, voltage, and temperature ranges, data retention, and susceptibility to disturbances.
[0049] The invention is particularly applicable to cases where the V TH_E - and V TH_P -Distributions of the memory cells are very broad, so that the separation between V TH_E and V TH_P very small and MW EFF is very narrow.
[0050] As described above, the ability to shift the center position of the V REF-Distribution by changing the ratio of erased to programmed cells in the plurality of cells used to generate the V REF -distribution is discussed in the context of overcoming the effects of unequal distributions of erased and programmed cells in an array. However, the same technique can also be used to compensate for other technological effects, such as variations in process, voltage, and temperature ranges resulting from manufacturing techniques or array operation, which may result in a change in the position and / or width of the V TH_E - and V TH_P distributions. In particular, the techniques described above provide the ability to always find the center of the V REF -Distribution exactly in the middle of the V TH_E - and V TH_P -distributions by setting an adjustable center position of the V REF -distribution is provided.
[0051] Furthermore, as described above, the components used to generate V REFused memory cells are selected so that they are evenly distributed across the memory array. One reason for this is that cell position is a factor that contributes to cell response under different operating conditions (i.e., varying process, temperature, and voltage ranges). Cells positioned close together on a chip are more likely to have a similar operating response compared to cells positioned far apart on a chip. One reason for this is that minor manufacturing defects are likely to be local to certain positions and will thus affect local cells equally. By using a variety of cells evenly selected from across different positions in the array (and thus across the chip), the response variation of memory cells near the selected reference cells is inherently taken into account when determining an average V REFThe term "uniformly distributed" means that the selected cells are positioned at approximately equal distances from each other in the array.
[0052] If parts of a chip are known to have a specific effect on the cell response, it is also possible to select a variety of cells that are less evenly distributed across an array (e.g., to avoid an area). For example, reference memory cells can be selected to provide a V REF that is more strongly correlated with specific sections of the array. This provides another degree of freedom in positioning the center of the V REF -Distribution ready.
Claims
[1] Flash memory, comprising: a first plurality of memory cells, each memory cell of the first plurality of memory cells being selectively connected to a first input of a comparator; a second plurality of memory cells (702) selectively connected to a second input of the comparator, wherein a first number of the second plurality of memory cells (702) are in an erased state; wherein a second number of the second plurality of memory cells (702) are in a written state; wherein each memory cell of the first plurality of memory cells and each memory cell of the second plurality of memory cells has a first cell capacitance; wherein the sum of the first number and the second number is at least three, and wherein each memory cell of the first plurality of memory cells and each memory cell of the second plurality of memory cells (702) has a first size comprising a first channel length and a first channel width. [2] The flash memory of claim 1, wherein the first plurality of memory cells and the second plurality of memory cells (702) are distributed in an array, and wherein the second plurality of memory cells (702) are located in positions evenly distributed throughout the array. [3] The flash memory of any preceding claim, wherein the second plurality of memory cells (702) is configured to generate a reference voltage. [4] The flash memory of claim 3, wherein the comparator is configured to compare a first voltage received at the first input with the reference voltage received at the second input. [5] The flash memory of claim 3 or 4, wherein the second plurality of memory cells (702) is configured to generate the reference voltage by integrating a signal current received by each memory cell of the second plurality of memory cells (702) for a predetermined time. [6] Flash memory according to one of the preceding claims, comprising a plurality of comparators, wherein the first plurality of memory cells is divided into a first plurality of bit lines, each connected to a respective first input of a respective comparator of the plurality of comparators; and wherein the second plurality of memory cells (702) is divided into a second plurality of bit lines connected in parallel to each respective second input of the plurality of comparators. [7] A method of operating a flash memory, the method comprising: Receiving a first signal from a first memory cell of a first plurality of memory cells at a first input of a comparator; Receiving a second signal from a second plurality of memory cells (702) at a second input of the comparator; wherein a first number of the second plurality of memory cells (702) are in an erased state; wherein a second number of the second plurality of memory cells (702) are in a written state; wherein each memory cell of the first plurality of memory cells and each memory cell of the second plurality of memory cells (702) has a first cell capacitance; and wherein the sum of the first number and the second number is at least three, wherein each memory cell of the first plurality of memory cells and each memory cell of the second plurality of memory cells (702) has a first size comprising a first channel length and a first channel width. [8] The method of claim 7, wherein the first plurality of memory cells and the second plurality of memory cells (702) are distributed in an array, and wherein the second plurality of memory cells (702) are in positions evenly distributed throughout the array. [9] The method of claim 7 or 8, comprising generating the second signal with the second plurality of memory cells (702), wherein the second signal is a reference voltage. [10] The method of claim 9, wherein generating comprises integrating a signal current received from each memory cell of the second plurality of memory cells (702) for a predetermined time. [11] A method according to any one of claims 7 to 10, comprising comparing the first signal and the second signal at the comparator.
Citation Information
Patent Citations
Memory array with read reference voltage cells
US20110194330A1
Reference voltage generation for memory circuits
US6687150B1
Memory device, method of performing read or write operation and memory system including the same
US20130148429A1
