Embodiments of the present invention relate to methods of testing and repairing a cell array having a plurality of memory cells and to memory devices.
Conventional non-volatile memory devices broadly fall into two different categories, floating gate and trapping layer. Trapping layer memory cells, such as SONOS-memories or NROM-memories, have an n-channel FET with the gate dielectric replaced by a trapping layer and two barrier layers sandwiching the trapping layer. The trapping layer is the storage element of the memory cell. The barrier layers inhibit direct tunneling of charge carriers from and to the non-conductive trapping layer. Floating gate memory cells include usually an n-channel FET with a floating gate sandwiched between a dielectric tunnel layer separating the floating gate from the transistors channel region and a dielectric barrier layer between the floating gate and the control gate that is connected to the address circuitry. The floating gate is the storage element of the memory cell. The tunnel dielectric and the dielectric barrier layer insulate the conductive floating gate. A memory cell may be programmed by injecting charge carriers into the trapping layer or the floating gate from either the control gate or the channel region. The memory cell may be erased by removing or compensating the previously injected charge. The embodiments of the current invention are not limited to these two different categories of non-volatile memories and may cover other types of solutions.
Memory cells may be based on a binary or a multi-level programming/sensing scheme, for example a 4-bit per cell programming/sensing scheme. According to the binary programming/sensing scheme, the n-channel FET of a trapping layer or floating gate memory switches from a non-conductive state to a conductive state, when a read voltage applied to the control gate exceeds a threshold voltage. The n-channel FET returns to the non-conductive state, when the read voltage falls below the threshold voltage. A negative charge that is stored in the storage element functions as a negative bias of the control gate and shifts the threshold voltage towards higher voltage values.
The state of the memory cell is detected by applying a suitable read voltage to the control gate and checking whether the FET is conductive or not. In this case, the read voltage is selected such that, on one hand, the read voltage is high enough to ensure that all erased memory cells are conductive and that, on the other hand, the read voltage is low enough to ensure that none of the programmed memory cells is non-conductive.
According to a 4-bit per cell programming/sensing scheme, three different amounts of charge are injected into the same area or equivalent amounts of charge are stored in different locations of the storage element. Three different threshold voltage levels are used to define four different ranges or states. Each of the four different ranges or states represents a unique arrangement of a pair of bits (e.g., 00, 01, 10 or 11).
The threshold voltage of each memory cell depends on, to some extent, geometric and physical properties of the respective memory cell. For example, channel length, channel doping profile, and/or barrier layer thickness vary from memory cell to memory cell, due to fluctuations or imperfections of process steps required in course of the fabrication of the memory device. As a result, the threshold voltage varies from memory cell to memory cell. For each cell array with a plurality of memory cells, a characterizing threshold voltage distribution results.
Conventionally, a memory cell that exhibits a threshold voltage outside a predetermined range either for the erased state or for the programmed state is classified as being defective and is replaced by a redundant repair cell if applicable. However, identifying memory cells that might affect the performance of the memory device during lifetime remains a general design and test requirement.
One embodiment refers to a method of testing a cell array. In a distribution section near a distribution edge, a distribution gradient of a threshold voltage distribution of the cell array is determined by varying a read voltage applied to the cell array. The distribution gradient in the distribution section is compared to a limit gradient. If the distribution gradient exceeds the limit gradient, an inner distribution edge is determined by a read voltage, at which the distribution gradient exceeds the limit gradient. Memory cells that exhibit a threshold voltage in a range between the distribution edge and the inner distribution edge may be classified as being defective.
The accompanying drawings are included to provide a further understanding of the present invention and are incorporated in and constitute a part of this specification. The drawings illustrate the embodiments of the present invention and together with the description serve to explain the principles of the invention. Other embodiments of the present invention and many of the intended advantages of the present invention will be readily appreciated as they become better understood by reference to the following detailed description. The elements of the drawings are not necessarily to scale relative to each other. Like reference numerals designate corresponding similar parts.
In the following Detailed Description, reference is made to the accompanying drawings, which form a part hereof, and in which is shown by way of illustration specific embodiments in which the invention may be practiced. In this regard, directional terminology, such as “top,” “bottom,” “front,” “back,” “leading,” “trailing,” etc., is used with reference to the orientation of the Figure(s) being described. Because components of embodiments of the present invention can be positioned in a number of different orientations, the directional terminology is used for purposes of illustration and is in no way limiting. It is to be understood that other embodiments may be utilized and structural or logical changes may be made without departing from the scope of the present invention. The following detailed description, therefore, is not to be taken in a limiting sense, and the scope of the present invention is defined by the appended claims.
By applying a test voltage lower than VLL to a cell array with erased and programmed memory cells, none of the memory cells, whether programmed or erased, is conductive and no information can be obtained from the cell array. Applying a test voltage higher than VLL but lower than VLH would cause only a portion of the erased cells to becoming conductive. By applying a test voltage higher than VHH, all programmed memory cells become conductive and no information can be obtained from the cell array. Applying a test voltage lower than VHH but higher than VHL would cause only a portion of the programmed cells to becoming conductive. For obtaining the correct information from each memory cell, the read voltage must be in the range of a “sense window” between VLH and VHL.
The diagram of
Thus here explained in detail for an upper distribution edge of a erased binary non-volatile cell array with memory cells of trapping layer memories, the same idea applies accordingly to a lower distribution edge, to programmed cell arrays, to floating gate memories and to multi-bit or multi-level memory cell architectures, by way of example.
A test method for memory devices according to an exemplary embodiment of the invention is illustrated in the flowchart of
A cell array with binary trapping layer memory cells may be initialized by erasing all memory cells or by programming all memory cells (100). The cell array may include a plurality of memory cells that are assigned to the same sense amplifier or erase section in order to eliminate adulterations of the threshold voltage distribution resulting from support circuitry. In an exemplary embodiment, initialization of the memory cells is omitted and a virgin cell array is tested, wherein manufacturing induced cell abnormalities may become more evident.
A test voltage Vtest may be initialized by being set equal to a start voltage Vstart incremented by a step voltage Vstep (101). According to an exemplary embodiment, the start voltage is selected such that by applying a test voltage higher than the start voltage all memory cells of the cell array remain conductive and are detected as being erased. Various methods for predicting a suitable start voltage are possible. For example, the start voltage may be obtained from a previous test procedure in course of which a sense window applicable for the cell array is determined. Alternatively, a technology dependent average read voltage may serve as start voltage. The start voltage may be readjusted by further process steps (not shown) if required, for example, if the gradient of the distribution gives a hint for an improper choice of the start voltage.
The step voltage is usually predetermined by the resources of the memory device or an external test apparatus. Usually a voltage source for supplying the read voltage is programmable in steps of for example 50 mV in order to facilitate an adjustment of the read voltage within a respective sense window and in order to track the read voltage to a shift of the sense window during the life cycle of the memory device.
The test voltage Vtest may be decremented then by the step voltage (102) such that in the following a test voltage Vtest equal to the start voltage Vstart is applied to the memory cells of the cell array (104). Other embodiments may provide a first decrementing (102) after a first comparison (110).
The cell array is read out by applying the test voltage to the control gates of each memory cell (104). The storage elements, for example the trapping layer or the floating gate of a virgin or of an erased memory cell should be discharged and the respective control gate should not be biased. Each memory cell is checked whether it is in the non-conductive state (110). If none of the memory cells has yet switched into the non-conductive state, the test voltage is again decremented by the step voltage (102). The cell array is rescanned for non-conductive memory cells (104). Decrementing the test voltage (102), applying the test voltage (104), and scanning for conductive memory cells (110) are repeated until at least one non-conductive memory cell is detected. If at least one non-conductive memory is detected, therewith implicitly an outer upper distribution edge of the threshold voltage distribution is detected, which is equal to the current test voltage. A step counter sc and/or a distribution tail detection flag may be initialized or cleared (111) and a current number count of non-conductive cells cnc may be determined (112). Determination of cnc may be omitted, for example, if the current number count cnc is yet determined in course of the scanning for non-conductive cells. A temporary, previously detected number count of non-conductive cells pnc may be set equal to the current number count (113).
The test voltage is further decremented by the step voltage and the step counter may be incremented by one (114). The current number count cnc may be updated by being set equal to the number of all memory cells having yet switched into the non-conductive state at the current test voltage (116). Alternatively, only the number of memory cells having switched to the non-conductive state in course of the respectively last incrementing step may be determined by either masking out the memory cells having switched into the non-conductive state during a previous incrementing step or by obtaining a new current number count cnc through counting the number of non-conductive memory cells at the current test voltage and subtracting the previously obtained number count pnc.
The information obtained so far is analyzed (120), wherein in either case the gradient of the threshold voltage distribution in a distribution section near the outer upper distribution edge is assessed. For this purpose, the difference between the current number count cnc and the previously obtained number count pnc may be evaluated. If the difference is too small and falls below a predetermined limit, then a distribution tail is detected as illustrated in the second threshold voltage distribution 24 of
If the difference is sufficiently large, the current number count cnc may be compared to a predetermined maximum count pmaxc (130). If the current number count cnc does not exceed the predetermined maximum count pmaxc, the previously number count pnc may be overwritten with the current number count cnc (113) and decrementing Vtest, incrementing sc (114), updating the current number count (116), and assessing the distribution gradient (120) may be repeated until a distribution tail is detected or until the predetermined maximum count pmaxc is surpassed by the current number count cnc. In the latter case, the test sequence is terminated without any memory cells being classified as being unreliable or defective (132). The predetermined gradient may be set in dependence on the current value of the step counter in order to get a more flexible approach.
If the difference between the current number count cnc and the previously obtained number count pnc is not sufficiently large and falls below the predetermined limit, then a distribution tail is detected. The current number count cnc may be compared to the predetermined maximum count pmaxc (140). If the current number count cnc exceeds the predetermined maximum count pmaxc, the test sequence may be terminated by classifying memory cells assigned to the previously obtained number count pnc as being defective (142). As long as the current number count cnc does not exceed the predetermined maximum count pmaxc, overwriting pnc with cnc (113), decrementing Vtest, incrementing sc (114), updating cnc (116), and assessing the distribution gradient (120) may be repeated. In this case, only a portion of a longer distribution tail may be determined, wherein as many conspicuous memory cells as possible are identified.
For avoiding the loss of a short distribution tail, a distribution tail detection flag may be set, when a distribution tail is detected at the first time (122). The distribution tail detection flag may be checked when no distribution tail is detected within the current loop, i.e. when the distribution gradient is qualified as not being conspicuous for the current step count. Once a distribution tail is detected, the test sequence may be terminated with the previously obtained non-conductive memory cells pnc as being classified as being assigned to the distribution tail and therefore as being defective (142). In this case, the complete distribution tail may be detected.
Alternatively or additionally, in a further step (not shown) the step counter may be assessed. If, for a predetermined limit of the step counter, the current number count does not exceed a further predetermined limit count that may be equal to the predetermined maximum count, the memory cells that has yet been detected as being non-conductive may be classified as being defective.
The method may be implemented both in an external test apparatus or in a built-in circuitry within the respective memory device. The non-conductive memory cells may be counted by a counter unit that is implemented as integral part of a test control unit being provided within the memory device. The test control unit may be a centralized one controlling the test sequence for all cell arrays within the memory device or a distributed one, wherein each cell array of the memory device is assigned to one test control unit and wherein each test control unit is assigned to one cell array. As a distribution tail concerns a comparable small number of cells, the required circuitry is comparable simple and does not require much resources with regard to space and test time.
A counter unit that is assigned to one sense amplifier may be a simple three or four digit counter.
While the method according to
The cell array may be completely programmed (200) and a test voltage Vtest may be initialized by being set equal to a lower start voltage Vstart decremented by the step voltage Vstep (201), wherein the lower start voltage is selected such that all memory cells of the programmed cell array remain non-conductive, when a test voltage lower than the lower start voltage is applied. The methods for predicting a suitable lower start voltage may correspond to that for predicting a suitable upper start voltage as discussed above.
The test voltage Vtest may then be incremented by the step voltage (202) such that in the following a test voltage Vtest equal to the start voltage Vstart is applied to the memory cells of the cell array (204). Other embodiments may provide a first incrementing (202) after a first comparison (210).
The cell array is read out by applying the test voltage to the control gates of each memory cell (204). The state of each memory cell is sensed (210). If none of the memory cells has yet switched into the conductive state, the test voltage is again incremented by the step voltage (202) and the cell array is rescanned for conductive memory cells (204). Incrementing the test voltage (202), applying the test voltage (204), and scanning for conductive memory cells (210) are repeated until at least one conductive memory cell is detected. The lowest test voltage, at which at least one conductive memory cell is detected, defines an outer lower distribution edge of the threshold voltage distribution. Operations (211) to (224) correspond to the respectively corresponding operations (111) to (124), wherein the current number count cnc and the previously detected number count pnc refer in this case to a number of conductive cells respectively and wherein the test voltage is further incremented instead of being decremented as long as the current number count cnc of conductive cells does not exceed a maximum fail count pmaxc.
Thus the gradient of the threshold voltage distribution in a distribution section near the outer lower distribution edge is assessed by evaluating the difference between the current number count cnc and the previously obtained number count pnc. Memory cells with a threshold voltage in the range of the detected distribution tail may be qualified as being defective or unreliable and the respective pass/fail information may be transmitted via a signaling interface of the device to a test apparatus. Alternatively or additionally, the information about conspicuous memory cells is stored in a bad block mapping register of the memory device.
Referring to operations (230), (240), the maximum fail count pmaxc may correspond to a number of available repair cells assigned to the tested cell array. A pass/fail information identifying conspicuous memory cells is alternatively or additionally transmitted to a repair unit. The repair unit controls the allocation of repair cells to address lines and data lines such that each detected conspicuous memory cell may be replaced by a repair cell (242). The repair unit may be provided completely within the memory device.
The flowchart of
The start voltage may be obtained by determining and averaging threshold voltage distributions of a plurality of cell arrays that are based on the same manufacturing process. The fit step voltage and the number of repetitions result from the requirement of obtaining suitable basic values that enable a good interpolation of the respective distribution edge and from the requirement of a short test time. In an exemplary embodiment referring to a typical distribution width of 1.5V, the counts for three to five test voltages may be determined, wherein the fit step test voltage may range from 100 mV to 300 mV. Alternatively, the step voltage may differ from step to step. Via a fitting algorithm, the basic values of the actual threshold voltage distribution are compared to an average voltage distribution that characterizes the respective cell array (311). An inner lower distribution edge VHLcalc is predicted for the actual threshold voltage distribution (312). The test voltage Vtest is set equal to the predicted lower distribution edge and applied to the memory cells (313). The cell array is scanned for conductive cells (313). Memory cells that are in the conductive state at the predicted lower distribution edge are harder to program than it could be expected. Thus, if conductive cells are detected, they are supposed to be conspicuous or unreliable in some way. The conductive cells are therefore classified as being defective and may be replaced to improve the reliability of the cell array and the memory device.
The basic values of the actual threshold voltage distribution may be obtained for the whole distribution between the lower distribution edge and the higher distribution edge, wherein both a lower VHLcalc and a higher VHHcalc distribution edge may be predicted. With respect to the reliablity of the memory cells, the assessment may be limited to the respective distribution edge confining to a “sense window”, i.e., for a cell array with binary memory cells VLHcalc and VHLcalc, wherein the basic values may be obtained only from the respective half of the threshold voltage distribution.
Since not a complete threshold voltage distribution, but only a comparable small number of basic values is determined and transferred to an off-chip test apparatus for calculating the predicted distribution edges, the method is comparable fast. According to another embodiment, the test method is executed via circuitry that is realized completely within the memory device. Then, according to another exemplary embodiment, alternative algorithms that may require less circuitry may be implemented in order to detect a distribution tail.
The memory device 4 may include further a programmable read voltage source 42 for supplying the read voltage, wherein the read voltage is alterable in steps defined by a step voltage. A test control unit 45 of the memory device 4 may be provided in a centralized manner, wherein each cell array 41 is assigned to the same test control unit 45, or in a distributed manner, wherein one test control unit 45 is provided per cell array 41. The test control unit 45 may be connected to the read voltage source 42 and includes a voltage control unit 451 being capable of controlling the read voltage source 42 in dependence on the results of a threshold voltage distribution assessment. The test control unit 45 further may include an array control unit 452 configured to controlling read operations concerning the memory cells 411 through a sense unit 44 and an address unit 43, at least one counter unit 453 for counting the memory cells 411 having a predefined state, e.g., all conductive or all non-conductive memory cells 411, and an analysis unit 454 configured to evaluating the memory cells 411 on basis of the respectively obtained number of memory cells 411 exhibiting a predetermined state.
The counter unit 453 may be assigned to a complete cell array 41 or to one of the sense amplifiers of the cell array 41. Then, as only a section of the threshold voltage distribution is evaluated that represents a comparable small number of memory cells 411, a small three or four-digit counter may be sufficient.
In an exemplary embodiment, the test control unit 45 may include a data storage unit 455 capable of temporarily storing at least the respective previously obtained cell count before the cell count is overwritten by a currently determined current cell count. The analysis unit 454 may be further suitable to compare the currently obtained cell count with the previously obtained cell count. The comparison may concern the cell counts directly or values that are deduced from the cell count in a way to qualify deviations with regard to an average distribution gradient in a suitable way.
According to another exemplary embodiment, the test control unit 45 includes alternatively or additionally to the storage unit 455 a step counter 456 that is capable of counting a step number of increments/decrements of the test voltage, wherein the analysis unit 454 may be configured to rating the cell array 41 by comparing the currently obtained number with a predetermined value assigned to the respective step number. The assessment of the distribution tail may then be based on more than one assessment criterion or on a very simple criterion. Alternatively or additionally, test cycle time may be minimized, if one of the criterions indicates a normal or non-conspicuous distribution. The test control unit 454 may further be configured to classifying memory cells assigned to the previously obtained number of memory cells as being defective.
While specific embodiments of the invention has been described in detail, it will be apparent to one skilled in the art that various changes and modifications can be made therein without departing from the spirit and the scope thereof. Although specific embodiments have been illustrated and described herein, it will be appreciated by those of ordinary skill in the art that a variety of alternate and/or equivalent implementations may be substituted for the specific embodiments shown and described without departing from the scope of the present invention. This application is intended to cover any adaptations or variations of the specific embodiments discussed herein. Therefore, it is intended that this invention be limited only by the claims and the equivalents thereof.