Efficient transfer of hard data and confidence levels in reading a nonvolatile memory

Abstract

An apparatus for data storage, includes circuitry and a plurality of memory cells. The circuitry is configured to store data in a group of multiple memory cells by writing multiple respective input storage values to the memory cells in the group, to read respective output storage values from the memory cells in the group after storing the data, to generate for the output storage values multiple respective confidence levels, to produce composite data that includes the output storage values, to test a predefined condition that depends on the confidence levels, upon detecting that the condition is met, to compress the confidence levels to produce compressed soft data, and include the compressed soft data in the composite data, and to transfer the composite data over an interface to a memory controller.

Description

TECHNICAL FIELD

Embodiments described herein relate generally to data storage, and particularly to methods and systems for efficient transfer of hard data and confidence levels in reading a nonvolatile memory.

BACKGROUND

In various storage systems, a memory controller applies soft decoding to encoded data retrieved from a memory device. To this end, the memory device transmits to the memory controller hard bits read from the memory and confidence levels associated respectively with the hard bits.

Methods for transferring hard data and corresponding confidence levels are known in the art. For example, U.S. Pat. No. 9,671,972 describes a method for data storage that includes storing data in a group of analog memory cells by writing respective input storage values to the memory cells in the group. After storing the data, respective output storage values are read from the analog memory cells in the group. Respective confidence levels of the output storage values are estimated, and the confidence levels are compressed. The output storage values and the compressed confidence levels are transferred from the memory cells over an interface to a memory controller.

U.S. Pat. No. 9,214,965 describes a method for improving data integrity in a non-volatile memory system, the method includes: accessing a non-volatile memory cell for retrieving hard data bits; generating soft information by capturing a reliability of the hard data bits; calculating syndrome bits by applying a lossy compression to the soft information; and generating a host data by executing a low density parity check (LDPC) iterative decode on the hard data bits and the syndrome bits.

SUMMARY

An embodiment that is described herein provides an apparatus for data storage that includes circuitry and a plurality of memory cells. The circuitry is configured to store data in a group of multiple memory cells by writing multiple respective input storage values to the memory cells in the group, to read respective output storage values from the memory cells in the group after storing the data, to generate for the output storage values multiple respective confidence levels, to produce composite data that includes the output storage values, to test a predefined condition that depends on the confidence levels, upon detecting that the condition is met, to compress the confidence levels to produce compressed soft data, and include the compressed soft data in the composite data, and to transfer the composite data over an interface to a memory controller.

In some embodiments, the circuitry is configured to detect that the condition is met based on a number of confidence levels among the multiple confidence levels having a selected confidence level value. In other embodiments, the circuitry is configured to detect that the condition is met in response to identifying that at least one of the confidence levels is indicative of a low level of confidence compared to another of the confidence levels. In yet other embodiments, the circuitry is configured to produce the compressed soft data independently of the output storage values.

In an embodiment, the circuitry is configured to produce the compressed soft data dependently on the output storage values. In another embodiment, the circuitry is configured to determine an index of a confidence level having a selected confidence value, to calculate a count of the output storage values having a selected storage value in a range of indices below the index, and to include in the compressed soft data a binary representation of the calculated count. In yet another embodiment, the circuitry is configured to determine an index of a confidence level having a selected confidence value, and to include in the composite data a binary representation of the index.

In some embodiments, the circuitry is configured to include in the composite data a control bit indicative of whether the soft data is included in the composite data. In other embodiments, the circuitry is configured to include the output storage values in the composite data but not the confidence levels, and to further include in the composite data an indication that the confidence levels are not included in the composite data. In yet other embodiments, the circuitry is configured to compress the confidence levels by applying a lossy data compression scheme to the confidence levels.

In an embodiment, the circuitry is configured to apply the lossy data compression scheme by limiting a number of confidence levels among the multiple confidence levels having a selected confidence level value, to a predefined limit number. In another embodiment, the circuitry is configured to store the confidence levels in a buffer of the apparatus, and to perform a combined copy and compression operation that produces the compressed soft data during the copy of the soft data from the buffer to an output buffer for transferring to the memory controller.

There is additionally provided, in accordance with an embodiment that is described above, a method for data storage, including, in a storage apparatus that includes a plurality of memory cells, storing data in a group of multiple memory cells by writing multiple respective input storage values to the memory cells in the group. Respective output storage values are read from the memory cells in the group after storing the data. For the output storage values multiple respective confidence levels are generated. Composite data that includes the output storage values is produces. A predefined condition that depends on the confidence levels is tested. Upon detecting that the condition is met, the confidence levels are compressed to produce compressed soft data, and the compressed soft data is included in the composite data. The composite data is transferred over an interface to a memory controller.

These and other embodiments will be more fully understood from the following detailed description of the embodiments thereof, taken together with the drawings in which:

BRIEF DESCRIPTION OF THE DRAWINGS

FIG. 1 is a block diagram that schematically illustrates a memory system, in accordance with an embodiment that is described herein;

FIG. 2 is a block diagram that schematically illustrates a scheme for producing hard data and confidence levels for efficient transfer to a memory controller, in accordance with embodiments that are described herein;

FIG. 3 is a flow chart that schematically illustrates a method for joint data compression of hard data and corresponding soft confidence levels, in accordance with an embodiment that is described herein;

FIG. 4 is a flow chart that schematically illustrates a method for data decompression of soft confidence levels that were jointly compressed with hard data, in accordance with an embodiment that is described herein;

FIG. 5 is a flow chart that schematically illustrates a method for data compression, in which soft confidence levels are compressed independently of corresponding hard data, in accordance with an embodiment that is described herein; and

FIG. 6 is a flow chart that schematically illustrates a method for data decompression of soft confidence levels that were compressed independently of corresponding hard data, in accordance with an embodiment that is described herein.

DETAILED DESCRIPTION OF EMBODIMENTS

Overview

Data is typically stored in memory cells of a nonvolatile memory by writing respective storage values to the memory cells. The storage operation programs each memory cell to one of several programming states, which represent respective data values. Data that is stored in memory cells may have varying levels of confidence, i.e., different likelihoods that the data read from the memory cells truly reflects the data that was stored in the memory cells.

Typically, the storage values (e.g., threshold voltages) of the memory cells that are associated with a given programming state have a certain statistical distribution. When the storage value distributions of different programming states overlap, storage values that lie in boundary regions between adjacent programming states might cause read errors. Such storage values may be regarded as having low confidence levels. Storage values that lie in the middle of the distributions, on the other hand, are more likely to be reliable.

The performance of data readout from the memory cells can be improved by considering the confidence levels of the different storage values. For example, in some data storage schemes, an Error Correction Code (ECC) unit encodes the data for storage with an ECC and decodes the ECC when retrieving the data. Some ECC decoders decode the ECC by operating on soft metrics. The confidence levels of the read storage values can be used to produce such soft metrics.

Transferring the confidence levels from the memory cells to the ECC decoder typically adds a considerable amount of communication traffic between the two. A typical read operation retrieves data from thousands of memory cells simultaneously. Transferring the confidence levels assigned to these read results may require an exceedingly high communication rate. The high communication rate can be especially problematic when the memory cells and the ECC decoder reside in separate devices.

Embodiments that are described hereinbelow provide improved methods and systems for data readout from memory cells. The methods and systems described herein estimate the confidence levels of the storage values. Composite data to be transferred to the memory controller includes the storage values, and conditionally includes a compressed version of the confidence levels. For example, the compressed confidence levels may be included in the composite data only when at least one of the confidence levels is indicative of a corresponding unreliable storage value. Since most of the storage values tend to be reliable, and only a small fraction of the storage values have low confidence levels, the compressed confidence levels are rarely transferred, which considerably reduces the amount of composite data transferred.

In the description that follows, the term “confidence levels” is also referred to as “soft data” and these terms are used interchangeably. Similarly, the term “compressed confidence levels” is also referred to as “compressed soft data” and these terms are used interchangeably.

The compression of confidence levels, when used, is typically efficient, again since most of the storage values tend to be reliable. In other words, confidence levels often exhibit little or no variability from one storage value to another, and therefore lend themselves to highly efficient compression.

Upon arrival at the ECC decoder, the confidence levels are recovered from the composite data and used for decoding the ECC.

Consider an embodiment of an apparatus for data storage, including circuitry and a plurality of memory cells. The circuitry is configured to store data in a group of multiple memory cells by writing multiple respective input storage values to the memory cells in the group, to read respective output storage values from the memory cells in the group after storing the data, to generate for the output storage values multiple respective confidence levels, to produce composite data that includes the output storage values, and to test a predefined condition that depends on the confidence levels. The circuitry is further configured to, upon detecting that the condition is met, compress the confidence levels to produce compressed soft data and include the compressed soft data in the composite data, and to transfer the composite data over an interface to a memory controller.

In some embodiments, the circuitry detects that the condition is met based on the number of confidence levels among the multiple confidence levels having a selected confidence level value. For example, the circuitry detects that the condition is met in response to identifying that at least one of the confidence levels is indicative of a low level of confidence compared to another of the confidence levels.

In some embodiments, the circuitry may produce the compressed soft data dependently on the output storage values. For example, the circuitry determines an index of a confidence level having a selected confidence value, calculates a count of the output storage values having a selected storage value in a range of indices below the index, and includes in the compressed soft data a binary representation of the calculated count. In other embodiments, the circuitry may produce the compressed soft data independently of the output storage values. For example, the circuitry determines an index of a confidence level having a selected confidence value, and includes in the composite data a binary representation of the index.

In some embodiments, the circuitry includes in the composite data a control bit indicative of whether the compressed soft data is included in the composite data. In an example embodiment, the circuitry includes the output storage values in the composite data but not the confidence levels, and further includes in the composite data an indication that the confidence levels are not included in the composite data.

The circuitry may compress the confidence levels using, for example, a lossy data compression scheme. In such embodiments, the circuitry may apply the lossy data compression scheme by limiting the number of confidence levels among the multiple confidence levels having a selected confidence level value (e.g., a low confidence level), to a predefined limit number. In an embodiment, the limit number may be set to two.

In an embodiment, the circuitry is configured to store the confidence levels in a buffer of the apparatus, and to perform a combined copy and compression operation that produces the compressed soft data during the copy of the soft data from the buffer to an output buffer for transferring to the memory controller.

Using the disclosed embodiments, a high readout throughput in transferring hard data and confidence levels can be attained, that improves over known schemes. Moreover, the disclosed embodiments have low complexity and are implementable within the memory device and memory controller.

System Description

FIG. 1 is a block diagram that schematically illustrates a memory system 20, in accordance with an embodiment that is described herein. System 20 can be used in various host systems and devices, such as in computing devices, cellular phones or other communication terminals, removable memory modules (“disk-on-key” devices), Solid State Disks (SSD), digital cameras, music and other media players and/or any other system or device in which data is stored and retrieved.

System 20 includes a memory device 24 that stores data in a memory cell array 28. The memory cell array includes multiple memory cells 32. The term “memory cell” is typically used to describe any memory cell that holds a continuous, analog level of a physical quantity, such as an electrical voltage or charge. Memory cell array 28 may include memory cells of any kind, such as, for example, NAND, NOR and CTF Flash cells, PCM, NROM, FRAM, MRAM and DRAM cells. Memory cells 32 may include Single-Level Cells (SLC) or Multi-Level Cells (MLC, also referred to as multi-bit cells). Alternatively, memory cells that store a higher number of bits per cell, such as Triple-Level Cells (TLC) and Quad-Level Cells (QLC) can also be used.

The charge levels stored in the memory cells and/or the analog voltages or currents written into and read out of the cells are referred to herein collectively as analog values or storage values. Although the embodiments described herein mainly address threshold voltages, the methods and systems described herein may be used with any other suitable kind of storage values.

System 20 stores data in memory cells 32 by programming the memory cells to assume respective memory states, which are also referred to as programming levels. The programming states are selected from a finite set of possible states, and each state corresponds to a certain nominal storage value. For example, a 2 bit/cell MLC can be programmed to assume one of four possible programming states by writing one of four possible nominal storage values to the cell. Alternatively, memory cells that store a higher number of bits per cell such as TLC memory cells that store three bits per cell and QLC memory cells that store four bits per cell can also be used.

Memory device 24 includes a reading/writing (R/W) unit 36, which converts data for storage in the memory device to storage values and writes them into memory cells 32. In alternative embodiments, the R/W unit does not perform the conversion, but is provided with voltage samples, i.e., with the storage values for storage in the cells. When reading data out of memory cell array 28, R/W unit 36 converts the storage values of memory cells 32 into digital samples having a resolution of one or more bits. The R/W unit typically reads data from memory cells 32 by comparing the storage values of the cells to one or more read thresholds. Data is typically written to and read from the memory cells in groups that are referred to as pages. In some embodiments, the R/W unit can erase a group of cells 32 by applying one or more negative erasure pulses to the memory cells.

The storage and retrieval of data in and out of memory device 24 is performed by a memory controller 40, which communicates with device 24 over a suitable interface 42. In some embodiments, memory controller 40 produces the storage values for storing in the memory cells and provides these values to R/W unit 36. Alternatively, memory controller 40 may provide the data for storage, and the conversion to storage values is carried out by the R/W unit internally to the memory device.

Memory controller 40 communicates with a host 44, for accepting data for storage in the memory device and for outputting data retrieved from the memory device. In some embodiments, some or even all of the functions of memory controller 40 may be implemented in hardware. Alternatively, memory controller 40 may include a microprocessor that runs suitable software, or a combination of hardware and software elements.

In some embodiments, R/W unit 36 includes a data compression module 45, which compresses some of the information that is to be sent to memory controller 40. The memory controller includes a decompression module 46, which decompresses the compressed information received from memory device 24. In particular, R/W unit 36 may produce confidence levels of the storage values read from memory cells 32, and data compression module 45 may compress these confidence levels and send the compressed confidence levels to memory controller 40. (In some embodiments, data compression module 45 can also be used for compressing other types of information, such as stored data that is retrieved from memory cells 32.)

In some embodiments, the memory device performs two read operations to produce two local readouts using two respective pre-assigned read thresholds. Based on the two local readouts, the memory device determines hard storage values and associated binary confidence levels to be transferred to the memory controller. For example, memory cells whose threshold voltages fall between the two read thresholds are considered as having low reliability, whereas memory cells whose threshold voltages fall below or above the two read thresholds are considered as having high reliability. In alternative embodiments, more than two read thresholds may be used to produce multiple groups of confidence levels.

The memory controller uses the storage values read from memory cells 32, and the associated confidence levels, to reconstruct the stored data. For example, memory controller 40 may include an Error Correction Code (ECC) unit 47, which encodes the data for storage using a suitable ECC, and decodes the ECC of the data retrieved from memory cells 32. ECC unit 47 may apply any suitable type of ECC, such as, for example, a Low-Density Parity Check (LDPC) code or a Bose-Chaudhuri-Hocquenghem (BCH) code. In some embodiments, ECC unit 47 uses the confidence levels to improve the ECC decoding performance. Several example methods for obtaining and compressing confidence levels, as well as for using the confidence levels in ECC decoding, are described, for example, in a U.S. Pat. No. 8,230,300, whose disclosure is incorporated herein by reference. (In the event of any inconsistencies between any incorporated document and this document, it is intended that this document control.)

The memory system configuration of FIG. 1 is an example memory system configuration, which is shown purely for the sake of conceptual clarity. Any other suitable memory system configuration can also be used. Elements that are not necessary for understanding the principles of the present invention, such as various interfaces, addressing circuits, timing and sequencing circuits and debugging circuits, have been omitted from the figure for clarity.

In the example system configuration shown in FIG. 1, memory device 24 and memory controller 40 are implemented as two separate Integrated Circuits (ICs). In alternative embodiments, however, the memory device and the memory controller may be integrated on separate semiconductor dies in a single Multi-Chip Package (MCP) or System on Chip (SoC), and may be interconnected by an internal bus. Further alternatively, some or all of the circuitry of the memory controller may reside on the same die on which the memory array is disposed. Further alternatively, some or all of the functionality of memory controller 40 can be implemented in software and carried out by a processor or other element of the host system. In some embodiments, host 44 and memory controller 40 may be fabricated on the same die, or on separate dies in the same device package.

In some implementations, a single memory controller may be connected to multiple memory devices 24. In yet another embodiment, some or all of the memory controller functionality may be carried out by a separate unit, referred to as a memory extension, which acts as a slave of memory device 24. Typically, memory controller 40 includes a general-purpose processor, which is programmed in software to carry out the functions described herein. The software may be downloaded to the processor in electronic form, over a network, for example, or it may, alternatively or additionally, be provided and/or stored on tangible media, such as magnetic, optical, or electronic memory.

Memory cells 32 of memory cell array 28 are typically arranged in a grid having multiple rows and columns, commonly referred to as word lines and bit lines, respectively. The memory array is typically divided into multiple pages, i.e., groups of memory cells that are programmed and read simultaneously. Memory cells 32 are typically erased in groups of word lines that are referred to as erasure blocks. In some embodiments, a given memory device includes multiple memory cell arrays, which may be fabricated on separate dies.

In the description that follows and in the claims, the term “circuitry” refers to elements of memory device 24, excluding interface 42 and memory cell array 28. In the example of FIG. 1, the circuitry includes R/W unit 36, including data compression module 45.

Schemes for Producing Composite Data for High Throughput Transfer

As described, for example, in U.S. Pat. No. 8,230,300, assigning confidence levels to hard bits of an ECC code word read from memory typically improves the decoding capability of the ECC code word, significantly. The hard data and the confidence levels are typically produced within the memory device and transferred to the memory controller. In some embodiments, to reduce the amount of information transferred over the interface that interconnects between the memory device and the memory controller, the memory device applies a suitable data compression scheme to the confidence levels and transfers the confidence levels to the memory controller in a compressed form. The memory controller recovers the confidence levels by applying a suitable decompression scheme and uses the hard data and the recovered confidence levels to perform an ECC soft decoding scheme to recover the unencoded data.

It is now demonstrated that composite data in which the data compression scheme applied to the soft data depends on the hard data may be represented with a smaller number of bits compared to composite data in which the data compression scheme is independent of the hard data. Let S_n denote a sequence of soft bits that get a value “1” with probability P and a value “0” with probability (1−P). The entropy of this sequence H(S_n) is given by:H(S_n)=−Log₂(P)·P−Log₂(1−P)·(1−P)[Bits]  Equation 1

The minimal number of bits required for representing a sequence of random hard bits H_n (whose entropy equals 1) and corresponding soft bits S_n is thus given by:N_S+H=1+H(S_n)=1−Log₂(P)·P−Log₂(1−P)·(1−P)[Bits]  Equation 2

Consider a sequence of composite data C_n that incorporates information from both sequences of hard bits and soft bits. The composite sequence has the following statistical properties:

$\begin{array}{cc}{C}_n=\{\begin{array}{cc}10& \mathrm{with}\mathrm{probbility}\left(1-P\right)/2\\ 01& \mathrm{with}\mathrm{probability}P\\ 00& \mathrm{with}\mathrm{probaility}\left(1-P\right)/2\end{array}& \mathrm{Equation}3\end{array}$

In Equation 3, the most significant bit (left bit) corresponds to the hard data and the least significant bit (right bit) corresponds to the soft data. The minimal number of bits required for representing C_n is given by:

$\begin{array}{cc}{N}_c=-{\mathrm{Log}}_2\left(P\right)·P-{\mathrm{Log}}_2\left(\frac{1-P}{2}\right)·\left(\frac{1-P}{2}\right)\left[\mathrm{Bits}\right]& \mathrm{Equation}4\end{array}$ and after rearrangement,N_C=N_S+H−P[Bits]  Equation 5

This means that N_c≤N_S+H, and therefore transmission of the composite sequence may require a smaller bandwidth compared to the transmission of the hard data and soft data independently of one another.

Next is described a class of low-complexity schemes that may be used for producing low-bandwidth composite data based on both the hard and soft data. Further below, example schemes of this class that attain compression rates close to the theoretical bounds will be described.

FIG. 2 is a block diagram that schematically illustrates a scheme for producing hard data and confidence levels for efficient transfer to a memory controller, in accordance with embodiments that are described herein.

The scheme in FIG. 2 may be used in implementing at least part of memory device of FIG. 1, including data compression module 45.

Memory device 24 in FIG. 2 includes a buffer 70 denoted “hard buffer” or “H_BUFF” that stores hard data bits read from memory array 28, and another buffer 74 denoted “soft buffer” or “S_BUFF” that stores confidence levels that R/W unit 36 respectively assigns to the hard data bits. A sequence of hard data bits is referred to as “hard data” and is denoted H. A sequence of soft data bits (the confidence levels) is referred to as “soft data” and is denoted S.

In FIG. 2, memory device 24 includes a composite data producer 78, which produces a sequence of composite data bits, denoted C, based on both the hard data H and the soft data S. As will be described below, composite data producer 78 conditionally applies a suitable data compression scheme to the soft data S.

The memory device stores the composite data C in an output buffer 82 and transfers the composite data from the output buffer to memory controller 40 over a suitable interface 86 (Interface 86 may implement interface 42 in FIG. 1).

Composite data producer 78 includes a data compression module 80 that when requested applies a suitable data compression scheme to the soft data S for producing compressed soft data denoted S. Data compression module 80 is also referred to as a “data compressor.” In some embodiments, data compressor 80 applies a joint data compression method that compresses the soft data S depending on the hard data H. In other embodiments, data compressor 80 applies an independent data compression scheme that compresses the soft data S independently of the hard data H. Composite data producer 78 includes the hard data H in C, and conditionally includes the compressed soft data S in C. The composite data producer additionally includes a control bit (denoted “CB”) that indicates whether the composite data C includes the compressed soft data S, or not.

In some embodiments, composite data producer 78 produces the composite data C as given by:

$\begin{array}{cc}C=\left\{\begin{array}{cc}\left[H,\mathrm{CB}=0\right]& N1=0\\ \left[H,\mathrm{CB}=1,\underset{\_}{S}\right]& N1>0\end{array}\right\}& \mathrm{Equation}6\end{array}$

In Equation 6, C denotes a composite data sequence derived from (i) a sequence H containing a number W of hard data bits, and (ii) a sequence S of W confidence levels that are respectively assigned to the W hard data bits in H. Further in Equation 6, S denotes compressed soft data derived from S (e.g., using data compressor 80), and N1 denotes a non-negative integer number of the low confidence levels in the uncompressed soft data S. In the present example, the confidence levels have two possible values denoted “0” and “1”, wherein a “0” value refers to a high confidence level and a “1” value refers to a low confidence level.

The number (N1) of confidence levels in S having a “1” value (low confidence) is typically much smaller than the number (W−N1) of confidence levels in S having a “0” value (high confidence).

The control bit (CB) in Equation 6 distinguishes between cases in which N1 equals zero (in which case S may be omitted from C), and cases in which N1 has a positive non-zero value (and S is included in C). When N1 equals zero, the uncompressed soft data S contains only high confidence levels. In this case C includes H and the control bit CB set to “0”, but S is omitted. On the other hand, when N1>0, S contains one or more low confidence levels, in which case C includes H, the control bit CB set to “1”, and a compressed version S of S.

Since most of the hard data bits are read reliably, N1 equals zero with high probability, and the compressed soft data S is rarely calculated and included in C. Consequently, the scheme in Equation 6 produces composite data that is much shorter (on average) than the length W of the hard data H plus the length of the compressed soft data S.

At the memory controller side, the memory controller receives the composite data C, and reproduces C to recover H and S. The memory controller extracts from C, the W hard bits of H and the control bit CB. When the control bit equals “0”, the memory controller recovers S by setting all the W confidence levels in S to a high confidence level. Otherwise, the memory controller extracts from C the compressed confidence levels S, and decompresses S to recover the W confidence levels in S, as will be described further below.

In some embodiments, data compressor 80 produces S by applying to S a lossy data compression scheme. In some embodiments, in performing this lossy data compression scheme, the number of confidence levels in S having a low confidence level is limited to a predefined maximal number denoted N1max. In example embodiments that will be described below, before applying data compression to S, the number of low confidence levels in S, is limited to N1max=2, so that when S contains more than N1max low confidence levels, the R/W unit zeros the excessive low confidence levels.

In some embodiments, hard buffer 70 and soft buffer 74 are inaccessible directly to interface 86, but only via output buffer 82. In a typical implementation, shortly before the actual transmission of a composite data sequence through interface 86, the hard data from the hard buffer and the data from the soft buffer are copied to the output buffer.

In some embodiments, at least part of the scheme for producing the composite data may be implemented by copying of fixed quotas of bits from the hard buffer followed by provisional quotas of compressed confidence levels from the soft data buffer. In some embodiments, to reduce readout latency, the memory device synchronizes between operations of (i) copy hard data from the hard buffer to the output buffer and (ii) copy soft data from the soft buffer to the output buffer while applying data compression to the soft data. In (ii), the memory device performs a combined copy and compression operation that produces the compressed soft data during the copy of the soft data from the soft buffer to the output buffer, for transferring (as part of the composite data) to the memory controller.

In some embodiments, during internal copy operations within the memory device, no other internal operations may be performed. This may degrade the readout throughput due to performing separate copy operations to hard data and soft data over distinct time periods. By performing the two copy operations in parallel to one another, as described above, the latency reduces considerably.

It should be noted that parallel copy of the hard and soft data to the output buffer as described above is not mandatory. In other embodiments, the copy and compression operation of the soft data may be carried out immediately after (or before) the copy operation of the hard data, with no (or minimal) time gap in between.

Low-density storage systems often operate in a pipeline mode, in which the next sense operation (reading from the memory array) is performed in parallel to transmitting previously read data over the interface of the memory device. This means that the sense operation may start after copying the hard data, soft data or both, to the output Buffer. In some embodiments, this is done efficiently by copying the hard data and soft data in parallel to one another, as described above.

An Example Joint Compression Scheme

As noted above, in some embodiments, data compressor 80 applies to the soft data S and hard data H a joint data compression scheme, in which the soft data S is compressed depending on the actual content of the hard data H.

A method for producing compressed confidence levels S dependently on the hard data H is now described.

FIG. 3 is a flow chart that schematically illustrates a method for joint data compression of hard data and corresponding soft confidence levels, in accordance with an embodiment that is described herein.

The method will be described as executed by data compressor 80.

The method begins at a reception stage 100, with data compressor 80 receiving uncompressed soft data S that includes a number W of binary-valued confidence levels. At a soft indices identification stage 104, the data compressor identifies in S a first index I₀ and second index I₁ for which the confidence levels have a value “1” (a low confidence level). When at stage 100 S contains a single confidence level having a “1” value, the data compressor sets I₀=I₁. At a zero sequences identification stage 108, the data compressor calculates Ĩ₀ as the number of “0” bits in H up to the index I₀ (e.g., having indices between 0 and Ĩ₀−1), and calculates Ĩ₁ as the number of “0” bits in H up to the index I₁ (e.g., having indices between 0 and Ĩ₁−1), e.g., using expressions given by:

$\begin{array}{cc}{\tilde{I}}_0=\sum _{k=0}^{I_0-1}!\left(H_k\right)& \mathrm{Equation}7\end{array}$ $\begin{array}{cc}{\tilde{I}}_1=\sum _{k=0}^{I_1-1}!\left(H_k\right)& \mathrm{Equation}8\end{array}$

wherein in Equations 7 and 8, H_k denotes the kth element in H, and the operator !(⋅) denotes a logical binary inversion operator.

At a conditional swapping stage 112, the data compressor calculates the total number W0 of “0” bits in H, and upon identifying that Ĩ₁−Ĩ₀>└W0/2┘, swaps roles between Ĩ₀ and Ĩ₁.

At a binary representation stage 116, the data compressor calculates numbers of bits b₀ and b₁ given by:b₀=┌Log₂(W0)┐b₁=┌Log₂(1+└W0/2┘)┐  Equation 9

and calculates S0 and as a binary representation of Ĩ₀ using a number b₀ of bits, and calculates S1 as a binary representation of [(Ĩ₁−Ĩ₀) modulo W0] using a number b₁ of bits.

At a final compression stage 120, the data compressor produces S as the concatenation of the two binary representations as given by S=[S0,S1]. Following stage 120 the method terminates.

FIG. 4 is a flow chart that schematically illustrates a method for data decompression of soft confidence levels that were jointly compressed with hard data, in accordance with an embodiment that is described herein.

The method will be described as executed by memory controller 40, e.g., by decompression module 46. The method may be applied to soft data compressed using the method of FIG. 3.

The method of FIG. 4 begins at an input stage 150, with decompression module 46 receiving compressed soft confidence levels S and hard data H that have been extracted from composite data C as described above. At a bit-number calculation stage 154, the decompression module calculates the total number W0 of “0” bits in H, and further calculates the numbers of bits b₀ and b₁ based on W0 as given in Equation 9 above. At a numbers of hard bits determination stage 158, decompression module 46 extracts from S a number of b₀ bits to produce Ĩ₀, and a number of b₁ bits to produce (Ĩ₁−Ĩ₀).

At an indices-calculation stage 162 the decompression module calculates the indices I₀ and I₁ based on H, Ĩ₀ and (Ĩ₁−Ĩ₀), and at a decompression stage 166 recovers S based on I₀ and I₁. Following stage 166 the method terminates.

Using computer simulations, the inventors tested the joint compression scheme described above, for a number W=63 of hard bits in H, and a probability of a “1” valued confidence level of P=0.02. The effective compression rate under these conditions is given by CRjoint=0.13. The compression rate (CRjoint in this example) is measured relative to the length of S and is given by [length (C)−length (H)]/length (S).

An Example Independent Compression Scheme

In this section, an example compression and decompression schemes are provided, in which the confidence levels S are compressed independently of the hard data bits H.

FIG. 5 is a flow chart that schematically illustrates a method for data compression, in which soft confidence levels are compressed independently of corresponding hard data, in accordance with an embodiment that is described herein.

The method will be describe as executed by data compressor 80.

The method begins at a reception stage 200, with data compressor 80 receiving uncompressed soft data S that includes a number W of binary-valued confidence levels. At a soft indices calculation stage 204, the data compressor identifies a first index I₀ and second index I₁ in S having a confidence level value “1” (a low confidence level). If S contains a single confidence level having a “1” value, the data compressor sets I₀=I₁.

At a conditional swapping stage 212, the data compressor checks whether I₁−I₀>[W/2], and if so, swaps roles between I₀ and I₁.

At a binary representation stage 216, the data compressor calculates numbers of bits b₀ and b₁ as given by Equation 9 in which W replaces W0, calculates S0 as a binary representation of I₀ using b₀ bits, and calculates S1 as a binary representation of (I₁−I₀) modulo W using b₁ bits. At a final compression stage 220, the data compressor produces S as the concatenation of the two binary representations as given by S=[S0,S1]. Following stage 220 the method terminates.

FIG. 6 is a flow chart that schematically illustrates a method for data decompression of soft confidence levels that were compressed independently of corresponding hard data, in accordance with an embodiment that is described herein.

The method will be described as executed by memory controller 40, e.g., by decompression module 46. The method may be applied to data compressed using the method of FIG. 5 above.

The method begins at an input stage 250, with decompression module 46 receiving compressed soft confidence levels S extracted from composite data C as described above. S includes a compressed version of uncompressed confidence levels S of length W.

At a bit-number calculation stage 254, the decompression module calculates the numbers of bits b₀ and b₁ based on W as given in Equation 9 above in which W replaces W0. At an indices-determination stage 258, decompression module 46 extracts from S a number of by bits to produce an index I₀, and a number of b₁ bits to produce an indices difference (I₁−I₀), and at a decompression stage 266 recovers S based on I₀ and (I₁−I₀). Following stage 266 the method terminates.

The performance of the independent compression scheme in the method of FIG. 5 is analyzed herein. The average compression ratio attained by the method of FIG. 5 is given by:

$\begin{array}{cc}\mathrm{CR}=\frac{1}{W}\left[P0\left(W+1\right)+\left(1-P0\right)\left(W+1+\mathrm{Length}\left(\underset{\_}{S}\right)\right)\right]& \mathrm{Equation}10\end{array}$

Wherein P0 is given by P0=(1−P)^W. The compression rate in Equation 10 is minimized when the length of S is minimized. Specifically, a condition for minimizing the length of the compressed confidence levels S is given by:

$\begin{array}{cc}\min \mathrm{length}\left(\underset{\_}{S}\right)=\lceil {\mathrm{Log}}_2\left(\sum _{k=1}^{\mathrm{Nmax}}\left(\begin{array}{c}W\\ k\end{array}\right)\right)\rceil & \mathrm{Equation}11\end{array}$

In Equation 11, Nmax denotes the maximal number of “1” values in a sequence of W confidence levels.

Based on Equations 9 and 10, the optimal compression ratio is given by:

$\begin{array}{cc}\mathrm{CRopt}=\frac{1}{W}\left[P0\left(W+1\right)+\left(1-P0\right)\left(W+1+\lceil {\mathrm{Log}}_2\left(\sum _{k=1}^{\mathrm{Nmax}}\left(\begin{array}{c}W\\ k\end{array}\right)\right)\rceil \right)\right]& \mathrm{Equation}12\end{array}$

It can be shown that for a length W=63, the minimal length of S is 11 bits. Moreover, the proposed independent compression scheme achieves an optimal compression rate as given in Equation 12.

Using computer simulations, under the conditions W=63 and P=0.02, the resulting compression ratio is given by CRindependent=0.14. This result demonstrates that the joint compression scheme of FIG. 3 above (having a compression ratio 0.13) outperforms the independent compression scheme of FIG. 5.

The embodiments described above are given by way of example, and other suitable embodiments can also be used.

Although the embodiments described herein mainly address compression of confidence levels related to data retrieved from a memory, the methods and systems described herein can also be used in other applications, such as in applying data compression to other types of sparse data.

It will be appreciated that the embodiments described above are cited by way of example, and that the following claims are not limited to what has been particularly shown and described hereinabove. Rather, the scope includes both combinations and sub-combinations of the various features described hereinabove, as well as variations and modifications thereof which would occur to persons skilled in the art upon reading the foregoing description and which are not disclosed in the prior art. Documents incorporated by reference in the present patent application are to be considered an integral part of the application except that to the extent any terms are defined in these incorporated documents in a manner that conflicts with the definitions made explicitly or implicitly in the present specification, only the definitions in the present specification should be considered.

Claims

1. An apparatus for data storage, comprising: a plurality of memory cells; andcircuitry configured to: store data in a group of multiple memory cells by writing multiple respective input storage values to the memory cells in the group;read respective output storage values from the memory cells in the group after storing the data;generate for the output storage values multiple respective confidence levels;test a predefined condition that depends on the confidence levels;upon detecting that the condition is met, compress the confidence levels to produce compressed soft data, and produce composite data including both the output storage values and the compressed soft data;upon detecting that the condition is not met, produce the composite data including the output storage values but not the compressed soft data; andtransfer the composite data over an interface to a memory controller.

2. The apparatus according to claim 1, wherein the circuitry is configured to detect that the condition is met based on a number of confidence levels among the multiple confidence levels having a selected confidence level value.

3. The apparatus according to claim 2, wherein the circuitry is configured to detect that the condition is met in response to identifying that at least one of the confidence levels is indicative of a low level of confidence compared to another of the confidence levels.

4. The apparatus according to claim 1, wherein the circuitry is configured to produce the compressed soft data independently of the output storage values.

5. The apparatus according to claim 1, wherein the circuitry is configured to produce the compressed soft data dependently on the output storage values.

6. The apparatus according to claim 1, wherein the circuitry is configured to determine an index of a confidence level having a selected confidence value, to calculate a count of the output storage values having a selected storage value in a range of indices below the index, and to include in the compressed soft data a binary representation of the calculated count.

7. The apparatus according to claim 1, wherein the circuitry is configured to determine an index of a confidence level having a selected confidence value, and to include in the composite data a binary representation of the index.

8. The apparatus according to claim 1, wherein the circuitry is configured to include in the composite data a control bit indicative of whether the compressed soft data is included in the composite data.

9. The apparatus according to claim 1, wherein the circuitry is configured to, upon detecting that the condition is not met, further include in the composite data an indication that the compressed soft data is not included in the composite data.

10. The apparatus according to claim 1, wherein the circuitry is configured to compress the confidence levels by applying a lossy data compression scheme to the confidence levels.

11. The apparatus according to claim 10, wherein the circuitry is configured to apply the lossy data compression scheme by limiting a number of confidence levels among the multiple confidence levels having a selected confidence level value, to a predefined limit number.

12. The apparatus according to claim 1, wherein the circuitry is configured to store the confidence levels in a buffer of the apparatus, and to perform a combined copy and compression operation that produces the compressed soft data while copying the confidence levels from the buffer to an output buffer for transferring to the memory controller.

13. A method for data storage, comprising: storing data in a group of multiple memory cells by writing multiple respective input storage values to the memory cells in the group;reading respective output storage values from the memory cells in the group after storing the data;generating for the output storage values multiple respective confidence levels;evaluating a predefined condition, which depends on the confidence levels, over multiple sets of output storage values and corresponding confidence levels;for one or more of the sets in which the condition is met, compressing the corresponding confidence levels to produce compressed soft data, and producing composite data including both the output storage values and the compressed soft data;for at least one of the sets in which the condition is not met, producing the composite data including the output storage values but not the compressed soft data; andtransferring the composite data over an interface to a memory controller.

14. The method according to claim 13, and comprising detecting for a given set that the condition is met based on a number of confidence levels among the multiple confidence levels having a selected confidence level value.

15. The method according to claim 14, wherein detecting that the condition is met comprises detecting that the condition is met in response to identifying that at least one of the confidence levels is indicative of a low level of confidence compared to another of the confidence levels.

16. The method according to claim 13, wherein compressing the corresponding confidence levels comprises producing the compressed soft data independently of the output storage values.

17. The method according to claim 13, wherein compressing the corresponding confidence levels comprises producing the compressed soft data dependently on the output storage values.

18. The method according to claim 13, and comprising determining an index of a confidence level having a selected confidence value, calculating a count of the output storage values having a selected storage value in a range of indices below the index, and including in the compressed soft data a binary representation of the calculated count.

19. The method according to claim 13, and comprising determining an index of a confidence level having a selected confidence value, and including in the composite data a binary representation of the index.

20. The method according to claim 13, and comprising including in the composite data a control bit indicative of whether the compressed soft data is included in the composite data.

21. The method according to claim 13, and comprising, upon detecting that the condition is not met, further including in the composite data an indication that the compressed soft data is not included in the composite data.

22. The method according to claim 13, wherein compressing the corresponding confidence levels comprises applying a lossy data compression scheme to the confidence levels.

23. The method according to claim 22, wherein applying the lossy data compression scheme comprises limiting a number of confidence levels among the multiple confidence levels having a selected confidence level value, to a predefined limit number.

24. The method according to claim 13, and comprising storing the corresponding confidence levels in a buffer, and performing a combined copy and compression operation that produces the compressed soft data while copying the corresponding confidence levels from the buffer to an output buffer for transferring to the memory controller.