WO2008087082A9

WO2008087082A9 - Method and apparatus for recording data into a matrix of memory devices

Info

Publication number: WO2008087082A9
Application number: PCT/EP2008/050213
Authority: WO
Inventors: Michael Drexler; Thomas Brune
Original assignee: Thomson Licensing; Michael Drexler; Thomas Brune
Priority date: 2007-01-15
Filing date: 2008-01-10
Publication date: 2008-09-18
Also published as: WO2008087082A1

Abstract

For recording or replaying in real-time digital high band- width video signals, e.g. HDTV, HD progressive or HD film capture signals, very fast memories are required. For the storage of streaming HD video data, NAND FLASH memory based systems can be used. Flash memory devices are physically accessed in a page oriented mode. Each flash device contains some static defect blocks which are already marked during production. Advantageously, by dealing with dynamic defect blocks during recording in a similar way, dynamic defects are handled like static defects. Thereby the file system remap tables can be searched to find spare blocks which are not marked as 'bad blocks'.

Description

Method and Apparatus for recording data into a matrix of memory devices

The invention relates to a method and to an apparatus for recording data into a matrix of memory devices.

Background

For recording or replaying in real-time digital high bandwidth video signals, e.g. HDTV, HD progressive or HD film capture signals, very fast memories are required. For the storage of streaming HD video data non-volatile memory based systems can be used, in particular NAND flash memory de- vices. Flash memory devices are physically write accessed in a page-oriented mode, whereby one 'page' includes e.g. 1024 data words and related error correction code (ECC) . Erase operations on a specific flash memory can be carried out on predetermined-size data physical blocks only. These data blocks are denoted by the term ^xflash block' in the follow^¬ ing. A flash block consists of e.g. 64 pages. For example, the following NAND flash memories are on the market: Samsung K9K2G16U0M-YCB000 (2Gbit, lβbit oriented), K9W4G08U0M-YCB000 (4Gbit, 8bit oriented), Toshiba TH58NVG2S3BFT00 (4Gbit, 8bit oriented), MICRON MT29G08AAxxx (2Gbit, 8bit oriented), MT29G16AAxxx (2Gbit, lβbit ori^¬ ented), Samsung K9K4G08 (4Gbit, 8bit oriented).

NAND flash memories have two basic disadvantages: - the write access is rather slow;

- they have unmasked production defects and acquire even more defects during their lifetime. The required error handling is under user responsibility. This is also true for equivalent memory types. Since a detection of defects in flash memory devices (e.g. NAND devices) takes place for instance during an erase op- eration, a defect in a page makes an entire flash block unusable. Such defect flash blocks must not be used by the file system. However, the handling of such defects is under user responsibility. It is known to use redundancy codes like Reed-Solomon for such error processing, but that has other disadvantages like high dynamic time consumption versus real-time behaviour.

An embodiment of the invention is based on a dynamic re- mapping of blocks as disclosed in EP-A-1808863. During re^¬ cording to flash devices, dynamic defects in flash blocks are recognised and the (small) blocks are re-mapped after^¬ wards to corresponding spare blocks. The file system uses a corresponding re-map table.

Invention

The inventive processing can be used for real-time recording of high definition streaming video data on non-volatile memory devices, in particular NAND flash memory based devices. Using flash devices while recording in real-time at high data throughput will result in big files or takes, respec^¬ tively .

A problem to be solved by the invention is to mark dynamic defect blocks of flash based memory devices in an advanta^¬ geous manner.

A further problem to be solved by the invention is to erase, following formatting, spare blocks of flash based memory devices with implemented dynamic re-mapping.

High-speed input data are written in a multiplexed fashion into a matrix of multiple flash devices. A list-based proc- essing is performed that is as simple and as fast as possi^¬ ble, and defect pages of flash blocks of single flash memo- ries are addressed within the matrix architecture. Each flash memory device contains some static defect sec^¬ tions or small blocks, even a fabric fresh device. During production these defects are marked by modifying distinctive locations inside the blocks. Users have to cope with this information. Advantageously, by doing something similar with dynamic defect blocks during recording, dynamic defects are handled the same way than static defects. The information of the dynamic re-mapping is considered in the file system. If the flash device is re-formatted (i.e. a new fresh file system is being generated) , the information about the previously re-mapped dynamic defects is lost. Dur^¬ ing the following recordings the dynamic defect algorithm will re-map these blocks again. But during memory lifetime the amount of dynamic defects will rise. The maximum number of dynamic defects, which can be handled per take, is lim^¬ ited. If the amount of dynamic defects is higher than this limitation, error-free recording can no longer be guaranteed. By marking the dynamic defect small blocks, dynamic defect small blocks will change to static defect small blocks. A re-formatting of the flash devices will no longer cause losing the defect-information.

If, for any reason, as mentioned above, the flash based re^¬ corder is to be re-formatted (i.e. a new fresh file system is to be generated) , the information of the previously re^¬ mapped dynamic defects is lost and will be re-build with the next recordings by detecting the dynamic bad small blocks again. But the spare small blocks, which will be used for the re-mapping, are no longer erased, since they were al^¬ ready used before the re-formatting. This leads to errors in the streaming data. According to the invention, it is worked through the file system tables to find the spare small blocks which are not marked as ^xbad blocks' , and to erase them. Advantageously, the processing can be performed by software or by hardware means .

Drawings

Exemplary embodiments of the invention are described with reference to the accompanying drawings, which show in:

Fig. 1 simplified internal organisation of a flash memory; Fig. 2 static and dynamic defect small blocks and spare area;

Fig. 3 re-defining a dynamic defect small block as a static defect small block;

Fig. 4 flash memory following re-formatting and erasing, wherein the dynamic defect small block information has been kept;

Fig. 5 fresh or empty flash memory;

Fig. 6 a static defect small block is re-mapped during for^¬ matting using one of the spare small blocks; Fig. 7 following recording of a take, there are the empty take and take 1, in which one small block has a dy^¬ namic defect;

Fig. 8 re-formatted flash memory wherein all take informa^¬ tion and dynamic defect information is lost; Fig. 9 erased flash memory following re-formatting, in which flash memory one spare small block is not erased.

Exemplary embodiments

As mentioned above, NAND flash devices are organised in flash blocks (i.e. big blocks), where the blocks are further structured in pages 0 ... n. The size of the pages is a mul- tiple power of ^X2' plus some spare bits for ECC (error cor^¬ rection code) or other information. During production, be- fore shipping the flash devices, all flash blocks are tested. Intact flash blocks are erased completely, whereby flash blocks containing invalid bits (static defect small blocks) are marked vendor-dependent. Most vendors do this by writing a different value than OxFF (this is the 'erased' value) to the first location of the spare area of the first two pages of each defect flash block, as shown in Fig. 1 by 'marked defect section' .

A flash based system, which implements a file system, sets up this file system during formatting by reading all the static defect small block locations. These small blocks are neither used for recording data nor erased, because this will delete the static defect small blocks information, too.

The inventive processing for marking dynamically defect small blocks, the getting-defect of which occurs during mem^¬ ory lifetime, uses this behaviour in the manner as depicted in Fig. 2 to Fig. 4.

If in a recorded take a dynamic defect is detected, the cor^¬ responding small block is re-mapped (by device driver soft^¬ ware or by a separate controller) to the spare area as shown in Fig. 2. A linked list provides information about empty takes. The originally defect small block stays normally un^¬ touched containing the recorded data with the defect inside. Each page of the flash block is recorded without touching the locations of the static defect information, which proc^¬ essing makes it possible to recover the defect information at any time. The small block was previously marked as ^xok' during production, and the dynamic defect recognition does not change this status. This leads to a non-defect small block again following a re-formatting.

According to the invention, this behaviour is changed in order to recognise the dynamic defects, too, during re- formatting. This can be achieved by modifying the static de^¬ fect locations of the dynamic defect small blocks during re^¬ mapping as depicted in Fig. 3. This makes the dynamic defect block recognisable during a re-formatting, because the dy- namic defect has been changed to, or re-defined as, a static defect. As stated above, each value different from OxFF, which is the value of the erased block, can be used for marking the defect. All known vendors use OxOO for marking static defects. To distinguish from the vendors marking, all values except OxOO and OxFF can be used for marking the de^¬ fects. If the block is erased before, it is possible to use the entire block for storing additional information, and this information will be kept during memory lifetime. Because of the defect, redundancy is to be used for the infor- mation. For instance, the number of dynamic defects or the date of entry could be stored in order to help product sup^¬ port estimating the flash memory quality.

Fig. 4 shows the flash memory system following a re-format- ting and erasing (i.e. an empty take). Advantageously, the dynamic defect block information has been kept.

Fig. 5 shows the physical structure of one device of a fresh formatted flash storage system from fabric with no defect blocks at all. A complete HD video storage system consists of several of such devices, each of which has a general re^¬ cording area containing mostly all of the flash blocks, ex^¬ cept the spare blocks. The file system builds a linked list of all these blocks, which is denoted a take. Before begin- ning storage, there is only one take called the ^xempty take', containing all the blocks of the recording area. The remaining spare blocks are foreseen for static and dynamic re-mapping, as explained above. The spare blocks are single blocks not contained in any list. All blocks are erased dur- ing production in the fabric. Fig. 5 shows the ideal condi^¬ tion of having no defects at all. But, as already mentioned above, in reality there are some defect blocks from fabric, called ^static defects'. Fig. 6 depicts how a static defect block is re-mapped during for^¬ matting using one of the spare blocks. Static defect blocks are identified during formatting by reading their contents, and they are marked as defect in fabric. Static defect blocks are never used for user-side storage.

Fig. 7 shows the status following recording of a take (take 1) . There are now 2 takes, the empty take and take 1. One of the blocks of take 1 has a dynamic defect.

Dynamic re-mapping is performed immediately after the re^¬ cording, using a further one of the spare blocks. Fig. 2 depicts the resulting file system.

At some future time instant, the flash memory device is re^¬ formatted because of e.g. malfunction, factory reset, firm^¬ ware update. All take-related information and all dynamic defect-related information is lost. Fig. 8 depicts this status. The static defect-related information was again read from the contents of the blocks and is restored. Following the re-formatting, the non-empty take is erased, which is shown in Fig. 9. However, a comparison of Fig. 9 with Fig. 6 makes clear that one of the spare blocks was not erased. This will lead to an erroneous data stream in case this spare block is used in future for a dynamic re-mapping. But erasing the entire spare area does not solve this prob^¬ lem. Inside the spare area there can be static defects, too, which are marked from production in fabric and are never erased because this would erase the static-defect informa^¬ tion, too.

The inventive processing for erasing the spare blocks car^¬ ries out the following steps or stages: For each flash memory device and for each spare block of the device the re-map table of the file system is analysed. The re-map table stores the static defect information from for^¬ matting. Taking this information into account ensures that no static defect block will ever be erased. Using these information items, one spare block after the other is erased. This will result again in a physical struc^¬ ture of a single flash memory device according to Fig. 6.

In an other embodiment for deriving the static-defect infor^¬ mation, the flash block content itself is read out and the static defect information which was written during production into that flash block is checked. But reading the flash block content is much slower than reading the re-map table stored e.g. in an SRAM. If the formatting which reads the flash memory content, too, is combined with the erasing, then this leads to an efficient way of erasing the spare area .

The invention can be used in any block-oriented storage that can be affected with defects occurring during memory life- time.

Claims

1. Method for recording input data into a non-volatile memory device, said memory device being arranged as at least one erasable block, each block containing multiple writable pages, to which pages said input data are written in a sequential manner, and wherein said pages may con- tain originally static defect small blocks which are marked correspondingly and for which by re-mapping spare small blocks are used as a replacement, these spare small blocks being arranged in a corresponding spare area assigned to each one of said pages, said method being char- acterisβd by: upon determining that in said recording one or more small blocks, other than said static defect small blocks, are dynamic defect small blocks, marking these dynamic defect small blocks as static defect small blocks, for re- mapping these dynamic defect small blocks to said spare area, too.

2. Method according to claim 1, wherein said memory device is a NAND flash memory device.

3. Method according to claim 2, wherein multiple ones of said NAND flash memory device form a matrix of memory devices for recording said input data in a multiplexed fashion, and wherein said input data includes high-speed video data, in particular HD video data.

4. Method according to one of claims 1 to 3, wherein said dynamic defect small blocks are marked with a code different than that for said static defect small blocks.

5. Method according to one of claims 1 to 4 wherein, upon a re-formatting of said memory device, a file system table storing information about said re-mapping is analysed and, using the static defect-related information stored in said table, spare small blocks not assigned to a static defect are erased.

6. Method according to claim 5, wherein said table is stored in an SRAM.

7. Apparatus being adapted for recording in a multiplexed fashion HD video data into a matrix of NAND flash memory devices, wherein each on of said memory devices is arranged as at least one erasable block, each block containing multiple writable pages, to which pages said HD video data are written in a sequential manner, and wherein said pages may contain originally static defect small blocks which are marked correspondingly and for which by re-mapping spare small blocks are used as a replacement, these spare small blocks being arranged in a corresponding spare area assigned to each one of said pages, said apparatus being characterised by: upon determining that in said recording one or more small blocks, other than said static defect small blocks, are dynamic defect small blocks, marking these dynamic defect small blocks as static defect small blocks, for remapping these dynamic defect small blocks to said spare area, too.