TW200820256A - Parallel threshold voltage margin search for MLC memory application - Google Patents
Parallel threshold voltage margin search for MLC memory application Download PDFInfo
- Publication number
- TW200820256A TW200820256A TW95139396A TW95139396A TW200820256A TW 200820256 A TW200820256 A TW 200820256A TW 95139396 A TW95139396 A TW 95139396A TW 95139396 A TW95139396 A TW 95139396A TW 200820256 A TW200820256 A TW 200820256A
- Authority
- TW
- Taiwan
- Prior art keywords
- read
- sum code
- memory
- sum
- programming
- Prior art date
Links
Landscapes
- Read Only Memory (AREA)
Abstract
Description
W2914 200820256, 九、發明說明: 【發明所屬之技術領域】 本發明是㈣於基於多層單元("Μ1χπ)之記壯 置’且_是有_用明取虹式記㈣裝k技^。 【先前技術】 =之快閃記憶體單元儲存—浮動閘構造或其他 何儲存構造上之電荷。_電荷 電壓(⑽。在1取動作中,—讀取電臨限 兄憶體單元之閘極,而此記憶體否此 電流h或者所料之電_旦η導通(例如傳導 狀恕。舉例而言,在讀取動作期間傳導相當 祆 憶體單元可能被指定””之數位數值,而在讀取動'作;严 傳導非常小或沒有傳輸電流 π乍肩間 〇奸之數位數值。電何倾加至此電射特構造以及 =何儲存構造被移除,収編程與抹除此記憶體單元 P ’將此儲存值從1改變至Q。此電荷係_ , 構造而被維持,直到此記怜體單 社'、' 何諸存 連績苑加電力的情況下維持此資料 而要 應用而言是非常受歡迎的。 …’後快閃記憶體 可二藉由提供選擇性不同的數量之_ =上來表儲存)夕重資料值之似係 本上,少許負電荷略微增加記怜許抑一 聊^展基 電荷更進-步增加附。—早70之Vth’且更多負 ,㈣作mx蚊此記憶體 5 200820256TW2914 單兀已被充電(編程)至何種狀態。舉例而言,在儲存2位 兀之貧料之四位準單元中,假設vth。表示在紅c &未被編 私或已被抹除(其可以是本質上不存在有電荷於電荷餘存 構造上之狀態)之臨限電壓,”比表示在相當小量之負電 荷已被傳輸至電荷儲存構造上時之臨限電壓,帅2表示在 更多負電荷已被傳輪至電荷儲存構造上時之臨限電壓,且 Vth3表不在更多負電荷已被傳輸至電荷儲存構造上時之臨 R電壓。施加在Vth。舆Vthl之間之一讀取(字線或閘)電壓 亚經由此裝置感測電流,將表示此裝置是否已被編裎,然 後’施加在vthi與Vth2之間之一字線電壓並感測此裝置是 、否已經導通’將表示此裝置是否已被編程成第〆位準或第 > =位準等等。或者,施加一固定字線電壓,而此單元所傳 導之電流係與三個參考電流同時比較。依此方式,可以在 —個讀取動作中感測此MLC所有四個位準。 、結合MLC之記憶體陣列通常以熟知方式讀取,此熟知 % 方式係施加一讀取電壓(Vt)至一選擇字線,然後感測耦合 至MLC區塊之位元線上之電流或電壓,ΜΙχ區塊係藉由使 用一排感測放大器而由字綠啟動。典型的讀取動作係為頁 面式。舉例而言,二十億位元("2Gb")之記憶體裝置(或 中之記憶體陣列)可被設計成〗28, 〇〇〇個兩千位元組 ^2ΚΒ")頁面。所感測之數值係载入至一資料閂鎖器或緩 衝器,如熟習快閃記憶體|置之技藝者所熟知的。舉例而 έ,請參見在2003年3月25日發證給parker的美國專 利第6, 538, 923號。一頁面係接著以在匹配感測放大器之 6 W2914 200820256, •數目之區塊上之操作順序被編程並讀取。感測放大器在多 f位準之連續讀取動作的情況下包含-基率,或用以平行 ⑸取此夕重位準之—組基準’此位元線上之電壓或電流係 ,、此基準或此組基準比較,藉以彳貞測此單元之批,因而 偵測此記憶體狀態。 而’在一陣列内與在單一頁面之陣列内之被編程成 特疋Zfe體狀怨之這些單元之Vth可以遍及臨限電壓之 刀佈内化。因此,被施加以讀取一 mlc之一讀取電壓或 •用來感測- MLC之輪出電流之—參考電流,必須落在在多 重編程位準之Vth電壓分佈間之間距之内 。此種間距係以 讀取電壓範圍或讀取容絲示。確認MIX裝置具有足夠讀 取電壓範圍是很重要的。 用以決定搜尋位於每個編程位準之每個頁面之每個 Vth位準之分佈之邊界之讀取容限之習知技術是費時且使 用很多儲存容量(測試器記憶體)以為不同編程位準登錄 _ 從晶片讀取之資料。因此,需要一種能避免這些習知技術 之問題之用以決定讀取容限之技術。 【發明内容】 根據本發明’提出一種操作一記憶體陣列之方法,記 憶體陣列包含MLC,每單元儲存多重位元。此方法用以決 定橫越過此陣列之多重閾值位準之讀取容限,並包含以下 步驟:執行一編程操作以儲存一已知資料集於此陣列中之 一受測區塊,而在此受測區塊中之每個單元儲存—組多重 200820256Tw29i4 位元碼中之—碼’其中在這組多重位元碼中之每個多重位 70碼係對應至可被編程於此單元中之多重_㈣^ 一。關於2位元碼,這級多會 /、 ^ 重仅%碼有四個2位元碼,祐 使用於記憶體早元之四個蘭宿乂 二 知資料集中之多重位元碼包^ 在已 、 )匕3對應於一第一閾值位準且 以如說明於此之一期望總和碼异一 α , ν ν 6表π之已知總數之碼,以;s 對應於一第二閾值位準之已知乂及 千I匕知總數之碼等等。在執行 程操作之後,使用一第一字绩吟广 仃此、扁W2914 200820256, IX. Description of the invention: [Technical field to which the invention pertains] The present invention is (iv) based on the multi-layer unit ("Μ1χπ), and _ is _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ [Prior Art] = Flash memory cell storage - floating gate structure or other storage of stored charges. _Charge voltage ((10). In the 1 action, read the gate of the battery body, and the memory does not have this current h or the electrical _ η conduction (for example, conduction). In other words, during the reading action, the equivalent value of the memory unit may be assigned a numerical value of "", while the reading is done; the transmission is very small or there is no transmission current π 乍 乍 〇 之 之 。 。 。 。. How to add to this electric structure and = the storage structure is removed, the programming and erasing of this memory unit P 'change this stored value from 1 to Q. This charge system _ , the structure is maintained until this It is very popular to use this document to maintain this information in the case of He Jicun and the company. In addition, the flash memory can be provided by providing different quantities. _ = up to the table to store) the value of the data on the basis of the date, a little negative charge slightly increased the memory of the pity, a chat, the base charge is more advanced - step increase attached. - early 70 Vth' and more negative, (d) for mx mosquitoes this memory 5 200820256TW2914 single 兀 has been charged (programmed) to What state. For example, in the four-level quasi-unit that stores the poor position of the two digits, vth is assumed. The red c & is not privately edited or has been erased (it may be essentially not present) The threshold voltage of the state of the charge on the remaining structure of the charge," the ratio indicates the threshold voltage when a relatively small amount of negative charge has been transferred to the charge storage structure, and handsome 2 indicates that more negative charge has been transmitted. The threshold voltage at the charge storage configuration, and Vth3 indicates that the R voltage is not present when more negative charges have been transferred to the charge storage structure. One of the readings between Vth.舆Vthl (word line or The voltage sensed by the device will indicate whether the device has been programmed, and then 'apply a word line voltage between vthi and Vth2 and sense if the device is turned on or not' will indicate the device Whether it has been programmed to the first level or the > = level, etc. Alternatively, a fixed word line voltage is applied, and the current conducted by the unit is compared with the three reference currents simultaneously. In this way, Sensing all four of this MLC in a read action A memory array incorporating MLC is typically read in a well-known manner. This known % mode applies a read voltage (Vt) to a selected word line and then senses the current coupled to the bit line of the MLC block or The voltage, ΜΙχ block is activated by the word green by using a row of sense amplifiers. The typical read action is page type. For example, a 2 billion bit ("2Gb") memory device ( Or the memory array) can be designed as 〖28, 两 two thousand bytes ^2ΚΒ") page. The sensed value is loaded into a data latch or buffer, such as familiar flash The memory is well known to those skilled in the art. For example, see U.S. Patent No. 6,538,923 issued to Parker on March 25, 2003. A page is then programmed and read in the order of operation on the block of matching sense amplifiers 6 W2914 200820256. The sense amplifier includes a -base rate in the case of a continuous read operation of multiple f-levels, or a voltage or current system on the bit line of the group reference d which is used to parallel (5) take the level of the level. Or this group of benchmark comparisons to detect the batch of the unit, thus detecting the state of the memory. The Vth of these cells, which are programmed in an array and in an array of single pages, are programmed to be internal to the voltage of the threshold voltage. Therefore, the reference current applied to read one of the reading voltages of one mlc or the sense of the current used to sense the -MLC must fall within the distance between the Vth voltage distributions of the multi-programming level. This spacing is shown in the read voltage range or read capacity. It is important to verify that the MIX device has a sufficient read voltage range. The prior art technique for determining the read margin of the boundary of the distribution of each Vth level at each of the programmed levels is time consuming and uses a lot of storage capacity (tester memory) for different programming bits. Quasi-login_ Data read from the wafer. Therefore, there is a need for a technique for determining read tolerance that avoids the problems of these prior art techniques. SUMMARY OF THE INVENTION In accordance with the present invention, a method of operating a memory array is provided, the memory array comprising MLCs, each cell storing multiple bits. The method is for determining a read margin across a plurality of threshold levels of the array and includes the step of performing a programming operation to store a known data set in one of the tested blocks in the array, where Each cell in the tested block stores - the code in the multiple 200820256Tw29i4 bit code - where each multiple bit 70 code in the set of multiple bit codes corresponds to multiples that can be programmed into the cell _ (four) ^ one. Regarding the 2-digit code, this level will be more than /, ^ is only the % code has four 2-digit code, you can use the multiple bit code package in the memory of the four elements of the memory of the early morning. 、3 corresponds to a first threshold level and has a known total number of codes π of the sum total number α, ν ν 6 as shown in this one; s corresponds to a second threshold bit Known 乂 and thousands of 匕 know the total number of codes and so on. After the execution of the operation, use a first word, 吟 仃 、, flat
φ 一 ^ 子線电壓讀取此受測區塊,決宏 表不被編程成一第一編程位準 、疋 1 +之—第一記憶體單元數目 之;第:=,並決定表示被編程成第二編叫 一弟二數目之早兀之_篦— 士 丁、 乐—匕頊總和。第一已讀總和碼在 與關於第一閾值位準之第一 P ▲ 貝^馬係 ^ 巳知總數比較,而第二已讀嫡 和碼係關於第二閾值位準之箓— 項〜、 總和碼比較資訊。字線電壓係達=之== 增量呈階梯狀變化,或達到此胜1 ^ j X運引此特定範圍之一部分呈階梯狀φ a ^ sub-line voltage reads the block under test, the macro table is not programmed to a first programming level, 疋 1 + - the number of first memory cells; the first: =, and the decision is programmed to The second part is called the first brother of a younger brother. _ 篦 - Sing, Le - 匕顼 sum. The first read sum code is compared with the first P ▲ 关于 关于 关于 关于 关于 , , , , , , , , 、 、 、 、 、 、 、 、 、 、 、 、 、 、 、 、 、 、 、 The sum code compares the information. The word line voltage is up to == the increment is stepwise, or the win is 1 ^ j X. One part of this specific range is stepped.
變化,而已讀總和碼及/或比較資訊係為每個階梯狀的電 壓登錄。於一特定實施例中,字線電壓係達到一特定範圍 之字線電壓,例如以100 mV之步階呈階梯狀變化。關於 此範圍之字線電壓之登錄之已讀總和碼及/或比較資訊, 係用以決定在第一編程位準與第二編程位準之間之一讀 取範圍,其中此讀取範圍係為一定範圍之字線電廢,而於 此範圍内,所施加之字線電壓位準產生數個總和碼,此些 總和碼係與在此陣列之受測區塊之内之已知總數相匹配。 在一特定實施例中,此記憶體陣列包含每單元兩個位 8 'W2914 200820256 元之多層單元(”MLCn),且資料集具有對應於一閾 位準之碼之一第三已知總數。於此情況下,第一已:總數 表不被編程成第一編程位準、第二編程位準二。 準之複數個MLC,而第二已知總數表示被= 立 位旱或弟二編裎位準之複數個。除了 士The change, while the read sum code and/or comparison information is registered for each stepped voltage. In a particular embodiment, the word line voltage is a word line voltage that reaches a particular range, e.g., stepwise in steps of 100 mV. The read sum code and/or comparison information regarding the registration of the word line voltage of the range is used to determine a read range between the first programming level and the second programming level, wherein the reading range is For a range of word lines, in this range, the applied word line voltage level produces a number of sum codes that are related to the known total number of blocks in the array under the array. match. In a particular embodiment, the memory array comprises two levels of 8 'W2914 200820256 elements per unit ("MLCn"), and the data set has a third known total of one of the codes corresponding to a threshold level. In this case, the first: total number table is not programmed to the first programming level, the second programming level is two. The plural number of MLCs, and the second known total number is represented by = standing position or brother II a multiple of the ranks.
與:二已讀總和碼以外’又計算一第三已讀總:::二: 已讀總和碼表示位於第三編程位準之一第二^ ”、、--已讀總和碼表示位於第一編程位準、第C數、目’第 第三編程位準之—第一 MLC讀取數目,而二= 表示位於第二編鞀^ 〃 昂一已碩總和碼 數目。第-4: 讀總和碼躲第itr係與弟—已知總數作比較,第二已 與第三已知總數。總數作比較,而第三已讀總和碼係 於-特定實二:Γ提供第一總和竭咖^ 層單元^^^己产 兀之受測區塊係為一頁面之多 列主題之另一部分體裝置,或一讀取動作之—記憶體陣 或比較資訊而在Μ且頁面讀取範圍係使用總和碼及/ 定。頁面讀取範程位準與第二編輕位準之間決 記憶體陣列之受测:為後來的頁面讀取動作儲存。或者, 部分。 、|區塊係一頁面之]^1^記憶體裝置為之一 在—特定實姑h 兩千位元組(214f -中,記憶體陣列之受測區塊係為一種 14>關於儲存於N I)頁面’而已讀總和瑪包含15位儿<〇 · 對於四位準單元準單元中之每-個N-1位準。因此, ,可產生三個已讀總和瑪,而對於2K位 9 200820256TW2914 ’ 元組頁面,對於總數為45位元之資料每個頁面(2K位元組 頁面儲存16Κ位元)可產生三個15位元已讀總和碼。因 此,更一般言,對於包含2ΝΜ位元之測試區塊,第一已讀 總和碼、第二已讀總和碼及第三已讀總和碼包含各個Ν位 元碼。用以供容限測試或容限搜尋用之分析,係如說明於 此地於此組Ν位元碼,而非於此組2W位元之資料上而完 成,實質上降低測試處理所需要之記憶體資源。 總和碼及/或比較資訊亦可選擇地用以確認正確之編 _ 程操作。舉例而言,如果已讀總和碼低於期望(已知)總和 碼,則可將另一編程脈衝施加至此記憶體陣列中之被選擇 之記憶體單元,並讀取新的總和碼,且將新的總和碼與期 望總和碼作比較。 應用於此之技術亦適合於單一位準單元式記憶體陣 於一特定實施例中,讀取位於第一字線電壓之每單元 包含兩個位元之MLC記憶體陣列之步驟,係包含同時比較 • 一記憶體單元之一單元電流與一第一基準值、一第二基準 值及一第三基準值,以產生兩位元資料輸出,其指示記憶 體單元之一編程狀態。 於一實施例中,具有一 MLC記憶體陣列之一積體電路 (n 1C")包含一内建自我測試("BIST")設計(包含邏輯),例 如一狀態機器與其他專用電路、由軟體所控制之一處理器 或處理器與專用電路之組合,其被設計成用以依據上述技 術來操作此MLC記憶體陣列。於一特定實施例中,BIST邏 200820256雙4 • 輯從1C提供一通過/失敗結果至一測試器,此1C指示此 陣列中之任何頁面是否符合特定最小讀取容限。 為讓本發明之上述内容能更明顯易懂,下文特舉一較 佳實施例,並配合所附圖式,作詳細說明如下: 【實施方式】 請參考第1-6圖,提供關於MLC式記憶體陣列中之讀 取容限搜尋之技術之詳細說明。And: the second read sum code is 'calculated a third read total::: two: the read sum code indicates that the second programming level is one of the second ^", --- the read sum code indicates that the first A programming level, a Cth number, a third 'programming level' - the first MLC reading number, and a second = indicating the number of the first total number of codes in the second compilation. The sum code hides the itr system and the younger brother - the known total is compared, the second has been compared with the third known total number, and the third read total code is tied to the specific one: the first total is provided ^ The layer unit ^^^ has been tested for another part of the body of a page, or a reading action - memory array or comparison information and the page reading range Use the sum code and /. The page read mode level and the second code light level between the memory array is tested: for subsequent page read action storage. Or, part., | block system one One of the pages of the ^1^ memory device is in a specific two-dimensional byte (214f -, the measured block of the memory array is one 14>About the Store on NI) page' and the read summa contains 15 digits<〇· for each of the four quasi-cell quasi-units N-1 level. Therefore, three read sums can be generated玛, and for the 2K bit 9 200820256TW2914 'tuple page, for each page with a total of 45 bits of information (2K Bytes page stores 16 Κ bits) can produce three 15-bit read sum codes. Therefore, more Generally speaking, for a test block comprising 2 bits, the first read sum code, the second read sum code, and the third read sum code comprise respective unit bit codes for tolerance testing or tolerance search. The analysis is performed as described in this group of bit codes, rather than the data of the 2W bits of this group, which substantially reduces the memory resources required for the test processing. Sum code and/or comparison information Alternatively, it may be used to confirm the correct operation. For example, if the read sum code is lower than the desired (known) sum code, another programming pulse may be applied to the selected one of the memory arrays. Memory unit and read the new sum code, and Comparing the new sum code with the expected sum code. The technique applied here is also suitable for a single level cell memory array. In a particular embodiment, reading each cell at the first word line voltage comprises two bits. The step of the MLC memory array includes simultaneously comparing: a cell current of a memory cell with a first reference value, a second reference value, and a third reference value to generate a two-dimensional data output, Indicating a programming state of one of the memory cells. In one embodiment, an integrated circuit (n 1C") having an MLC memory array includes a built-in self-test ("BIST") design (including logic), for example A state machine and other dedicated circuits, a processor controlled by software or a combination of processors and dedicated circuits, are designed to operate the MLC memory array in accordance with the techniques described above. In a particular embodiment, the BIST logic 200820256 dual 4 • provides a pass/fail result from 1C to a tester indicating whether any page in the array meets a certain minimum read tolerance. In order to make the above description of the present invention more comprehensible, a preferred embodiment will be described below with reference to the accompanying drawings, and the following detailed description is given as follows: [Embodiment] Please refer to Figures 1-6 for the MLC type. A detailed description of the techniques for read margin search in memory arrays.
⑩ 第1圖係為顯示每單元儲存兩個位元之一例示的MLC 產品之讀取電壓Vt分佈之一圖100。縱轴表示在被編程成 一閾值位準之一記憶體陣列中之單元數目,而橫轴係為被 施加至此些MLC之此些字線之Vt,其將克服此單元之此電 荷儲存構造上之傳輸電荷並允許此單元傳導電流。四個分 佈係顯示為位準L0、LI、L2、L3(狀態或資料值)。為了方 便討論之目的,位準L0表示11之一數位資料值,L1表示 10之一數位資料值,L2表示01之一數位資料值,而L3 ⑩ 表示⑽之一數位資料值;然而,這些表現是可選擇的, 而其他數值定義亦是可能的。 MLC之狀態(數值)可藉由施加一讀取電壓至此MLC之 此字線,並經由此單元感測電流而決定。舉例而言,如果 此MLC被編程成閾值位準LI、L2或L3,則在一第一 Vt容 限102(例如REF1)内之一讀取電壓將不會導通此MLC(亦 即,到達經由此單元而將在一基準位準之上的電流傳導至 一位元線之狀態)。當經由此MLC之電流係與一基準作比 11 200820256TW2914 ’較(例如藉由一感測放大器執行)時,使用者知道MLC係位 於最低位準L0(其常表示一抹除單元)。利用一類似的方 式,在第二Vt容限104與第三容限1〇6之vt範圍 (Window)内施加電壓至此字線來辨別位準L1與[2,以及 位準L2與L3。 第2A圖顯示一 MLC記憶體產品之兩個不同頁面之臨 限電壓分佈。曲線L0、LI、L2、L3表示此晶片在特定範 圍内之臨限電壓之分佈。曲線204、206、208與210表示 馨 關於不同編程位準,而在在此晶片内之一頁面中被發現之 讀取電壓之分佈。舉例而言,頁面0(η Ρ0Π )具有關於位準 L〇之臨限電壓之分佈204,以及關於位準L1之臨限電壓 之分佈206。在P0上之L0與L1之間的Vt範圍(讀取範圍) 孫為在分佈204之高邊界與分佈206之低邊界之間的電壓 差異WL一01 一P0。同樣地,WL一 12-P0表示在位準L1之臨限 電壓之分佈206之高邊界與P0上之位準L2之臨限電壓之 分佈208之低邊界之間的Vt範圍,而WL 一 23JP0表示在關 φ 於位準L2之臨限電壓之分佈208之高邊界,以及關於P0 上之位準L3之臨限電壓之分佈210之低邊界之間之Vt範 園。 第二頁面P4具有各個位準L0、LI、L2、L3之不同分 佈212、214、216、218。如上所述,P4之此些Vt範圍係 為乳一0〇>4^12一?4以及札一23—?4。於此簡化例子中, 任兩個位準之間的任何頁面之最窄讀取範圍係為 WLJ2_P0。MLC記憶體產品一般具有一最小讀取範圍規格 12 200820256娜 , (例如300 mV)。如果WL_12_P0小於此最小讀取範圍規格 (亦即小於300 mV),則此產品測試失敗。一 MLC記憶體產 品中之不同頁面可能且通常一定具有不同的Vt範圍,且 具有在不同的組之數值間之不同的讀取電壓範圍。 習知之讀取容限測試方法,係藉由將此字線位準從關 於一特定範圍之一最低值步進至一最高值,並記錄於此階 梯狀的字線電壓所感測之結果來搜尋Vt容限(讀取範 圍)。此些結果係與預先知道的編程資料集作比較。這些 記錄的結果會受到分析,以找到此容限之上下邊界。頁面 式讀取/步進WL處理係重複三次於每個頁面每單元MLC產 品(每個讀取範圍一次)之兩個位元,直到整個MLC記憶體 陣列已被讀取與容限登錄為止。 這個方法有數個缺點。每個頁面之每個位準之邊界係 被搜尋以計算此讀取範圍,其花費相當多的時間。需要一 個大型測試器記憶體來儲存此整個晶片之位元資訊(例如 在上述例子中為二十億個位元)。來自此整個晶片之登錄 • 資料係被評估以計算每個讀取單元之邊界,而此必須為每 個讀取範圍執行(例如在一四個位元MLC裝置中為三次)。 第2B圖係為顯示決定一 MLC式記憶體之此些讀取電 壓範圍(Vt容限搜尋)之一習知技術方法200之流程圖。一 基準值(例如基準位準1("REF1"))係被設定以表示一第一 資料值(步驟220),並連接至一感測放大器陣列。施加至 此頁面之字線位準係被設定至一特定範圍中之最低位準 (步驟222)(或者至最高位準或一任意位準)。此頁面係被 13 200820256tw2914 , 讀取(步驟224),而此資料係被輸出並登錄在此測試器記 憶體中(步驟226)。如果此全晶片並未被讀取(分支228), 則此頁面數會增加(k+1),且重複這個迴圈直到此記憶體 中之所有頁面於這個字線位準被讀取為止(分支23〇)。如 果此字線電壓並非位於此最大允許值(分支232),則此字 線電壓係增加(j + 1),且針對每個字線電壓值重複此迴圈 (分支228),直到達到此最大字線電壓為止(分支。 如果此基準值並非此最大基準值(例如ref3)(分支 ⑩236),則此基準值係步進mi)(例如從_至REF2或 遞至剛),並針對每师賴每個字線鳄重複迴圈 228與232。在此字線電壓已對於所有三個基準值呈階梯 狀變化(分支238)之後,分析整個晶片之記錄資料(步驟 240),並決定晶片之最小讀取範圍(步驟 取 之任何頁面上之任何兩個數值之間之最 窄項取範圍。 • 彳目較於習知技術之下,本發明之實祕使用已讀總和 響碼(jS-read麵code)以決定MLC式記憶體產品之最小讀 範圍而非王曰曰片紀錄為使用於第2b圖之處理。已讀 總和碼表示絲個讀取㈣而於每個頁面巾讀取之位於 既定閾值位準之單元之數目。在一兩個位元紅中,有三 個讀取電麼範圍,一個在^與“之間,第二個在li^ L自2之間,而第三個在12與u之間。因此,使用三個已讀 二=碼。其他實施例具有更多或更少資料位準,而讀取電 壓範圍之數目對應增加或減少。為了討論之便利性起見, 200820256w29i4 兩位元MLC將被使用於一例子中,且將說明三個已許娘 碼 SUM1、SUM2、SUM3。 第3圖顯示在MLC式記憶體陣列中相對於編程^立準平 行產生之總和碼。SUM3係為於編程位準三L3所感測到 單元數目。舉例而言,如果一頁面係兩千位元組,而每單 元有兩個位元,則關於SUM3[0 ·· 14],我們將具有位一 寬度’此乃因為編程成位準L3之最兩數目單元係為1(其 思指此頁面中之所有MLC已被編程成L 3)。層三之最】^ ⑩元數目係為〇(其意味著所有頁面並不具有任何編程成 之 MLC) 〇 SUM2係為於編程位準L2與L3所感測到之單元數目 兩千位元組頁面可能沒有被編程成L2或L3之頁面之 MLC(亦即,頁面上之每個位元係位於li或L〇),於 、10 Figure 1 is a graph 100 showing the read voltage Vt distribution of an MLC product exemplified by one of two bits stored per cell. The vertical axis represents the number of cells in a memory array programmed to a threshold level, and the horizontal axis is the Vt applied to the word lines of the MLCs, which will overcome the charge storage structure of the cell. Transfers charge and allows this unit to conduct current. The four distributions are displayed as levels L0, LI, L2, L3 (status or data values). For the purpose of discussion, the level L0 represents 11 one-bit data value, L1 represents 10 one-digit data value, L2 represents 01 one-digit data value, and L3 10 represents (10) one-digit data value; however, these performances It is optional, and other numerical definitions are also possible. The state (value) of the MLC can be determined by applying a read voltage to the word line of the MLC and sensing the current via the cell. For example, if the MLC is programmed to a threshold level LI, L2, or L3, then one of the read voltages within a first Vt tolerance 102 (eg, REF1) will not turn on the MLC (ie, arrive via via This unit conducts current above a reference level to the state of a bit line). When the current through the MLC is compared to a reference ratio of 11 200820256TW2914 ' (for example, by a sense amplifier), the user knows that the MLC is at the lowest level L0 (which often represents an erase unit). In a similar manner, voltages are applied to the word line in the vt range 104 of the second Vt tolerance 104 and the third tolerance 1〇6 to identify levels L1 and [2, and levels L2 and L3. Figure 2A shows the threshold voltage distribution for two different pages of an MLC memory product. The curves L0, LI, L2, L3 represent the distribution of the threshold voltage of the wafer within a certain range. Curves 204, 206, 208, and 210 represent the distribution of read voltages found in one of the pages within the wafer with respect to different programming levels. For example, page 0 (η Ρ 0 Π ) has a distribution 204 of threshold voltages for level L , and a distribution 206 of threshold voltages for level L1. The Vt range (read range) between L0 and L1 on P0 is the voltage difference WL_01-P0 between the high boundary of the distribution 204 and the low boundary of the distribution 206. Similarly, WL-12-P0 represents the Vt range between the high boundary of the threshold voltage distribution 206 of the level L1 and the low boundary of the threshold voltage 208 of the level L2 on P0, while WL-23JP0 Represents the high boundary of the distribution 208 of the threshold voltage φ at the level L2, and the Vt range between the low boundaries of the distribution 210 of the threshold voltage with respect to the level L3 on P0. The second page P4 has different distributions 212, 214, 216, 218 of the respective levels L0, LI, L2, L3. As mentioned above, the Vt ranges of P4 are milk-zero 〇>4^12 one? 4 and Zhayi 23-?4. In this simplified example, the narrowest read range for any page between any two levels is WLJ2_P0. MLC memory products typically have a minimum read range specification of 12 200820256 Na, (eg 300 mV). If WL_12_P0 is less than this minimum read range specification (ie less than 300 mV), then this product test fails. Different pages in an MLC memory product may and often must have different Vt ranges and have different read voltage ranges between different sets of values. The conventional read tolerance test method searches for the word line level from the lowest value of one of the specific ranges to a highest value, and records the result of sensing the stepped word line voltage. Vt tolerance (read range). These results are compared to a previously known set of programming data. The results of these records are analyzed to find the lower bound above this tolerance. The page read/step WL process is repeated three times per cell per cell MLC product (once per read range) until the entire MLC memory array has been read and tolerantly logged. This method has several drawbacks. The boundary of each level of each page is searched to calculate this read range, which takes a considerable amount of time. A large tester memory is needed to store the bit information for this entire wafer (for example, two billion bits in the above example). Logins from this entire wafer • The data is evaluated to calculate the boundaries of each read unit, which must be performed for each read range (e.g., three times in a four-bit MLC device). Figure 2B is a flow chart showing one of the prior art methods 200 for determining such read voltage ranges (Vt tolerance search) for an MLC type of memory. A reference value (e.g., reference level 1 ("REF1")) is set to represent a first data value (step 220) and is coupled to a sense amplifier array. The word line level applied to this page is set to the lowest level in a particular range (step 222) (either to the highest level or an arbitrary level). This page is read (step 224) by 13 200820256 tw2914, and this data is output and registered in this tester memory (step 226). If the full wafer is not being read (branch 228), the number of pages will increase (k+1) and the loop will be repeated until all pages in the memory are read at the word line level ( Branch 23〇). If the word line voltage is not at this maximum allowable value (branch 232), then the word line voltage is increased (j + 1) and this loop is repeated for each word line voltage value (branch 228) until this maximum is reached The word line voltage is up (branch. If this reference value is not this maximum reference value (eg ref3) (branch 10236), then this reference value is stepped mi) (eg from _ to REF2 or handed to just) and for each division Lay each word line crocodile repeats loops 228 and 232. After the word line voltage has changed stepwise for all three reference values (branch 238), the entire wafer's recorded data is analyzed (step 240) and the minimum read range of the wafer is determined (steps are taken from any of the pages) The narrowest term between the two values is taken. • Under the prior art, the reality of the present invention uses the read sum code (jS-read face code) to determine the minimum of the MLC memory product. The read range, rather than the Wang Bian film record, is the process used in Figure 2b. The read sum code represents the number of cells that are read (4) and read at each page towel at a given threshold level. In the bit red, there are three reading ranges, one between ^ and ", the second between li^ L and 2, and the third between 12 and u. Therefore, three are used. Read two = code. Other embodiments have more or less data levels, and the number of read voltage ranges corresponds to increase or decrease. For the sake of convenience of discussion, 200820256w29i4 two-element MLC will be used in an example. Medium, and will explain three confirmed mother code SUM1, SUM2, SUM3 Figure 3 shows the sum code generated in the MLC-type memory array relative to the programming quasi-parallel. SUM3 is the number of cells sensed at the programming level three L3. For example, if a page is two thousand bits Group, and each unit has two bits, then for SUM3[0 ·· 14], we will have a bit width 'this is because the most two number of units programmed into level L3 is 1 (think of this page) All MLCs have been programmed to L 3). The best of layer 3 is ^ 10 number is 〇 (which means that all pages do not have any programming MLC) 〇 SUM2 is for programming level L2 and L3 The number of sensed units of two thousand bytes may not be programmed into the MLC of the page of L2 or L3 (ie, each bit on the page is located in li or L〇),
下,此數目係為0,或所有MLC係為L2或1^3,於此狀、^1 下,此數目係為16K。因此,SM2亦具有15位元寬声/ 如果SUM 2並非為〇,則位於L2之MLC之數目可由=】偷 減去SUM3所決定。 ^ SUM 1係為於編程位準l 1、L2與L3所感測到之單元 數目。再者,SUM1具有15位元寬度,而被編程成[I、u 與L3之單元之數目可能使用SUM1、SUM2、SUM3來決定。 如果關於SUM1之位元數目係為零,則位於此頁面上之所 有MLC係被假設為位於L0。可僅使用45位元(三個仅準各 15位元)來表示整個兩千位元組頁面。 SUM1、SUM2、SUM3之期望值係從將被編程成Μΐχ陣 15 200820256TW2914 資料载入至記憶體陣列之前或之後 續取之^ =主總和碼係與在MLC頁面已被編程之後被 。如果期望總和瑪匹配已讀總和碼, 字線(直係在“正心取已產生。換言之’藉由使用施加至 子線(其係在硬取容限範圍之内)之一讀取電壓來讀取頁 ® ° 、 此内部邏輯控制MLC記憶體裝置之測試讀取過程,用 參以執㈣複的讀取循環,直到一負載緩衝器係充滿一個頁 面=貧訊(參考第5A圖關於讀取結構之一例)為止。由此 二《貝取> 料值(MLC位準)所決定之頁面之總和碼係與那個 頁面之期望總和碼作比較。舉例而言,如果編程頁面具 有(SUM3 ’ SUM2,SUM1) = (15, h OBFF,15, h 33FF,if h 3C00) ’ 則 L3 有 3071(3Κ-1)位元;L2 有 10240 (10Κ) 位元;L1有2049 (2K-1)位元;且L0有1024 (1K)位元。 所以 ’ SUM3 = (3Κ-1) ; SUM2 = (13Κ-1) ; SUM1 (15Κ)。 _ 然而’在一讀取動作之後,我們可獲得新的總和碼(SUM3, SUM2,SUM1) (15,h 0BFD,15,h 33FF,15,h 3C00)。 新的總和碼顯示於L3有兩個位元,其係在此讀取偏壓狀 態(字線電壓)下被讀取為L2。第三層基準位準(1^^3)是錯 誤的,且其意味著如果我們想要獲得正確資料,就應將 REF3調得較低。 然而,在一讀取動作之後,如果對於相同的資料集, 我們獲得已讀總和碼(SM3,SM2,SUM1) = (15, h 0BFF, 16 200820256_4 • 15,h 33FF,15,h 3C02),則我們將知道在此讀取偏壓 狀態(字線電壓)下,於L2存在有一個位元,其係被讀取 為L1,而於L0存在有兩個位元,其係被讀取為L1。於此 例中,REF2與REF1兩者應被調整,以便讀取整個頁面之 正確資料值。 編程位準L1具有一編程確認位準PV1 ("下vt極限π) 以及一 L1之上限EV1,編程確認位準PV1通常位於被編程 成L1之記憶體陣列之MLC之可允許Vth分佈之下端。相 _ 似地,編程位準L2具有被編程成L2之MLC之一下限PV2 以及一上Vt極限EV2。第三編程位準L3具有類似的極限, 為簡化圖例之便,其並未詳細表示。 類似技術可被應用至SLC式、每單元一位元之記憶體 陣。 以下說明依據本發明之一實施例之用以測試讀取容 限之MLC式記憶體陣列之操作方法。第一與第二期望總和 碼係從陣列資料(亦即,將被編程或已被編程為一 MLC測 • 試區塊之資料,例如一 MLC記憶體裝置之一選擇之頁面) 計算出。第一期望總和碼表示將被編程成至少一第一編程 位準(亦即’被編程成一第一編程位準與一第二編程位準 之單元之總數)之MLC之數目。第二期望總和碼表示將被 編程成一第二編程位準之數目。 在記憶體陣列(例如頁面)之MLC係於各種字線電壓 項取’且计算弟一與弟二已Ί買總和碼。於一實施例中,初 始字線電麼係最低之允許字線電壓,而在讀取MLC之後, 17 2〇〇820256rW2914 " 字線電壓會增加。或者,字線電壓開高走低,或另外被改 變成獲得期望比較資訊。第一已讀總和碼表示以第一編程 位準或第二編程位準讀取之MECS之數目(亦即,以l 1與 L2讀取之單元之總數)。第二已讀總和碼表示以第二編程 位準讀取之MLC之數目。 於每個字線電壓之已讀總和碼係與相對應的期望總 和碼作比較。於某些實施例中,特別用以感測所以編程位 準之整個範圍之字線電壓係受到評估。或者,字線範圍之 。卩为文到才估。通常表示讓已讀總和碼係與期望總和碼 相符之字線電壓之總和碼比較資訊係被儲存。在第一編程 伋準與第二編程位準之間之一讀取範圍(vt容限)係依據 總和碼比較資訊而決定。 第4圖顯示依據本發明之一實施例之MLC式記憶體陣 列之操作方法411之流程圖,用以表示晶片是否通過一讀 取容限測試,此操作方法最好是由晶片上BIST邏輯執行。 馨 ▲種才欢查、、&寿方去係用以基於從此頁面讀取之資料,來比 ^ 45位元總和竭(在2K位元組頁面中是每單元兩位元)與 f之45位το總和喝,且此檢查總和方法係被使用以決 ^頁面之取動作是否已經通過或失敗。通過/失敗 貝訊係為45位兀(每一個SUM1、SUM2、SUM3為15位元), ,非=使用於習知之方法之每個頁面有2K位元組之資 σί1 &可改善測試時間,減少需要用以執行Vt容限測試 之^己饭、體貝源’並促進具有MLC式記憶體陣列之1C之内 建自我測試(nBlSTn)。 18 ;W2914 200820256, 依據顯示於第4圖之方法,從將被編程為一頁面之 MLC陣列之頁面資料值計算出期望總和碼,或此總和碼係 以其他方式提供。期望總和碼係被讀取與儲存以供由測試 邏輯存取(步驟412)。字線位準係被設定成—初始數值(例 如特別供一讀取動作用之最低字線電壓)(步驟414)。頁 面係被讀取以產生讀取資料(步驟416)。利用此讀取資 料,可決定已讀總和碼並將已讀總和碼與载入至步驟Below, this number is 0, or all MLC systems are L2 or 1^3, in this case, ^1, this number is 16K. Therefore, SM2 also has a 15-bit wide sound / If SUM 2 is not 〇, the number of MLCs located at L2 can be determined by == stealing SUM3. ^ SUM 1 is the number of cells sensed by programming levels l 1 , L2 and L3. Furthermore, SUM1 has a 15-bit width, and the number of cells programmed to [I, u, and L3 may be determined using SUM1, SUM2, and SUM3. If the number of bits for SUM1 is zero, then all MLCs located on this page are assumed to be at L0. The entire two thousand-byte page can be represented using only 45 bits (three only 15 bits each). The expected values of SUM1, SUM2, and SUM3 are from before or after the data is to be loaded into the memory array. Continued ^^ The main sum code is after the MLC page has been programmed. If the summation mar is expected to match the read sum code, the word line (directly in the "positive center has been generated. In other words" by using a voltage applied to the sub-line (which is within the hard-tolerance range) to read the voltage Read page ® °, this internal logic controls the test read process of the MLC memory device, and uses the read (four) complex read cycle until a load buffer is full of a page = poor (refer to Figure 5A for reading) Take the structure as an example). The sum code of the page determined by the second "beats" (MLC level) is compared with the expected sum code of that page. For example, if the programming page has (SUM3 ' SUM2, SUM1) = (15, h OBFF,15, h 33FF,if h 3C00) ' Then L3 has 3071 (3Κ-1) bits; L2 has 10240 (10Κ) bits; L1 has 2049 (2K-1) Bit; and L0 has 1024 (1K) bits. So ' SUM3 = (3Κ-1) ; SUM2 = (13Κ-1) ; SUM1 (15Κ). _ However, after a read action, we can get new The sum code (SUM3, SUM2, SUM1) (15, h 0BFD, 15, h 33FF, 15, h 3C00). The new sum code is shown in L3 with two bits. , which is read as L2 under this read bias state (word line voltage). The third level reference level (1^^3) is wrong, and it means that if we want to get the correct data, REF3 should be adjusted lower. However, after a read action, if for the same data set, we get the read sum code (SM3, SM2, SUM1) = (15, h 0BFF, 16 200820256_4 • 15, h 33FF,15,h 3C02), then we will know that under this read bias state (word line voltage), there is one bit in L2, which is read as L1, and there are two in L0. The bit is read as L1. In this example, both REF2 and REF1 should be adjusted to read the correct data value for the entire page. The programming level L1 has a programming acknowledgment level PV1 (" Vt limit π) and an upper limit EV1 of L1, the programming confirmation level PV1 is typically located below the allowable Vth distribution of the MLC of the memory array programmed to L1. Similarly, the programming level L2 has been programmed to L2 One of the lower limit of the MLC PV2 and an upper Vt limit EV2. The third programming level L3 has a similar limit, The legend is not shown in detail. A similar technique can be applied to an SLC-style memory cell array of one unit per cell. The following describes an MLC type for testing read tolerance according to an embodiment of the present invention. The method of operation of the memory array. The first and second desired sum codes are calculated from the array data (i.e., the data to be programmed or programmed into an MLC test block, such as a page selected by one of the MLC memory devices). The first desired sum code represents the number of MLCs to be programmed into at least one first programming level (i.e., the total number of cells programmed to a first programming level and a second programming level). The second desired sum code represents the number to be programmed into a second programmed level. The MLC in the memory array (e.g., page) is taken from various word line voltage terms and the sum code of the brothers and brothers is calculated. In one embodiment, the initial word line voltage is the lowest allowed word line voltage, and after reading the MLC, the 17 2 〇〇 820256rW2914 " word line voltage is increased. Alternatively, the word line voltage is turned high or low, or otherwise changed to obtain the desired comparison information. The first read sum code represents the number of MECSs read at the first programmed level or the second programmed level (i.e., the total number of cells read at l 1 and L2). The second read sum code represents the number of MLCs read at the second programmed level. The read sum code pattern for each word line voltage is compared to the corresponding expected sum code. In some embodiments, the word line voltage, particularly to sense the entire range of programmed levels, is evaluated. Or, the word line range. I am only able to estimate it. It is generally indicated that the sum code comparison information of the word line voltages for which the read total code system matches the expected sum code is stored. A read range (vt tolerance) between the first programming level and the second programming level is determined based on the sum code comparison information. 4 is a flow chart showing an operation method 411 of an MLC-type memory array according to an embodiment of the present invention, for indicating whether a wafer passes a read tolerance test, and the operation method is preferably performed by BIST logic on a wafer. . Xin ▲ species only check, and Shoufang to use based on the information read from this page, than the total of ^ 45 bits (in the 2K byte page is two yuan per unit) and f The 45-bit το sum is consumed, and this check sum method is used to determine whether the page take action has passed or failed. The pass/fail beta system is 45 digits (each SUM1, SUM2, and SUM3 is 15 digits), and non = = 2K bytes for each page used in the conventional method σί1 & can improve test time , to reduce the need to perform Vt tolerance testing, and to promote the built-in self-test (nBlSTn) of 1C with MLC-type memory array. 18; W2914 200820256, according to the method shown in Figure 4, the expected sum code is calculated from the page data value of the MLC array to be programmed as a page, or the sum code is otherwise provided. The desired sum code is expected to be read and stored for access by the test logic (step 412). The word line level is set to an initial value (e.g., the lowest word line voltage for a particular read operation) (step 414). The page is read to produce a read material (step 416). Using this read data, you can determine the read sum code and load the read sum code into the step
中之數值作比較,以為每個字線位準提供一比較結果(步 驟418)。關於字線位準之比較結果係被儲存(步驟42〇)。 參見第5C圖’可以看出比較結果可被儲存為單純的一位 元通過/失敗數值,其乃關於此單元之每個閾值位準之每 個字線位準。如果整個晶片並未以此字線位準評估(分支 422),則增加此頁面數值且迴圈繼續直到全晶片已 被評估為止(分支424)。 90 予線位準料到評估,而如果沒有位於關於此讀取動 作之取大允許的字線位準(分支426),則此面係 一個字線位準(1 + 1)受_估,以增加類似第5C圖之-資 t集。或者,此字線數值開騎録大允許的數值並於每 -人迴圈通過時減少,或此字線數值開 7每次通過時’改變至另-數值,直到有興;= ,數值已被評估為止。類似地,在繼續至下—個頁面之 爾’可以以有興趣之全部字線數值來評估—頁面。其他實 ^例具有替代順序之步驟。、 在以有興趣之所有字線數值來評估全晶片之後(分支 19 200820256,· 428),計算最小讀取範圍(步驟43〇),以下參考第5A—5c 圖作更進一步之說明。 第5A圖係為被使用來在每單元中同時讀取兩個位元 之感測結構500之例子。位元線電壓負載電路5〇从經由 箝位電路504B提供負载電流至ΜΙχ單元5〇2,並以類似方 式提供負載電流至二個參考單元卿〗、REF2、腳3。於此 單7G之位70線上之負載電路5〇4A係由一感測致能信號 SENB所控制,而此些參考單元位元線上之此些負載電路 • 5·總是於此例子中設定。參考單元係使用一電路測試 為、電路編程器、晶片上基準編程電路減他技術來進行 編輊,俾能使它們每個具有不同的臨限電壓。當一字線電 壓(WL)被施加至此些參考單元與ΜΙχ單元5〇2時,每個將 產生不同的參考電流IREF1、lRm、丨鹏,且MLC單元502 將產生ICell 〇 讀取動作表示MLC單元502已被編程成這四個 編程位 準(L0、LI、L2、L3)中之哪一個。參考單元係受到偏壓, 俾能使每個參考單元傳導一參考電流,其與相關的編程位 準之Vt成反比。 感測放大器506、508、510比較ICell與I_、hm及 I酬以分別產生資料值])〗、D2與D3。如果經由此MLC單 元502之電流係大於一參考電流,則其表示MLC單元之 係 於參考单元之Vfh。舉例而言,如果ICELL大於IREF1, 則感測放大器506輸出邏輯"丨"之di數值。其他感測放大 器以類似方式操作,而來自這些感測放大器5〇6、508、510 20 200820256 二适;r細a/u · FW2914 之資料值(亦即,邏輯”1”或邏輯"0")D1,D2、D3係被平行 提供至一資料計算區塊512。The values in the comparison are compared to provide a comparison result for each word line level (step 418). The result of the comparison regarding the word line level is stored (step 42〇). Referring to Figure 5C, it can be seen that the comparison result can be stored as a simple one-bit pass/fail value for each word line level for each threshold level of the cell. If the entire wafer is not evaluated by this wordline level (branch 422), then the page value is incremented and the loop continues until the full wafer has been evaluated (branch 424). 90 The line position is expected to be evaluated, and if there is no word line level (branch 426) that is allowed for this read action, then the face is leveled by a word line (1 + 1). In order to increase the similarity of the 5C chart - the set of funds. Or, the value of this word line is recorded and the maximum allowed value is reduced and decreased when each person passes the loop, or the value of the word line is opened 7 each time to 'change to another value, until there is a leap; It is evaluated. Similarly, the page can be evaluated with the value of all word lines of interest as it continues to the next page. Other examples have steps to replace the order. After evaluating the full wafer with all the word line values of interest (branch 19 200820256, · 428), calculate the minimum read range (step 43〇), which is further explained below with reference to Figures 5A-5c. Figure 5A is an example of a sensing structure 500 that is used to simultaneously read two bits in each cell. The bit line voltage load circuit 5 〇 supplies the load current from the clamp circuit 504B to the ΜΙχ cell 5 〇 2 and supplies the load current to the two reference cells REF, REF 2, and foot 3 in a similar manner. The load circuit 5〇4A of the 70G line of the single 7G is controlled by a sensing enable signal SENB, and such load circuits on the reference unit bit lines are always set in this example. The reference cells are programmed using a circuit test, a circuit programmer, and a on-wafer reference programming circuit to enable them to have different threshold voltages. When a word line voltage (WL) is applied to such reference cells and ΜΙχ cells 5〇2, each will generate a different reference current IREF1, lRm, 丨, and the MLC cell 502 will generate an ICell 〇 read action representation MLC Unit 502 has been programmed to which of the four programming levels (L0, LI, L2, L3). The reference cells are biased such that each reference cell conducts a reference current that is inversely proportional to the associated programming level, Vt. The sense amplifiers 506, 508, 510 compare the ICell with I_, hm, and I to generate data values], D2, and D3, respectively. If the current through the MLC unit 502 is greater than a reference current, it represents the Vfh of the MLC unit that is tied to the reference unit. For example, if ICELL is greater than IREF1, sense amplifier 506 outputs a di value of logic "丨". Other sense amplifiers operate in a similar manner, from these sense amplifiers 5〇6, 508, 510 20 200820256; r fine a/u · FW2914 data values (ie, logic "1" or logic "0" ;) D1, D2, D3 are provided in parallel to a data calculation block 512.
資料計算區塊產生兩位元資料輸出Dm,其指示此MLC 單元係被編程成哪一位準。舉例而言,如果MLC單元係位 於編程位準L0,則ICell係大於它們的ΙκΕΠ,大於IREF2,並 大於Im3,而資料計算區塊輸出DoiTT = 11(“壹壹,,)。類似 地’ D〇UT = 1〇 (”壹零”)表示此MLC單元係被編程成L1等。 表1顯示在感測放大器輸出與DoiJT之間之代表關係: 表1 D1 D2 D3 D〇UT 1 1 1 11 0 1 1 10 0 0 1 01 0 0 0 00 第5B圖顯不依據一實施例之一例示的總和碼計算區 總和碼計异區塊使用標準邏輯式比較器與加法器區 因此省略詳細之操作說明。從一組資料計算區塊(例 D 512 亦睛參見第5A圖,512,D_)之資料輸出(例如 52?°係被提供至供複數個單元用之三個比較器儲存體 哭^ 524、526。於此例子中,係使用料個並聯感測放大 塊 塊 評估64個單元。可以熟習本項技 从无、來平打操作其他數目之感測放大器。 21 200820256 一-«麵 m · rW2914 L2或L3位準之單元。比較器輸出係被提供至一 器528以產生讀,其係儲存於立元暫存哭53〇中 第二比較器儲存體524比較複數個資料輸出與Q1或〇〇數 值,其表示被編程成L2或L3位準之單& ·/ 出至二弟二加法器534以產生SUM2,其係儲存於一第二 16位元暫存器536。第三比較器错存體526比較複數個資 料輸出與〇〇數值,其表示此些單元被編程成L3位準;並 提供此些輸出至一第三加法器538來產生SUM3,其係儲存 • 於一第三16位元暫存器540。 第一,第二與第三比較器儲存體522、524、526平行 操作,並配合加法器528、534、538以平行產生SUM1、SM2、 SUM3。被編程成L3位準之單元之數目係以SUM3表示。被 編程成L2位準之單元之數目係由將SUM2減去SUM3所決 定(參見第3圖),而被編程成L1位準之單元之數目係由 將SUM1減去SUM2所決定。剩下的單元係位於l〇(抹除) 位準。換言之,在64個MLC之單一讀取期間,總和碼SUM1、 響 SUM2、SUM3表示64個MLC中有多少個是位於U、L2、L3。 得知在此測試區塊中之單元之總數亦允許吾人決定位於 L0之單元之數目,換言之,單元之總數減去SUM1。使用 這些技術,可使用單一的讀取順序來確認被編程成不同位 準之MLC陣列。並不需要為每個編程位準而以許多字線電 壓來重複讀取陣列(比較第2B圖,標號236)。 第5C圖顯示使用說明於此之此技術作容限測量之真 值表,其乃關於1C中之一頁面之MLC記憶體。於此圖表 22 PW2914 200820256 之第一列’顯示之字綠恭γ 並以議毫伏特增加4料開始於4.5V並結束於6.1V 記憶體陣狀-特定範^ ^至^/之顧表示關於此 nREFl錯誤?,,之第一歹j之子線電塵。於此圖表中之標為 較資訊,而第—基準位準1顯示第一基準位準之總和碼比 位準是錯誤的。或者,:示關Γ字纏之REF1之 其表示位準U縣㈣示絲之單元之數目, 配从SUM1表示之期望數目,而”〇"The data calculation block generates a two-dimensional data output Dm indicating which of the MLC units is programmed. For example, if the MLC unit is at the programming level L0, then the ICell is greater than their ΙκΕΠ, greater than IREF2, and greater than Im3, while the data calculation block outputs DoiTT = 11 ("壹壹,,"). Similarly 'D 〇UT = 1〇(“壹壹”) indicates that this MLC unit is programmed to L1, etc. Table 1 shows the representative relationship between the sense amplifier output and DoiJT: Table 1 D1 D2 D3 D〇UT 1 1 1 11 0 1 1 10 0 0 1 01 0 0 0 00 Figure 5B shows a summary code calculation area according to one of the embodiments. The sum code block uses a standard logic comparator and an adder area, thus omitting detailed operations. Description: Data output from a set of data calculation blocks (Example D 512 also see Figure 5A, 512, D_) (for example, 52 ° ° is provided to the three comparators for multiple units of crying ^ 524, 526. In this example, 64 parallel units are evaluated using a parallel sense amplification block. It is familiar with this technique to operate other numbers of sense amplifiers from none to level. 21 200820256 一-«面m · rW2914 L2 or L3 level unit. Comparator output is provided to The 528 is operative to generate a read, which is stored in the epoch temporary crying 53 第二 the second comparator bank 524 compares the plurality of data outputs with a Q1 or 〇〇 value indicating a single & of the L2 or L3 level programmed And / out to the second two adder 534 to generate SUM2, which is stored in a second 16-bit register 536. The third comparator error 526 compares a plurality of data output and 〇〇 values, which represents The cells are programmed to the L3 level; and the outputs are provided to a third adder 538 to generate SUM3, which is stored in a third 16-bit register 540. First, second, and third Comparator banks 522, 524, 526 operate in parallel and cooperate with adders 528, 534, 538 to produce SUM1, SM2, SUM3 in parallel. The number of cells programmed to the L3 level is represented by SUM3. Programming to L2 The number of cells is determined by subtracting SUM2 from SUM3 (see Figure 3), and the number of cells programmed to the L1 level is determined by subtracting SUM1 from SUM2. The remaining cells are located at l〇 (Erase) Level. In other words, during a single read of 64 MLCs, the sum code SUM1, rang SU M2 and SUM3 indicate how many of the 64 MLCs are located in U, L2, and L3. It is known that the total number of units in the test block also allows us to determine the number of units located in L0, in other words, the total number of units minus SUM1. Using these techniques, a single read sequence can be used to identify MLC arrays that are programmed to different levels. There is no need to repeatedly read the array with many word line voltages for each programming level (compare Figure 2B, Reference numeral 236). Fig. 5C shows a truth table using the technique described herein for tolerance measurement, which is related to the MLC memory of one page in 1C. In the first column of the chart 22 PW2914 200820256, the word 'Green' γ is displayed and the increase of 4 volts starts at 4.5V and ends at 6.1V memory matrix - the specific range ^ ^ to ^ / This nREFl error? ,, the first 歹j sub-line electric dust. The figure in this chart is labeled as the comparison information, and the first reference level 1 shows that the sum of the first reference levels is wrong. Or, the number of units of the REF1 that indicates the level of the REF1, which indicates the level of the U County (four), is matched with the expected number indicated by SUM1, and “〇"
、# :目互相匹配。在此圖表上之標為"REF2錯誤?π ::一歹丄9係顯不關於第二基準位準之總和碼比較資訊, ,、、不_ 之位準對於那個字線電壓是錯誤的。或者, 從最低字線電壓(於此例子中係為4·5ν)開始,執行 一頁面讀取動作,並比較關於所有三個位準之總和碼與此 期望碼。關於此低字線電壓,所有三個總和碼之比較將表 示故障,使人聯想到REF1、REF2與REF3應被調整(REn 可能太高)或此字線電壓係離開適當的容限。當字線電壓 係提高(於此例子是到達4· 8V),SUM1將變成正確(REFi錯 誤=〇(偽)),因為正確之被編程成LI、L2與L3之绝和之 Γΐΐ?表示讀取之數目單元,其表示位準L2並未匹配 /、SUM1所表示之期望之數目,而"〇”表示此些數 目相互匹配。此圖表上標為„REF3錯誤?„之第三列,係顯 示關於第三基準位準之總和碼比較資訊,其表示鹏之 碑關於那個字線電壓是錯誤的。或者,"1,,第三列表示 讀取之單元之數目’其表示位準L3並未匹配由s腦、漏 與SUM1表示之期望數目,而”〇"表示此些數目相互匹配。 23 FW2914 200820256 數目MLC之總數將以此Vt被讀取。然而,不存在有在以 不同位準(參見第3圖,SUM1)編程之單元之數目之間之區 別。SUM2與SUM3將是錯誤的(Rej?2錯誤=1,REF3錯誤= 1) ’假設在此頁面有某些MLC位於L2與L3。當字線電>1 繼續提高至闕於一頁面讀取動作之最大容許數值,可獲得 顯示於真值表500之結果。數值與條件係僅作例示的目 的,並僅提供作為圖例與討論之目的。 在每個位準之間之讀取電壓範圍對關於此頁面之每 個基準’係可從顯示第5C圖之圖表之資料決定。從第5C 圖可見’關於字線電壓之讀取範圍對REF1係從4. 8V至 5· 2V,其意味著在L0與L1之間有400 mV。類似地,關於 字線電壓之讀取範圍對REF2係從5· 0V至5· 4V,其意味著 在L1與L2之間有400 mV。關於字線電壓之讀取範圍對 REF3係從5· 3V至5· 6V,其意味著有3〇〇 my在L2與L3 之間。這種處理係為其他頁面之MLC記憶體陣列重複("全 晶片讀取”),並獲得最小讀取電壓範圍。 BIST可籍由提供特疋之最小讀取範圍數值至一测试 器而執=二在晶片上之BIST邏輯自動執行全晶片讀取動 作’、同時字線電壓係橫越過其範圍呈階梯狀變化。μ 輯或測試H儲存讀取絲,計算這些總㈣ = ίΓτΐί見第6圖)與可允許(肢)之最小讀取範7 BIST邏輯可選擇地傳遞此IC晶片之通過 圍 者,可將個別頁面之通過/失敗資訊提 可將MLC書己憶體陣列設定成”被鎖面失^ ’ 〈夭敗頁面而無 24 200820256 二達編抓· i’W2914 法被消費者使用。雖然這會減少消費者於IC上可利用的 之總§己憶體,但是其允許無法通過最小讀取範圍規格之具 有一個或多個頁面1C之使用。 第6圖顯示用以測試一 1C 602之一測試系統600之 例示圖,而1C 602具有内建之自我測試BIST邏輯6〇4A、 604B、604C(為簡化說明測試流程之便,僅顯示為多重功 能區塊)。舉例而言,BIST邏輯可被内嵌至矽或其他1(:半 導體材料中,或暫時载入至IC之一可編程邏輯部。在頁 •面讀取動作期間BIST邏輯隱控制MLC記憶體陣列讓 與字線電壓,其由—測試器_提供或於晶片上提供。數 個頁面之MLC P車列606係以這些可允許之字線電屋 估’如上所述。 WST邏輯604Β儲存這些讀取結果,計算三個總和 碼亚比^較從這些讀取動作獲得之這些總和碼與特定(可 允許)之最小讀取範圍。可選擇的BIST邏輯6〇牝提供通 φ 匕/失敗貝料至測试益608。或者,這些總和碼數值係被提 供至㈣mu,其比較這些總和魏值與㈣總和碼數 值i並關於此些頁面或此IC用之最小讀取電壓範圍。 於二特^實施例中’如果在MLC陣列中之任何頁面無 去7G成任何最小讀取範圍規格,則BIST邏輯⑼扣提供一 失敗數值。 相較於基於第2A與2B圖所討論之習知技術技術,本 發明之實施例具有數個優點。第_,並不需要的尋找每個 位準之廷些字線電壓邊界,此乃因為此讀取範圍係直接地 25 IW2914 200820256 從這些總和碼決定。第二,並不需要的儲存關於整體晶 (整個記憶辦列)之此位元資訊’而只有儲存此期望她和 碼’以與已讀總和碼作比較。第三’登錄資料是總和:元 比車父資訊’而非整體晶片仇元資訊。第四,並不需要對著 三個不同的基準值重複此讀取動作三次,此乃因為供給所 有三個基準位準用之平行讀取係被執行以產生總和碼。本 發明之實施例提供以較少測試次數達成vt容限之可靠的 測試,而某些實施例包含BIST。 說明於此之實施例係基於MLC技術而應用至多重位 元單元。亦可基於SLC技術,來將此技術應用至單—的位 早 。 、綜上所述,雖然本發明已以一較佳實施例揭露如上, =並非用⑽定本發明。本㈣所屬技術領域中具有通 ,在减離本發明之精神和範_,#可作各種 =與潤飾。因此,本發明之保 當, # : The eyes match each other. Marked as "REF2 error on this chart? The π::1 歹丄9 series shows no comparison information about the sum of the second reference levels, and the level of _, _ _ is wrong for that word line voltage. Alternatively, starting from the lowest word line voltage (in this example, 4·5 ν), a page read operation is performed, and the sum code for all three levels is compared with the expected code. For this low word line voltage, a comparison of all three sum codes will indicate a fault, reminiscent of REF1, REF2 and REF3 should be adjusted (REn may be too high) or the word line voltage will leave the appropriate tolerance. When the word line voltage is increased (in this case, it reaches 4·8V), SUM1 will become correct (REFi error = 〇 (pseudo)), because the correct one is programmed to be the absolute sum of LI, L2 and L3? Take the number of cells, which means that the level L2 does not match /, the expected number represented by SUM1, and "〇 indicates that these numbers match each other. This chart is marked with „REF3 error? The third column of „ shows the comparison of the sum code of the third reference level, which indicates that the word of Peng is wrong with the word line voltage. Or, "1, the third column indicates the unit of reading The number 'it indicates that the level L3 does not match the expected number represented by s brain, leak and SUM1, and "〇" indicates that these numbers match each other. 23 FW2914 200820256 The total number of MLCs will be read with this Vt. However, there is no difference between the number of cells programmed at different levels (see Figure 3, SUM1). SUM2 and SUM3 will be wrong (Rej?2 error=1, REF3 error=1) ‘Assume that there are certain MLCs located on L2 and L3 on this page. When the word line > 1 continues to increase to the maximum allowable value for a page read operation, the result displayed on the truth table 500 can be obtained. Numerical values and conditions are for illustrative purposes only and are provided for purposes of illustration and discussion only. The range of read voltages between each level can be determined from the data showing the chart for Figure 5C for each reference on this page. It can be seen from Fig. 5C that the reading range for the word line voltage is from 4.8V to 5.2V for REF1, which means that there is 400 mV between L0 and L1. Similarly, the read range for word line voltage versus REF2 is from 5.0 V to 5.4 V, which means 400 mV between L1 and L2. The reading range of the word line voltage is from 5.3 V to 5.6 V for REF3, which means that 3 〇〇 my is between L2 and L3. This processing is repeated for other pages of MLC memory arrays ("full wafer readout)) and obtains a minimum read voltage range. BIST can be implemented by providing a minimum read range value to a tester. = 2 The BIST logic on the wafer automatically performs a full-wafer read operation', while the word line voltage is stepped across its range. μ or test H stores the read wire, and calculates these totals (4) = ίΓτΐί See section 6. Figure) and allowable (limb) minimum read mode 7 BIST logic can optionally pass the pass of the IC chip, the individual page pass/fail information can be set to set the MLC book array to "be The lock surface is lost ^ ' 夭 页面 页面 而 而 2008 2008 2008 2008 2008 2008 2008 2008 2008 2008 2008 2008 2008 2008 2008 2008 2008 2008 2008 2008 2008 2008 2008 2008 2008 2008 2008 2008 2008 2008 2008 2008 2008 2008 2008 2008 2008 2008 2008 2008 2008 2008 2008 2008 2008 While this reduces the total amount of memory available to the consumer on the IC, it allows for the use of one or more pages 1C that cannot pass the minimum read range specification. Figure 6 shows an illustration of a test system 600 for testing a 1C 602, while the 1C 602 has built-in self-test BIST logic 6〇4A, 604B, 604C (to simplify the description of the test flow, only shown as multiple Function block). For example, BIST logic can be embedded in 矽 or other 1 (: semiconductor material, or temporarily loaded into one of the programmable logic sections of the IC. BIST logic implicitly controls the MLC memory array during page-to-face read operations The word line voltage is supplied by the tester or provided on the wafer. The number of pages of the MLC P train 606 is estimated by these allowable word lines. As described above, the WST logic 604 stores these reads. Taking the result, calculate the sum of the three sum codes. The sum of the sum codes obtained from these read actions and the specific (allowable) minimum read range. The selectable BIST logic provides the pass φ 匕/failed bucks. To test benefit 608. Alternatively, these sum code values are provided to (iv) mu, which compares these sum and (4) sum code values i and the minimum read voltage range for such pages or this IC. In the embodiment, the BIST logic (9) provides a failure value if any of the pages in the MLC array are not 7G into any minimum read range specification. Compared to the prior art techniques discussed based on Figures 2A and 2B, Embodiment of the present invention There are several advantages. The _, does not need to find the voltage boundary of each word line of each level, because this reading range is directly determined from these sum codes. I secondly, it is not needed. Store this bit information about the overall crystal (the entire memory list) and only store the expected her and the code ' to compare with the read total code. The third 'login data is the sum: the yuan is the parent information' instead of The overall chip does not need to repeat this read operation three times for three different reference values, because the parallel read system for all three reference levels is executed to generate the sum code. Embodiments of the invention provide reliable testing that achieves vt tolerance with fewer test times, while some embodiments include BIST. The embodiments described herein are applied to multiple bit cells based on MLC technology. Also based on SLC technology In order to apply this technique to a single bit early, in summary, although the present invention has been disclosed above in a preferred embodiment, = is not defined by (10). The present invention has In the spirit and scope of the present invention, # can be used for various = and retouching. Therefore, the protection of the present invention
專利範圍所界定者為準。 〈甲。月 26 'W2914 200820256, 【圖式簡單說明】 第1圖顯示例示的MLC產品之多重閾值位準之Vth分 佈。 第2A圖顯示兩個不同的頁面之MLC記憶體產品之臨 限電壓單元分佈。 第2B圖顯示決定mlc式記憶體之讀取電壓範圍之習 知方法。 第3圖顯示用來說明總和碼產生之圖表。 第4圖係為依據本發明之一實施例之mlc式記憶體陣 列之操作方法之流程圖。 第5A圖係為依據一實施例之感測結構之例子。 第5B圖顯示依據一實施例之例示總和碼計算區塊 520 〇 第5C圖顯示關於在一 iC中之一頁面之Μ1χ記憶體之 總和碼比較資訊。 第6圖顯示用以測試具有B丨s τ邏輯之丨c之測試系統 之例示圖。 27 200820256rw2914 • 【主要元件符號說明】 LO、LI、L2、L3 :曲線/位準 100 :圖 102 :第一 Vt容限 104 :第二Vt容限 106 :第三Vt容限 200 :方法 204、206、208、210 :曲線 ⑩ 212、214、216、218 :分佈 220 、 222 、 224 、 226 、 230 、 234 、 238 、 240 、 242 : 步驟 228、232 :迴圈 411 :操作方法 412、414、416、418、420、424、428 :步驟 500 :感測結構 502:MLC 單元 • 504A :負載電路 504B :箝位電路 512:資料計算區塊 522 :第一比較器儲存體 524 :第二比較器儲存體 526 :第三比較器儲存體 528:第一加法器 530:暫存器 28 200820256rW2914 ‘ 534 :第二加法器 536 :第二16位元暫存器 538 :第三加法器 540 ··第三16位元暫存器 6 0 0 :測試糸統The scope defined by the patent scope shall prevail. <A. Month 26 'W2914 200820256, [Simplified Schematic] Figure 1 shows the Vth distribution of the multiple threshold levels of the illustrated MLC products. Figure 2A shows the distribution of the threshold voltage cells of the MLC memory products for two different pages. Figure 2B shows a conventional method of determining the read voltage range of the mlc type memory. Figure 3 shows a graph used to illustrate the generation of the sum code. Figure 4 is a flow chart showing the operation of the mlc-type memory array in accordance with an embodiment of the present invention. Figure 5A is an example of a sensing structure in accordance with an embodiment. Figure 5B shows an exemplary sum code calculation block 520 according to an embodiment. Figure 5C shows the sum code comparison information for the memory of one page in an iC. Figure 6 shows an illustration of a test system for testing 丨c with B丨s τ logic. 27 200820256rw2914 • [Main component symbol description] LO, LI, L2, L3: curve/level 100: Fig. 102: first Vt tolerance 104: second Vt tolerance 106: third Vt tolerance 200: method 204, 206, 208, 210: curve 10 212, 214, 216, 218: distribution 220, 222, 224, 226, 230, 234, 238, 240, 242: steps 228, 232: loop 411: method of operation 412, 414, 416, 418, 420, 424, 428: Step 500: Sensing Structure 502: MLC Unit • 504A: Load Circuit 504B: Clamp Circuit 512: Data Calculation Block 522: First Comparator Bank 524: Second Comparator Storage 526: third comparator storage 528: first adder 530: register 28 200820256rW2914 '534: second adder 536: second 16-bit register 538: third adder 540 · · Three 16-bit scratchpad 6 0 0: test system
602 : 1C 604A、604B、604C :自我測試 BIST 邏輯 606 : MLC記憶體陣列 / 29602 : 1C 604A, 604B, 604C : Self Test BIST Logic 606 : MLC Memory Array / 29
Claims (1)
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
TW95139396A TWI308335B (en) | 2006-10-25 | 2006-10-25 | Parallel threshold voltage margin search for mlc memory application |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
TW95139396A TWI308335B (en) | 2006-10-25 | 2006-10-25 | Parallel threshold voltage margin search for mlc memory application |
Publications (2)
Publication Number | Publication Date |
---|---|
TW200820256A true TW200820256A (en) | 2008-05-01 |
TWI308335B TWI308335B (en) | 2009-04-01 |
Family
ID=44770105
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
TW95139396A TWI308335B (en) | 2006-10-25 | 2006-10-25 | Parallel threshold voltage margin search for mlc memory application |
Country Status (1)
Country | Link |
---|---|
TW (1) | TWI308335B (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
TWI595490B (en) * | 2012-09-24 | 2017-08-11 | 東芝股份有限公司 | Memory device |
-
2006
- 2006-10-25 TW TW95139396A patent/TWI308335B/en active
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
TWI595490B (en) * | 2012-09-24 | 2017-08-11 | 東芝股份有限公司 | Memory device |
Also Published As
Publication number | Publication date |
---|---|
TWI308335B (en) | 2009-04-01 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US7580302B2 (en) | Parallel threshold voltage margin search for MLC memory application | |
KR101903932B1 (en) | Verify operations with different sense node voltages in a memory device | |
EP2332146B1 (en) | Data state-based temperature compensation during sensing in non-volatile memory | |
US7457178B2 (en) | Trimming of analog voltages in flash memory devices | |
US7254071B2 (en) | Flash memory devices with trimmed analog voltages | |
US8887011B2 (en) | Erased page confirmation in multilevel memory | |
KR101502104B1 (en) | Reducing the impact of interference during programming | |
US7558149B2 (en) | Method and apparatus to control sensing time for nonvolatile memory | |
TW201007744A (en) | Non-volatile memory with adaptive setting of state voltage levels and the method therefor | |
CN103219041A (en) | Adaptive programming and erasure methods for analog memory cells | |
US8737125B2 (en) | Aggregating data latches for program level determination | |
US8792281B2 (en) | Read threshold estimation in analog memory cells using simultaneous multi-voltage sense | |
CN103765519A (en) | Intelligent shifting of read pass voltages for non-volatile storage | |
TW200849263A (en) | Alternating read mode | |
CN113450856B (en) | Memory device and method of operating the same | |
WO2020205016A1 (en) | Multi-state programming in memory device with loop-dependent bit line voltage during verify | |
CN105009218A (en) | Programming method to tighten threshold voltage width with avoiding program disturb | |
US10366729B2 (en) | Sense circuit with two-step clock signal for consecutive sensing | |
KR20160140506A (en) | A non-volatile memory comprising a current sensing checking block and the corresponding program verify method | |
CN114730597A (en) | Hybrid erase mode for high data retention in memory devices | |
CN114730599A (en) | NAND string precharge by hole injection through substrate during programming | |
CN109119114B (en) | Sensing circuit with two sensing nodes for cascaded sensing | |
CN104067348B (en) | Programming and erasing scheme for analog memory unit | |
TW200820256A (en) | Parallel threshold voltage margin search for MLC memory application | |
US10964402B1 (en) | Reprogramming memory cells to tighten threshold voltage distributions and improve data retention |