processing and presentation tools for memory array analysis

Analysis You Never Dreamt Before

BarnieMAT is a unique software tool providing all necessary functionality to process memory array measurement data in a single, fully integrated environment. Using BarnieMAT, engineers can obtain fast and reliable results with minimal effort and can focus on their analysis task rather than spending time with manual processing or by developing custom solutions. BarnieMAT users discover the possibility of new analysis methods they have never even dreamt before.Latest generation testers are able to create huge quantity of data, which can be a digital readout or an analog measurement on each of eventually billions of cells. The processing of the measurement data and the extraction of summary information is essential in the memory technology development, device characterization and failure analysis, to understand the physical phenomena and to reach the right conclusion.

BarnieMAT exploits the features of the BARNIE environment.

• Graphical framework shared with the data collection and correlation tools.
• Source data can be the BARNIE database, STDF or XML files, output of the commonly used testers or custom format. Data can be also received from testers on-line via TCP-IP.
• The results of the array analysis can be passed directly to further correlation analysis.
• Complex analysis flows can be defined using a powerful scripting language.
• Standard processing functions are implemented as “stored procedures” and contained by Windows DLLs. User can add his/her own functions by implementing the code in MFC C++ language, supported by an extensive library.

BarnieMAT:

• Processes bitmaps (analog and digital), cell lists, distributions, galaxies, tables.
• Embeds fast optimized algorithms for bitmap processing.
• Deals with large size measurement data (64 Gbits of cells, 16-bit cell data, multilevel cells).

Extensive Set of Processing Functions

A wide range of processing functions is available for extracting key information from the measurement data.

• Conversion functions transform an object into a second (map to distribution, analog to digital map, cell list identification …).
• Operation functions execute math functions on the object items (differential map, map scaling …).
• Overlays merge measurement data from various areas of the same device (block fail density …).
• Compressions carry out data reduction (fail number per page …).
• Pattern recognition functions support the identification of specific topologic patterns in a map (ex.: neighbor bit failures).
• Trace-back enable to go back to the source data starting from the result of the processing (ex.: once a distribution is created from a map and the user is interested in the topologic position of the outlier cells, it is possible to show the original map highlighting the cells belonging to the tail).

Innovative Filtering Engine

Processing functions are supported by a powerful filtering engine, to define the cells to consider and how to group them.Filter can be a map or a cell list (the data of the map or list defines the grouping rule, ex.: a fail map can be used to consider only failing cells when processing a map).Filters can also be described with the intuitive Filter Definition Language (FDL). With FDL it is possible to define the electrical or topologic position of the cell and/or measured values.

• “SPLIT=(WL%64)” è the cells will be split in 64 groups based on their wordline position.
•  “BL%2 = 0 AND DATA>5.5” è cells having a high threshold value on even bitlines.

Best in Class Viewers

Results of the measurements and processing can be consulted by dedicated presentation tools.

• Topologic view of the content of the array in 2D and in 3D, with color coded values.
• Parallel zoom and pan of two or more maps.
• Bookmarking of critical cells on a map.
• Consultation of 2D and 3D (galaxy) distributions with on-the-fly statistical data calculus.
• EXCEL-like histogram view of tables.

Map Viewer

The Map Viewer helps in the consultation of the measurement data, shown in true topologic format.

• The cell value (which can be the result of a digital or analog measurement) is coded by color using a linear or logarithmic scale.
• The zoom option allows to see each single detail. At macro level, when more than once cell is represented by a single pixel, the user can select the merge option (min, max, average) to identify easily critical cells.
• A navigator shows always which part of the device is in the view area.
• Cells can be bookmarked by a second measurement data (like a fail list) or by the same data (selecting the criteria like high Vt cells). There is an easy navigation between the bookmarked cells.
• Map area can be displayed also in 3D form making easier visual pattern recognition.

Distribution Viewer

The Distribution Viewer can display the splits of one or more distributions.

• The user can enable/disable the display of each split independently, having a clean view of the interesting curves.
• It is possible to zoom into the critical distribution area.
• The viewer integrates statistical curve analysis (edges, gaps, standard deviation, …).
• With the trace back possibility it is possible to show the physical location of a tail in the map viewer. 

Galaxy Viewer

The Galaxy Viewer is an efficient graphical tool to consult galaxy data. Galaxy is a three-dimensional distribution, ideal to understand the behavior of the cells during disturb and programmability experiments. A cell-by-cell measurement is done before and after disturb or program pulses and the number of cells is shown in function of the two measurement data.

• The 2D view codifies the number of cells in function of the two measurements coded by color on a linear or logarithmic scale.
• The third dimension of the 3D view codes the number of cells in a graphical way.
• Zoom / pan features are available.
• With the trace back possibility it is possible to show the physical location of a tail in the map viewer. 

Table Viewer

Using the Table Viewer, the data can be displayed in table or in graphical format. Many graphical options (histograms, charts) are available for easy consultation.

Local Object Storage

Measurement and processed data is stored in a local Object Registry. One data object can be split in multiple groups based on topology or other classification (ex: wordline based distribution of a single measurement).

• Maps contain a single measurement data for each cell. The data is coded in 8 or 16 bits per cell based on a linear scale. Maps can be ordered electrically or topologically based on the measurement method and conversion functions can translate from one form to the other. Maps can have multiple segment related to multiple measurement areas, saving space in case of non contiguous areas.
• Lists like maps, store measurement data per cell but only for a selected set of cells. Lists can be generated by the tester (fail list) or can be the result of a cell identification function (like the cells having threshold voltage between 2.0 and 2.3 V). Lists might store the expected pattern which allows the reconstruction of a map, avoiding the storage of not meaningful information (exact value and position of critical cells and approximate value of the not critical cells).
• Distributions are results of data reduction, providing the number of cells for measurement ranges. Due to the easy consultation, distributions are the favorite objects for understanding the device behavior (like erase/program window after a program operation). BarnieMAT is able to store multiple curves inside one distribution object which is sometimes essential for understanding the results (separating the distribution of cells which should be programmed or erased, the eventual overlap of the two curves can be clearly seen).
• Galaxies represent 3D distributions, counting the cells in function of two measurement ranges. Galaxies are used often for disturb experiments (seeing which part of the distribution moved where) or for correlations (between sectors, VT-Icell, …). Galaxies, like distributions, can be split by filtering conditions.
• Tables are the most condensed results and are typically the final data of a processing / data reduction sequence. An example can be the Vt shift after a sequence of programming pulses separated by wordline: the bi-dimensional table, indexed by the program pulse count and by the wordline, contains the Vt shift values.

Triggers can be set to object event (like creation, update, …) invoking processing functions once the event has been detected.

Data Flow

BarnieMAT objects can be received on-line, injected directly from the tester equipment via TCP-IP interface, or off-line, loaded from the BARNIE test result database or imported from a file (XML, CSV, Octave, proprietary format, custom importer tools). Importers are available for several memory testers (MOSAID, KALOS).

In case of TCP-IP connection, the display of the measurement data during measurement is a good way to verify the correct execution of long test operations (ex: a device map is displayed block by block updating the picture after each block is measured).

BarnieMAT data can be extracted for post-processing with third party tools. Exportation can be in CSV, Octave, in proprietary format or by using custom exporting tools.

Interactive Processing

Every processing function can be selected from a list and invoked from a dialog box. Easy data input is supported by the possibility to select between last inputs, by combo boxes offering the selection of the related object types and by an online help system describing the function of the operations and of the description of the parameters.

Analysis Flow Automation

Analysis flows can be defined by a scripting language.

The language combines the need of easy definition (no need of programming knowledge) and the flexibility necessary for the analysis flows (variables, flow control, subroutines, script libraries). The script implementation is supported by a powerful built-in editor providing a set of user friendly features:

• context sensitive help
• word extension
• dialog box for command parameters
• syntax checker
• syntax coloring.

Graphical Framework

All viewer, editor and control tools are accessible from a single framework. The framework allows the display of any number of objects. The view of objects can be linked (available from v3) for parallel zooming of maps …

The layout of the screen can be defined by the user. Multiple windows can be arranged on the screen for correlating measurements or views or a single measurement can be consulted in full-screen mode to consult details.

Device Knowledge

BarnieMAT needs to know about the electrical and topologic characteristics of the device under analysis. This information is supplied by an external Windows DLL exporting a specific interface (BTAPI). The available tools and libraries support the implementation of these DLLs which requires typically a few hours of effort.

Open Architecture for Custom Processing