Software for deciphering proprietary binary data files

Scientific instruments produce proprietary binary data containing primary data and metadata. MARBLE helps domain scientists decipher these files and generates Python scripts that translate the instrument measurements.

MARBLE is open-source and free software. The source code is available in the repository, with documentation hosted on GitHub Pages.

Contributors

• Steffen Brinckmann (IEK-2, FZJ) [Principal investigator]
• Volker Hofmann (IAS-9 and HMC, FZJ)
• Fiona D'Mello (IAS-9 and HMC, FZJ)

Introduction into proprietary binary data-files and MARBLE

In proprietary binary files, data can be grouped sequentially in sections, like sequential chapters in a book, and sections can have very different lengths. One section may contain only the operator's name, while another contains thousands of temperature values. These files are called binary because they are not human-readable, and proprietary because the instrument vendor designed them for a particular company or instrument. MARBLE helps scientists decipher these files.

MARBLE reads the proprietary binary files and - with the help of the scientist - outputs a python converter. This python converter can then be used to translate all proprietary binary files from this instrument into an hdf5-file format, which can be easily read by any computer language. The python converter also acts as verification tool: if a binary file A can be converted by this specific converter, then this file A comes from this instrument. This verification ability is helpful in finding files from a particular instrument.

In MARBLE, data in proprietary binary files is grouped into classes:

• Metadata is data that describes the experiment. Examples are the name of the operator, the instrument vendor, the start time of the experiment. This metadata is commonly stored in key-value-pairs. For instance, "operator" is the key and "Peter Smith" is the value as both form an inseparable pair. Generally, the first parts of proprietary binary files contain lots of metadata.
• Primary data is the data that the operator wanted to measure and this primary data has a form of a list. Lets say we want to measure the temperature at our house every 1min and store this information; then temperature is a primary data and stored in a long list. Generally, the instrument also saves the time increment after the measurement start and stores these time increments in a separate list, which is also primary data. Primary data can be of two types floating point numbers with normal or high precision.
• Undefined sections are those sections of the file which the scientist and MARBLE have not identified yet. Some of these sections might be important or unimportant. Unimportant sections are those where the programmer at the instrument vendor was lazy and did included garbage or empty space. These might also be linked to specific languages the instrument vendor used for programming.

MARBLE is semi-automatically deciphering files because it has different methods to automatically identify sections in binary files. For some files, the methods work perfectly while for other files they fail. These failures are where the scientist uses his domain knowledge to augment the automatic methods or corrects MARBLE's mistakes:

• The scientist always labels data:
  • specifies the key for a value: if he/she reads a name, the key would be "operator"
  • specifies the terms, units and IRI: links to terminology servers
• Especially primary data has often to be corrected by shifting the starting position or by changing the length. Although these changes feel strange at first, it becomes easier soon after understanding the rule: "different primary data sections generally have the same length".
• The user identifies which sections are important as these should be saved in the output. Additionally, the user defines how certain he/she is about specific sections.
• The user can also remove wrongly identified sections and then use other algorithms to identify those.

How to install and run MARBLE

Install MARBLE with

pip install pymarble

To start graphical user interface (GUI)

marbleGUI

How to use the GUI

1. A dialog opens asking you for a binary file to load.
2. One row will be automatically filled: everything of the file is unknown. Right-click and select "Automatically identify" depending on your data. A lot will be marked in most cases: click the F5 key to only focus on the non-binary data (see also statusbar).
3. Now go through the sections and label them
   • by entering a "key" and "unit" where applicable. You can do that directly in the table columns that are grey: click to enter the key, value, probability, importance.
   • if you double click on the start and length column a form will open which gives you more options
   • If you want that the converter uses certain data, ensure that the "important" checkbox is ticked.
   • For keeping track of your own progress, you can use the "certainty". If you are unsure use red, medium-sure is yellow and very sure is green.
4. Intermediate saving and reloading can be done with the .tags file. Use the F4 and F2 key for convenience.
5. Once you are done, click the save to save a python converter into the directory of the proprietary binary file that you analysed.
6. Go into the directory and use python converter to convert all files from this instrument by executing the command "python <converter.py> <proprietary_binary_file.dat>". You can add the "-v" option to the command to make the converter more verbose during translation. If all the conversions are successful, you deciphered this proprietary binary file successfully.

How to use the command line interface (CLI)

The CLI allows you to learn more about proprietary binary files and MARBLE. If you want to become more advanced, follow the tutorials in this section.

marbleCLI

There are several tutorials for the CLI:

• Tensile machine
• Tensile machine, large file
• Image data
• Data of multiple tests in one file

You can read them and follow those commands. However, you can also just execute them with the argument "m", without the quotation marks. All of these tutorials are in the form of linux scripts, which are used for verification of the code at each development step.