Error-Correcting Codes Laboratory
Learn iterative decoding by building it: belief propagation, LDPC, Turbo and concatenated codes, EXIT charts and density evolution, through computer projects.
Iterative decoding transformed error-correcting codes using surprisingly elementary ideas. This project-based lab starts with channels and capacity, then builds Belief Propagation and applies it to LDPC, Turbo and serially concatenated codes. Students write simulations in a language of their choice, some running for hours or days. Tests are absent, and a final project often involves original research.
What You Will Achieve
- Formulate decoding problems for channels such as Gaussian and erasure
- Implement Belief Propagation decoding
- Simulate LDPC, Turbo and serially concatenated codes
- Use EXIT charts and Density Evolution to predict performance
Program Overview
Channels and Probabilities
The opening sessions define the problem.
- Channels and capacity
- Repetition code and prior probability
- Gaussian channels and parity product code
Belief Propagation
The central algorithm is then developed and applied.
- The Markov property
- Linear codes
- LDPC codes
- Belief propagation on trees
Turbo Codes and Beyond
Other code families follow, then student work takes over.
- Convolutional codes
- Turbo codes
- Serial concatenated codes
- Student presentations
Course Completion
Job Outlook
This background suits work in communications engineering, data storage, wireless systems and information theory research.
Good for: Mathematics and engineering students who like probability and would sooner learn by writing simulations.
Keep in mind: From spring 2004, so later code families are missing. The Fan textbook is separate. Notes and project briefs are published, but the graded write-ups and feedback that drove the class are not.
Bottom line: A distinctive, hands-on route into modern coding theory, best for readers willing to write the code themselves.
Frequently Asked Questions
Which languages can I use?
Any computing environment works, though the instructor recommends C, Java or MATLAB and warns that some experiments run for hours or days.
Are there tests or exams?
None at all. All the work is project-based: weekly short labs, two involved lab projects with write-ups, and one final project.
What did final projects look like?
Samples cover turbo equalization, irregular repeat accumulate codes, a high-performance serially concatenated code and improved LDPC codes over the erasure channel.



There are no reviews yet.