Development and verification of a high data rate and low latency communication system using Xilinx GTY transceivers in Virtex UltraScale+ FPGA. Designed for inter-FPGA transmission with close to maximum line rates and minimum possible latency. Verified in hardware using Xilinx VCU128 Evaluation Kit.
Languages | VHDL, Tcl |
Tools | Vivado |
Target | Virtex UltraScale+ |
Time frame | 10.2022 – 04.2023 |
Commercial | ✔ |
Project Owner | FastLogic |
Project goals
- Design a link layer point-to-point communication system utilizing GTY transceivers
- Self-synchronous communication – the data stream contains both the data and the clock
- Communication channel supports between 1 and 4 individual line (up to full Quad)
- Each line operating at 25 Gbps line rate
- Minimize data latency as much as possible
- Automatic channel initialization
- Auxiliary low-speed UART connection for initialization
- Low Bit Error Rate
Single channel connection
Channel initialization stages
Block Synchronization:
- Performed for each line individually
- Changing data alignment until expected value detected
- Continuous monitoring
Channel Bonding:
- Different lines’ latency compensation
- Uses custom elastic buffers and dedicated controller
- Lowest possible delay
Dynamic Scrambling:
- Additional transmitted data encoding
- The same data is scrambled differently each time
GTY transceiver
Achieved results
- Hardware verified using Xilinx VCU128 Evaluation Kit
- Automatic hardware verification using custom Tcl scripts
- 0 bit errors detected during 36 000 restarts
- Approximately 80% lower latency compared to the Aurora 64/66B Protocol
My responsibilities
- General architecture design and implementation
- IP cores generation and configuration
- Simulation verification
- Design optimization
- Debugging
- Debug functionalities implementation
- Hardware verification
- Refactoring/Clean-up (warnings removal etc.)
- Automatization scripts implementation in Tcl
- Concept, implementation and verification documentation in LaTeX