Document ID
APT-DS-SLVS-RX-0001
Revision
2.1
IP version
3.4.1
Release date
November 2026
Chapter 5: Example Designs and Software

5.6Verification

Release 3.4.1 is verified on the AXI4S and NATIVE outputs with two independent benches. A VHDL testbench with its own SLVS-EC transmitter model runs a matrix over every lane count, pixel depth and payload protection setting, with embedded data lines at the start and the end of the frame, and checks every image line and every embedded data line on both outputs. A UVM 1.2 testbench drives the lanes from a transmitter model and AXI4-Lite from a register model, and a reference model checks every line and every status change, with protocol monitors on the AXI4-Stream rules on every clock.

4,846
matrix runs passing on AXI4S and NATIVE: 1 to 8 lanes, RAW8 to RAW16, CRC and ECC, with embedded data
264
UVM regression runs passing: 22 tests, 3 seeds, 4 builds
0
mismatches against the Reed-Solomon reference decoder over 30,000 vectors
  • Line Length sweeps at 4, 6 and 8 lanes cover every case where a line ends on a partial beat, in both builds.
  • Errors are injected and accounted for: corrupted header copies, payload CRC32 errors, symbol errors in every Reed-Solomon block, on image and embedded data lines alike. Every error is corrected or counted, and the CRC error registers are read back at the end of every run.
  • ECC results are checked bit for bit against a C reference decoder, for both ECC Options, up to two errors beyond the correction limit per block.
  • The output stages are run with stream clocks from 1.1 to 12 ns, sink stalls, overflow and recovery, and stream resets, on the image and the embedded data stream.
  • The AXI4-Lite interface is stress-tested with random backpressure and multiple outstanding requests.
  • The receiver has run in hardware with the Sony IMX530 on Zynq UltraScale+ MPSoC.