Engineering services

FPGA Design Services

Contract RTL design and custom IP core development on AMD/Xilinx platforms, by the engineer who writes the cores.

Most FPGA contract work stops at integrating vendor IP. The interface cores on this site were written from scratch: CoaXPress 2.1 Host and Device, SLVS-EC RX, and NVMe. That means when a link will not train, or a transceiver will not lock, or a datasheet and a sensor disagree, the debugging happens at the protocol level rather than against a black box.

Engagements run from a bounded core integration through full design ownership: architecture, RTL, constraints, timing closure, simulation, and hardware bring-up on real silicon.

Custom IP and RTL

  • Protocol cores written from specification, not adapted from reference designs
  • CoaXPress 2.1 Host and Device, SLVS-EC RX, NVMe, 12G-SDI, Ethernet
  • AXI4 / AXI4-Stream / AXI4-Lite interconnect and register maps
  • GenICam-compliant camera control and XML description
  • IP packaging and encryption for redistribution to your customers

High-speed serial

  • GTH, GTY, and GTP transceiver configuration and DRP runtime reconfiguration
  • Multi-lane alignment, link training, and rate negotiation
  • Clock domain crossing and elastic buffering across recovered clocks
  • Equalisation and eye-margin debug against real cable and backplane loss

Platforms and closure

  • Kria K26 and K24 SOM, Zynq UltraScale+ MPSoC, Artix and Kintex
  • Vivado from block design through timing closure and bitstream
  • Constraint authoring: timing, placement, and I/O planning
  • Simulation and verification benches, ILA-based hardware debug
Representative work

Built, shipped, and debugged on real hardware

CoaXPress 2.1 Host and Device

A full CXP-12 implementation covering both link directions, with GenICam control, packet layer, and cable discovery. Shipping as a licensed core.

SLVS-EC RX

Sony SLVS-EC receiver handling multi-lane alignment and sensor-specific framing, targeting sensors where no vendor IP exists.

NVMe Host

An NVMe host controller for direct-to-storage capture, so high-rate sensor data lands on an SSD without a CPU in the path.

How engagements run

Scope that fits the problem

  • Fixed-scope core integration and bring-up on your hardware
  • Custom protocol core developed to your specification
  • Design rescue: timing closure, link stability, or interface debug on an existing project
  • Ongoing retained design capacity

Related work on this site

These are shipping products from the same bench. They are the clearest evidence of what the service delivers.

Discuss a project