Katana: customizable, embedded, flexible and optimal ecosystem





Katana started as a project for high performance computing (HPC) on the ZCU104 board with a Zynq Ultrascale+ FPGA and evolved as a flexible and modular architecture for accelerating any algorithm on FPGA.

The basic elements of a Katana accelerator are:

  • A main (or interfacing) processor, like the PS in Zynq or Microblaze in Artix or Kintex. This processor acts as a system controller and does no data processing.
  • A number of processing blocks, implemented mostly in HLS that perform specific functions like FFT, DWT or Convolution.
  • A number of DMA engines that move data from external RAM to the hardware blocks.


  • Modular: Add the operational blocks and functions you need where you need them
  • Flexible: Each block can be individually tuned to optimize any parameter
  • Customizable: Katana can integrate any custom operational block using AXI-Stream
  • Portable: Katana is usable in all Xilinx FPGAs
  • Embedded: As small as a credit card. And not much bigger most times.


Although Katana can be used for any computational process done in an FPGA, its outstanding characteristic of flexibility and modularity makes it ideal for applications where optimization is critical, either for the lowest cost or power or the maximum data throughput. Just a few are:

  • Radar Signal Processing
  • Real-Time pattern recognition
  • Computer vision
  • Image segmentation, object recognition
  • Convolutional Neural Networks
  • Machine Learning
  • Financial forecasting


I am Joan Abelaira, born in Barcelona (Catalonia) where I grew up with an interest for Science and Technology. I later graduated by UPC (Catalonia Polytechnic University) and did my Final Thesis in the University of Birmingham as an Erasmus student.

I worked in a number of companies as well as running my own consultancy company in Barcelona until 2012 where I moved to the UK. I have been interested an involved in almost all fields of electronics but I have always had a preference for FPGA.

Nowadays I am focused on FPGA applications and especially those related to Machine Learning.


Thanks a lot for your interest in my work. If you want to reach out for me, please use the email address at the footer.