The Transmogrifier-3a Field-Programmable System

Introduction

The Transmogrifier-3 is a next-generation field-programmable system built at the University of Toronto. The goal of this project is provide a Field-programmable system with a usable logic capacity in excess of 3 Million gates, on one board, with video input and output capability. We are using it do develop various applications
in vision and graphics research. For example, we have successfully implemented frame-rate stereo vision with excellent accuracy. This has applications is many computer vision problems and 3-D imaging.

The Board

The TM-3a consists of a single board which contains:

Four Xilinx V2000E FPGAs. These contain

35,000 4-input lookup table/FF logic elements
160 blocks with 4K bits each of dual port SRAM

more than 8Mbytes of RAM.
a video-input interface (which connects to standard video cameras)
a video-output interface.

Here is a picture of the TM-3 board, which uses Virtex V1000s:

Status

TM-3a is operational and has been used in a several projects relating to Stereo Vision, Ray Tracing, and high speed searching of the human genome.
We now have two working TM-3a board

For more details on the TM-3 architecture, chips and use please see the TM-3 Home Page .

The People

The architecture of the board was developed by David Galloway, Marcus van Ierssel, Jonathan Rose , and Paul Chow , with some specification input from personnel at ATI, including Rob Bicevskis and Ron White. The board was built by Marcus van Ierssel, and debugged with the aid of David Galloway. David Galloway wrote the software that supports users.

Acknowledgements

Funding for this project has come from NSERC , Micronet , Xilinx, ATI , and Ricoh .

Support, in the form of equipment contributions from Cypress Semiconductor are gratefully acknowledged.

Research on FPGAs at the University of Toronto.

Return to Jonathan Rose's Page .Computer Group.