How PMC FPGA Boards Excel at Image Processing Application Note
Acromag has engaged in a number of image processing applications based upon implementations of Camera Link running on a Virtex-5 FPGA module. In several of these instances, LVDS signals are used to collect the image data through front and/or rear I/O connections on the PMC FPGA module. Other applications use a custom interface developed for selected imaging devices to capture data through the PMC module’s front and rear I/O interfaces. To select the appropriate FPGA module for image processing applications engineers must consider a number of pertinent criteria.
Finding an appropriate Commercial-Off-The-Shelf (COTS) FPGA module for image processing depends on a number of key factors: I/O interface, frame speed, frame size, on-board processing requirements, and host communication speed/volume. The Xilinx Virtex-5 FPGA devices found on Acromag’s latest COTS PMC Modules offer blazing 550MHz processing speed and a variety of high-performance memory options that are well-suited to image processing functions.
Table 1: Acromag PMC modules with re-configurable Virtex-5 FPGAs
|Model Number||Logic Cells||Total Block RAM(Kbits)||DSP48E Slices||DDR2 SDRAM Capacity
|Dual-Port SRAM Capacity
Note 1: Available increments are 32M, 64M, and 128M (contact factory)
Note 2: Available increments are 256K, 512K, and 1024K (contact factory)
Video image capture is most frequently performed using direct access to the LVDS data lines available through Acromag’s AXM-D04 front mezzanine card or through the rear I/O P4 PMC connector. Alternatively, some applications use Camera Link interface chips on either a custom rear transition module or a custom AXM mezzanine module. The objective of these interfaces is to capture one whole raw image frame into storage on the PMC Module. There are three modes of Camera Link (base, medium, and full) which differ in the number of bits utilized in the raw image frame capture and provide for standardized hardware connections as defined in the Camera Link specification. Upon successful acquisition of a raw image frame, several processing objectives are possible: event detection, storage, enhancement/storage, or overlay/enhancement/storage.