02-01-2023, 05:49 PM
(This post was last modified: 02-01-2023, 05:54 PM by ppppenguin.)
I implemented a standards converter in a Xilinx XC2S200 which has just over 5000 logic cells. I think I was being fairly clever about the interpolation multiplier, multiplexing all 4 lines through it at 4x clock. The design used SDRAM framestores. All interpolation was after the framestores since I had enough memory bandwidth to make it look like a multiport RAM. If you're using block RAM to do internal linestores then you're likely to run out of that resource before anything else.
The mapping report said:
Logic Distribution:
Number of occupied Slices: 1,072 out of 2,352 45%
Number of Slices containing only related logic: 1,072 out of 1,072 100%
Number of Slices containing unrelated logic: 0 out of 1,072 0%
*See NOTES below for an explanation of the effects of unrelated logic
Total Number of 4 input LUTs: 1,198 out of 4,704 25%
Number used as logic: 1,041
Number used as a route-thru: 145
Number used as Shift registers: 12
Number of bonded IOBs: 202 out of 284 71%
IOB Flip Flops: 149
IOB Latches: 8
Number of Block RAMs: 8 out of 14 57%
Number of GCLKs: 3 out of 4 75%
Number of GCLKIOBs: 3 out of 4 75%
Number of DLLs: 2 out of 4 50%
I think a synthesised 8x8 multiplier uses about 64 cells. Needs a fair bit of internal pipelining to work at video speeds.
I've attached all the VHD files from my design but I think I sent you these long ago. It will take a fair but of dredging my own memory banks to remember much more.
This was the design I later adapted to do 405 to 625 conversion. The hardware is a board I designed for a professional cliient some years ago. I happen to have a few around.
PS: Don't try to design a multiplier yourself. I've done this on older FPGAs before there were tools like Coregen to do the job for you. It's a lot of work to design and debug.
PPS: I've totally lost touch with the FPGA market. Haven't done any real design work for years. Xilinx and Altera have both been bought by huge companies (Intel and AMD) who I suspect aren't interested in small devices anymore. So you're loking at 2nd tier makers like Lattice.
The mapping report said:
Logic Distribution:
Number of occupied Slices: 1,072 out of 2,352 45%
Number of Slices containing only related logic: 1,072 out of 1,072 100%
Number of Slices containing unrelated logic: 0 out of 1,072 0%
*See NOTES below for an explanation of the effects of unrelated logic
Total Number of 4 input LUTs: 1,198 out of 4,704 25%
Number used as logic: 1,041
Number used as a route-thru: 145
Number used as Shift registers: 12
Number of bonded IOBs: 202 out of 284 71%
IOB Flip Flops: 149
IOB Latches: 8
Number of Block RAMs: 8 out of 14 57%
Number of GCLKs: 3 out of 4 75%
Number of GCLKIOBs: 3 out of 4 75%
Number of DLLs: 2 out of 4 50%
I think a synthesised 8x8 multiplier uses about 64 cells. Needs a fair bit of internal pipelining to work at video speeds.
I've attached all the VHD files from my design but I think I sent you these long ago. It will take a fair but of dredging my own memory banks to remember much more.
This was the design I later adapted to do 405 to 625 conversion. The hardware is a board I designed for a professional cliient some years ago. I happen to have a few around.
PS: Don't try to design a multiplier yourself. I've done this on older FPGAs before there were tools like Coregen to do the job for you. It's a lot of work to design and debug.
PPS: I've totally lost touch with the FPGA market. Haven't done any real design work for years. Xilinx and Altera have both been bought by huge companies (Intel and AMD) who I suspect aren't interested in small devices anymore. So you're loking at 2nd tier makers like Lattice.
www.borinsky.co.uk Jeffrey Borinsky www.becg.tv







