Zeph -
I'm only speaking for myself, but I'd implement that a bit differently. The OP has indicated that the design is a bit resource-constrained (what project isn't), so I'd ditch the shift register. Instead, I'd configure the output side of the dual-port RAM for a 1-bit data width, and follow that with a multiplexer that selects between a '1', a '0', and the output from the SRAM.
Nice find on that Ethernet board. One of the reasons that I dropped my earlier project was that the 10 Mb/s Ethernet PHY chip that I was using (LXT905) was pretty much obsolete even at the time. That board seems to be fairly reasonably priced, hopefully it will continue to be available for some time to come.
What language are you designing with? VHDL or Verilog or ??
Thanks, Phil - that's exactly the kind of "alternative design" that I was hoping to learn from. I didn't know that the RAM could be programmed for one bit output - that's major flexibility! Given that, one could get by with a 3 bit shift register in the approach I was suggesting (rather than 24 bit), but if implementing and driving a three input multiplexor takes fewer resources that would be better still. Looking forward to hearing from OP.
I'm used to "software is far more flexible for the complicated and changeable things", and while I know at the general level that FPGA's challenge that paradigm, I'm not very well aware of how far they change those tradeoffs.
Zeph wrote:Thanks, Phil - that's exactly the kind of "alternative design" that I was hoping to learn from. I didn't know that the RAM could be programmed for one bit output - that's major flexibility! Given that, one could get by with a 3 bit shift register in the approach I was suggesting (rather than 24 bit), but if implementing and driving a three input multiplexor takes fewer resources that would be better still. Looking forward to hearing from OP.I'm used to "software is far more flexible for the complicated and changeable things", and while I know at the general level that FPGA's challenge that paradigm, I'm not very well aware of how far they change those tradeoffs.
Each family of parts has different characteristics. I'm more familiar with the Xilinx family than the Altera family, but the data sheet for that Altera part says that the RAM ports can be configured for different widths.
David_AVD wrote:Can you explain what is meant by this?
I am using vixen 2.5. If you have sixteen channels and output them to Ethernet, the packet will contain the header plus the 16 bytes of channel data. So for output to the universe, I only have 16 bytes to send. If there are 75 channels then the packet sends the header plus the 75 bytes for the channels and so forth.
Phil wrote:What language are you designing with? VHDL or Verilog or ??
I am using VHDL. I will consider using the RAM instead of the shift registers. However, there are header bytes for DMX and Renard, so I need some way of sending these bits.
Zephl wrote:Examples ??
I have used both Xilinx and Altera software. I would recommend the Altera, specifically v9.1. You will have to look for it on their site. All of the software has free versions. The reason I like this version is for the simple simulator for code. Also, Altera has a library of the old 74xxx parts and you can just wire them up in a schematic for a design. This graphical method is a good way to start learning. Altera also has a wealth of online videos for training. If you are used to software, this is a shift in thinking and will take some time. Example code for a counter in VHDL is shown below:
cnt_proc : process(clk,ss_cnt_sclr,ss_cnt_en)
begin
if (clk'event AND clk = '1') then
--
if (ss_cnt_sclr = '1') then ss_cnt <= "00000"; --sclr
elsif (ss_cnt_en = '1') then ss_cnt <= ss_cnt + 1; --count enable
else ss_cnt <= ss_cnt; --else hold count value
end if;
--
end if;
end process cnt_proc;
