[time-nuts] 5 MHZ PIC PPS Divider?
hmurray at megapathdsl.net
Wed Apr 16 14:15:53 EDT 2008
> The ECL part you've selected seems to simplify interfacing with its
> synchronous enables.
I think you are missing the decimal point.
A synchronous enable will have setup/hold times. They will (generally) be
less than the cycle time (1 ns). You aren't going to get that level of
timing accuracy out of a PIC.
You will have to include a (small?) cloud of high speed logic around the
basic counter chip to make it work right.
If you want to use ECL counters, my straw man would be
a free running counter
a register to grab a copy of the counter
logic to load the register on the rising edge of the PPS
and generate a data-ready signal for the PIC
That lets you grab the counter once each second. Subtract the previous value
to get the number of ticks in this second.
Make the counter wide enough so you can figure out the high bits. Or feed
the top bit to a counter/timer in the PIC. Or use it for the PIC's clock.
If you like ECL, here is another approach:
Start with a shift register runing at 1 GHz
Add a holding register that grabs a copy of the shift register every N
The holding register goes into an FPGA that runs at 1/N GHz
You have to make 2 clocks, and you have to make sure that the holding
register meets setup time at the FPGA. That determines how big you have to
The FPGA would look for rising edges. If not, it bumps a counter by N. If
it finds one (maybe by table lookup), it adds N-x to the counter, copies the
counter to a holding register, and reloads the counter to x. Pipeline as
Some FPGAs have high speed serial links. The contain a PLL and a big shift
register and lots of logic to do 8B/10B decoding and recognize sync patterns
and ... There is usually an option to disable that logic. So you could do
everything in the FPGA. Unfortunately, they tend to be the (very) expensive
These are my opinions, not necessarily my employer's. I hate spam.
More information about the time-nuts