[time-nuts] Arnold's counter design
buehl at superlink.net
Fri Jan 13 11:37:30 EST 2006
How are you progressing on your counter design?
Would be best to take this discussion off-list.
Advise your email address;
Tom buehl at superlink.net
At 02:09 PM 10/16/2005, you wrote:
>I am very much interested on your works and results, it sounds to
>be a interesting way, nice project. Unfortunately I do understand
>nothing of these new devices.
>I would like to keep in touch with you, thank you.
ON 16 OCT TOM BUEHL WROTE:
Some more info regarding use of Xilinx FPGA for counting.
I recently completed a design using Xilinx Spartan II FPGA. This FPGA is
only specified for maximum 200 MHz internal clock.
My skills in this area are very limited, and this is my first for FPGA
design. When finished, it counted reliably to above 210 MHz. Input pulses
about 2 nS wide. So the performance of the Xilinx parts are excellent.
This was a pre-set counter: Enter desired count number into registers/
memory, and the counter would output pulse when this count
finished. Counting could be triggered or could automatically begin
again. Therefore the OUTPUT could be every 10 pulses, up to every
999,999,999 pulses, with resolution of 1 pulse.
Although this was not interfaced with time reference, circuit could be
modified to compare with a reference.
I have a plan to do the next one with higher speed Xilinx, and also know
how I will be able to get counting at twice the internal clock rate of the
FPGA. This would make the FPGA with minor additional circuitry able to
count at greater than 500 MHz.
Just as other time_nuts cautioned, the problem is handling the input
signals at this frequency. Even at the 200 MHz rate, the very high
transition time of the input pulses required that the circuit be "clean" at
1 GHz. You really need to design the input circuit first, so that the
input signal is properly "conditioned" to provide clean data pulses to the
FPGA. This was my biggest mistake on the first design. The FPGA counted
fine if driven directly from a generator, but amplifying the input signals
to adjust gain and trigger threshold was a VERY big design task.
As you preceed, I would be glad to give you my comments on your work.
Tom Buehl buehl at superlink.net.
More information about the time-nuts