[time-nuts] TEC party file format?

[Chris Albertson]

On Wed, Jun 29, 2011 at 10:54 AM, Hal Murray <hmurray at megapathdsl.net> wrote:
>
>> How does the pulse trigger the capture ? If some hardware line is polled,
>> how frequent is that polling ? The counter units may well be nanoseconds,
>> but the inherent uncertainty of the polling instant must be taken into
>> account.
>
>> If instead there is no polling, but it is a hardware triggering, then could
>> you please give more details ? Thanks.

The hardware design dates from the old IBM PC in the 1980's  The pin
on the serial port (and one on the old Parallel printer port)  is tied
to a CPU interrupt line.  The micro-code inside the Intel chip polls
this is very high rate, roughly a few GHz.  (Actually we don't know if
it is microcode or clocked logic and it may depend on the exact CPU
model.)

The interrupt jumps to a small bit of code inside the Kernel.  This
code reads a hardware counter.  The count is saved and the units are
ns.  But acuracy is not at the ns level.   But is certainly at the sub
us level.

I suspect all of this happens MUCH faster then you could implement
using 7400 series TTL logic.  The 5ns gate delay of a TTL chip is an
"eternity" for a modern CPU.     Even for me who works with computers
all the time it's hard to imagine a CPU that can issue multiple
instructions per cycle with a cycle rate over 3GHz.  and I've got
eight of those processors inside this computer.

The user level program does not need to run in "hard" real time, it
only has the read each time tag before the next tag is saved.  The
tags are sequence numbered so the user program can see if it missed a
tag so missing one is not a serious error.

I don't know the final level of accuracy but NTP uses the same system
for the pulse per second ref. clock and if you take care it can run at
about the 1uS level with respect to UTC.

Top do better I think you'd need to use one of the HP Universal
Counters and a GPIB interface or maybe the "Pic Tic II" to measure
zero crossings.  Or maybe some one can build a counter that runs off a
10MHz lab standard.  But Pic Tic seems to have already do this. So I'd
say if  1uS is not acceptable then use the external counter

Again, what are the requirements?   Can't evaluate a proposed design
with out knowing what we are looking for.


Chris Albertson
Redondo Beach, California