Re: [LEAPSECS] Precision vs. resolution

From: Tom Van Baak <tvb_at_leapsecond.com>
Date: Thu, 1 Jun 2006 11:21:46 -0700

> I should perhaps explain that I was interested in an internal
> representation for durations, which I am now representing as a triple of
> months, minutes, and seconds (the number of minutes in a month is not
> predictable, nor the number of seconds in a minute given leap seconds,
> but all other relationships are predictable: 10080 minutes/week, 12
> months/year, 100 years/century, etc.) To this I would add a fourth

I'm curious what motivated this particular representation
given that one could perhaps also use years/days/seconds,
or weeks/hours/seconds, or months/hours/seconds, or like
GPS, weeks/seconds, or like MJD, days/fractions...

> nonnegative integer representing "clock resolution units" and wanted to
> make sure I had the terminology correct.

Take this with a grain of salt since I'm still confused by
resolution vs. granularity myself, but isn't the phrase
"clock resolution units" redundant? If resolution is about
the minimum unit of measure it would seem the phrase
"clock resolution" is sufficient, no?

Further, if you are counting clock cycles to subdivide
integer seconds, then consider words like clock rate,
or clock period, or clock granularity, or clock cycles,
or just plain clock tick. That gives you
"months/minutes/seconds/ticks" which, to me at least,
sounds better than
"months/minutes/seconds/clock resolution units".

If your clock resolution is ms or us or ns then it's even
simpler, e.g., "months/minutes/seconds/ns".

This representation would accommodate, for example,
a typical 1.193 MHz clock -- since the resolution is
1 ns while the granularity is 838 ns.

/tvb
Received on Thu Jun 01 2006 - 11:51:27 PDT