[time-nuts] How close can you trim a Cs?
cfharris at erols.com
Wed Mar 9 19:13:31 EST 2005
John Ackermann N8UR wrote:
> Hi Chuck --
> But if the bias is constant, it won't affect the results because I'm
> looking at the change in time interval over time, not the absolute value
> of the reading. So, if the initial time difference reading is 1.100us
> and 24 hours later it's 1.200us, the offset is 100 nanoseconds in 86,400
> seconds, or 1.16x10e-12.
> The fact that there's a bias in both the initial and the final values
> doesn't matter because it's "common mode" to both the starting and
> ending readings and thus cancels out (assuming, of course, that the bias
> doesn't change during the course of the measurement).
The problem is the "bias" does change with voltage, temperature, humidity,
phases of the moon, etc. So in a short term, it can be quite significant.
However, I seem to have gotten my mind wrapped around this problem in a
wrong way, I keep thinking you are trying to get stable indications of your
offset, in the 10E13 range, using 100 second time intervals, but of course,
you are not. Instead you are looking at trends that go over much longer
periods of time. For those longer periods, your technique (and counter)
should be adequate to the task.
More information about the time-nuts