Bob wrote:
Again, since it’s a Kalman, even the people who designed it were a bit unclear about just what was going on inside. Was 0.1C data useful vs 1C ?…. nobody really knew. Based on many measurements of the TC of the DAC / Ref on the TBolt, it could indeed use some “help” when in holdover.
Even if what is "useful" was uncertain at the design stage, a factor of 100 difference in reported temperature resolution [~0.01C vs 1C, not ~0.1C vs 1C] should cause significant differences in both training and holdover performance. Differences that could not easily be ignored.
I agree that Tbolt holdover performance is nothing for Trimble to be very proud of, in any case.
Didier wrote:
The only controlled experiment I am aware of was run by Mark Sims in March 2009 and he measured a 100% difference in PPS change per hour between the old chip and the new chip in holdover in the same Thunderbolt. Big enough to be statistically significant.
Perhaps, if all variables were at least reasonably tightly controlled. But in the test, they were not. In particular, the Kalman training, which is likely to be by far the largest variable, was not controlled (and bear in mind that this also influences the Kalman response, which is likely to be the second largest variable). With the two largest variables uncontrolled, I do not believe the result can be considered statistically significant.
Best regards, Charles _______________________________________________ time-nuts mailing list -- [email protected] To unsubscribe, go to http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com and follow the instructions there.
