Richard B. Gilbert wrote: > David J Taylor wrote: >> "Richard B. Gilbert" <> wrote in message >> news:-zkdnagr0u362bnxnz2dnuvz_ogdn...@giganews.com... >> [] >>> Do you disagree with the 500 PPM limit? 500 PPM works out to 43 >>> seconds per day. A clock that gains or loses that much time every >>> day has to be considered seriously broken! Plus or minus 50 PPM, or >>> less, is typical. >>> >>> I suppose that someone could change the limit to 1000 PPM or even >>> 10,000PPM. Would that be the right thing to do? Why? >> >> No, I don't disagree or agree with 500ppm - simply asking whether or >> not people felt there was any mileage in using a somewhat larger value. >> >>> The 500 PPM limit may be completely arbitrary but I suspect that it >>> includes a vast majority of computer clocks. >> >> I think this is an excellent point. It would be fascinating to see a >> histogram of computer clock frequency error, so that one could make a >> statement like "500ppm satisfies 98% of computer clocks" with some >> degree of authority. I'm not aware of any such data, though. >> > > Dave Mills MIGHT have some relevant data.
Dave Mills has said that NTP can become unstable if more than 500ppm of correction is allowed. (Note though, that on any reasonable Unix, you can adjust the effective clock frequency in steps of 100ppm, to bring it to within 50ppm, before running ntpd.) > > Since computer time of day clocks are not critical for most users I > suspect very little money or effort is spent on making them accurate. > Most people are probably satisfied if the computer has the right date > while they are printing checks! > > If most PC users really wanted accurate time, it could probably be done > for less than $25 per machine (S.W.A.G. based on the cost of a cheap > wrist watch.) More like 50 cents. The main part of the cost is the time taken by a person to twist a trimmer to zero beat. Traditionally, at least, the time of year clock basically used digital watch hardware, just lacking the trimmer capacitor, for find adjustment. PCs, don't of course, have time of day clocks as hardware. The timing, when powered up uses software to do most of the work. _______________________________________________ questions mailing list questions@lists.ntp.org https://lists.ntp.org/mailman/listinfo/questions
