jeh@crash.CTS.COM (Jamie Hanrahan) (11/14/87)
In article <8711130234.AA12248@ucbvax.Berkeley.EDU> "ERI::SMITH" <smith%eri.decnet@mghccc.harvard.edu> writes: >By the way, someone named George Miller who used to work here had the >idea--I don't know if it was original--that power companies should tweak >their 60 cycle frequency in such a way as to produce a continuous, >automatic adjustment of synchronous wallclocks to solar time. It sounded >like a good idea to me, but probably in the same general category as >calendar reform... It's my understanding that this is done now, and has been done for years. Although the moment-to-moment accuracy of the "60 cycles per second" isn't very good -- I used to have a reed-type AC line freq meter and it would often show variances of plus or minus a Hertz or two -- over the long term the power companies are supposed to pump out 60 cycles per second. They do this by running what amounts to a digital clock from their own power, comparing that with either a local cesium-beam atomic clock or the WWV signal, and adjusting over the long term to compensate for errors. Oh, you wanted correction to solar time, by which I gather you mean CUT (leap-second corrections and all), not to the flat "seconds" that come out of the cesium-beam clock? Well, er, I guess somebody decided that 60 cycles per standard second was more useful. You see, a second has a standard length; we don't shrink or stretch them to match the earth's inconstant spin. Instead we occasionally change the definition of which second is about to come next (by adding leap seconds). So it wouldn't make sense to fine-tune the 60 Hz to mean "60 Hz per mean solar second". (Note that "leap seconds" are only added on rare occasions; every few years or so, I think. I wouldn't worry about them.) If your AC line-driven clock is off, it's because it's responding to noise on the power line as if it was extra cycles (if fast), or because of brief power outages that it rides through on the filter caps (if slow). My 5314- based clock (still running after all these years) was always gaining time due to line noise -- a common problem with the 531x chips. I added a PLL circuit to make it insensitive to clocking other than at or near 60 Hz, and all was (and still is) well. (Note: A cesium-beam clock doesn't need calibration; it's a primary standard.)