THE FOLD / GLITCH / HEISENBUG / THE CLOCK DRIFT
THE CLOCK DRIFT
both clocks are correct and they still disagree
1 WHAT IT IS · WHAT IT DOES · FACT OR FICTION
A quartz oscillator is specified in parts per million, which sounds like a rounding error until you multiply it by a day.
LIT verified live. At 50 ppm — an ordinary commodity crystal — a clock drifts 4.32 seconds per day and 1,577.85 seconds per year, which is over twenty-six minutes. It takes 20,000 seconds to accumulate a single second of error. Two machines specified ±50 ppm can be 100 ppm apart from each other and diverge at 8.64 seconds per day — exactly twice the single-clock figure, because error against a reference and error against a peer are different quantities.
LIT verified live. At 50 ppm — an ordinary commodity crystal — a clock drifts 4.32 seconds per day and 1,577.85 seconds per year, which is over twenty-six minutes. It takes 20,000 seconds to accumulate a single second of error. Two machines specified ±50 ppm can be 100 ppm apart from each other and diverge at 8.64 seconds per day — exactly twice the single-clock figure, because error against a reference and error against a peer are different quantities.
2 HOW IT WAS WEAVED · AI + HUMAN
Datasheet drift is why NTP exists at all, and why every distributed protocol that assumes bounded skew must say what bound it assumes.
AVAN (AI) checked the doubling rather than only tabulating rates. The pairwise figure is the one designs actually need and the one most often taken from the single-clock column — a system tolerant of 4.32 s/day between a node and UTC may still be broken by 8.64 between two nodes that are each within spec.
AVAN (AI) checked the doubling rather than only tabulating rates. The pairwise figure is the one designs actually need and the one most often taken from the single-clock column — a system tolerant of 4.32 s/day between a node and UTC may still be broken by 8.64 between two nodes that are each within spec.
3 ONE DIMENSION
Parts per million, into seconds per day and per year.
4 TWO DIMENSIONS · INTERACTIVE
Pick a crystal grade and let it run.
5 THREE DIMENSIONS + AVAN’S INVERSE
The green forward object: two clocks, each in spec, walking apart.
AVAN’s addition (the inverse-companion): the forward reading is that clocks drift and must be disciplined. The inverse is that both of those clocks are correct. Each is inside its published tolerance, neither is faulty, and they still disagree by nine seconds a day — so the disagreement is not an error state anybody can detect locally or repair by being more careful. Read backwards, “the clocks are wrong” is a category mistake: the specification permits this, and any protocol that assumed otherwise was assuming something nobody ever promised.
LIT at 50 ppm a clock drifts 4.32 seconds per day and 1,577.85 per year, taking 20,000 seconds to accumulate a single second of error, while two machines each specified plus or minus 50 ppm can be 100 ppm apart and diverge at 8.64 seconds per day - exactly twice the single-clock figure, because error against a reference and error against a peer are different quantities
FIG Datasheet drift is why NTP exists and why every distributed protocol assuming bounded skew must say what bound it assumes. AVAN checked the doubling rather than only tabulating rates. The pairwise figure is the one designs actually need and the one most often taken from the single-clock column - a system tolerant of 4.32 s/day between a node and UTC may still be broken by 8.64 between two nodes that are each within spec.
FIG Datasheet drift is why NTP exists and why every distributed protocol assuming bounded skew must say what bound it assumes. AVAN checked the doubling rather than only tabulating rates. The pairwise figure is the one designs actually need and the one most often taken from the single-clock column - a system tolerant of 4.32 s/day between a node and UTC may still be broken by 8.64 between two nodes that are each within spec.
◆ sealed .dlw.fold → folded to ROOT_0 · a sphere of HEISENBUG · David Lee Wise (ROOT0), with AVAN