| | Windows Time Server: Appendix H |
 | | While f for a particular clock is a random variable with respect to the population of all clocks, for any one clock it ordinarily changes only slowly with time and can usually be assumed a constant for that clock. |
 | | The clock-filter algorithm determines the selected clock offset theta (peer.offset), together with the associated roundtrip delay delta (peer.delay) and filter dispersion epsilon sub sigma, which is added to the associated sample dispersion epsilon sub i to form the peer dispersion epsilon (peer.dispersion). |
 | | When a number of clocks are involved, it is not clear beforehand which are correct and which are not; however, as cited previously, there are a number of techniques based on clustering and filtering principles which yield a high probability of detecting and discarding incorrect clocks. |
| www.bytefusion.com /windows-time-server/xp-2000-time-server/ntp-windows-time-server/appendixh.htm (3025 words) |