KCCS
Precision Timing
Application Note AN-002 · By KCCS Engineering Team · Updated 2026 · 8 min read
When a PTP grandmaster loses its timing source, the network switches to holdover mode. How long it stays accurate depends entirely on the internal oscillator. Specifying holdover as "±1 μs over 24 h" without breaking down where that error comes from leads to over-design (paying for an OCXO you do not need) or under-design (field outages).
1. Temperature-induced drift. This is usually the largest term. If the oscillator has frequency vs temperature stability Δf/f and the equipment room changes by ΔT during holdover, the frequency shift is Δf/f × ΔT. Over a holdover window of T hours, the accumulated time error is:
Error = (Δf/f × ΔT) × T
Example: an OCXO with 0.1 ppb/°C sensitivity in a room that swings 10 °C over 24 h accumulates 0.1 ppb × 10 = 1 ppb frequency shift. Over 24 h (86,400 s), that is 86.4 μs of time error. This is why "±1.5 μs/24 h at 25 °C" becomes a very different number in an unconditioned cabinet.
2. Aging. A typical OCXO ages at 1E-10/day. Over one day that is 8.6 ms. This is small compared to temperature drift in most real-world deployments, but it dominates if the equipment room is tightly controlled.
3. Initial offset and measurement noise. The residual frequency error at the moment holdover starts, plus the short-term noise integrated over the window. A well-disciplined oscillator leaves less than 1E-12 offset; ADEV contributes a few hundred nanoseconds over 24 h.
| Term | Value | 24 h Error |
|---|---|---|
| Temperature drift (0.1 ppb/°C × 5 °C swing) | 0.5 ppb | 43 μs |
| Aging (1E-10/day) | 0.1 ppb | 8.6 ms |
| Initial offset (1E-12) | 0.001 ppb | 0.09 μs |
| Total worst-case | ~52 μs |
If your network requires ±1 μs over 24 h, a 0.1 ppb/°C OCXO in a swinging environment will not meet spec. Options: (a) thermally stabilize the enclosure, (b) step up to a ±0.01 ppb temperature-stability OCXO like KOT series, or (c) use a CPT/CSAC atomic reference that does not have the same temperature sensitivity.
- Know your actual temperature profile. A lab-grade ±1.5 μs/24 h spec is not the same as field holdover across -20 °C to +60 °C.
- Account for warm-up. If the device loses power and restarts, the oven warm-up time matters more than holdover itself.
- Design for the 99th percentile, not the typical case.
Send us your temperature range, holdover window and allowed time error. We will recommend the right KDO series or atomic clock.
Talk to Our EngineersOur technical team can help you choose the optimal oscillator for your specific application requirements.