Network time servers and other time synchronization applications, rely on an accurate reference clock input, such as GPS. Holdover is a mechanism whereby a device can continue to provide accurate time even when a reference clock input fails or becomes unavailable.
The Need For Holdover
Timing applications require a reference clock input to maintain accurate time. Common sources of reference clocks include GNSS, GPS and other satellite PVT (Position, Velocity and Time) navigation systems.
However, there may be rare occasions when a reference clock becomes unavailable. Cabling issues, antenna failures or signal interference can be possible causes of loss of availability. In such circumstances, holdover mechanisms allow a device to continue to provide accurate time by using an internal clock.
When a time synchronization device relies on an internal clock, it is known as holdover. The clock acts like a flywheel that continues to operate at a constant speed, maintaining time. Often, holdover is referred to as flywheel mode.
Essentially, holdover provides continuous operation until communication with a reference clock input can be re-established.
The amount of drift of a devices internal clock with respect to a reference clock input is known as holdover performance.
Holdover Performance
A devices internal clock will drift when a reference clock input becomes unavailable (flywheeling). The amount of drift is known as ‘Holdover Performance’.
A number of factors can affect holdover performance, such as the quality of the clock components used. Additionally, environmental conditions, specifically temperature variation can have a great affect.
Oscillators are used in synchronisation applications to provide a holdover mechanism. Oscillator manufacturers will often specify holdover performance as frequency stability over temperature. Specifically in Parts per Million (PPM) or Parts per Billion (PPB) over a specified temperature range. E.g. 2 parts per million over -20C to +85C.
The amount of drift over a specific time period can easily be calculated:
TimePeriod (secs) / frequency stabilityFor example, 2PPM over 24 hours equates to 86,400secs * (2 / 1,000,000) = ±0.173 secs or ±173msecs.
An oscillator with a frequency stability of 2PPM will therefore maintain time within +-173msecs after 24 hours. This will hold true only over the specified temperature range.
However, if the ambient temperature is reasonably stable, much better holdover performance can often be achieved.
Higher quality components often provide improved holdover, generally at greater expense. Oscillators can be broken down into four sub groups, increasing in cost: Quartz Crystal, TCXO, OCXO and Rubidium.
Quartz Crystal Oscillators (XO)
Quartz Crystal Oscillators are low cost components commonly found in clocks, watches and consumer electronics. They rely on the oscillation of a quartz crystal to provide a stable frequency.
A common abbreviation for Quartz Crystal Oscillator is ‘XO’. They are the amongst the lowest cost of oscillator components, but only offer limited stability.
A Quartz Crystal oscillators stability is typically specified at 25C. It generally does not represent the oscillators accuracy across a temperature range.
20PPM at 20C equates to 1.73 secs over 24 hours. This is only at a constant temperature. Any temperature variation over this period will significantly affect performance.
Temperature Controlled Crystal Oscillators (TCXO)
By factory calibrating frequency over a number of temperature points, a Temperature Controlled Crystal Oscillator can guarantee extremely tight stability over temperature.
A TCXO’s temperature coefficient is calibrated and corrected over temperature with an active
correction circuit. It results in an extreemly tight frequency variation over a wide temperature range. While more expensive than XO’s, TCXO’s provide a fantastic price – performance compromise.
The typical stability of a TCXO is in the range 1 to 5 PPM over -20 to +85C. However, if the ambient temperature where a device is located is much more stable, such as in a typical data centre, much better holdover performance can be achieved.
Oven Controlled Crystal Oscillators (OCXO)
By enclosing a Crystal Oscillator in a temperature controlled oven (or small chamber), much improved holdover performance can be achieved. Such devices are often referred to as Oven Controlled Crystal Oscillators (OCXO).
As mentioned earlier, temperature variation greatly affects crystal oscillator stability. An OCXO provides greater stability by maintaining a crystal at a constant temperature, regardless of environmental temperature. OCXO oscillators are expensive, but provide a very stable frequency output.
Typical stability: 1 PPB to 500 PPB over -20 to +85C
Rubidium Oscillators (Rb)
Based on atomic clock technology, Rubidium oscillators are amongst the most expensive of oscillators. They work by exciting Rubidium (Rb) gas stored in a chamber using microwave energy to control an OCXO. Rubidium oscillators provides an extremely stable frequency output, but at very high cost.
The Rubidium Standard is used in PVT (Position, Velocity and Time) navigation satellite systems and telecommunications systems.
Typical stability: 0.002 PPB
Summary
Holdover provides continuation of operation for a period of time in the event of loss of a reference clock input, such as GPS. Sometimes called flywheel mode, it allows time for a reference input to be re-established.
Holdover performance is dependent on the stability of the oscillator used in the device. Higher quality oscillators provide better stability and holdover performance, but at a cost.
TimeTools uses a high stability TCXO in their T550 NTP Network Time Server to provide holdover. TCXO’s are relatively cost effective and provide sufficient holdover performance for a reasonable period. In short, they provide a great price – performance compromise.
Related Information
The Fundamentals Of Time Synchronization
The Importance of Accurate Time on Computer Networks
A Guide To GPS NTP Servers For Network Time Synchronization
Additional Resources
Synchronization and Holdover in Telecommunication
Holdover in synchronization applications
Holdover Performance > Clocks, Time Error, and Noise
| About Andrew Shinton Andrew Shinton is the joint founder and Managing Director of TimeTools Limited. He has a BSc (Hons) degree in Computer Science. Andrew has over 20 years experience of GPS systems and Network Time Protocol (NTP) in the Time and Frequency Industry. |