Conclusions

 A useful analytical tool for modelling the behaviour of the clocks used to achieve synchronisation in SDH networks has been obtained. This was based on implementing a model proposed by ETSI. In the course of the work, it was found that certain problems exist with the standards at the time of writing. This has made it necessary for certain recommendations to be made as to modifications that are felt necessary to the standards documents.

The ETSI model as it stands seems inadequate in reflecting the true characteristics of synchronisation clocks. An important property of the slave clocks used in SDH synchronisation networks is their ability to clean up an incoming noisy synchronisation signal. This is typified by the low pass filtering characteristic inherent in phaselock loop transfer functions. It was found that simply cascading models of PRC's, SSU's, and SEC's by adding the noise generated by each model did not result in a simulated maximum reference chain satisfying the pr4escribed limits for wander within a SDH synchronisation network. The extension of the model (from an idea presented in [26]) to incorporate the slave clock noise filtering property has enabled a maximum reference chain to be more properly modelled.

It was found that different values were required for the proposed model to satisfy the standards for noise generation of individual slave clocks. The maximum values proposed for the gain elements K_f and K_w in order that the ETSI noise model satisfies the masks for wander in locked mode are shown in table 9.1.

 

Clock Type K_f K_w

Primary Reference Clock 10000 3.0

Synchronisation Supply Unit 4500 2.75

SDH Equipment Clock 1425 3.0

These figures differ from those propounded in the standards [4], [5], [14]. However, the model has only been proposed as a tool to allow parameters to be changed to reflect the noise generation of particular clock types. These parameters have been changed by others [14] for precisely this reason, so the discrepancy is not a major worry.

The inaccuracy in [4] and [5] as to the ETSI model being used to derive the standards for individual clock wander (known as masks) should be resolved. This is a logical nonsense given that the model parameters have been changed to allow the model to satisfy the masks. Because of this, a revised explanation as to the origin of the wander masks is required. The standards should make clear that the shape of the SDH Equipment Clock wander mask is inappropriate in that it does not represent the asymptotic behaviour of a SEC of bandwidth $1 \leq B\leq 10Hz$. It should also be reported that in certain circumstances, the minimum 30Hz sampling frequency for the collection of time error data to calculate TDEV may be relaxed without adversely affecting the quality of results. In order for the approximation $H(S)\approx H_A(s)$ to hold, the transfer function for filter A in the noise model should be revised to

$$H_A(s)=\prod_{n=1}^8{\sqrt{\frac{\alpha_{1}}{2 \pi\cdot 7 \sqrt[4]{7}}}\cdot\frac{s+\sqrt{7}\alpha_n}{s+\alpha_ n}}$$

where $\alpha_{n+1}=7\alpha_n$, $\alpha_8=2\pi \cdot 6.72 rad/s$, and $f_0=\frac{\alpha_{1}}{2 \pi\cdot\sqrt[4]{7}}$.