Calculating Time Error Limits for Boundary Clocks

Deployed synchronization networks often consist of chains of boundary clocks, known as T-BCs, which have specific devices and defined network time error specifications. This test case validated the performance of such a chain based on the quantity of class D T-BCs within it.
This year, in contrast to last year, we only tested this for Class D Telecom Boundary Clocks (T-BCs).
Class D T-BCs have only a single metric, max|TEL|, defined, for which the accumulation is not specified. max|TEL| may be considered as all cTE, all dTEL, or a combination. Considering all as cTE gives a worst-case limit of N∙ max|TEL|, i.e., N∙5ns, with N being the number of T-BCs in the chain, as the full allowance is being considered as cTE and therefore linearly summed. If considered as only dTEL, the limit could be calculated, i.e., as √N ∙ x²; our test passed this tight limit.
In this test, the Microchip TimeProvider® 4500 was used as the Telecom Grandmaster (T-GM), and six Class D T-BCs were used. According to the formula, the limit was 30ns.
The GM populated PTP and SyncE downstream, and all T-BCs were locked. Each T-BC was connected to one of the time error measurement devices, either the Calnex Paragon-Neo PAM4 or the Keysight Time Sync Analyzer. By doing so, we could not only see the time error at the end of the chain but also the time error introduced by each T-BC in the chain.

Figure 124

Figure 124: Calculating Time Error Limits for Boundary Clocks - Setup

The results were remarkable, as we only measured a maximum absolute low-pass filtered time error of 4.677ns at the end of the chain, the H3C S12500R. The time error results of all other devices in the chain can be seen in the following three images.

Figure 125

Figure 125: Time error measured at the H3C SR12500R

Figure 126

Figure 126: Time error measured at the Ciena 5134

Figure 127

Figure 127: Time error measured at the Arrcus S9600-72XC, Ericsson Router 6671, Huawei ATN 910D-A, and Juniper Networks PTX10002-36QDD

One interesting observation that can be made in the third image is the behavior of the Juniper Networks PTX10002-36QDD. As one can see, T-BC-1, T-BC-2, and T-BC-3's behavior shows slightly oscillating movement, periodically going slightly up and down; the PTX10002-36QDD is the fourth T-BC in this chain but does not reflect T-BC-1/2/3's pattern but rather cleans up this behavior and stays fully stable, which is remarkable in itself but also good for the following BCs.
    
In addition to the maximum absolute low-pass filtered time error, also the constant time error (cTE) and the filtered dynamic time error (dTE) were measured at the end of the chain even though there are no specifications yet for cTE and dTEL of Class D T-BCs. The measured cTE was -1.588ns, and the average dTEL was -0.874ns.