Test results

In the beginning, we directly connected in our test setup two Ordinary Clocks. We used IEEE1588 enhanced plug-in modules on our modular Ethernet switches (MICE series). To put the protocol through its paces, we added a high network load with an Ethernet packet generator.

So that we could examine the difference between local clock and reference clock as close-to-applications as possible, we enhanced both units with a Pulses Per Second (PPS) signal output and connected them to an oscilloscope. Thus we could seize now very elegantly the deviation of the two signals and also represent the frequency distribution of the deviation. The synchronization accuracy which we reached lay within ± 100 ns (max. jitter). The measurement ran for a period of 84 hours.

The following figure shows the frequency distribution of the offset values between master clock and slave clock in nanoseconds. The standard deviation amounts to 23.95 ns and the average value -4.248 ns.

The drift values of the oscillators limits the synchronization accuracy with the available prototypes. Quartz frequency of 50 MHz (± 50 ppm) results in a dissolution of 20 ns. Thus the system can adjust the drift within the range of ± 20 ns per second. If you now regard the relative drift between the local clock and the master clock during two other following synchronization telegrams, then it becomes clear that the short-term stability of the oscillators substantially determines the synchronization accuracy in the in-swing condition.