delivery which is much more difficult to meet. 2. Test cases 1 - 6 all have similar results in terms of MTIE. The reason is that there is no variable delay. All them exhibit a fixed frequency offset of around 1 ppb for short observation windows that saturates for larger observation windows. This saturation value is larger when the fixed delay is very large. I do not have a clear explanation for this effect but it is most probably caused by Net.Storm. The impairment generator is asynchronous and this fact could be more noticeable when it has to store packets for longer periods of time as it happens when the delay is configured to be in the range of milliseconds. 3. Test cases 7 - 9 represent uniformly distributed delay in the forward or backward paths or in both at the same time. As expected, MTIE is larger if there is variable delay in the network. It is seen that a few tens of microseconds are enough to generate a FAIL result. That means that any network operator willing to deliver time over PTP has to control carefully delay variation. There is a remarkable point: delay variation in the backward path does not affect the MTIE result. We compute MTIE only in the Sync packets and these go from the GM to the slave only. 4. Test cases 10 - 12 are similar to cases 7 - 9 but with much higher delay variations. I have not generated the MTIE graphics for cases 10 and 12 because it is too large to have any meaning. Test case 11 shows again that variable delay in the backward path does not have any effect in the MTIE result. This is a quite unnatural situation however.