I was going through my old presentations looking at frequency and phase requirements for LTE-A and HetNets. The slide above is some years old but it does summarise the requirements well. There is also an interview by Martin Kingston & Andy Sutton of EE on this topic which is available here. I would think that with 5G latencies often quoted as less than 1ms (but in practice it may be up to 10ms) would have very critical frequency and phase timing requirements.
ThinkSmallCell recently held a webinar on this topic. The write-up is available here and slides/video is embedded below. Here is something I found interesting:
In the past, a central Grand Master supplied a common signal that was hardwired throughout the network. Today, we now see distributed master clocks appearing almost everywhere. Typical requirements are for 50ppb frequency and 1.5us phase timing over the air, driven from 16ppb and 1.1us into the base station.
Frequency sync requires a Primary Reference Clock (PRC), whereas Timing sync requires a Primary Reference Time Clock (PRTC). The latter must come from a satellite GNSS source, such as GPS, and be traceable to Universal Co-ordinated Time (UTC).
The end-to-end Inter-Cell time error budget of 1.5us (1500nanoseconds) is split into three parts:
- A time source, with an error of up to 100n
- The transmission network, with up to 1000ns
- The small cell (eNodeB), with up to 400ns
The transmission network may have up to 10 boundary clocks with a combined total of 500ns error. The remaining allowance is split equally between dynamic time errors and network asymmetry. It is especially important that packets travelling in each direction (uplink/downlink) incur similar amounts of delay variation – if the time taken to send and receive packets varies differently, then phase timing errors would mount up rapidly.
It is this asymmetry of packet delay variation which is the biggest problem with engineering phase timing throughout a large network.
The ITU has defined two different time profile standards related to transmitting the phase sync signal.
G.8275.1, which relies on full on-path support. Each node in the backhaul transmission network must be fully aware of the phase timing component and actively support its transmission. Each router or node would have its own boundary clock that synchronises and re-generates the timebase locally. This may be feasible for new product but would otherwise require replacement or upgrade for existing routers and backhaul transmission equipment.
G.8275.2 was recently consented and only requires partial on-path support. One or more boundary clocks are installed at the most effective points in the backhaul path, with many legacy routers/nodes being unaware of the special importance of the PTP packets.
It is crucial to take into account the existing technical infrastructure and also cost for deployment. As part of this effort, it is critical to engineer the network so that asymmetry correction can be considered.
In cases where full on path support is deployed, the mitigation of uplink versus downlink asymmetries are extremely important and usually requires a manual calibration of each link which is extremely costly.Here are the slides with Video in the end. Video can also be directly viewed on Youtube here.
*** Edited 11/04/16 - 10.30 ***
RTT has just published an article on related topic titled 'A second look at time', available here.