@Nigel 11 -- A few question re your TDM proposal.
The whole concept of the Young, Townsend et al 'Quantum information to the home'/FTTH paper is indeed extremely intriguing. Your reply implies the method as presented within the paper is more complex than necessary and I think I agree with you.
Synchronously interleaving entangled qbits at an RXed power of about -88 dBm [p7 & Fig 7, p10] in what is effectively the intermodulated noise (Raman scattering) region of the fibre seems to me to be a daunting if almost impossible task. Thus questioning if there are simpler/alternative ways of tackling the QKD/key problem makes sense. Moreover, the effective data rate of the key distribution is ridiculously low--a little over 1kbps--given its context. Although the QKD data rate is sufficient for purpose, it nevertheless is exceedingly close to margins of unacceptability--at least in practical engineering terms--hence another reason to look for alternative methodologies.
Perhaps their approach was adopted to maintain FTTH compatibility/standards thus your TDM proposal would be unacceptable from this perspective. I may be way off-beam here (sorry no pun intended) but I'm unclear how your 'Time-coherence between the ends is a problem solved by NTP and/or GPS' is actually supposed to work in practice.
1. With your TDM system it's unclear to me why one needs to adapt time-coherence in the same way as that proposed in the paper.
2. If so however, then how would NTP/GPS time-coherence fit into the picture given the Inter-Symbol Interference (ISI) is limited by the ~35ps jitter of the single-photon detectors which operate within a window period of only 100ps? (The short-term GPS jitter at the RX-end would be substantially more than 100ps given proximity effects, TX/RX/pulse jitter, dialectic variations and ionospheric delays over the long TX/RX path length. Remember, c(v) is ~0.3m/ns, and correction schemas such as Carrier-Phase Enhancement (CPGPS) would, if used here, be ineffective through latency/delays.
3. Why would you use a small mark-space ratio of only 10%? If TDM dedicates clear-channel time to QKD exchange then why couldn't the exchange take place in orders of magnitude less time?
'Tis a fascinating issue, perhaps you may wish to reply.