I would also been intersted in knowing which (if any) gateways with GPSs deals with the leap seconds issue correctly.
When a GPS is first powered on an gets a fix, then the secons boundary, from 1PPS, should be within tens of nS. However the time the GPS shows will only be correct when its had an over the air update of the current value of leap seconds. This can take up to 12.5 minutes from the GPS powering on, and there is always to potential for that update to be missed.
Sorry to be picky about that, but building NTP devices is some kind of sports and there are many factors to note. I know, you guys want to build location services for LoRa, but even for a Indoor Gateway, there will be too much temperature differences to get accurate results over time.
I was aware of the variations in velocity factors in cable, I spent 15 years installing LANs, when I started I built my own TDR gear for measuring cables.
The leap second âproblemâ is not an issue that should affect TTN localisation, as long as the timing goes from the 1PPS edge.
Although my self I wonder if it will be possible to time arrival of packets (to do localisation) to with say the required 333nS to get even 100m accuracy.
Ouch⊠ok, sorry. That should be less a problem. Why we discuss about PPS nanosecond accuracy, when we need absolute timestamps, gained probably from UART? I think for this the Mikrotik Gateways could be a step forward, because they have OS support for GPS disciplined NTP on board, although not pps. Also Multitech has NTP
And for your demands NTP on the receiving gateway over the backhaul network like pool.ntp.org should be accurate enough.
And not to forget about raspberry pi based gateways under some linux. There you can use pps. But there is much tuning needed according to temperature and workload. Good examples here.And Here.
Well, allright - but is there a suitable gateway out of the box which sends timestamps, generated either by GPS with PPS (but not without), or NTP? Besides MatchX 1701?
may be you can try this one, they have an amazing firmware with integrated LoRa server on it (but no GPS). Iâm always wondering if adding cheap USB GPS device could works as expected?
Last but not least, $149 and if you want 4G add $100, not bat
There are two issues: first the firmware doesnât support it, though you could build your own firmware from their open source repos for earlier kits which would.
The other is that you need the PPS connection from the GPS to the SX1301 chip. Probably you could modify the board to insert that.
And then you probably donât want a laggy USB interface, but rather to use a hardware UART. The console one is definitely exposed, it looks like one of the others is but you probably have to rebuild the kernel to make it actually work: Introduce uart · RAKWireless/WisCore Wiki · GitHub
Why we discuss about PPS nanosecond accuracy, when we need absolute timestamps, gained probably from UART? 
I think for this the Mikrotik Gateways could be a step forward, because they have OS support for GPS disciplined NTP on board, 
