Little adventure for today, Settling a hunch about Starlink and presumably other LEO internet access. The latency of the connection is asymmetric. Is that a bad thing? Not really, but it's interesting to observe that it shares a similar latency gap that DSL with interleaving does. How did we (me and a flatmate) test this, especially since I live in a flat in London? We simply drove out to Rainham with a £15 eBay 12V inverter and I hooked it up to a global GNSS time source and ran my split ping utility https://blog.benjojo.co.uk/post/ping-with-loss-latency-split
one honk maybe more
benjojo
replied 14 Jan 2023 18:54 +0000
in reply to: https://benjojo.co.uk/u/benjojo/h/Hyff4VfLKDTDf36vJ4
Here is the chart of latency data, We didn't capture for very long as it was getting dark and we didn't want to get locked in the car park. Also the car battery was likely not enjoying the 60W or so load the starlink+inverter was giving.
benjojo
replied 14 Jan 2023 19:00 +0000
in reply to: https://benjojo.co.uk/u/benjojo/h/Cv96TtFpQvC51h49XR
To summarise, since RX and TX in the context of bi directional connections is a odd one. Starlink has more (on avg about 10ms) latency from going You->LEO->World than World->LEO->You. This has some interesting implications for things like netcode in online games, but nothing new since a lot of connections (DOCSIS/DSL/etc) have had asymmetric latency for ages. I need to try and get longer term data though, if anyone on starlink is around in the UK, and you are okay with being a test for a few weeks, I can ship a GNSS latency test unit :)
benjojo
replied 14 Jan 2023 19:09 +0000
in reply to: https://mastodon.social/users/Edent/statuses/109689123740138613
@Edent I totally non-informed theory is there is a lot more error correction on the uplink than the downlink. Since the downlink can be quite powerful (they are in space facing down after all), the uplink has to fight with more noise.