And yet another roaming story from my recent trip to North America: For voice and video calls to family and friends, I use my Prosody XMPP server in Germany, a voice stream reflector that is also located in the country and the ‘Conversations’ messenger app. By having all components close together, the speech path delay is minimal. But when roaming to North America, the round trip delay time to and from Germany is around 300 milliseconds when connected to an LTE network. That’s quite a bit longer than what I would have expected, as the RTT time from one of my servers in Germany to ‘archive.org’ that is hosted in San Francisco is around 160 ms.
But let’s stick with the 300 ms round trip time for now, which means there is a 150 ms one way delay, i.e. the delay between mouth to ear. However, if both ends of the conversation are at the far end of North America, e.g. in California, while the voice reflector is in Europe, the mouth to ear delay doubles, i.e. it is around 300 milliseconds. ITU G.114 recommends a mouth to ear delay of less than 150 ms for conversational calls, so the 300 ms are way beyond this recommendation. So how does that sound like in practice?
To find out, I established a ‘Conversations’ voice call over two LTE/5G ENDC connections between two smartphones and observed the delay. And indeed, the delay is quite obvious, but it is still possible to have a normal conversation, albeit with a bit of silence when the direction of the conversation changes.
If I was at the North American west coast for a longer time than just for a week or two and wanted to use ‘Conversations’ for local calls, I would obviously change my setup: SIM cards with LTE P-GWs on the continent and a media reflector somewhere here as well. In other words: The Edge is still important!