@CarlIT, maybe we are not in so much disagreement as I thought. A few clarifications...
I never suggested that echo cancellation added to delay, I only mentioned it as an oddity that When speaking into an iPhone I heard nothing (silence) on the landline, in same room.
I already said that core network includes IP, but rarely find that the packetisation delay (which must exist) to be noticeable or annoying. Packetisation also requires a further delay, reassembling received IP packets that may be out of sequence. I think that is what you called the jitter buffer which again must cause a further delay, however short, it cannot be ‘none’.
In contrast, I called my bank yesterday at a UK call centre. In this case, as was very common with such call centres, there was noticeable tendency for the two parties to keep interrupting one another, which is a common consequence of poor quality packetisation delay. I would speculate that the call centre was implemented using VoIP over a less than PSTN quality network.
I also explained (or tried to) that whilst I do not like the delays inherent to VoIP I am prepared to live with them at PSTN level ( ie if implemented well). The delay exhibited by mobiles, which will include packetisation with possibly more compromised quality, is what I find quite intrusive. There may be other factors at play with mobile that add further to the delay but regardless, I nearly always find the mobile delay to be highly intrusive, and prefer to use a landline, even if it entails VoIP somewhere in the mix.
As an aside, people may have noticed a specific thing that no longer works, on packrtised services, it is the conversation along lines of....
“Thanks for calling, Goodbye”
“Yes, Bye.”
Then a split second later, one party remembers something ele
“Hang on, one more thing, are you still there?”
“” silence, as although still holding the handset, the other party does not respond, and hangs up. Whereas in days of analogue, I swear the “Hang on” would be heard....
...It would take physchological studies to confirm, but I have a theory that the delay is somehow enough for the remote party’s brain to command the ear to stop listening, whereas with POTS you could beat that brain command. Just a daft theory.