Nw as I came to think of, it might indeed be that there are some people who do not know the internals of cellular networks.

It might be that I am glossing over some bits due to the very usual mistake that people make, assuming that other people already have the basic knowledge because everyone around me knows these things.

So, here is really quick introduction to the different ETSI-GSM generations;

2G, original GSM
First wholly digital cellular radio system. Originally designed for voice only, later enhanced by services like SMS messaging and IP calls.
Air interface is based on a slow frequency hopping scheme, with TDMA interleaving with 8 timeslots.
Much of the network intelligence resides in the core side, BSC and MSC elements. BTS is fairly dumb device which only really connects the air interface to T1/E1 SDH links. The backhaul is SDH and ATM.
The reason why it is not practical to have simultaneous CS and PS calls is primarily the limited bandwith of the air interface and secondarily the routing of the backhaul T1/E1 to specific network element that is used to route the PS call to internet.

2.5G, enhanced "Edge" GSM
This basically adds possibility of allocating more than one timeslot per UE, so that downlink data rates can be doubled or quadrubled.
Of course this also limits the number of UE's in the cell accordingly.

3G, WCDMA UMTS
3G is a logical continuation of the ETSI spec and can coexist with 2G well but it completely overhauls both air interface and backhaul.
Air interface is based on fast frequency hopping by code spreading. Signalling is very complex and there are huge amount of different options and tunables.
BTS is more intelligent than on GSM, transferring some BSC functions to the edge. Backhaul is ATM and the rich signalling allows simultaneous use of several logical channels which means for example CS and PS calls.

3.5G, HSDPA and HSUPA extensions
Small enhancements which are basically just software tweaks in signalling to group logical channels so that data rates can be increased. HSDPA (=High Speed Downlink Packet Access) is the more common 3,5G technology, HSUPA which is the same thing for uplink direction is a bit rarer feature.

4G, LTE (Long Term Evolution)
This again revamps both air interface and backhaul, where CDMA is now dropped for ODFMA and ATM is finally put to rest and IP goes directly down to BTS.
Signalling is actually simpler than on 3G which was really an elephant on that respect, still more intelligence is distributed towards the edge and the BTS is now handling quite lot of things by itself.
Because SDH/ATM has been dropped from the backhaul CS calls are no longer possible and indeed thare were emerging operators that had no telco legacy and implemented VoLTE-only networks.

5G
This is really more or less an umbrella standard covering a wide range of modulation and spectrum allocation technologies for different use cases. (everything between very low-power/low-bandwith IoT gadgets to super-low latency factory automation and autonomous vehicle control and super-high bandwith for media terminals) Like 4G this is also All-IP network and again more intelligaence is pushed toward the edge, so BTS sites will have cloud-computing resources for some nice new functionalities and services.

To recap; all evolution of the ETSI-GSM networks has been to the direction of pushing more and more functionality and intelligence from the core towards the network edge. Because the air interface is the "last mile", everything can be done faster when the processing is nearest to UE and signalling is done in the BTS to minimize the loops done via the backhaul.

I understand that the network is capable of doing that. But it also means that the phone must be capable of doing that too, right? Which is kinda incompatible with this claim.

You cannot have it both ways. Either "voice is always on GSM" or "voice is not always on GSM"

Sorry to be a nuisance, but I'm still confused

In case anyone is wondering why the history starts from 2G, the first generation mobile phones were anaologue. With no scrambling or encryption. You could listen to at least one side of the conversation (because different frequencies were used for up- and downlink) if you had a radio receiver that could be tuned to the relevant frequency. Don't ask me how I know

And some call even older technologies the 0G (zero G).

Well, think of it this way.

With all ETSI-GSM family networks, voice call is always "digital", however there are two ways to set up the voice call.

• 1.) The "traditional" digital telephony way, which is pretty much adaptarion from digital switching of the old-style non-mobile land telephony, usually called CS (Circuit Switched) call. This relies on SDH/ATM backbone and establishes a virtual switched circuit between the UE's. (This means there is a dedicated channel and timeslot for the call which is reserved for the duration of the call)
• 2.) The newer Packet Switched (PS) call which is just digitized voice call carried over IP/UDP packet stream, no preallocated routes needed.

The 2G and 3G networks have SDH/ATM backhaul and can establish CS calls, 4G and 5G cannot do this.

The 3G, 4G and 5G networks have IP backhaul and can establish PS calls. (and actually at least theoretically also 2G could do that because it could use VoIP over data connection altough I don't think that has ever been released in a public network)

The reason 3G is such a beast is because it has dual backhaul and also the additional burden of managing the power control over multiple cells which is required by the CDMA modulation scheme... compared to that 4G signalling is a walk in the park!

5G again gets a whole load of new stuff which ramps up the complexity (ultra-short feedback loops over air interface, adaptive antenna-array control, management of several modulation techniques simultaneously, etc...)

Thanks, juiceme, but unfortunately, instead of simplifying it for me, you complicated it even more. Maybe we are talking about different things. All this time I was talking about the communication between the handset and the nearest cell tower. I understood the "voice is always on GSM" claim in that context and that context alone. Maybe I was wrong and needed to take a more holistic approach.

Indeed we could be talking past each other, happens to me all the time!

I was mainly pointing out the differences in backhaul, which in reality are the more important things, air interface is just air interface

So, when concentrating on the air interface, between the UE and BTS, there is not really that much difference between whether the call is CS or PS. All the difference is in signalling; how to establish the routing of the data in backhaul, what codec, channel configuration and data rate to use, etc...

So the question "is CS call always GSM?" is a bit misleading; "GSM" is not a very specific term really.

Some news regarding Jolla and VoLTE, found in, of all places, the latest Ubuntu Touch Q&A summary:

Obviously this information is Ubuntu Touch and PinePhone-centric, but it is interesting that Jolla got a mention and that they all meet on these issues.

Did a imei test recently and t-mobile says what we already figured out.

Here are the screen grabs of the results from https://www.t-mobile.com/resources/bring-your-own-phone

x

But nice is that even your phone is definitely not supported T-Mobile offers you as next step to buy a SIM!?