Harald wrote:
In AoIP there's no BSC-colocated transcoder... while in classic TDM based A interface there normally is, and you have PCMA/PCMU.
I decided to dive into this rabbit hole once again, and I discovered some details which differ from what has always been taught in Osmocom land. The purpose of this post reply is to share these findings with the community.
As it turns out, in Nokia's implementation of GSM BSS the transcoder (bank of TRAUs) is not enmeshed inside the BSC, nor is it enmeshed inside the MSC. Instead they introduced a third interface named Ater, joining the ranks of A and Abis. The voice path looks like this:
MSC <---A i/f---> TCSM <---Ater i/f---> BSC <---Abis i/f---> BTS
The bank of TRAUs (the term TRAU strictly means a single voice/CSD channel handler) is called TCSM in Nokia land, which stands for Transcoder and Submultiplexer. This TCSM sits between MSC and BSC in the voice path, while its physical placement can be either at the MSC site or the BSC site. (Locating the TCSM at the MSC site results in 4x savings on TDM leased line costs, hence I would imagine this choice was the popular one, but what do I know, I wasn't there...) Whatever the physical placement of the TCSM, the A interface is a bunch of E1/T1 circuits, and so is Ater. But the codecs are different: on A each call takes up a full timeslot in G.711, on Ater each call is 1/4 of a timeslot in the same format as on Abis.
I have yet to learn how Siemens and Ericsson implemented this part in their respective BSS architectures, but at least old Nokia gear has a sane & sensible design in this regard.
I am now looking to acquire a Nokia TCSM, either TCSM2 or TCSM3i. Based on my reading of the docs found online, it should be possible to fire up a TCSM by itself, having neither an MSC nor a BSC of TDM kind. What would one do with a TRAU by itself, without those other network elements? Answer: play with TFO! Without TFO, a TRAU is quite UNinteresting: it doesn't do anything more than what is contained in reference C sources for HR/EFR/AMR codecs from ETSI, i.e., practically free code which we can run and study as we please. But if a TRAU supports TFO and if that feature is enabled, then it is a complex wonder with a stealthy in-band messaging protocol (messages transferred by stealing the lsb of every 16th sample), lots of state machines and timers - this is the kind of stuff where it would be impractical to develop a new implementation and have any confidence about its correctness without being able to test against someone else's existing implementation.
If anyone here has any leads to Nokia surplus GSM equipment dealers, I would greatly appreciate some pointers. I know there are people who run Nokia InSite/MetroSite/UltraSite BTSes with Osmocom - perhaps the same dealers who sell that BTS gear would also have some components from higher up the signal chain?
M~