Back in 1992, the GSM world was simple. A carrier was divided into 8 timeslots and in those that were not used for broadcast information signaling, one voice call was carried. After some time, half rate channels were invented to put two calls in a single timeslot. At first, there was the HR codec which was inferior in speech quality but would use only use alternating instances of an assigned timeslot. But then, advances in coding technology gave rise to the Adaptive Multi Rate Codecs (AMR), and those codecs which fit into half a timeslot were, from a sound quality point of view, quite up to the Enhanced Full Rate voice quality. A nice move as voice capacity of the network effectively doubled.
And now, companies in 3GPP are attempting to again double the number of voice calls per timeslot to up to four! The work item is called MUROS (Multiple Users Reusing One Slot) and you can find some interesting papers about the concept HERE. The folder unfortunately also contains contributions to other topics discussed during this ad-hoc meeting so it’s a bit time consuming to find the papers on MUROS. For a start have a look at the following ones:
080007 – Details on Downlink DARP and Interference
080019 – Short Into to MUROS in Uplink and Downlink
080041 – A first draft of the 3GPP Technical Report which will result from this work
These papers give a first idea on how four voice calls could potentially be put into a single timeslot. Instead of further splitting up the timeslot in time, as was done for AMR half rate, two signals are to be emitted by the base station simultaneously. From a mobile station point of view one signal contains the user data and the other signal is perceived as noise. The mobile phone then has the task to filter out the noise (i.e. the data stream which is unwanted). In uplink direction, it is also foreseen that two mobiles transmit at the same time and that the base station uses interference cancellation and multi-user detection schemes to separate the two signals, potentially based on the use of different training sequences in the middle of the slot.
While at the moment different means are discussed of how the two signals can be told from each other on the receiving end it is clear that MUROS will not work with current mobile devices and also not with current mobile chipsets. Features like SAIC (Single Antenna Interference Cancellation) and DARP (Downlink Advanced Receiver Performance), both also discussed in 3GPP, will probably become the cornerstones for realizing MUROS.
While making this work on the radio layer is surely a formidable challenge, it will also prove to be tricky for radio resource management. Mobile can only tell the two signals apart under radio conditions that are favorable to them. The task of radio resource management will be to dynamically instruct the mobile to handover to ‘less used’ timeslots when signal conditions deteriorate. In the worst case, the mobile could even have to be handed over to a timeslot which it uses exclusively until radio conditions improve.
Should this feature be accepted and implemented it will keep GSM radio layer-, radio network, and chipset designers busy for quite some time to come.