One approach to deploying LTE without packet switched voice call functionality at the beginning is to instruct mobile devices to use a 2G and 3G network when the user makes or receives a voice call and return to LTE afterwards. This solution is referred to as CS fallback and has been specified in 3GPP TS 23.272. As it's likely that it will be deployed over time in quite a number of networks and used over many years, I thought I have once again a closer look at the specs and write a little primer about it. A little warning: This is somewhat of a propeller head post which requires some background knowledge on the circuit switched core network of GSM and UMTS and how LTE works.
SGs, a new interface between the circuit switched core and the LTE network
In principle CS fallback connects the Mobile Switching Center to the LTE Mobility Management Entity (MME) via the new SGs interface. The name and functionality of this interface is similar to the already existing optional Gs interface between an MSC and a 2G or 3G SGSN. In some networks this interface is used for paging and location updating synchronization between the circuit switched and packet switched core network to reduce the signaling load and, in case of GSM, to be able to signal an incoming voice call to a mobile device that is currently busy on the packet switched side.
When a mobile device is CS fallback capable it initially performs a combined CS+PS attach to the LTE network. This informs the network that the mobile device wants to use circuit switched services in addition to IP based services over the LTE network and is capable of falling back to a 2G or 3G network for incoming and outgoing voice calls. The MME then performs a location update on behalf of the mobile device over the SGs interface with the Mobile Switching Center and the HSS. This MSC is referred to as the SGs MSC below to distinguish it from other MSCs that might also become involved during the CS-fallback procedure. If successful it signals to the mobile device that it has been registered in the circuit switched network as well and that incoming calls will be signaled to it.
Incoming Calls – Mobile Terminated Calls
When an incoming call for the subscriber arrives at the Gateway-MSC, the HSS is interrogated for the location of the subscriber. The HSS then returns the information to the G-MSC that the subscriber is currently served by the SGs MSC. The call is then forwarded to the SGs MSC. The SGs MSC will then send a paging message over the SGs interface to the MME which will in turn inform the mobile device and require it to leave the LTE network to accept the incoming voice call in a 2G or 3G cell. The mobile device will then do as instructed and start communication over a GSM or UMTS cell.
Moving from one radio technology to another can be done in several ways. The basic scenario is a redirect in which the network gives the mobile device an instruction to select a different radio network. The instruction can contain information about the target cells to reduce the time it takes the device to find a suitable cell and establish communication. In a more advanced scenario, a full Inter-radio access technology (IRAT) packed switched domain handover from LTE to UMTS or GSM is performed which is prepared in the network and thus the interruption time is lower. In this scenario, the network can instruct the mobile device to perform radio measurements. The results of those measurements are then used by the network to select a suitable target cell and give the mobile device precise instructions of how to quickly get to this cell to minimize the handover time.
The thing with the location area
If the GSM or UMTS cell is in a location area that is different from that in which the mobile device is currently registered, a circuit switched location update procedure is required before the call can be forwarded. This could be the case, for example, if the SGs MSC connected to the MME is not controlling the GSM or UMTS cell to which the UE is transferred. This could happen in case the mobile device has selected a cell other than the one intended by the network, e.g. in areas with a location area border, or in case only a single MSC is SGs capable and hence acts as a mere relay for signaling messages rather than a switching center to which cells are directly connected.
In case of UMTS or GSM Dual Transfer Mode, the packet switched context can also be moved so any packet switched communication can continue while the voice call is ongoing. This is important, for example, so email push and other applications can continue running in the background while the voice call is ongoing. Also, this allows users to continue using other web based applications during the phone call, e.g. searching for some information on the web during the call, etc.
In case the SGs MSC does not control the 2G or 3G cell it is necessary that the SGs MSC redirects the voice path that has been established between the Gateway-MSC and itself to the MSC controlling the cell. This is done with the help of the location update procedure which is invoked as described above. Part of the location update procedure is to inform the Home Subscriber Server (HSS) of the change in location. The HSS in turn informs the SGs MSC that the subscriber has changed the MSC area. The SGs MSC then informs the Gateway MSC of this change with a ‘roaming retry’ message as specified in 3GPP TS 23.018. The Gateway MSC then removes the speech path to the SGs MSC and establishes a new link to the new serving MSC. What I m not quite sure about is whether the roaming retry is actually used today in practice for other purposes already. If you have some details about this, I'd be quite interested.
If a location update was necessary the mobile terminated call is delivered immediately after the procedure has finished. To make the MSC aware that a circuit switched call is waiting for the mobile device after the location update it includes a CS-fallback (CSMT) flag in the location update message. This flag allows the MSC to link the location update and the call delivered by the Gateway MSC. In case no location update is required, the mobile sends a paging response message to the MSC, which has the advantage that the call can be established more quickly.
Outgoing Calls – Mobile Originated Calls
When the user initiates a mobile originated call, the mobile device contacts the network with an Extended Service Request message which contains a CS fallback indicator. The network then decides based on its capabilities and that of the mobile device to either perform:
- A packet switched handover to a GSM or UMTS cell, which is the fastest way to move the mobile device to a radio infrastructure from which the circuit switched call can be initiated.
- An RRC release with redirect to GSM or UMTS, optionally with information about possible target cells to decrease the time necessary to find the cell. This is somewhat slower than a handover as the mobile device is required to reestablish contact to the UMTS network on its own without help of the LTE network.
- An inter-RAT cell change order to GSM. Optionally, the network can include information on potential GSM cells in the area (Network Assisted Cell Change, NACC)
Contacting the network prior to leaving the LTE network is necessary so the mobile device’s context in the LTE base station (eNodeB) can be deleted and to get additional information on potential target cells to speed up the process.
The fallback mechanism to GSM or UMTS is also used for supplementary services based on Unstructured Supplementary Service Data (USSD) messages which are used for modifying parameters such as call forwarding settings, or querying the amount of money left on a prepaid account.
Delivery of SMS messages, however, does not require a fallback to a GSM or UMTS network. This is because SMS messaging is not based on USSD and is not a service implemented in the MSC. Instead, the SGs MSC can forward an SMS message it has received from the SMS Service Center (SMSC) to the MME via the SGs interface. The MME will then deliver it to the mobile device via MME to mobile device signaling messages that are transparent to the eNodeB. Mobile originated SMS messages can be delivered in the same way in the other direction.
As CS fallback is not a Voice over IP technology, it is likely that it will mostly be used in LTE networks before VOLTE becomes available. Furthermore, CS fallback can be used as a backup solution in roaming scenarios in which voice capable LTE devices are roaming in a foreign LTE network in which VOLTE is not available or in case no roaming agreement is in place for IMS voice services.
Pros and Cons of CS fallback
The main advantage of CS fallback is that it will enable network operators and device manufacturers to introduce LTE devices with a single cellular radio chip before VOLTE becomes available and network are deployed widely enough to prevent having to hand over the call to UMTS or GSM too often (how that is done is another story).
The downside of the approach is the increased call setup time required due to the change of the access network and the potential location update procedure that is required in case the new cell is in a different location area before the normal call establishment signaling can proceed. For LTE to LTE CS fallback voice calls, the extra call establishment delay is even doubled. In other words, the extra call establishment time is likely to be noticed by the customer who expects new technologies to work better than the previous ones and not worse.
An alternative to CS fallback is to use dual radio mobile devices that use LTE for data only while it is available and a legacy network for voice calls and also for data once the user roams out of the LTE coverage area. Verizon, for example, is doing this today, perhaps because it was one of the first LTE networks and needed LTE capable mobile devices including voice early on to relieve the strain on its aging CDMA network. If this approach works well enough they just might hold on long enough until their LTE network is dense enough to introduce VOLTE and skip CS fallback altogether.
For UMTS network operators, however, there is little incentive at the moment to go to dual radio devices as they still have ample data capacity in their UMTS networks. As a consequence, they have launched their LTE networks mostly with data only devices. For them, CS fallback might be the better alternative unless the additional call setup delay time becomes annoying. There are several deployment options that influence the additional time required to set-up the call so it's going to be interesting who will do what to reduce the delay.
CS fallback sounds easy but from the description above I think it is quite clear that it is not quite that. A new interface to be implemented in the MSC software and the MME, the use of roaming retry functionality that is not used so far (please correct me if I'm wrong) and the new CS fallback flag in the location update message will keep network and device engineers busy for a while. A lot of effort for a "temporary" solution.