 While we are genuinely concerned about how efficient can we utilize a spectrum to increase high data rates and to sustain those data rates both for the users close to the cell center and those at the edge of the network. It is also important to understand that high data rates can work best if the delay or the latency for connection establishment and data transfer is also low. In this module, we try to look at a very interesting aspect of these 4G networks with regards to voice over IP because voice over IP is a very important application and service provided by 4G networks and how this voice over IP actually dictates the latency requirements. Let's start with the broad categorization of delays or latency with regards to the cellular networks for providing broadband connectivity. We could either have control plane delay or we could have user plane delay. The control plane delay actually consists of all the delays which relate to signaling that is connection establishment and connection release. Now when we say connection establishment, it actually means it is the total time which is consumed for transitioning an idle user from being idle to becoming active and actually starting to send data. The recommendations in 4G specifications are that it has to stay around a hundred milliseconds. As far as the user plane is concerned, it actually means user data. So the user data plane requires the delay not to exceed 10 milliseconds under an unloaded condition. The unloaded condition actually is a completely off-peak R situation in which a single user that is sending small packets in a single data stream does not experience more than 10 milliseconds of delay in terms of data transfer. Now with these two kinds of delays, let's look at the relationship that voice over IP would hold with this latency. First of all, voice over IP in terms of the service and application is central to the LTE or the LTEA networks. First of all, the most important feature of the LTEA networks is that it is all IP based. It means the sending side and the receiving side actually encapsulate their user data into IP packets. And then the IP addressing scheme is at work and we see the concept of TCP IP layering at play here. With this packet switched technology, voice over IP would require some quality of service guarantees. So in terms of end to end latency, it is described as 150 milliseconds, which is highly suited to have a smooth playback of live and streaming service. But if it starts to exceed more than 400 milliseconds, it becomes unacceptable and we would experience unnecessary jitter. Now, although we have just looked at the requirements, these requirements are now going to influence the overall architecture from the conventional architecture of the 2G networks and 2.5G data networks such as GPRS and the Edge networks. We were accustomed to network elements like we'd have a mobile station that would talk to a base transceiver that would be controlled by a base station controller, a mobile switching center, the SSGN and the GGSN nodes. But once we look at the 4G network, it actually has to avoid delays at these intermediate nodes. So the design of 4G networks is required to be relatively flat. That is why we see that intermediate network elements somehow are skipped and the base station is actually connected to the centralized gateway. That is why if we just compare 4G to its immediate predecessor as 3G, we saw the radio network controllers in 3G networks, but in 4G, these RNCs have been removed. So utilizing limited number of devices allows us to have better latency experience.