Radio is always prone to disturbance. 10-6, is an acceptable bit error rate in radio. This means that at 70 megabits per second we can have 70 errors per second. In the first video, we have seen that coding is used at the physical layer to correct some of these errors. But, how can we handle those that have not been corrected? This is one of the functions of the MAC layer, “Media Access Control”. Another function of the MAC layer is to let several services communicate with the same mobile at the same time. This function is called “Multiplexing”. In the second part of the video, we will present how it is managed by L. T. E. L. T. E uses error detection codes – called CRC : Cyclic Redundancy Check. These codes are calculated by the sender and appended at the end of the message. The receiver makes the same calculation. If it gets the same result, it knows that the data are correct. There are 3 possible cases: first, if the message is correct, then the receiver confirms the reception with a message called an “acknowledgement” or ACK. This message is sent on the control channel we spoke of in the previous video. In the second case, the data has been corrupted during transmission. The receiver informs the transmitter with a non-acknowledgement message, or NACK. The transmitter then resends the message and continues to do so until it receives a positive acknowledgement. In the last case, the transmitter that has sent a message has not received an acknowledgement. So, either the message or the acknowledgement got lost. Either way, something went wrong. So, if the receiver doesn’t receive any acknowledgement after a certain time, it tries to send the data again. If the message is still not received after a given number of attempts, the MAC layer gives up. Protocols on a higher level will probably make new attempts later on, but, for the MAC layer, these are considered as new data. This mechanism is called “Send and Wait A. R. Q.” “Send and Wait,” because the transmitter sends the data and waits for the response. And A. R. Q for “Automatic Repeat ReQuest.” This is the mechanism that retransmits automatically if no acknowledgment is received. Processing a message or an acknowledgement may take a certain time. That’s why the norm states that the recipient sends its acknowledgement 4 sub-frames after receiving the message. In the same way, after receiving this acknowledgement, the transmitter waits 4 sub-frames before either repeating the message or sending a new one. The total cycle lasts 8 sub-frames, or, 8 milliseconds. During this time, the transmitter does not transmit. This is the principle of "Send and wait.” To optimize resource consumption, L. T. E actually runs 8 Send and Wait processes in parallel. While the first process is blocked, waiting for its acknowledgement, another process can transmit data. And so on, until we have 8 processes in parallel. This is called “PARALLEL Send and Wait.” The problem with parallel Send and Wait is that it can cause de-sequencing. In this example, blocks 1,3 and 4 were received correctly, but block 2 was lost. Block 2 will be re-sent 8 sub-frames later. So, the sequence delivered by the MAC layer to the higher level will be block 1, block 3, block 4, and finally block 2. The MAC layer does not deal with the re-sequencing. We will see in the next video that that is the role of the R. L. C layer which is right above the MAC layer. If you look at L. T. E. E documentation you will probably see the term HA. R. Q instead of A. R. Q. It is not too important but let’s take a quick look at this slight difference. The idea is that to increase error correction performances, the receiver stores all erroneous messages. This accumulation increases redundancy and thereby the possibility of getting the correct data. HA. R. Q stands for Hybrid A. R. Q. “Hybrid” because the MAC layer has to collaborate with the physical layer to let it know that the data it processes are actually repetitions of the same transport block. Now, let’s speak about another function of the MAC layer. You use your smart-phone for voice conversations as well as receiving emails or browsing web pages. And you may want these at the same time. But there is a single communication channel between the base station and your mobile. So, all this traffic has to be mixed over this pipe. This is what we call “MULTIPLEXING”. Before getting further, let me introduce some vocabulary. In the classical layered model of networks, data at the top of a layer are called “SDU” or “Service Data Units” and the data at the bottom are called “PDU” or “Protocol Data Units.” In our case, the MAC layer receives the MAC-SDUs and produces MAC-PDUs. Now, let’s talk about multiplexing. Data handled by the MAC layer are provided by the upper layer, R. L. C, they're provided through message queues. These queues are called logical channels. We will see in the next video that there are several of them, corresponding to different levels of service. Here, for example, is a channel dedicated to signaling (L. T. E internal control), another to voice and a last one for general data. The output of the MAC layer, the MAC_PDUs are the transport blocks that will be transmitted by the physical layer. To form the transport blocks, the MAC layer assembles the SDUs from the different logical channels. This is referred to as multiplexing because there are several incoming streams and only one outgoing. To help the receiver break down, or de-multiplex, the MAC layer adds headers that indicate where each message starts and to which logical channel they belong to. Padding can also be added to ensure that the size of the MAC_PDU is exactly the same size as the transport block to be transmitted. In short: as well as allocating resources, as seen in the previous video, the MAC layer ensures the reliability of transmission using HA. R. Q. It ensures retransmission in case of error. The MAC layer can also merge different data from different logical channels into the same transport block.
