If the voice signal is modulated and transmitted directly after channel coding, due to the condition changes in mobile communication channel, a deep of the fading will influence a successive string of bits and cause high bit error rate.
If the bits of a successive string are interfered or lost, the other end of the communication can not decode the interfered or lost bits. To solve this problem, some technique or method to separate the successive bits are required. Thus the successive bits in a message can be transmitted dispersedly so that the bit error should be discrete. In this way, even if errors occur, the errors are only concerned with a single or very short bit stream, which will not lead to that the whole burst or the whole message block cannot be decoded. In this case, the channel coding will take effect and recover the bit errors. This method is called interleaving technique. The interleaving method is the most effective coding method for dispersion of bit errors.
The key point of interleaving is to disperse some bits( suppose there are  “b” bits) of the code into some ( suppose to be “n” bursts) burst sequences so as to change the adjacent relationship between bits. The higher the value of “n” is, the better the transmitting works. However, the transmission delay is higher too. Therefore, a balanced consideration is needed, the interleaving is related to the purpose of the channel. In the GSM system, the second interleaving is applied.
After channel coding, the extracted 456 bits are distributed into 8 groups with each group containing 57 bits. That is the first interleaving, also called internal interleaving as shown in the above diagram. Through the first interleaving, the successive messages in the groups are dispersed. One burst carries two segments of voice information composed of 57 bits. Obviously, if the two groups of 57 bits information from the first interleaving of a successive 20ms voice blocks are inserted to the same burst sequence, the loss of the burst will lead to loss 25% bits in the 20ms voice block. Therefore, one more interleaving is needed between two voice blocks, which is called the inter-block interleaving or second interleaving.
Suppose that  voice block B is divided into 8 groups: perform inter-block interleaving to the first four groups (B0, B1, B2 and B3) of block B and the last four groups (A4, A5, A6 and A7) of the previous voice block A ; thus, four bursts are constituted: (B0, A4), (B1, A5), (B2, A6) and (B3, A7); to break the adjacency relationship between successive bits, bits of block A occupy the even position of the burst while bits of block B occupy the odd position of the burst. For example, B0 occupies the odd bit of the burst while A4 occupies the even bit. Similarly, perform interleaving to the last four groups of block B and the first four groups of the next block C. After the second interleaving, a 20ms voice block is inserted into 8 different burst sequences respectively and then transmitted one by one. Even if a whole burst is lost during transmission process, only 12.5% of each voice block is influenced and the errors can be corrected through channel coding at the other end. In addition, the second interleaving for the control information is different. The interleaving mode is (B0, B4), (B1, B5), (B2, B6) and (B3, B7).