IBM 8260 manual Clock Recovery, Phase Jitter

The other important point about the Differential Manchester coding is that it uses a higher baud rate (the number of state changes on the transmission media) than the actual data transfer bit rate on the ring, to provide the benefit described above. In fact, the baud rate on the token-ring is twice the data transfer bit rate. On a 4 Mbps token-ring, the baud rate is 8 megahertz. A 16 Mbps token-ring runs at 32 megahertz.

8.1.5 Clock Recovery

The other function that needs to be done in a token-ring environment (which is characterized by a transmission media shared by many stations) is that each receiving station needs to recover a very precise timing from the received bit stream in addition to just reconstructing the bit stream. To do that in a token-ring network, each station could have been designed to have its own independent clock and a large buffer. In this case, every station would have waited to receive the whole frame in its buffer, then it would have used its own clock to generate a new signal which would have been sent to the next station. The receiving station would have repeated this process and so on. The result of this design would have been a very large ring latency which would have made it unsuitable for most applications.

In order to minimize ring latency and eliminate the need for a large buffer on each station, the characteristics of Differential Manchester coding (lots of transitions on the transmitted signal) are used so that each station can derive a highly accurate timing source from the received data stream. This timing source is then used by each station to synchronize its clocks with the other stations on the ring. The source timing for all the stations is provided by one of the stations on the ring which is called the active monitor.

Note: In any operating ring, one of the stations is elected as the active monitor and the other stations act as standby monitors prepared to take over if the active monitor fails. The process of electing an active monitor is called token-claiming process. This process is described in 8.1.2.2, ªToken-Claiming Processº on page 131.

Because of at least one state transition per bit in Differential Manchester coding, the receiving stations can derive a very stable source timing from the received bit with minimum circuitry and buffering. This results in a minimal delay in each station and hence in the ring. Also, the stability of the derived timing minimizes the phase jitter and thus allows more stations to be connected to a single ring segment. However, the phase jitter is not totally removed and requires special attention.

8.1.6 Phase Jitter

Jitter is the generic term given to the difference between the correct timing of the received bit and the timing detected by the received station.

Despite techniques such as Differential Manchester coding, it is impossible for the timing detected by the station to be exactly the same as the correct timing of the received signal. This is because of the characteristics of the transmission media and the high speed (4 or 16 Mbps) of the ring. Some bits will be detected slightly early and some others slightly late. This means that the timing may vary randomly by a small amount on either side of the correct timing.

In some networks, these small differences in the bit timing do not make any kind of difference. But, in the case of token-ring network, the jitter is accumulated as

Chapter 8. 8260 Token-Ring Support 133