Editing Bipolar encoding (section)

== Synchronization and zeroes ==

Bipolar encoding is preferable to [[non-return-to-zero]] whenever signal transitions are required to maintain synchronization between the transmitter and receiver. Other systems must synchronize using some form of out-of-band communication, or add [[frame synchronization]] sequences that don't carry data to the signal. These alternative approaches require either an additional transmission medium for the clock signal or a loss of performance due to overhead, respectively. A bipolar encoding is an often good compromise: runs of ones will not cause a lack of transitions.

However, long sequences of zeroes remain an issue. Long sequences of zero bits result in no transitions and a loss of synchronization. Where frequent transitions are a requirement, a self-clocking encoding such as [[return-to-zero]] or some other more complicated [[line code]] may be more appropriate, though they introduce significant overhead.

The coding was used extensively in first-generation [[Pulse-code modulation|PCM]] networks, and is still commonly seen on older [[multiplexing]] equipment today, but successful transmission relies on no long runs of zeroes being present.<ref name="allyouwanted">[http://www.dcbnet.com/notes/9611t1.html "All You Wanted to Know About T1 But Were Afraid to Ask", Bob Wachtel, retrieved on 25 January 2007]</ref>
No more than 15 consecutive zeros should ever be sent to ensure synchronization.

There are two popular ways to ensure that no more than 15 consecutive zeros are ever sent: [[robbed-bit signaling]] and [[bit stuffing]].

T-carrier uses robbed-bit signaling: the least-significant bit of the byte is simply forced to a "1" when necessary.

The modification of bit 7 causes a change to voice that is undetectable by the human ear, but it is an unacceptable corruption of a data stream. Data channels are required to use some other form of pulse-stuffing,<ref name="digitallink"/> such as always setting bit 8 to '1', in order to maintain a sufficient density of ones. Of course, this lowers the effective data throughput to 56&nbsp;kbit/s per channel.<ref name="dictionary">[http://www.faxswitch.com/Definitions/telecom_dictionary_b.html Telecom Dictionary, retrieved 25 January 2007]</ref>

If the characteristics of the input data do not follow the pattern that every eighth bit is '1', the coder using alternate mark inversion adds a '1' after seven consecutive zeros to maintain synchronisation. On the decoder side, this extra '1' added by the coder is removed, recreating the correct data. Using this method the data sent between the coder and the decoder is longer than the original data by less than 1% on average.