21.4.3 Clear channels for data applications

As already mentioned, users of Tl facilities are restricted from accessing the full 64kbit/s bandwidth of the PCM channel on any one circuit of the DS1, whether for voice or data applications, because the eighth bit of each word is unavailable. In voice trans­mission, the eighth bit has been used for in-band signalling (remem­ber that only the sixth and twelfth frames are really affected). In speech, this generally does not cause any problem, but the same condition cannot be guaranteed for data. The implication has been the deployment of 56kbit/s facilities and a loss of 8kbit/s of usable bandwidth. The maximum data rate in a DDS channel (Bell's Dataphone Digital Service, see Section 21.8) is thus 57kbit/s, with 7 bits per frame used for customer data and the eighth bit reserved for network control. This causes complications, particularly for users interfacing with an international facility, e.g. a 64kbit/s satel­lite channel.

In many early repeaters that were installed along the transmission cables linking the Tl devices, the synchronisation circuitry was very simple. An oscillator was run at a frequency close to the line pulse stream. The pulses were then used to pull the oscillator on fre­quency. This method relied upon the availability of pulses in the data stream. Because there could be many repeaters over a Tl span, the cumulative effect of a drift in frequency could cause an unac­ceptable amount of error in the transmission. Therefore the number of pulses, or the ones density, on the line is important. For voice transmission this is not a problem. However, for data it is a different story. To be able to maintain the ones density, zero code suppression schemes are used. These schemes consist of replacing long se­quences of zeros by a special pattern (for details on zero code suppression schemes, refer to Section 21.7). The assumption on the DS1 stream was that at least one eighth, i.e. 12.5%, of the incoming bits were Is. If that ratio was not maintained, synchronisation would be lost and an error would result. When the DS1 contains voice channels, work around tech­niques like changing the eight 0 to a 1 do not create major problems, since the distortion in the voice message is minimal and the human ear would not notice it. In contrast, changing a customer data bit is not acceptable, as that may imply a substantial difference in the received data (e.g. 0001 and 1001 are quite different numbers). Therefore, one bit out of eight in the customer channel is reserved in each frame to inject a 1. This bit is called the network control bit and is not made available to the user. While this resolves the ones density problem, it imposes the speed restriction alluded to above.

The approach embodied in the current generation of equipment, of substituting a 1 in the LSB position of an all zeros word, may be acceptable for voice, because the difference between a 00000000 coded sample and a 00000001 coded sample is imperceptible to the ear. When the bits of a DS1 channel represent directly input digital data, such as DDS or ISDN, however, one is not at freedom to alter the data, and a technical solution must be achieved.

A possible solution has already been introduced. This concept, also referred to as the primary rate interface or PRI, is one of the central physical components of ISDN. Its implementation involves taking all of the signalling information of the individual 23 Tl channels and combining it into a 24th channel. Often referred to as the "23B + D" format, this framing and signalling scheme allows 23 Tl channels to carry up to 64kbit/s of voice, data, video, or other information, while the D channel transmits all of the control data, including sophisticated signalling techniques described in CCITT Recommendation I.431/Z.921/Q.931. (The "B" in this designation stands for bearer; the "D" stands for delta, i.e. the channel that transmits change information).