Background

T1 is a high speed digital network (1.544 mbps) developed by AT&T in 1957 and implemented in the early 1960's to support long-haul pulse-code modulation (PCM) voice transmission. The primary innovation of T1 was to introduce "digitized" voice and to create a network fully capable of digitally representing what was up until then, a fully analog telephone system.

Perhaps the way to really begin this discussion is to discuss the AT&T Digital Carrier System referred to as "ACCUNET T1.5". It is described as a "two-point, dedicated, high capacity, digital service provided on terrestrial digital facilities capable of transmitting 1.544 Mb/s. The interface to the customer can be either a T1 carrier or a higher order multiplexed facility such as those used to provide access from (fiber optic) and radio systems."

So in the basic definition there is the discussion that there is a "higher order" or hierarchy of T1. There is T1 which is, as we have discussed, a network that has a speed of 1.544 Mbps and was designed for voice circuits or "channels" (24 per each T1 line or "trunk"). In addition, there is T1-C which operates at 3.152 Mbps. There is also T-2, operating at 6.312 Mbps, which was implemented in the early 1970's to carry one Picturephone channel or 96 voice channels.

There is T-3, operating at 44.736 Mbps and T-4, operating at 274.176 Mbps. These are known as "supergroups" and their operating speeds are generally referred to as 45 Mbps and 274 Mbps respectively.

The general T-Carrier hierarchy appears in Figure 1 and is detailed in Chart 1.

For mathematical reasons, a voice channel was selected to be at 64 Kbps. 24 of these channels is a composite of 1.536 Mbps, not 1.544 Mbps! Why is there a difference? The reason is that after a byte (8 bits) of data is sent from each channel (24 * 8 = 192 bits) there is an extra bit used for synchronizing called a Frame bit - hence 193 bits are sent and this increase of 1 bit per 192 causes the speed to increase to 1.544 Mbps.

The fundamental frame of T1 is shown in Figure 2.

Well, you might ask, 1.544*2 = 3.088 Mbps and not 3.152 Mbps for T1C, how come? Well, the answer is that the T1C frame is made up of 1272 bits and is quite different from the 193 bit frame of the T1 data stream. It should be pointed out that the frame length of T1C and higher signals are not related in any technical way to the T1 stream which is treated simply as a string of bits. The simplistic diagram in Figure 1 is correct from an organizational point of view and does not show the relationship of the formatted data.

Now I have been using the term "T1 data stream". To be consistent with AT&T parlance, a "T1 data stream" is called a "DS1". Equally, a T1C stream is referred to as "DS1C", etc. Another summary chart to show the relationship is in Figure 3:

A convenient way to think of T1 is from the first two layers of the ISO (International Standards Organization) OSI(Open System Interconnect) model: the Physical and Logical layers. The Physical layer focuses on the electrical characteristics such as signal shape, voltage levels, etc. The logical layer deals primarily with the format issue - how is the data extracted from the low-level protocol?

The designation "DS" in Figure 3 refers to "Digital Signals" and describes the physical layer. The designation "T" refers to the type of carrier that is being used. Often these are used interchangeably but that technically is not correct.

On the topic of standards, T1 has been specified first by AT&T and second, by ANSI (American National Standards Institute). The European equivalent of T1 is called CEPT and is a CCITT standard. As a point of interest, the CEPT standard is at 2.048 Mbps and does not use a "master clock". In the U.S., the three major carriers each have a single "master T1 clock" from which all the others are derived. In the U.S., all T1 clocks are "slave" to this master clock. The problem that occurs is when someone wants to interconnect a T1 network provided by MCI to a T1 network provided by Sprint. This requires what is known as an elastic buffer and this is built into most T1 devices.

When someone says they are running T1, they may be saying several different things: The may mean that they have a network that is passing data at 1.544 Mbps; they may mean that they have a network that conforms to the T1 electrical interface specification (DSX-1), or that they have a network that passes data that conforms to one of the several framing formats (D4, ESF, etc.). More likely than not, they mean all three but their concentration may be on only one of these items. The confusion in the user community is a result of the interchangeability of words and the confusing requirements for connection to the AT&T system.

Services and Quality

AT&T through ACCUNET T1.5 offers several services besides the already mentioned point-to-point service. There are four "transfer arrangements" that can be purchased:

1. Customer ability to change terminating location of T1 link with AT&T assistance (either signal or dial)

2. M24 Multiplexing allowing the user to connect up to 24 channels to individual switched and non-switched services offered by AT&T.

3. M44 Multiplexing allowing the user the capability to combine 2 T-1 lines, each carrying up to 22 channels to 1 T1 line using Bit Compression Multiplexing (BCM).

4. Customer Controlled Reconfiguration (CCR) allowing the customer to dynamically allocate circuits without AT&T assistance.

These services allow the user to have T1 trunks in several cities and allow data transfer to each. This along with the T1-Mux (to be discussed later) forms the modern T-1 network.

Associated with the lower costs of T1, the guaranteed quality of the network is also superior to leased lines. By specification, AT&T states that the performance objective is 95% Error Free Seconds (EFS) on a daily basis and the availability objective is 99.7% on a yearly basis.

Channel Banks and Formats

A digital source, or terminal, is the equipment that generates digital signals for transmission through the digital network. The large majority of digital sources now produce a DS-1 signal. The D4 Channel Bank is an example, although it can produce signals at other rates as well.

The reference to the term "Channel Bank" is made quite often in the T-1 language. The type of Channel Bank is important since it defines the type of formatting that is required. For example, a D4 Channel Bank must have a DS-1 signal with data formatted in accordance with the D4 format.

The purpose of a Channel Bank in the telephone company is to form the foundation of multiplexing and demultiplexing the 24 voice channels (DS0). The D-type Channel Bank is used for digital signals. There are five kinds of Channel Banks that are used in the System: D1, D2, D3, D4, and DCT (Digital Carrier Trunk).

A transmitting portion of a Channel Bank digitally encodes the 24 analog channels, adds signalling information into each channel, and multiplexes the digital stream onto the transmission medium. The receiving portion reverses the process. As these were designed as voice circuits, the assumption is that the digital data is PCM voice and that the voice is companded and expanded through the use of CODECs. D1 banks (later called D1A) were first installed in 1962 and their success led to modifications of D1B and D1C. The original D1A,B, and C banks used 7 bits for each voice sample and one bit in each code word for carrying the signalling (off hook, ring, etc). When it became desirable to connect several T1 transmission spans together, the performance was not too good. In addition, it was realized that providing signaling information in every code word was wasteful since 8,000 bits per second was not required to provide the signaling information for a channel; the signalling information simply did not change that quickly.

As a result of these conditions, another modification to the D1 series (D1D) and the new D2 channel bank were developed. The D2 bank uses all eight bits of every time slot to encode the analog signal except for selected frames. Supervisory and signalling information is sent by using the least significant bit from the code word in each channel every sixth frame. The companding characteristic also was changed to give better performance. The D2 bank increased the packing density to 96 channels in the same space as the 72 channels for a D1 bank.

D3 and D4 banks were motivated by advances in ICs, allowing packaging of 144 channels in a single bay. Following the D4 bank, advances in technology resulted in the development of the Digital Carrier Trunk unit, or DCT. It was developed by the Bell System to be smaller, lower cost, and easier to maintain than the D4 channel bank.

The D1 type channel bank (D1A,B,C) placed alternate 1's and 0's in the 193rd bit position. It was assumed that random data would not contain this pattern, in bits spaced exactly 193 bits apart, for any significant length of time. The receiving device would find the 193rd bit by using a simple search technique. This algorithm had the advantages of circuit simplicity and speed. In the early 1960's, there were few commercially available ICs for building complex logic functions, and elementary designs cost less. The disadvantages of this technique were rapidly uncovered when equipment was installed in actual customer sites. Certain standard analog tones, such as the 1000 Hz test tone, applied to one or more voice channels and digitized by Channel Bank, created an alternating one and zero pattern every 193 bits in one or more voice channels. It was possible for the terminal to lock up on the incorrect pattern. This condition, affecting all 24 channels, could last until the test tone was removed. The 1000 Hz tone has been changed to a 1004 Hz test tone.

By the time this problem became apparent, it had been decided to use T-carrier for toll quality telephony, which required more precise coding techniques. D1 channel banks used seven bit encoding for voice signals, and an eighth bit for signalling. The new format provided for eight bit coding most of the time (5/6 frames) and seven bits only in one frame out of six. This is known as 7 5/6 coding with "robbed bit" signaling and was first implemented in the D2 channel bank (D1D is a retrofit of D1 channel banks with D2 capability).

Besides the "false frame" problem, D2 bank designers were faced with a new set of problems. The new format required two steps; first, find the 193rd bit, and second, find the sixth and 12th frame in a 12-frame sequence. The time required to find the proper bit sequence rises exponentially as the number of bit positions between frame bits increases. Although we still use every 193rd bit, it is time-shared between the terminal framing pattern (odd numbered frame bits) and the superframe alignment pattern (even numbered frame bits). Finding the 193rd bit position was still based on an alternating 1's and 0's pattern, but now it only appeared in every other 193rd bit.

The new technique provided for increased "false frame" protection. The downside of the technique was that the time to reframe was much longer. With the D2 format the maximum average reframe time (MART) would be about 200 milliseconds. This was too much time to be out of service so new algorithms were developed that decreased the time to 50 msec which is now the specification standard. Succeeding channel bank equipment (D3 and D4) used the same framing sequence as D2. In fact, the Superframe Format is most often referred to as the D4 frame format even though it began with D2. This sequence defines a "superframe" consisting of two interleaved patterns. The terminal framing pattern ("F" bit) is a repeating ones and zeros in odd numbered frames and the superframe alignment pattern ("S" bit) is "001110" in the even numbered frames. This results in a 12-bit superframe pattern of:

The D4 Format is shown in Figure 4 below. Notice that the "F" bit and the "S" bit are all called "S bits". While this is confusing, it is a terminology remnant of the time when there were only "S" bits (vis-a-vis D1 format).

As early as 1979, AT&T proposed the Extended Superframe Format be implemented on its T1 circuits in order to provide in-service diagnostic capabilities as well as improved false frame protection. With ESF, the 193rd bit is now time shared by three functions: frame synchronization bits; CRC-6 bits; and Facility Data Link (FDL) bits. Frame synchronization bits are carried in six of the 24 bit positions provided by the 193rd bit. These are in the 4th, 8th, 12th, 16th, 20th, and 24th positions and the pattern is "001011". This simple six-bit pattern performs both the "F bit" and "S bit" functions of the D4 superframe. "False frame" sensitivity is eliminated by using the CRC-6 error checking bits to determine which of several "candidates" for the frame bit are the actual 193rd bit. CRC-6 uses a mathematical algorithm to check the contents of the entire superframe (all 4632 bits) and obtains a 6-bit (hence its name) coded "signature" for those data bits. The FDL may be used for any purpose, but is ideally suited for communicating ESF performance information from local, remote, and intermediate equipment along a facility and for sending control commands for protection switching, network and remote equipment configuration, etc. In essence it is a 4 Kbps channel embedded in the T1 format. Bellcore documement TR-TSY-000194 (Extended Superframe Format Interface Specification - December 1987), ANSI T1.403-1989, and AT&T Publication 54016 describes how this channel may be used. This includes the format of the messages , commands, and responses. Most CSU's today interpret these commands and execute the appropriate responses. The ESF Format is shown is Figure 5.

The chart shown in Figure 6 shows the differences between D1 through ESF formats. As most equipment today is either D4 or ESF, the data for D1 and D2 is displayed only for completeness.