T1 and Fractional T1

A T1 is used as a high speed digital data connection, normally at rates of 1.536 Mb/s.

The T1 (or DS1) was originally developed by the phone company to bundle several telephone conversations between central offices on one digital telephone line. The original T1 specification called for 8000 frames per second, each frame containing one framing bit and 24 channels. Each channel contained a signalling bit and a seven bit digital voice sample. Each voice channel was therefore carried on a 56kb/s (7 * 8000 = 56000) digital data stream. Each frame has a total of 193 bits, so the clock rate of a T1 is 1.544 MHz (193 * 8000 = 1544000).

For the purposes of a data circuit, we are not concerned with voice samples or signalling bits, and unless we are talking about a fractional T1, we don't even care about channels. This means we can use 192 of the 193 bits in a T1 frame, affording us 1.536 Mb/s.

Today there are a couple different kinds of T1's and they can be configured in several different ways. There are differences in framing, ESF allows us to use 8 bits per channel, D4 or SF allow us to use only 7 bits per channel. This translates to 64k or 56k channels. There is also bit encoding, such as B8ZS or AMI. B8ZS is better suited to the needs of data transmission, while AMI is fine for voice. For data purposes, we normally order ESF/B8ZS circuits.

Often, we will choose to use a "fractional" T1, which means that only some of the 24 channels are available for use. This is often a matter of expense. Long distance companies charge per channel for clear-line circuits, frame relay providers charge for different size ports, and Internet providers charge different rates based on the amount of bandwidth available to the customer. The size of a fractional T1 is normally configured at the CSU/DSU, by telling it how many and which ports are to be used.

A CSU/DSU is a device which converts a T1 line into a serial data stream, for use with a synchronous serial port. You can think of it as a modem for a T1, except that there isn't any modulation going on. The acronym actually stands for "Channel Service Unit" and "Data Service Unit".

A T1 is delivered over four wires, two pairs. Each pair completes an electrical circuit used to transmit a constant series of bits (this circuit forms a loop of wire, hence the term "local loop" refering to the part of circuit between the central office and the customer premises). One pair is used to transmit data, the other pair is used to receive data. The transmit and receive functions of the data circuit are actually quite independent. It is possible to send 1.536 Mb/s of data at the same time you are receiving 1.536 Mb/s of data.

A properly configured T1 will go for months without a single bit error. This may seem surprising to someone who is used to connecting through a modem, but T1's are extremely reliable (until someone digs them out of the ground with a backhoe). I say this because one might be tempted to tolerate a small number of errors on their circuit. Low error rates are becoming more and more common as phone companies play with new technologies to stretch the distance between repeaters. The original T1 spec called for a repeater every 1.6 miles, but new technology similar to that used for DSL has allowed T1s to be delivered much longer distances over a single pair of copper.

Loopbacks and Diagnostics

The primary diagnostic tool for a T1 is to place it into a "loopback" state at some point along the data path (a loopback is often referred to simply as a "loop", not to be confused with the local loop). In essense, this connects the transmit and receive pairs together. Thus, anything sent over the transmit pair should be seen coming back over the receive pair. By comparing the data you receive to the data you sent, you can verify the integrity of the data path from the test point to the loopback and back to the test point.

This is normally done by sending a series of test frames. There are about a half dozen standard types of test frames that can be sent. A common practice is to send a stream of one test frame for five minutes, then go on to another test. Over a half hour, the circuit should be thoroughly stress tested.

Loopbacks can be set at a variety of points along the circuit. Most (if not all) CSU/DSUs can be set to loopback. The demarc (the customer premises equipment belonging to the phone company which makes up the point of demarcation) normally can do this. Often the router can do this. Phone switches along the way can do this. By setting the loopback in different places and determining which locations fail, you can narrow down the part of the circuit which is not functioning correctly.

In theory, at least. In theory, the tests should fail on a circuit that is not working. In practice, they usually don't fail. I will never have an explanation for this, unless the people who perform these tests are instructed to lie. You will get a circuit installed. It will have some problem, like a high error rate. The telco will test the circuit, and it will pass with no errors. I think they want you to give up at this point. If you are smart you will insist that a problem exists and make them continue to search for the problem. Eventually they will find the problem and fix it. If, after they have identified the problem, you ask them how the tests could have passed with that problem, they will tell you that the tests should have failed and they don't know why they didn't. It's the darnedest thing, and I've probably seen this happen a hundred times during my career.

T1 vs. DS1, T3 vs. DS3

In the Internet industry, for some reason, you will be looked down on if you talk about a "T3". "You mean a DS3", they will say. I suppose it's a little like refering to beef as cow, and probably comes from resentment against the popular media for attempting to sound knowledgable by throwing the T3 buzzword around (I was appalled by a Nash Bridges episode where they narrowed down their suspect by determining that he had a "T3 line" to his home).

Usually, this happens to me when I'm talking to a salesman. Of course, it annoys me when this happens -- salespeople aren't supposed to pretend to be smarter than me. My normal response is to ask them what the difference is, and their answer is always as vague as their understanding of the conversation they had with one of their own technicians when they learned that there was a difference. "It has to do with the way ..." something or other.

I bring this up because the same distinction applies to T1 versus DS1, though oddly the people who chide me for talking about T3s don't seem to talk about DS1s.

The full name of a T1 is a "T1 carrier", and the name refers to the 193 bit frame, the 1.544 MHz clock frequency, and the collection of things which transmit and repeat the electrical signal from point A to point B.

The term DS1 refers to the payload. Each channel is called a DS0, and has a bandwidth of 8 bits per frame (64 kb/s). The combination of all 24 payload channels yields a DS1, capable of transmitting 1.536 Mb/s of payload data.

So, the difference pretty much boils down to 1 bit per frame, or 8 kb/s. A T2 combines four T1 frames and adds some additional framing (I'm not sure how much). Likewise, a T3 combines seven T2 frames plus some additional overhead.

So a DS1 is normally delivered over a T1 data circuit. A DS3 can be delivered over a T3 data circuit, however this is becoming less common. Often some kind of fiber technology is brought to the customer's premises, which carries an aggregate of the customer's DS3 and other payload for other customers. An example of this is SONET.

Anyway, none of this is terribly important. Sooner or later, you will encounter both terms and wonder what the difference is. As long as you seem to know the difference, you won't be harassed too much.

T1 for Voice

It is worth noting, that T1's are used for other things than data circuits. An ISDN PRI is delivered on a T1. Either T1's or ISDN PRI's can be used for voice traffic. The modem servers we prefer to use will accept either T1 or PRI "trunk lines" (phone lines) for receiving modem connections. When used this way, we benefit from less cable (one 2-pair cable instead of twenty-four 1 pair cables) and digital quality.