T1, DS1 & fractional T1

T1 versus DS1

The two terms are used interchangeably in practice, and that is mostly fine, but they describe different layers. T1 is the physical carrier system — the pair of copper pairs and the electrical specification. DS1 is the digital signal that rides on it. You will hear both used to mean "the 1.544 Mbps circuit," and in a migration conversation that is what matters.

Where 1.544 Mbps comes from

A DS1 carries 24 channels, each called a DS0, each running at 64 kbps. That is 1.536 Mbps of payload. The gap up to 1.544 Mbps is the framing overhead: one framing bit added after each round of 24 channel bytes.

The 8,000 samples per second is not arbitrary. It is the Nyquist rate for the 4 kHz channel the analog telephone network was built around, and it is the reason a DS0 is 64 kbps rather than any other number. Every later design inherited it.

Framing: D4/SF versus ESF

Those framing bits carry information, and how they are organized determines what you can see about the circuit.

• D4 / Superframe (SF) groups 12 frames. All 12 framing bits are used for frame alignment and signaling. There is no in-service error monitoring — to test a D4 circuit you generally have to take it out of service.
• ESF (Extended Superframe) groups 24 frames and divides the framing bits three ways: 2 kbps of frame alignment, 2 kbps of CRC-6 error checking, and 4 kbps of Facility Data Link. The CRC lets you detect errors while the circuit is carrying live traffic, and the FDL gives the carrier a management channel.

Line coding: AMI versus B8ZS

T1 receivers recover their clock from the data stream, which requires a minimum density of pulses. A long run of zeros starves the clock and the circuit loses framing.

• AMI (Alternate Mark Inversion) is the original scheme. It has no mechanism to guarantee ones density on its own, so it is paired with a constraint — commonly bit-7 stuffing, which forces a one into the seventh bit position of an all-zero byte. That works for voice, where corrupting one bit of a PCM sample is inaudible, and reduces a data channel to 56 kbps.
• B8ZS (Bipolar with 8-Zero Substitution) replaces a run of eight zeros with a deliberate bipolar violation pattern that the far end recognizes and reverses. It preserves all eight bits, giving a genuine clear channel 64 kbps DS0.

This is the origin of the 56k-versus-64k distinction that still appears in older documentation. If you inherit a circuit inventory listing 56 kbps channels, you are looking at AMI with bit stuffing, and the circuit is old.

Robbed-bit signaling

On channel-associated signaling (CAS) T1s, supervision — on-hook, off-hook, ringing — is carried in-band by "robbing" the least significant bit of each channel's sample in designated frames. In D4 that is frames 6 and 12; in ESF, frames 6, 12, 18 and 24.

The signaling states this produces are the familiar ones: loop start, ground start, E&M wink start, and immediate start. The important consequence for migration is that robbed-bit signaling is why a CAS T1 cannot carry clean 64 kbps data, and why the industry moved to ISDN PRI, which puts signaling on its own channel. See ISDN PRI & BRI.

Fractional T1

A fractional T1 is a subscription to a subset of the 24 channels rather than the whole 1.544 Mbps. A 256 kbps fractional service is four DS0s; 512 kbps is eight.

The detail that matters operationally: the service provider assigns specific timeslots. You are not given "any four channels" — you are given, for example, timeslots 1 through 4, and your CSU/DSU must be configured to use exactly those. A mismatch between the timeslots the carrier provisioned and the timeslots your equipment is configured for produces a circuit that comes up cleanly and passes no traffic, which is one of the more frustrating faults to chase.

The full T1 is physically delivered either way. Fractional service is a billing and provisioning arrangement, not a different circuit. That has a practical implication during migration: growing from fractional to full is a provisioning change, not an installation.

Channelized, unchannelized, and the wider hierarchy

• Channelized — the 24 DS0s are individually addressable, which is what you want for voice.
• Unchannelized (clear channel) — the full 1.536 Mbps presented as a single pipe, used for data.

Reading the alarms

When a T1 fails, the alarm tells you which direction the problem is in — which is the fastest way to work out whose fault it is.

A yellow alarm at your end and a red alarm at theirs is the classic one-direction failure. Both ends red is usually a framing or line-coding configuration mismatch.

What carries over to SIP, and what doesn't

That last row is the one that catches teams out. On a T1, the circuit either works or it does not, and quality is not really a variable. On SIP, everything can be nominally operational while customers report that calls sound terrible. Migration plans need a measurable quality target from the start.

Related reading

• ISDN PRI & BRI — what happens when you give signaling its own channel.
• TDM to IP migration — sizing and sequencing the replacement.
• Voice quality: MOS, R-factor & POLQA — how to set a quality target you can actually measure.