SIP Trunk vs PRI: Channels, Bandwidth, and When to Switch

Quick answer

"SIP trunk vs PRI" is a capacity and transport comparison. PRI buys fixed digital channels on a TDM circuit; SIP buys concurrent sessions over IP. Wikipedia's PRI summary and carrier planning docs agree on the channel math: T1 PRI = 23B+D, E1 PRI = 30B+D [Primary Rate Interface, checked 2026-10-21].

ATT's SIP bandwidth tip converts peak concurrent call paths into IP bandwidth and notes one SIP trunk can replace multiple PRIs when the WAN link is sized correctly [ATT SIP trunk bandwidth tips].

Side-by-side comparison (C1)

ATT's worked example: 200 peak calls → ~16 Mbps G.711 or ~6.4 Mbps G.729a on the IP port; the same peak on PRI would need nine T1 PRIs (9×23). Treat those kbps figures as planning floors, then add headroom for data CoS and bursts.

Decision rule

Stay on PRI if the carrier still mandates TDM for a site, elevators/alarms need copper, or your PBX has no certified SIP/SBC path. Move to SIP when peak CCPs are climbing in steps of 23, WAN QoS is proven, and DID porting plus emergency-address records are scheduled.

Hybrid is normal: SIP for overflow and new sites, one PRI retained for failover. Measure MOS/jitter after cutover; if voice quality collapses under load, fix QoS before canceling the last PRI. Related quality symptoms are covered in our VoIP jitter guide after publish.

Channel math you can defend in a budget meeting

Primary Rate Interface (PRI) is a TDM access that presents a fixed number of bearer channels plus a signaling channel. On a North American T1 PRI you typically get 23 bearer (B) channels and one signaling (D) channel; on an E1 PRI you typically get 30 B plus one D. That channel count is a physical ceiling until you add another PRI circuit. [Source: Wikipedia Primary Rate Interface overview, checked with this article's prior review notes on 2026-10-21]

SIP trunks instead present concurrent-call capacity as a logical limit enforced by license, SBC, or provider policy. You can often grow from 10 to 50 sessions without truck-rolling a second copper span, but you must size Internet or private IP bandwidth for the codec and packetization you actually use. G.711 at 20 ms frames needs more bits than a compressed codec; HD voice and video raise the bar again.

A practical planning sequence: (1) peak concurrent calls from the busiest week's CDR, (2) codec and packetization from the PBX dial-plan, (3) RTP bandwidth plus overhead from your carrier or SBC sizing note, (4) headroom for bursts and retries. Do not reuse a "80 kbps per call" rule of thumb if your vendor published a different worksheet for your software release.

Operational differences beyond the price quote

PRI failures are often binary: the span is up or down, and the carrier NOC speaks TDM language (LOS, AIS, slips). SIP failures are layered: DNS, TLS certificates, SBC ACLs, NAT timers, and the ISP path can each break dial tone while the office Wi-Fi still "works." Your runbook must include certificate expiry and SBC failover drills that PRI sites never needed.

Number porting, caller-ID presentation, and emergency-location handling also change shape. SIP makes it easier to host numbers in a cloud PBX, but you must confirm how the provider handles E911/MLTS obligations in your jurisdiction. PRI tied to a physical address sometimes made location obvious; SIP geo diversity can obscure it if admin data is stale.

Hybrid periods are common: keep PRI on one site while new branches open on SIP. During hybrid, train the help desk to ask "PRI or SIP path?" before troubleshooting, because jitter and one-way audio exist on SIP and not on the TDM bearer. See VoIP jitter for the IP-side checklist.

Migration checklist (PRI → SIP)

Inventory every DID, hunt group, fax device, alarm panel, and elevator line still living on copper. Fax and analog adapters are the usual landmines. Decide whether fax moves to an AT fax service, stays on a residual analog line, or rides T.38 on the SIP trunk with a tested ATA.

Stand up the SIP trunk in parallel, verify inbound/outbound to a pilot user group, then port numbers in waves. Keep the PRI warm until inbound porting completes and emergency-test calls succeed from each floor plan location your policy requires. Document the rollback: re-point the PBX route partition to PRI if SIP call completion drops below your acceptance rate.

Contract language to review: concurrent-call overage pricing, codec restrictions, fraud liability, SLA credits, and whether the quote assumes customer-provided Internet or a managed private circuit. A cheap SIP trunk on an unmanaged cable modem without QoS can erase the savings via failed sales calls.

Decision matrix (short)

Choose or keep PRI when you need predictable channel math, have analog/fax devices that already work on copper, or lack managed IP with QoS. Choose SIP when you need elastic capacity, multi-site dial plans, faster number provisioning, or cloud PBX hosting, and you can fund SBC/firewall and monitoring work.

Choose hybrid when headquarters still sits on PRI while branches open on SIP, or when regulators or landlords still terminate copper for elevators and fire panels. Budget the dual-run period explicitly; "temporary" hybrids that linger three years are common.

Whatever you pick, write acceptance tests in concurrent-call and completion-rate language, not marketing adjectives. Tie SIP tests to the same CDR and alarm sources operations already watches.

Cost model worksheet

Build a three-year model with rows for: PRI circuit fees, PRI channel fees, SIP session fees, SBC appliances or cloud SBC, firewall upgrades, engineer hours for cutover, fax mitigation, and estimated lost-call cost if QoS is deferred. Many SIP quotes look cheaper until engineer hours and fax are included.

Sensitivity-test concurrent-call growth at +20% and +50%. PRI requires another span at the ceiling; SIP usually requires a contract change and bandwidth. That asymmetry is the strategic reason to prefer SIP for growing contact centers when the IP path is engineered.

Include a line for monitoring: MOS/jitter dashboards are not optional after PRI goes away. Budget tooling beside the trunk.

Revisit the model annually when carrier renewals arrive. Trunking markets move; a decision that was correct in 2023 may be wrong on price in 2026 even if architecture remains sound.

Keep a live inventory of which sites still terminate PRI copper for elevators, fax, or alarms so the SIP project does not strand life-safety lines.

After cutover, schedule a thirty-day hypercare window with daily trunk completion-rate reports before you cancel the PRI circuits.

Record the SBC software version and certificate expiry dates in the same runbook as the trunk credentials so renewals do not become outages.

FAQ