How SIP Trunking Works

The copper that business phone lines have run on for decades is on its way out. In March 2026, the FCC adopted an order streamlining copper network retirement, which gives carriers a shorter path to shutting down legacy lines and moving customers to IP. If your company still has a PRI plugged into the back of a PBX, that circuit won’t be around forever, even if nobody has told you the date.

SIP trunking is what replaces it. This guide explains what SIP trunking for voice applications is, how a call gets from your phone system to the public network and back, and which deployment and pricing choices you’ll face along the way. We’ve kept it to the concepts every project shares: PBX setup, capacity math, signaling internals, and provider selection. Each has its own article, and a table near the end of this article describes them.

1What Is SIP Trunking?

SIP trunking uses SIP trunk technology to connect a phone system to the public switched telephone network (PSTN) over an Internet Protocol (IP) connection instead of physical phone lines. Your PBX, contact center platform, or application speaks SIP (Session Initiation Protocol) to a provider. The provider carries your calls to and from every other phone network. You keep the phone system. The provider replaces the phone company’s circuits.

SIP trunking isn’t a phone system, and it doesn’t give you extensions, voicemail, or ring groups. Those live in the PBX. The trunk is the on-ramp to the outside world, which is a big part of why developers like it: it does one job and stays out of the way of whatever you’ve built on top of it.

Trunks, Channels, and Numbers Are Three Different Things

On the old network, the words aligned with the hardware. A PRI was a physical circuit that carried 23 voice channels, so a company that needed 40 simultaneous calls bought two PRIs and lived with 46. SIP borrowed the vocabulary and dropped the hardware, and the terms have been slippery ever since.

A SIP trunk is the logical connection between your system and the provider. A channel is one concurrent call on that connection. A phone number, or DID, is an address that tells the provider which calls belong to you. None of the three depends on the others: fifty numbers can share a trunk that carries only four calls at once. How many channels you can run depends on your provider’s pricing model and your own bandwidth, and our guide to how many calls a SIP trunk can handle works through the math.

Origination means inbound calls and termination means outbound, named from the carrier’s side of the wire. Providers often price and route the two separately, which the article on VoIP origination and termination explains in detail.

Where the PBX Fits

The PBX is the piece you own. It can be a box in a closet, a virtual machine, an open-source install, or a hosted contact center platform, and the trunk doesn’t much care which. What matters is that it speaks SIP. If it does, you point it at the provider and exchange a few settings. If it predates SIP, a gateway can translate between the old interface and the new one, at the cost of another device to manage.

2How Does a SIP Trunking Call Flow Work?

Every call over a SIP trunk is two conversations happening at once. Signaling is the control conversation: who is calling whom, whether they answered, and when someone hung up. Media is the audio. SIP handles the first, and a separate protocol called RTP handles the second. Keeping them separate in your head makes troubleshooting easier, since a call that connects with no audio and a call that never connects are different failures with different causes.

A call that connects with no audio and a call that never connects are different failures with different causes.

The Steps of an Outbound Call

Here’s what happens, and the codes, when someone on your PBX dials an outside number.

1. INVITE: Your PBX sends the provider an INVITE naming the number dialed. Inside it is a Session Description Protocol (SDP) offer: the codecs your side supports and the address where you want to receive audio.
2. Authentication: The provider confirms the request is yours. It either recognizes your IP address or challenges the INVITE with a 407 response and waits for your PBX to resend it with credentials.
3. 100 Trying: The provider tells your PBX it has the request, then routes the call toward the destination carrier.
4. 180 Ringing: The far end alerts. Some networks send 183 Session Progress instead and play the ringback tone as early media.
5. 200 OK: Someone answered. This response carries the SDP answer, which settles the codec and tells your PBX where to send audio.
6. ACK and media: Your PBX acknowledges the answer, and RTP audio starts flowing in both directions.
7. BYE: Whoever hangs up first sends a BYE, the other side replies 200 OK, and the call is over.

Signaling (SIP)Media (RTP)Dashed arrows appear only with credential-based authentication

Notice that the audio in step 6 doesn’t have to follow the signaling. SIP messages go to the provider’s signaling address, and RTP can travel to a different media address altogether, which is why a firewall can pass one and block the other. When a call connects and one person can’t hear the other, look at the media path first. For a closer look at the two layers, see our article on SIP signaling and media.

Inbound Calls Need One More Decision

An inbound call runs the same exchange in reverse. A caller dials one of your DIDs; the PSTN hands the call to your provider, which then sends an INVITE to your PBX. The open question is how the provider knows where your PBX is.

With SIP registration, your PBX logs in to the provider at regular intervals and says, in effect, “send my calls here.” With host-based routing, you give the provider a fixed IP address or hostname, and the provider sends calls there without waiting for a login. Registration copes with changing IP addresses. Host-based routing removes passwords from the equation and tends to perform better behind fussy firewalls. Go deeper into SIP registration vs host-based routing and when to use each.

3Which SIP Trunking Deployment Fits Your Setup?

There isn’t a single way to deploy a SIP trunk, mostly because there isn’t a single kind of phone system. The trunk looks the same from the provider’s side in every case. What changes is what is on your end of it.

On-Premises PBX

This is the classic migration. You have an IP PBX in a server room that used to connect to a PRI and now connects to a SIP trunk over your internet circuit. Your dial plan, extensions, and handsets stay where they are. The work is mostly on the network: enough bandwidth, a firewall that treats SIP kindly, and a plan for what happens when the circuit drops.

Cloud or Virtualized PBX

Move the same PBX software to a cloud instance, and the trunk follows it. Voice traffic now runs between two data centers instead of over your office connection. This removes the office internet link from the critical path for inbound calls. A cloud host also gives you a stable public address, so host-based routing becomes easy.

Bring Your Own Carrier

Many contact center and unified communications platforms let you connect your own SIP trunk in place of the calling plan they sell. You keep the platform’s features and choose the carrier to use. Teams do this to control numbers, routing, and per-minute cost. It also avoids relying on a single vendor for both software and dial tone.

Legacy PBX With a Gateway

An older digital or analog PBX can still use SIP trunking through a gateway that converts between its native interface and SIP. It works, and it buys time before an expensive upgrade has to happen. You give up some flexibility, and you’ve added a box that can fail, so treat it as a bridge to a replacement.

Whichever shape yours takes, the configuration steps live in our SIP trunking setup guide.

4What Does Your Business Need Before It Migrates?

Most SIP trunking problems that show up after cutover were visible before it. A provider can deliver a clean call to the edge of your network and still watch it fall apart in the last connection. Check these areas before you port a single number.

A Network That Can Carry Voice

Plan on roughly 80 to 100 kbps per concurrent call, in each direction, using the G.711 codec. About 64 kbps of that is the audio itself, and the rest is packet overhead. Multiply by your peak number of simultaneous calls: ten calls at once need roughly 800 kbps to 1 Mbps each way.

G.729 compresses harder and needs about a third of that, with some loss in audio quality. On a modern connection, raw bandwidth is rarely what hurts you. Bad packet management is. Voice packets that get queued behind a large upload arrive late, and late audio is as bad as lost audio, so set Quality of Service (QoS) rules on your router to prioritize voice.

A Firewall That Understands SIP, or Leaves It Alone

Network address translation (NAT) is the classic cause of one-way audio. Your PBX advertises a private address in its SDP, the far end sends audio there, but the packets go nowhere. Many routers ship with a feature called SIP application layer gateway, or SIP ALG, that tries to fix this by rewriting SIP messages in transit, but it often causes the very problems it’s meant to prevent. Turn it off unless your provider tells you otherwise. Whether you also need a session border controller of your own depends on your company size, security policy, and PBX, so work through that question with your provider.

Security Controls You Can Verify

A SIP trunk is a billable service exposed to the internet, and attackers know it. Toll fraud, where someone hijacks your trunk to place expensive international calls, is the threat that costs real money. Ask what the provider gives you to limit damage, such as:

• Authentication options: IP-based authentication, plus the ability to disable outbound SIP credentials entirely
• Signaling encryption: TLS for SIP signaling
• Spending limits: a maximum outbound rate, destination whitelists, and alerts on unusual activity
• Call authentication: STIR/SHAKEN signing, so carriers can verify your caller ID

Emergency Calling and Number Porting

Two administrative jobs can trip up teams. The first is 911. US rules require multi-line phone systems to let users dial 911 directly, notify someone on site, and send a dispatchable location with the call. On a SIP trunk, that means registering an address for each number that can place an emergency call and updating it when people move. Use this as general information and have counsel confirm what applies to you.

The second administrative job is porting. You can bring existing numbers with you, but the transfer runs on the old carrier’s schedule as much as the new one’s. Gather a recent bill and the exact account details early, because a mismatch in the paperwork is the usual reason a carrier rejects a port.

5SIP Trunking, Hosted VoIP, or a Voice API: Which One Do You Need?

People constantly compare these three, but they aren’t really competitors. They work at different layers. SIP trunking is connectivity for a phone system you run. Hosted VoIP is a phone system someone else runs for you, with the connectivity bundled in. A voice API gives your software programmatic control over call routing and number configuration.

For an office that wants working phones and has no one who enjoys reading SIP traces, hosted voice is the right answer, and adding a self-managed PBX onto a team like that creates work without creating value. SIP trunking is effective when you already run the phone system, when voice is part of your product, or when you want control over routing and cost that a bundled seat license won’t give you.

The API question usually isn’t either/or. Plenty of teams run SIP trunks for the calls themselves and use an API to provision numbers and steer inbound routing. Our article The Difference Between SIP Trunking and Voice API lays out the architecture.

6How Do SIP Trunking Cost Models Compare?

Two providers can quote the same per-minute rate and send very different invoices. The reason is the pricing model that defines each call. There are two main models: metered and channel-based.

Channel-based plans are a holdover from the PRI, where capacity came in blocks of 23 channels per PRI. You paid for the block whether you used it or not. They still suit a call floor that’s busy all day. Metered pricing fits the way developers work: You can test without a commitment, a traffic spike doesn’t hit a ceiling, and a quiet month costs less than a busy one.

Two providers can quote the same per-minute rate and send very different invoices.

Whichever model a provider quotes you, the per-minute rate is only one line. Read the rest of the rate sheet:

• Monthly charges per phone number, with toll-free priced separately
• E911 fees per registered number
• Caller ID name lookups on inbound calls
• International rates for the countries you call
• One-time porting charges, if any

A provider that publishes all of these is easier to budget for than one that leads with a low rate and leaves the rest for the contract. Our guide to choosing SIP trunk providers covers what else to compare.

7Where Should You Go Next?

Everything here applies to any SIP trunking project. The specialist articles take over from here: two on how calls work, two on building and sizing a deployment, and three to help you choose an architecture and a vendor. Find the row with the question closest to the concern you have right now.