Connecting VICIdial to a SIP trunk: the carrier settings that matter

VICIdial runs on Asterisk, so connecting it to a SIP trunk is ordinary Asterisk work behind VICIdial’s admin screens. There are only a handful of settings, and getting each one slightly wrong is what produces the failed calls, the silent calls and the wrong caller ID.

Where the trunk lives

In VICIdial the trunk is a carrier, under Admin → Carriers. One carrier entry holds what would otherwise be split across Asterisk’s SIP configuration and its dialplan, and VICIdial writes it into Asterisk on each server for you. Use one carrier entry per provider route.

The examples below use the classic chan_sip format that most VICIdial installs still run. Substitute your provider’s details for the placeholders.

The registration string

Fill this in only if your provider authenticates by registration. If they authenticate by your server’s IP address, leave it empty — a registration you do not need is one more thing that can fail at three in the morning.

register => youruser:yourpassword@sip.yourprovider.example:5060

The account entry

This is the peer definition — how Asterisk reaches the provider:

[yourtrunk]
type=peer
host=sip.yourprovider.example
port=5060
disallow=all
allow=alaw
allow=ulaw
dtmfmode=rfc2833
insecure=port,invite
qualify=yes
nat=force_rport,comedia
context=trunkinbound

• Put alaw first for UK traffic. G.711 A-law is the UK and European standard; u-law is North American. Getting the order wrong forces transcoding somewhere along the path — more on codecs here.
• dtmfmode=rfc2833 carries keypresses properly through IVRs and is what most providers expect.
• context=trunkinbound is VICIdial’s standard context for inbound calls, so calls to your DIDs land where VICIdial expects them.
• If your provider authenticates by username and password, add defaultuser= and secret= with the values they give you.

The globals string

YOURTRUNK = SIP/yourtrunk

A variable the dialplan refers to. It means switching a campaign to a different route later is one edit rather than a search through the dialplan.

The dialplan entry

exten => _944X.,1,AGI(agi://127.0.0.1:4577/call_log)
exten => _944X.,n,Dial(${YOURTRUNK}/${EXTEN:1},,tTo)
exten => _944X.,n,Hangup()

• The campaign dials with a prefix of 9, followed by the number in international format starting 44. The pattern _944X. matches those calls.
• ${EXTEN:1} strips the leading 9, so the provider receives 44…. Check which format your provider wants — 44…, 0044…, +44… or national 0… — and adjust.
• Keep the AGI line. It is how VICIdial logs the call. Remove it and calls stop appearing properly in VICIdial’s reports, which you will only notice when you need them.
• For the US, the same idea with a pattern such as _91NXXNXXXXXX sends 1… to the provider.

Campaign settings that depend on the trunk

• Dial prefix must match the dialplan pattern — 9 in this example. A mismatch means every call fails to find a route.
• Campaign CallerID is your presentation number. It must be a valid number you hold, and your provider must pass it through unchanged — see which number to show.
• Dial timeout. Give phones time to ring. Cut calls off after a few seconds and you abandon people who were reaching for the phone, and it reads as a poor answer rate.
• Auto-dial level and drop rate are compliance settings as much as productivity ones — the UK rules are here.

Testing before you go live

1. In the Asterisk console, sip show peers should show the trunk as OK. If you register, sip show registry should show it as registered.
2. Manually dial a test number from an agent screen and check audio in both directions.
3. Watch the call in asterisk -rvvv. If it fails, the SIP response tells you why — here is what the common ones mean.
4. Check the caller ID that actually arrives on a mobile.
5. Confirm the call appears in VICIdial’s call reports, which proves the AGI line is working.

The usual failures

• 403 or 401 on every call — your IP is not authorised with the provider, or the credentials are wrong.
• 404 or 484 — the number format is wrong: the prefix was not stripped, or the provider expects a different format.
• One-way audio — almost always NAT. Check the nat setting, your RTP port range on the firewall, and Asterisk’s external address settings.
• Calls that drop at about 30 seconds — usually a SIP acknowledgement not making it back through NAT, so the far end gives up on the call.
• The wrong caller ID — the campaign CallerID is not set, or something between you and the provider is overwriting it.