# Call Does Not Connect

> Diagnose a call that never reaches in_progress: trunk registration, SIP auth, number format, ring timeout, and Q.850 cause codes, with fixes.

The call was placed but it never went live. Your `Call` stayed in `dialing` or
`ringing`, then landed on `no_answer` or `failed`. Audio never bridged. Each
of these outcomes means the same thing: the call **never reached
`in_progress`**. That is the only status that guarantees a live media path.
This page is a symptom → cause → fix map for that window. It covers causes
from "the trunk is not registered" to "the carrier rejected the number". It
also lists the Q.850 cause codes that tell you exactly why.

Start from the outcome. Identify the status the call *ended* on. Then find
that row below. `no_answer` and `failed` have different causes, so the ending
status already cuts the search in half.

> **NOTE:**
> This page is about calls that never connect. If the call reaches `in_progress`
> but you cannot hear anything, that is a media problem, not a signalling
> problem. Go to [one-way audio](/modalities/voice/troubleshooting/one-way-audio).
> If the trunk itself does not come up at all, start at
> [SIP trunk issues](/modalities/voice/troubleshooting/sip-trunk).

## First, read the ending status

Before anything else, look at where the call ended. The
[call lifecycle](/modalities/voice/transport-telephony/call-lifecycle) explains
how each status maps to a SIP moment. For triage, this is the short version:

| Ended on      | What it means                                                        | Look at |
|---------------|----------------------------------------------------------------------|---------|
| `dialing`     | The `INVITE` never got past the gateway. There was no route, or the trunk was down. | Trunk registration, number format, routing. |
| `ringing`     | The far end alerted but never answered.                              | Ring timeout, `no_answer` causes. |
| `no_answer`   | Ring timeout, busy, or unavailable before any answer.               | Ring timeout, Q.850 16/17/18/19. |
| `failed`      | The call was rejected, or media setup failed.                       | SIP auth, ACL, number format, Q.850 1/21/28/34/38. |

`completed` means the call *did* connect and then ended. That is not this
page. `no_answer` and `failed` both mean the call **never connected**, and the
split matters. `failed` is almost always a rejection that you can fix (auth,
format, ACL). `no_answer` is usually the far end (busy, no person there, ring
timeout).

## Symptom → cause → fix

> **NOTE:**
> You do not read raw SIP off the wire in normal use. The gateway speaks SIP for
> you and collapses it into `CallStatus`. The SIP codes below appear in the
> per-call event timeline and CDR (via the `cleared` event's Q.850 cause). Read
> them in [call traces](/modalities/voice/observability/call-traces). Use them
> to find the correct row, not as something you handle in code.

| Symptom | Likely cause | Fix |
|---|---|---|
| The outbound call falls straight to `failed`, with no ring. The CDR shows no route. | **The trunk is not registered, or it is down.** A register-mode trunk lost its registration, so there is no path to the carrier. | Check that the trunk is up and registered on [SIP trunking](/modalities/voice/transport-telephony/sip-trunking). Check the credentials and `register_expires_sec`. IP-authenticated trunks do not register — check that the carrier allow-listed your gateway's source IP instead. |
| `failed` immediately. The trace shows `401`/`407`, then a give-up. | **SIP digest auth fails.** The `sip_username`/`sip_password` is wrong, or the nonce is stale. | Check the trunk's SIP username and password again. They are stored sealed, so enter them again — do not guess. The registrar rejects replayed nonces, so a client that reuses an old challenge also gets `401`. Make sure the trunk authenticates again on each challenge. |
| An inbound call from your carrier is dropped. You never see it. | **The source-IP ACL or spam filter blocks it.** The gateway accepts INVITEs only from configured trunk source IPs. It refuses unknown sources before routing. | Add the carrier's signalling IP to the trunk source-IP list / [SIP ACL](/modalities/voice/troubleshooting/sip-trunk). Trunk INVITEs bypass the registrar ACL, but the source IP still must match a known trunk. |
| `failed` with a `4xx` on the first response. The trace shows `484`/`404`. | **The number format is wrong.** The dialed number is not E.164, or the carrier expects a different digit string. | Dial full **E.164** (`+` and country code, for example `+15551234567`). Some carriers want no `+`, or a specific technical prefix. Set that on the trunk's dial rules. Do not change the number in your app. |
| An inbound INVITE arrives but routes nowhere → `failed`. | **No trunk/DID match, or no route decision.** The gateway could not resolve the trunk (by DID, then by source IP), or the routing engine returned no destination. | Check that the DID is provisioned and mapped to a trunk and an agent/flow — see [number provisioning](/modalities/voice/transport-telephony/number-provisioning). Check that the inbound rule points at a real agent, skill, or VDN. |
| The call rings, then goes to `no_answer` at a fixed number of seconds. | **The ring timeout fired.** No `200 OK` arrived before `ringTimeoutSec` elapsed. | This is the expected result when no person answers. If it fires too early, raise `ringTimeoutSec` (server-clamped **5–120**, default **30**). If the call *always* ends exactly at the timeout, the far end possibly black-holes the INVITE — check the routing and the destination. |
| `no_answer` well before the timeout. The trace shows `486`. | **The far end is busy** (Q.850 17). | There is nothing to fix at the gateway. The callee is on another call. Retry with backoff, or route to voicemail/callback. |
| `no_answer`. The trace shows `480`. | **The callee is unavailable or not registered** (Q.850 18/20). For an agent leg, the SIP phone is not registered. | For carrier legs, retry later. For a human-agent SIP phone, check that it is registered (its AOR is live) — see [PBX & ACD](/modalities/voice/transport-telephony/pbx-acd). |
| The answer occurs (`200 OK`) but the call drops instantly at pickup. | **A `180`-without-SDP carrier teardown, or an SDP/codec mismatch.** Some carriers send an unsolicited `BYE` if early media was not offered correctly. | The gateway answers with **`183 Session Progress` that carries SDP** (not a bare `180`) to prevent this. If you bridge your own carrier and see this, check that your SBC/carrier does not strip the `183`+SDP. See the warning in [call lifecycle](/modalities/voice/transport-telephony/call-lifecycle#media-setup-where-audio-actually-comes-up). |
| Outbound campaign calls fail in bulk, but single calls work. | **Concurrency or pacing limits, or the per-trunk channel cap.** There are too many simultaneous INVITEs for the trunk's licensed channels. | Lower `calls_per_second` / `max_concurrent_calls` on the [paced dialer](/modalities/voice/transport-telephony/outbound). Check the carrier's channel limit. Individual `no_answer`/busy results in a campaign are normal. |

## Q.850 cause codes

When a call clears, the gateway records an **ITU-T Q.850** cause code with the
status. It is the most precise "why" you have. It rides the `cleared`
lifecycle event and the CDR. Read it in
[call traces](/modalities/voice/observability/call-traces). These are the
codes you will see on a call that never connected:

| Q.850 | Name | What it tells you | Typical `CallStatus` |
|------:|------|-------------------|----------------------|
| **1** | Unallocated number | The dialed number does not exist, or the carrier cannot route it. | `failed` |
| **16** | Normal clearing | A clean hang-up. You see this on a call that *did* connect, not on a failure. | `completed` |
| **17** | User busy | The callee is on another call. | `no_answer` |
| **18** | No user responding | The INVITE reached the far end but got no reply (black hole). | `no_answer` |
| **19** | No answer (alerted) | The call rang, and no person answered before the timeout. | `no_answer` |
| **20** | Subscriber absent | The endpoint is not registered, or the device is off (common for agent SIP phones). | `no_answer` |
| **21** | Call rejected | The far end refused the call (`403`/`603`) — often policy or auth. | `failed` |
| **28** | Invalid number format | The number was not in the form the carrier expects (E.164 / prefix). | `failed` |
| **34** | No circuit / channel available | The trunk is out of channels — a concurrency cap. | `failed` |
| **38** | Network out of order | A carrier-side network fault. | `failed` |
| **41** | Temporary failure | A transient carrier issue. It is safe to retry. | `failed` |

> **TIP:**
> **Q.850 17 / 18 / 19 / 20 all mean "the far end did not answer"** and show as
> `no_answer`. Retry logic is the correct response. **Q.850 1 / 21 / 28 / 34**
> are rejections that show as `failed`. A retry of the same call in the same way
> does not help. Fix the number, the auth, the ACL, or the channel cap first.

## Step-by-step: isolate a call that won't connect

  1. **Read the ending status and Q.850 cause**
Pull the call's timeline in
    [call traces](/modalities/voice/observability/call-traces) (or the
    `cleared` event / CDR). Note the terminal `CallStatus` and the Q.850
    cause. That pair alone points you at a row above. `failed` + cause 21/28
    → your side (format or auth). `no_answer` + 17/18/19 → the far end.
  2. **Confirm the call left the building**
If the call never got past `dialing`, the `INVITE` did not route. For
    **outbound**, check that the
    [trunk is registered / IP-authed](/modalities/voice/transport-telephony/sip-trunking)
    and that the trunk resolved. For **inbound**, check that the
    [DID is provisioned](/modalities/voice/transport-telephony/number-provisioning)
    and that the source IP matches a known trunk. The gateway refuses unknown
    sources before routing.
  3. **Check the number format**
Dial **E.164** (`+<country><number>`). A `484`/`404`/`28` almost always
    means the digit string was wrong for that carrier. Set a technical prefix
    or `+`-stripping on the trunk's dial rules. Do not edit the number in your
    app.
  4. **Rule out auth and ACL**
A `401`/`407` loop that ends in a give-up is digest auth. Enter the trunk
    `sip_username` / `sip_password` again. They are sealed — set them again,
    you cannot read them back. An inbound INVITE that vanishes with no
    response is usually the source-IP ACL. Add the carrier's signalling IP.
    See [SIP trunk issues](/modalities/voice/troubleshooting/sip-trunk).
  5. **Right-size the ring timeout**
If the call rings and dies at a fixed second count, that is
    `ringTimeoutSec` (default 30, clamp 5–120). Raise it for destinations that
    answer slowly. If the call hits the timeout *every* time, the far end
    black-holes the INVITE. Go back to routing, not the timeout.
  6. **For campaigns, back off the pacing**
Bulk `failed` results with cause 34 (no channel) mean a concurrency
    ceiling. Lower `calls_per_second` and `max_concurrent_calls` on the
    [paced dialer](/modalities/voice/transport-telephony/outbound) to stay
    under the trunk's licensed channels.

## Setting a sane ring timeout

`ringTimeoutSec` caps the `ringing` stage. If no `200 OK` arrives in time, the
gateway stops the wait, and the call lands on `no_answer` instead of a ring
that never ends. You get a deterministic terminal status that you can branch
on.

  
```ts
import { Voice } from "clutchcall_connect/voice";

const v = new Voice({
  baseUrl: "https://engine.firstbyte.fast",
  apiKey:  process.env.CLUTCHCALL_CREDENTIALS!,
  orgId:   "org_abc",
});

const call = await v.calls.originate({
  to:      "+15557654321",   // E.164 — full country code, no hand-mangling
  from:    "+15551234567",
  trunkId: "trunk_default",
  agent:   "sales-qualifier",
  ringTimeoutSec: 45,        // clamp 5–120; → no_answer if unanswered in 45s
});

console.log(call.status);    // "dialing" — not yet connected

const now = await v.calls.get({ sid: call.sid });
if (now.status === "in_progress") {
  // connected — media is live
} else if (now.status === "no_answer") {
  // rang out / busy — retry with backoff or route to voicemail
} else if (now.status === "failed") {
  // rejected — read the Q.850 cause before retrying the same way
}
```
  
  
```python
from clutchcall.voice import Voice

v = Voice(
    base_url="https://engine.firstbyte.fast",
    api_key=os.environ["CLUTCHCALL_CREDENTIALS"],
    org_id="org_abc",
)

call = v.calls.originate(
    to="+15557654321",         # E.164
    from_="+15551234567",
    trunk_id="trunk_default",
    agent="sales-qualifier",
    ring_timeout_sec=45,       # clamp 5–120; → no_answer if unanswered
)

print(call.status)             # "dialing"

now = v.calls.get(sid=call.sid)
if now.status == "in_progress":
    ...                        # connected
elif now.status == "no_answer":
    ...                        # rang out / busy — retry or voicemail
elif now.status == "failed":
    ...                        # rejected — check Q.850 cause first
```
  

> **WARNING:**
> Do not retry a `failed` call with cause **21 (rejected)**, **28 (bad
> format)**, or **1 (unallocated)** without a change. It fails in the same way.
> Fix the cause first. Cause **34 (no channel)** and **41 (temporary)** *are*
> safe to retry with backoff. Any `no_answer` is also safe to retry with
> backoff.

## Related

  - **[Call lifecycle](/modalities/voice/transport-telephony/call-lifecycle)** — How SIP signalling maps to each CallStatus, and where media comes up.
  - **[SIP trunk issues](/modalities/voice/troubleshooting/sip-trunk)** — When the trunk does not register, authenticate, or route.
  - **[One-way audio](/modalities/voice/troubleshooting/one-way-audio)** — The call connected, but audio flows in one direction only.
  - **[Number provisioning](/modalities/voice/transport-telephony/number-provisioning)** — Provision a DID and map it to a trunk, agent, or flow.
  - **[SIP trunking](/modalities/voice/transport-telephony/sip-trunking)** — Bring your own trunk or use a managed one; registration and IP auth.
  - **[Call traces](/modalities/voice/observability/call-traces)** — Read the per-call event timeline and Q.850 cause after the fact.