One-Way Audio and Dropped Calls: NAT, SIP ALG and Firewall Fixes
One-way audio, where you can hear the caller but they can't hear you, and calls that drop after half a minute are two of the most common VoIP faults. The phone system is rarely to blame. The cause is almost always the network in between: NAT, a SIP ALG on the router or a firewall rule. This guide explains why these break calls, which settings to change and how to find the fault step by step.
Symptoms and their usual causes
Start by describing the fault precisely. The pattern usually points straight at the cause.
| Symptom | Most likely cause | Check first |
|---|---|---|
| One side hears nothing | Audio is sent to a private or wrong address, or blocked on the way in | SIP ALG, the PBX's NAT settings, RTP port rules |
| No audio in either direction | RTP blocked both ways, or the PBX advertises the wrong public address | Firewall rules for the RTP range, external address setting |
| Call drops about 30 to 32 seconds after answer | The SIP ACK confirming the answer never arrives | SIP ALG, private IP in the Contact header |
| Call drops at a fixed time, such as 15 or 30 minutes | A session timer refresh fails, or a NAT mapping times out | Session timer settings, firewall UDP timeouts |
| Inbound calls stop arriving after a few idle minutes; outbound works | The NAT mapping created by the registration has expired | Registration interval, keep-alives |
| Audio stops after hold or transfer | The new media path set up by a re-INVITE is blocked or rewritten | SIP ALG, direct media settings |
| Fine in the office, broken from home | The remote phone sits behind a second NAT, often with SIP ALG on | Use the vendor's tunnel, an SBC or a VPN |
Choppy, robotic or delayed audio is a different problem, caused by jitter, latency or packet loss. See fixing VoIP call quality for that.
Why NAT breaks SIP calls
NAT (network address translation) lets every device in your office share one public IP address. Your router swaps private addresses such as 192.168.1.20 for its public one on the way out, and remembers the mapping so replies can come back.
VoIP uses two kinds of traffic. SIP is the signaling that sets up and ends the call. RTP carries the audio itself, on separate UDP ports chosen for each call. The trouble is that SIP messages carry addresses inside them: the SDP body (session description) states the IP address and port where each side wants to receive audio, and the Contact header says where to send later messages such as the ACK and BYE. NAT rewrites the packet headers but not these addresses inside the message.
So a PBX or phone that doesn't know it's behind NAT tells your provider to send audio to 192.168.1.20. That address means nothing on the internet, the audio disappears, and you get one-way or no audio.
Why calls drop at 32 seconds
When a call is answered, the answering side sends a 200 OK and waits for an ACK in return, retransmitting the 200 OK until one arrives. Under RFC 3261, if no ACK arrives within 64 times the T1 timer (500 ms by default), or 32 seconds, the session should be ended with a BYE. If the ACK is sent to a private address or blocked by the firewall, the call dies about half a minute after it was answered. In a SIP trace, you'll see the 200 OK repeated with no ACK.
Fix 1: Disable SIP ALG
A SIP ALG (application layer gateway), sometimes labeled SIP helper, SIP fixup or SIP passthrough, is a router feature that tries to rewrite the addresses inside SIP messages to fix NAT. Many implementations get it wrong, and they clash with PBXs and providers that already handle NAT themselves. The IETF's NAT behavior requirements, RFC 4787, say ALGs for UDP-based protocols should be turned off, and 3CX's firewall guide tells you to use a router or firewall on which SIP ALG can be disabled.
- On most small business routers it's a single checkbox under NAT, firewall or advanced settings.
- Some business firewalls handle SIP through a VoIP profile or session helper, and may need more than one change. Check the vendor's documentation for your model.
- Some ISP routers hide the setting. Ask the ISP to disable it, or put their router in bridge mode and use your own firewall.
After the change, reboot the router or clear its connection table so old mappings are removed, then re-register the phones and trunk. Skipping this step is why many people conclude "it didn't help".
Fix 2: Tell the PBX it's behind NAT
An on-premise PBX must put its public address in SIP and SDP when it talks to the outside world, and its private address when it talks to phones on the LAN. That depends on two settings: its public IP or hostname, and the list of networks it should treat as local.
- Asterisk and FreePBX: the Asterisk NAT guide names three PJSIP transport options,
local_net,external_media_addressandexternal_signaling_address, plusdirect_mediaon the endpoint, which decides whether audio may flow directly between phones. FreePBX shows the same settings in its Asterisk SIP Settings page. - 3CX: discovers its public IP through a 3CX service and uses an FQDN. Its guide also asks for split DNS (hairpin NAT) so the same name works inside and outside the office.
- Other platforms: look for a NAT section with fields for the public IP or domain and the local networks.
Watch for double NAT: an ISP router in front of your own firewall, both translating addresses. The PBX must advertise the outermost public IP, and ports must be forwarded through both devices. Putting the ISP router in bridge mode removes the problem entirely.
Fix 3: Open the right ports, and only those
An on-premise PBX needs its SIP port and its RTP range reachable from your provider. As an example, 3CX's guide lists:
- 5060 UDP and 5060 to 5061 TCP for SIP
- 9000 to 10999 UDP for RTP audio, noting that each call uses two RTP ports, so you need twice as many ports as simultaneous calls
- 5090 UDP and TCP for the tunnel used by 3CX's remote apps
The forwarded range must match the range configured on the PBX exactly. A firewall that forwards 10000 to 20000 to a PBX using 9000 to 10999 produces the most confusing fault of all: audio that works on some calls and not on others.
Then lock it down. Allow SIP and RTP only from your SIP provider's published IP addresses where you can. A SIP port open to the whole internet is found by automated scanners that try to guess extension passwords, a common starting point for toll fraud; our guide to preventing toll fraud and SIP attacks covers the rest of that checklist. With a hosted PBX, phones register outbound and you normally forward nothing, but SIP ALG and UDP timeouts still matter. After any change, run your PBX's firewall test if it has one; 3CX includes a Firewall Checker.
Fix 4: Keep NAT mappings alive
When a phone or trunk registers, the router creates a mapping so the PBX or provider can reach it with incoming calls. If no packets pass for a while, the router deletes it, and inbound calls fail until the next registration. RFC 4787 requires UDP mappings to last at least two minutes and recommends five or more, but many routers use shorter timers.
- Set the re-registration interval below the router's UDP timeout.
- Turn on keep-alives (SIP OPTIONS or small NAT keep-alive packets) on phones and trunks, typically every 20 to 30 seconds.
- Raise the UDP timeout for SIP traffic on the firewall, if it allows that.
- Consider SIP over TCP or TLS, which keeps a connection open, if your provider supports it.
Fix 5: Session timers and drops at fixed intervals
Session timers, defined in RFC 4028, let one side refresh an active call periodically to prove it's still alive. The RFC recommends a 30-minute interval, with the refresh sent halfway through. If a refresh times out, the sender hangs up; if no refresh arrives at all, the other side sends a BYE shortly before the interval expires. A call that dies at the same point every time, often around 15 or 30 minutes, is the signature.
The refresh usually fails for the same reasons as the ACK: it goes to the address in the Contact header, so a wrong NAT setting or a SIP ALG breaks it. Fix the NAT problem first. If you need a quick test, align or disable session timers on the trunk and see whether the drops stop.
Remote phones and home workers
A desk phone at someone's home is behind their router, often with SIP ALG switched on. Exposing SIP to the internet so it can register is the riskiest fix. Better options:
- The vendor's app or tunnel: 3CX recommends an SBC (session border controller) or a router phone in front of remote IP phones, or its apps, which have a tunnel built in.
- A remote access service: Yeastar offers remote calling without port forwarding; see Yeastar Linkus Cloud Service.
- A site-to-site or client VPN: the phone then behaves as if it were on the office LAN.
A step-by-step diagnosis
- Pin down the pattern: who can't hear whom, inbound or outbound, how many seconds until the drop, which phones, trunks and locations.
- Make an internal call between two phones on the same LAN. If it works, the fault lies between the PBX and the outside world.
- Capture the SIP traffic on the PBX, using tcpdump or sngrep on Linux-based systems or the PBX's built-in capture tool.
- Read the SDP: check the
c=line (the audio address) and them=audioport in messages your PBX sends out. A private address heading to your provider means a NAT setting is wrong. If the message leaving the PBX is correct but the provider receives something different, the router is rewriting it. - Check the RTP: confirm audio packets flow in both directions. If they leave but never arrive, look at the firewall.
- Look for 200 OK retransmissions with no ACK: that's the 32-second drop.
- Change one thing at a time, clear the router's sessions and retest.
Common mistakes
- Disabling SIP ALG on your firewall but not on the ISP router in front of it.
- Blaming the provider before taking a packet capture. Their support team will need one to investigate, and it often shows the fault is local.
- Blaming the codec: a codec mismatch normally makes the call fail outright (SIP 488 Not Acceptable Here) rather than producing one-way audio.
- Adding QoS to fix missing audio. QoS for VoIP helps choppy audio on a busy link, not audio that's going to the wrong address.
For how these pieces fit into a complete phone system, see our business VoIP phone system guide, and SIP trunking explained for how providers connect to your PBX.
How BtoBSkills can help
Our VoIP support and configuration service covers trunk and extension setup and troubleshooting of technical issues like these, on 3CX, Yeastar, FreePBX and other Asterisk-based systems. Describe the symptom and your network setup (router or firewall model, on-premise or hosted PBX), and we'll work through the NAT, SIP ALG and firewall settings with you. Request a quote to get started.
Frequently asked questions
Why can I hear the caller but they can't hear me?
Does SIP over TLS avoid SIP ALG problems?
Does a VPN fix one-way audio for remote users?
Can I just forward all UDP ports to the PBX?
Is one-way audio my SIP provider's fault?
Related services
-
VoIP Support Configuration $200.00
Related articles
- Fixing VoIP Call Quality: Jitter, Latency and Packet Loss
Target values for latency, jitter and packet loss, a symptom-to-cause table, a step-by-step troubleshooting method and the fixes that solve most VoIP quality problems.
- VoIP Security: How to Prevent Toll Fraud and SIP Attacks
How attackers use hacked phone systems to run up call charges, the controls that stop them on 3CX, Yeastar and FreePBX, and what to do if you're hit.
- QoS for VoIP: How to Prioritize Voice on Your Network
Where voice actually gets delayed, which DSCP markings to use, how to set priority queues and shaping on your router, and a simple test to prove it works.
- Business VoIP Phone Systems: The Complete Guide for SMEs
How a business VoIP phone system works, the deployment and licensing choices you'll face, what your network needs, and how to switch without losing calls or numbers.
- SIP Trunking Explained: How It Replaces Traditional Phone Lines
What a SIP trunk is, how channels, numbers and authentication work, what it replaces, how providers package it and what to check before you sign.
Sources
- IETF RFC 3261: SIP: Session Initiation Protocol (section 13.3.1.4, 2xx retransmission and ACK) (accessed 2026-11-19)
- IETF RFC 4787: NAT Behavioral Requirements for Unicast UDP (REQ-5 mapping timer, REQ-10 ALGs) (accessed 2026-11-19)
- IETF RFC 4028: Session Timers in SIP (accessed 2026-11-19)
- 3CX: Firewall & Router Configuration (ports, SIP ALG, Firewall Checker) (accessed 2026-11-19)
- Asterisk Documentation: Configuring res_pjsip to work through NAT (accessed 2026-11-19)

