IP telephony: the 10 mistakes that ruin call quality, and how to avoid them

“Since we moved to VoIP, calls keep dropping.” When we take over a struggling IP telephony installation, the PBX is rarely the culprit: the network carrying it almost always is. The short answer: start by disabling SIP ALG, prioritising voice with QoS and isolating phones on a dedicated VLAN. These three fixes solve a large share of the problems we encounter. Here are the ten mistakes we find most often, the symptoms that give them away, and how to fix them, whether your PBX is 3CX, Yeastar or another IP PBX.

Why voice is so sensitive to the network

An email arriving a second late bothers no one. A conversation does. Voice over IP cuts speech into small packets sent continuously: if they arrive late (latency), irregularly (jitter) or not at all (loss), the ear hears it immediately as choppy audio, echo or silence. A download that slows down a little is not a problem; a voice packet that arrives too late is lost.

That is why a network that “works fine” for office use can make telephony painful. Diagnosis therefore starts with the symptoms.

Symptom	Likely cause	First check
One-way audio, call dropped after a few seconds	SIP ALG, misconfigured NAT	Disable SIP ALG, check RTP ports
Choppy audio at peak hours	No QoS, saturated link	DSCP 46 marking, link usage
Phones unregistering	SIP ALG, short NAT timeouts, consumer box	PBX logs, NAT settings
Drops on WiFi handsets	Saturated or poorly covered WiFi	Radio survey, voice SSID
Phones rebooting on their own	PoE budget exceeded	Switch PoE consumption
Whole company unreachable	Single internet link down	4G or 5G failover, automatic switching

1. SIP ALG enabled on the router

This feature, meant to help voice cross NAT, rewrites SIP packets and often breaks registrations: one-way audio, ringing that continues after pickup, phones that unregister. Fix: disable SIP ALG on the carrier router and on the firewall, then check NAT rules and RTP ports. It is the first setting we check.

2. No voice prioritisation

Without QoS, a Windows update or a cloud backup crushes voice packets. Fix: mark voice traffic DSCP 46 and prioritise it end to end, on the firewall, switches and WiFi, with about 100 kbit/s reserved per simultaneous G.711 call. Marking honoured by only one device is useless: the whole chain must be consistent.

3. Phones and computers on the same network

Broadcasts, ARP storms, or simply an infected laptop saturating the network, and voice suffers. Fix: a dedicated voice VLAN with its own filtering rules, which also simplifies security, as described in the 10 FortiGate settings after install.

4. The PBX behind a consumer box

Dynamic IP, sloppy NAT, no repair guarantee. Fix: a business link with a fixed IP, or a PBX hosted in a well-connected data centre close to users, and in every case a firewall you control, see business fiber in Morocco. A PBX hosted too far away adds latency to every call.

5. A single internet link

Fiber cut, telephony cut, and customers reach voicemail or a busy tone. Fix: 4G or 5G failover with automatic switching, and voice routed over the most stable link with SD-WAN, as explained in FortiGate SD-WAN for multiple sites. Also plan automatic forwarding to mobiles if the whole site is cut off.

6. WiFi phones on office WiFi

WiFi shared with laptops and smartphones causes jitter and packet loss, especially when moving between access points. Fix: prefer wired for desk phones, and for mobiles a prioritised voice SSID on properly sized access points, with fast roaming, see high-density enterprise WiFi.

7. Insufficient PoE budget

Twenty phones on a PoE switch rated for twelve: the last ones plugged in reboot at random. Fix: add up each handset’s rated consumption, keep a comfortable margin, and choose PoE+ for phones with colour screens or expansion modules. A UPS on the switches also keeps calls going through a brief power cut.

8. The wrong codecs

G.729 saves bandwidth at the cost of quality, G.711 does the opposite, Opus adapts to network conditions. Fix: G.711 or Opus internally and over fiber, G.729 reserved for constrained links. Align codecs between handsets, PBX and carrier: every transcoding step degrades audio and consumes resources.

9. No monitoring

You learn about the problem from customers who can no longer reach you. Fix: monitor the PBX, SIP trunk registration and call quality, measuring jitter, latency and loss, with alerts above 30 ms jitter or 1% loss. We cover the useful indicators in IT monitoring: the KPIs to track.

10. A poorly chosen SIP carrier

A cheap SIP trunk with poor international routes, or no reachable technical support. Fix: a carrier certified for your PBX, with portable Moroccan numbers, responsive support and spending limits to protect against toll fraud.

Our diagnostic method, step by step

1. Collect the symptoms: which phones, what times, internal or external calls, desk or mobile. A problem affecting only external calls points to the link or carrier; a general problem, to the internal network.
2. Review the configuration of the router, firewall and PBX: SIP ALG, NAT, QoS, VLANs, codecs.
3. Measure during calls: traffic captures, jitter, latency and loss, at quiet and peak times.
4. Fix in order: SIP ALG, QoS and voice VLAN first, then the rest according to the measurements.
5. Verify and monitor: new measurements after the fixes, alerts in place to catch any relapse.

Voice quality checklist

• SIP ALG disabled on the carrier box and on the firewall.
• Voice marked DSCP 46 and prioritised across the whole chain.
• Dedicated voice VLAN with its own filtering rules.
• Business link with fixed IP, plus 4G or 5G failover with automatic switching.
• Desk phones wired; dedicated voice SSID for mobiles.
• PoE budget calculated with margin, switches on a UPS.
• Codecs aligned between handsets, PBX and carrier.
• Monitoring of jitter, latency, loss and SIP trunks.
• Certified SIP carrier, with spending caps.

How we do it

A telephony audit measures these ten points on your real network, with traffic captures during problem calls. We give you a clear report: what degrades your calls, in order of priority, and the proposed fixes. We then apply the settings on the firewall, switches and WiFi, and measure again to prove the improvement. In operation, the PBX and call quality are monitored 24/7, and a critical incident is handled in under 15 minutes.

That is the starting point of our IP telephony offer. Choosing the PBX itself is covered in 3CX, Yeastar or Teams Phone, and our case studies show how we have put existing installations back on track. If your calls keep dropping, request your free initial audit: we reply within 24 business hours.