Why Do My VoIP Calls Sound Choppy? A Field Guide to Fixing Call Quality
Published on August 10, 2026
"The new phones sound like a robot underwater." When VoIP call quality goes bad, it's almost never the phone and rarely the provider's core network — it's one of a half-dozen findable, fixable problems between your handset and the far end. We operate PBX infrastructure carrying hundreds of thousands of calls a month; here are the causes we actually see, in the order we check them.
The three numbers that determine call quality
Voice is real-time. Unlike email or web traffic, packets that arrive late might as well not arrive. Three metrics decide everything:
| Metric | Good | Trouble | What it sounds like |
|---|---|---|---|
| Latency (one-way delay) | < 150 ms | > 250 ms | Talking over each other, walkie-talkie effect |
| Jitter (variation in delay) | < 20 ms | > 40 ms | Robotic, warbled, "underwater" audio |
| Packet loss | < 0.5% | > 1–2% | Choppy audio, dropped syllables, gaps |
Every fix below is really about improving one of these three.
Cause 1: Bandwidth contention (the most common)
A single call needs only ~100 Kbps each way — trivial. The problem is what happens when someone starts a cloud backup, a 4K video stream, or a large file upload on the same connection. Upload saturation is the classic killer: consumer internet plans have a fraction of their download speed as upload, and when the upload pipe fills, your outgoing voice packets queue behind the backup traffic. The far end hears you break up while you hear them fine — that asymmetry is the fingerprint of upload saturation.
Fix: Enable QoS (Quality of Service) on your router and tag voice traffic (SIP and RTP) as highest priority. On business-grade routers this is a checkbox-and-a-half. If your router can't do QoS, that router is costing you more than its replacement would.
Cause 2: Wi-Fi
Wi-Fi adds jitter by design — it's a shared medium with retransmissions. A softphone on a laptop three rooms from the access point will never sound as consistent as a wired desk phone.
Fix: Wire desk phones with Ethernet, always. For mobile apps, expect occasional roughness and keep the important calls on wired or cellular. If you must run phones on Wi-Fi, give voice its own SSID/VLAN with WMM enabled.
Cause 3: The route, not the destination
Sometimes your network is clean and the far-end is clean, but an internet carrier in between is dropping packets. We've diagnosed cases where calls between the U.S. and Asia degraded because an upstream transit provider's route developed persistent packet loss — nothing on either endpoint was wrong, and the fix was getting traffic rerouted onto a different backbone.
Fix (yours): Run mtr or a continuous ping toward your provider's media server during a bad call and save the output. Fix (ours): A provider that runs its own infrastructure can open tickets with upstream carriers, shift media servers, or reroute traffic. This is a real difference between providers that operate their network and providers that resell someone else's.
Cause 4: Media server geography
If your phone system's media servers are on another continent, physics adds latency you can't tune away. Every 1,000 miles is roughly 10–15 ms round trip, before processing. A U.S. office relaying audio through media servers in the wrong region can pick up 200+ ms and constant talk-over. We've fixed "mysterious" choppy audio for customers simply by ensuring their tenant's media path matched where their users and PBX actually were.
Fix: Ask your provider where the media servers handling your calls physically are. The answer should match where you and your callers are.
Cause 5: SIP ALG — the router "feature" that breaks phones
Many routers ship with SIP ALG enabled. It rewrites SIP packets in flight, intending to help NAT traversal, and instead causes one-way audio, dropped calls, and phones that randomly unregister.
Fix: Disable SIP ALG. This is step one of VoIP deployment on almost every router.
Cause 6: Failing hardware and cabling
A flaky Ethernet cable, a dying switch port, or an old PoE switch browning out under load will produce intermittent chaos that looks like a network problem because it is one.
Fix: Swap the cable and port first — it's a 60-second test that resolves a surprising share of "provider problems."
A 10-minute diagnostic you can run yourself
- Does the problem follow one phone or hit everyone? One phone → cable/port/handset. Everyone → router, bandwidth, or upstream.
- Is it one direction? They hear chop, you don't → your upload. You hear chop → your download or the far side.
- Does it correlate with time of day? Business-hours-only degradation is bandwidth contention.
- Wired vs. Wi-Fi — same call, both ways. If wired is clean, you have your answer.
- Run a speed test during a bad call and note the upload number.
What your provider should be doing
Call quality is a shared responsibility: your LAN and internet on one side, the provider's media servers and carrier routes on the other. A good provider monitors per-call statistics (MOS scores, loss, jitter), can tell you which leg of a bad call degraded, and has engineers who will actually pull the packet captures. That's how we run our network at Yondertele — and it's why "it just sounds better" is the review we hear most after migrations. If your current system sounds like a walkie-talkie, talk to us — over a call that sounds like a call.