Why a longer, loaded test matters
Most call quality complaints don't happen all day. They happen when the office is busy — ten people on calls while someone uploads a file and a computer downloads updates. A short test on a quiet line misses that. This test adds the bandwidth of your calls, runs long enough to catch bursts, and reports the worst 10 seconds, because that's the moment a customer hears.
How to choose the settings
- Simultaneous calls — count people who could be on the phone at your busiest time, not total phones. A 20-phone office often peaks at 6–10 calls.
- Codec — most business systems use G.711. Use G.729 only if your provider says so.
- Length — 2 minutes is a good first look. Use 5 minutes if problems come and go.
- When — run it during business hours with normal office traffic. That's the point.
Reading the results
| Result | What it means |
|---|---|
| Good MOS 95%+ of the time, worst 10 sec ≥ 3.6 | The line handles this many calls. |
| Good average, bad worst 10 seconds | Intermittent congestion — other traffic is crowding calls. QoS, SQM or a voice-only connection fixes it. |
| Latency climbs steadily during the test | Queues are filling — bufferbloat or not enough upload. |
| "Could not keep up" warning | The connection can't carry this much call traffic plus everything else. Add bandwidth or a second line. |
Frequently asked questions
How many VoIP calls can my office internet handle?
Run the office test above with your expected number of simultaneous calls during business hours. As a rule of thumb, each G.711 call needs about 87 kbps in each direction, and you should keep 25% headroom — but the real limit depends on what else shares the connection, which is why a loaded test is more reliable than math.
Why do our calls go bad only when the office is busy?
When the connection fills up with uploads, video meetings, backups and updates, voice packets wait in line behind them. That adds delay and jitter and can drop packets. QoS that prioritizes voice, Smart Queue Management on the router, or a separate voice connection prevents it.
How long should a VoIP test run?
A 25-second test is fine for a first check. Problems that come and go need longer: 2–5 minutes catches most bursts, and running the test at the times calls go bad matters more than the length.
What is a VoIP hammer test?
A hammer test places many real calls at once to a phone system to find its breaking point — trunk limits, carrier overflow, PBX overload and audio problems. TestCall.com from Infotel Systems runs hammer tests with up to 500 simultaneous calls and a report showing MOS, packets received and lost, and disconnect codes for every call.
Should I count phones or simultaneous calls?
Simultaneous calls. Bandwidth is used only while a call is active, so a 30-phone office that never has more than 8 people on calls at once needs capacity for about 8 calls plus headroom.
Does this test place real phone calls?
No. It sends web traffic at the same bit rate as the calls you choose and measures quality while it does. It tests the internet connection. To test your phone numbers and phone system end to end, use a call-based service such as TestCall.com.
Sources & references
- IETF RFC 3550 — RTP: A Transport Protocol for Real-Time Applications
- ITU-T G.107 — The E-model (R-factor)
- Bufferbloat.net project
- TestCall.com — Infotel Systems voice testing
Written by Eric · Reviewed November 2026