Search the blog
With the analogue phone network closing on 31 January 2027, every business and most homes will soon make their calls over broadband. The question we are asked most often is "is our line fast enough?" and the honest answer is that speed is rarely the problem. A voice call uses less bandwidth than a single web page. What breaks calls is timing: packets arriving late, arriving unevenly, or not arriving at all.
This guide explains the four numbers that matter, what the main platforms say they need, how to test your own line properly, and what to do if it fails. It is the practical companion to our landline switch-off checklist and our VoIP and switch-over page.
How much bandwidth a call really uses
Traditional telephony digitises speech as 8,000 samples a second at 8 bits each, the G.711 coding that internet telephony inherited from the phone network (IETF, 2003). Multiply those out and you get 64 kbit/s of speech data. The audio is sent in packets every 20 milliseconds (IETF, 2003), and each packet carries its own IP, UDP and RTP headers, so on the wire a G.711 call is nearer 80 kbit/s in each direction.
Modern platforms squeeze harder. The Opus codec used by most app-based calling scales "from low bitrate narrowband speech at 6 kbit/s" upwards, with 16–20 kbit/s recommended for wideband speech (IETF, 2012). Microsoft lists 58 kbit/s up and down as the recommended rate for a Teams audio call, with a bare minimum of 10 kbit/s and 76 kbit/s for best performance (Microsoft, 2026a). Zoom quotes 60–80 kbit/s for meeting audio and 60–100 kbit/s for a Zoom Phone call (Zoom, 2026).
So a 20-person office with every handset in use at once needs about 2 Mbit/s each way for voice. An ordinary full-fibre line has that many times over. If your line fails a VoIP test, bandwidth is almost never the reason.
| Activity | Microsoft Teams (recommended) | Zoom |
|---|---|---|
| Audio call | 58 / 58 | 60–80 |
| Phone call over the platform | 58 / 58 | 60–100 (Zoom Phone) |
| One-to-one video | 1,500 / 1,500 | 600 (high quality); 1,200 (720p) |
| Group video meeting | 2,500 / 4,000 | 1,000 / 600 (high quality); 2,600 / 1,800 (720p) |
| Screen sharing in a meeting | 2,500 / 2,500 | 50–150 |
The four numbers that decide call quality
Bandwidth is the width of the pipe, and for voice it is rarely the constraint. The other three describe how reliably packets travel through it.
Latency is the time a packet takes to get from you to the other end. Microsoft's connectivity test fails a line unless the measured latency to the Teams service is "lower than 100ms" (Microsoft, 2026b). The international telephony standard, ITU-T G.114, says a one-way delay of 400 ms "should not be exceeded for general network planning" but warns that highly interactive tasks such as voice calls "can be affected by much lower delays" (ITU-T, 2003). Once the delay is long enough for people to talk over each other, the call feels broken even if every word is clear.
Jitter is the variation in that delay. Voice packets leave every 20 ms; if they arrive bunched up and then spread out, the receiving phone holds them in a buffer to smooth them, and when the variation exceeds the buffer you hear syllables drop out. Microsoft's test requires jitter "lower than 30ms" (Microsoft, 2026b).
Packet loss is the share of packets that never arrive. Voice runs over UDP, which does not resend lost data; there is no time. Microsoft's pass mark is packet loss "lower than 1.00%" over a ten-second test call (Microsoft, 2026b). Its network guidance describes the symptom well: calls have "static and cut out, or voices sound like robots" (Microsoft, 2026a).
| Metric | Pass |
|---|---|
| Latency (UDP, client to Teams front door) | under 100 ms |
| Jitter | under 30 ms |
| Packet loss (10-second audio test) | under 1% |
Upload matters more than download
Broadband packages are sold on the download figure, and on older copper lines the upload can be a small fraction of it. Voice is symmetrical: you send as much as you receive. Video is worse for the person on camera. Notice in Table 1 that Zoom's "high quality" group meeting needs 1,000 kbit/s up but only 600 kbit/s down (Zoom, 2026). When someone reports that "video is fine until I turn my camera on", the upload is usually the culprit, and no download upgrade will fix it. When you look at a package, look at the second number.
How to test properly
A single speed test at nine in the morning proves very little. The measurements that matter are latency, jitter and loss, taken when the line is busy.
- Test from a wired computer. Plug a laptop into the router with an Ethernet cable so you are measuring the line, not your Wi-Fi.
- Use a test that reports jitter and loss. Microsoft's free connectivity test at connectivity.m365.cloud.microsoft runs a ten-second UDP audio test and reports all three figures against the thresholds in Table 2 (Microsoft, 2026b). Most VoIP providers offer an equivalent.
- Test at your busy time. Run it several times across a working day, including when the office is backing up or on video calls together. Keep the worst results, not the best.
- Check the upload. Compare the upload figure with the totals in Table 1 for the number of simultaneous calls you expect.
- Test the Wi-Fi separately. Repeat from where people actually sit. If the wired test passes and the wireless one fails, the fix is inside the building.
Ofcom's checker at checker.ofcom.org.uk will also tell you which technologies are available at your postcode, which matters for what follows.
Wi-Fi, cables and QoS
Desk phones belong on a cable. Microsoft's guidance is blunt that Wi-Fi networks "aren't necessarily designed or configured to support real-time media", and recommends the 5 GHz band over 2.4 GHz, band steering so devices prefer it, and non-overlapping channels on neighbouring access points (Microsoft, 2026a). Most modern handsets take power and data over the same Ethernet cable, so the desk needs one socket and no plug.
Quality of Service (QoS) is the router setting that lets voice jump the queue. Microsoft recommends it "on all segments of a managed network" and notes that even a well-provisioned line benefits, because it protects calls from "unanticipated network events" such as someone starting a large upload (Microsoft, 2026a). On a small business router this is usually a single option to prioritise voice or particular devices. It only works inside your building, since the public internet ignores the markings: QoS cannot rescue a bad line, but it stops a good line being ruined by your own traffic.
Two other points from the same guidance bear repeating. Do not route calls through a VPN unless it is split-tunnelled, because VPNs add latency and often handle UDP badly; and make sure the firewall does not block or rewrite UDP, because Teams then falls back to TCP and "audio and video will be impaired" (Microsoft, 2026a; 2026b).
Where the UK stands on full fibre
The reason all this is getting easier is the fibre roll-out. Ofcom's Spring 2026 update reports full fibre available to 24.9 million UK homes, 82% of the total, with gigabit-capable broadband of any kind at 89%. Take-up lags availability: 12.4 million premises are actually connected, 47% of those that could be (Ofcom, 2026). On mobile, 4G reaches 96% of the UK landmass from at least one operator, and 5G coverage outside premises ranges from 76% to 94% depending on operator (Ofcom, 2026).
For most businesses in and around London the question is no longer "can we get fibre?" but "have we ordered it?".
What to do if your line fails the test
| What the test showed | Likely cause | What to do |
|---|---|---|
| Wired test passes, Wi-Fi test fails | Building Wi-Fi | Cable the desk phones; move to 5 GHz; add or reposition access points |
| Passes when quiet, fails at busy times | Your own traffic saturating the line | Enable QoS on the router; schedule backups outside hours; consider a second line for voice |
| High loss or jitter at all times on copper broadband | Line quality or exchange congestion | Ask the provider to test the line; order full fibre if the checker shows it is available |
| Full fibre not available | Location | Order SoGEA (copper broadband with no phone line) as an interim; use 4G/5G as backup or primary |
| Everything passes but calls still poor | Handsets, headsets, firewall or VPN | Check UDP is open, split-tunnel the VPN, replace consumer-grade kit |
Two of those options need a word of explanation. SoGEA is the broadband-only product Openreach sells over its existing fibre-to-the-cabinet network with no analogue phone line attached; Openreach says it is available to 98.5% of UK premises (Openreach, 2025). It is the sensible interim where full fibre has not yet arrived, and it is what your existing copper broadband becomes anyway once the phone part is withdrawn. 4G or 5G backup is a small router with a SIM that takes over if the main line fails. At the coverage levels Ofcom reports it is a realistic fallback for a phone system in most of the country, and for a very small office it can be the primary connection.
What this means for you
A voice call needs about 100 kbit/s in each direction and almost any broadband sold today has that. What it also needs is latency under 100 ms, jitter under 30 ms and loss under 1%, measured during your busy hour from a cable, not from the sofa. Test before you choose a phone system, not after, and test the upload as carefully as the download.
If the line passes, put the desk phones on cables, switch on QoS, and move on. If it fails, the fix is usually full fibre, a second line or sorting out the Wi-Fi, and the cost of finding out is an afternoon.
We run this test as the first step of every VoIP switch-over, for businesses and home users alike, and we say plainly if the line is not good enough before anyone signs anything. If you would like the results interpreted rather than just a screenful of numbers, get in touch.
Sources
- IETF (2003) RFC 3551: RTP Profile for Audio and Video Conferences with Minimal Control. Internet Engineering Task Force. https://www.rfc-editor.org/rfc/rfc3551.html (accessed 25 August 2026).
- IETF (2012) RFC 6716: Definition of the Opus Audio Codec. Internet Engineering Task Force. https://www.rfc-editor.org/rfc/rfc6716.html (accessed 25 August 2026).
- ITU-T (2003) Recommendation G.114: One-way transmission time — Summary. International Telecommunication Union. https://www.itu.int/dms_pubrec/itu-t/rec/g/T-REC-G.114-200305-I!!SUM-HTM-E.htm (accessed 25 August 2026).
- Microsoft (2026a) Prepare your organization's network for Microsoft Teams. Microsoft Learn. https://learn.microsoft.com/en-us/microsoftteams/prepare-network (accessed 25 August 2026).
- Microsoft (2026b) Microsoft 365 network connectivity test tool. Microsoft Learn. https://learn.microsoft.com/en-us/microsoft-365/enterprise/office-365-network-mac-perf-onboarding-tool (accessed 25 August 2026).
- Ofcom (2026) Connected Nations update: Spring 2026. Office of Communications, 13 May 2026. https://www.ofcom.org.uk/phones-and-broadband/coverage-and-speeds/connected-nations-update-spring-2026 (accessed 25 August 2026).
- Openreach (2025) Openreach announces price changes to encourage digital adoption of newer, more reliable and better value technology. Press release, 12 November 2025. https://www.openreach.com/news/openreach-announces-price-changes-to-encourage-digital-adoption-of-newer-more-reliable-and-better-value-technology/ (accessed 25 August 2026).
- Zoom (2026) Zoom system requirements: Windows, macOS, Linux — Bandwidth requirements. Zoom Support. https://support.zoom.com/hc/en/article?id=zm_kb&sysparm_article=KB0060748 (accessed 25 August 2026).