Voice

Voice Codecs

Pick a codec and packetization (ptime) and I compute the bandwidth per call (with IP/UDP/RTP and layer-2 overhead), compared against other codecs by bitrate, MOS and delay. Multiply by the number of calls for the trunk.

Voice codec

The bitrate is the encoded voice rate. The MOS (1–5) estimates perceived quality.

Packetization (ptime)

How many ms of voice travel per packet. Fewer ms = more packets/s (pps) = more header overhead.

Layer-2 overhead

On top of the IP+UDP+RTP = 40 B headers you add the layer-2 frame (Ethernet, VLAN, MPLS…).

Simultaneous calls

The trunk needs BW per call × number of calls.

redzilla.cl — codec

BW per call

—

Trunk total

—

Quality (MOS)

—

Per call —

Total —

Packet —

Estimated quality (MOS)

— scale 1–5

—

Codec comparison

BW per call with the chosen packetization and overhead. The highlighted row is the selected codec. Tap a row to pick it.

Codec	Bitrate	Sampling	Frame	MOS	Delay	BW/call

How it is computed · payload, headers and pps

1. pps = 1000 / ptime. With 20 ms of voice per packet that is 50 pps.

2. payload (B) = bitrate_kbps × ptime_ms / 8. G.711 at 64 kbps with 20 ms = 160 B.

3. Per-packet headers: IP 20 + UDP 8 + RTP 12 = 40 B plus layer 2 (Ethernet 18 B, +VLAN 22 B, MPLS 4 B).

4. BW per call (bps) = (payload + 40 + L2) × 8 × pps. The total = BW per call × number of calls.

5. MOS and delay are reference values per codec; they do not depend on packetization but help you trade quality against bandwidth.

Runs locally in your browser · no sign-up · nothing leaves your browser

How it works

The tool compares the voice codecs most used in IP telephony (G.711, G.729, G.723.1, G.726, iLBC, GSM-FR, Opus, AMR-WB and Speex) by bitrate, sampling rate, frame size, estimated MOS quality (1 to 5 scale) and algorithmic delay. It also calculates the real IP bandwidth per call based on the codec, the packetization interval (ptime) and the layer 2 overhead you choose.

The calculation follows the classic VoIP sizing formula: pps = 1000 / ptime, payload = bitrate × ptime / 8 in bytes, and each packet adds the IP 20 + UDP 8 + RTP 12 = 40 B headers plus the layer 2 frame (Ethernet 18 B, Ethernet with 802.1Q 22 B, or MPLS 4 B). Bandwidth per call is (payload + 40 + L2) × 8 × pps, and the trunk total comes from multiplying by the number of simultaneous calls.

Example: how much bandwidth does a G.711 call use?

G.711 encodes at 64 kbps; with a 20 ms ptime the payload is 64 × 20 / 8 = 160 B per packet, at 50 pps.
Over Ethernet, each packet adds 40 B of IP/UDP/RTP plus 18 B of layer 2: 218 B on the wire.
Bandwidth per call: 218 × 8 × 50 = 87,200 bps ≈ 87.2 kbps. For a 10-call trunk: 872 kbps.

Frequently asked questions

What is the difference between G.711 and G.729?

G.711 encodes at 64 kbps without compression and delivers traditional telephony quality (MOS 4.1) with minimal delay; G.729 compresses down to 8 kbps with MOS 3.9. Watch the overhead: with a 20 ms ptime over Ethernet, G.711 uses about 87 kbps per call and G.729 about 31 kbps, because the headers weigh proportionally much more against the small G.729 payload.

What happens if I raise the ptime from 20 to 40 ms?

Fewer packets travel per second (25 pps instead of 50), so the 58 B header cost is paid half as often and bandwidth per call drops. The cost is more packetization delay, and each lost packet carries more voice (40 ms instead of 20 ms), which degrades quality more under packet loss.

What does MOS mean and from what value is a call considered good?

MOS (Mean Opinion Score) rates perceived quality from 1 (bad) to 5 (excellent). From 4.0 upward it is called toll quality, equivalent to traditional telephony; between 3.5 and 4.0 it is acceptable for most uses. Wideband codecs (16 kHz sampling, such as Opus or AMR-WB) sound more natural than narrowband ones.

Why does a call use more than the codec bitrate?

Because the bitrate only measures the encoded voice. Every RTP packet carries 40 B of IP/UDP/RTP headers plus the layer 2 frame, and that overhead is paid per packet: at 50 pps it adds 23.2 kbps in headers alone over Ethernet. That is why G.729 at 8 kbps ends up using about 31 kbps of real bandwidth per call.

Was this tool useful?

Disclaimer

We take great care to keep every tool accurate and review it thoroughly; even so, we can't guarantee it is free of errors or take responsibility for how the results are used. We recommend double-checking anything critical.

Found an error? Let us know →