Audio-over-IP (Dante)

Pick channels, sample rate (48/96/192 kHz) and bit depth and I compute the raw audio bandwidth and the network estimate (with RTP/IP overhead), packets per second and the Dante latency.

Audio channels flows ×

Total simultaneous audio channels. Each PCM channel takes Fs × bits per second.

Sample rate Fs

48 kHz is the professional standard. Doubling the sample rate doubles the bandwidth.

Bit depth bits/sample

Dante carries PCM. 24 bit is the norm in production and live sound.

Per-flow latency Dante

Lower latency = more packets per second (more CPU/network load). 1 ms is the Dante default.

Scenarios

redzilla.cl — dante

Network BW · estimate

—

with RTP/IP overhead

Audio BW · raw

—

PCM, no headers

Packets / sec

—

per flow

Samples / packet

—

per channel

Raw —

Network —

Pkt/s —

Latency —

1 GbE link usage

— 1000 Mbps

Network estimate against a Gigabit Ethernet link (1000 Mbps).

Bandwidth by channels

Channels	Raw	Network	% of 1 GbE

current setup · at the chosen sample rate and bit depth.

How it is calculated · audio over IP

1. Raw audio bandwidth (PCM samples only): BW = channels × Fs × bits, with Fs in Hz. E.g. 1 × 48000 × 24 = 1.152 Mbps per channel at 48 kHz/24 bit.

2. Network bandwidth: we multiply the raw figure by an overhead factor (×1.3 by default) that approximates the RTP 12 + UDP 8 + IP 20 + Ethernet headers. The real value depends on packing (samples per packet): more latency means more samples per packet and lower relative overhead.

3. Packets per second per flow: pps = 1000 / latency_ms. At 1 ms that is 1000 packets/s; at 0.25 ms, 4000.

4. Samples per packet (per channel): Fs × latency_ms / 1000. At 48 kHz and 1 ms that is 48 samples.

5. This is an estimate for sizing the network. Dante and AES67 pack each flow somewhat differently and add control traffic (mDNS, PTP for clock sync). The fixed per-packet overhead weighs more with few samples per packet (low latencies) and few channels per flow.

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

How it works

The calculator estimates the bandwidth an audio-over-IP system such as Dante or AES67 consumes from four inputs: audio channels, sample rate (44.1 to 192 kHz), bit depth (16/24/32) and the per-flow latency setting (0.15 to 5 ms). It returns the raw audio bandwidth, the network estimate with overhead, packets per second and samples per packet, plus the share it takes of a Gigabit Ethernet link.

The math starts from the uncompressed PCM these protocols carry: raw BW = channels × Fs × bits. The network estimate multiplies that by a ×1.3 overhead factor, which approximates the RTP (12 bytes) + UDP (8) + IP (20) + Ethernet headers. Packets per second per flow come from 1000 / latency_ms and samples per packet from Fs × latency_ms / 1000: lowering the latency raises the packet rate and the relative weight of the headers.

Example: 64-channel console at 48 kHz / 24-bit with 1 ms

Raw: 64 × 48000 × 24 = 73,728,000 bps ≈ 73.7 Mbps of PCM audio.
On the network: 73.7 × 1.3 ≈ 95.8 Mbps including RTP/UDP/IP/Ethernet headers.
Packets: 1000 / 1 ms = 1000 packets/s per flow, with 48000 × 1 / 1000 = 48 samples per channel in each packet.
Link usage: 95.8 Mbps is about 9.6 % of a 1 GbE link, with plenty of headroom.

Frequently asked questions

How many Dante channels fit on a 1 Gbps link?

At 48 kHz / 24-bit each channel takes about 1.5 Mbps on the wire (1.152 Mbps raw × 1.3 overhead), so several hundred channels fit comfortably in 1 GbE: 256 channels sit around 38 % of the link. In practice keep usage under roughly 70 % to leave room for control traffic (PTP, mDNS) and other services on the network.

What happens if I lower the latency from 1 ms to 0.25 ms?

The audio bandwidth barely changes, but packets per second quadruple: from 1000 to 4000 per flow, with only 12 samples per packet at 48 kHz. That increases CPU load on the devices and the relative weight of the headers. Use low latencies only when the design demands it (live monitoring) and on networks with few hops.

What is the difference between raw bandwidth and the network estimate?

Raw counts only the PCM samples (channels × Fs × bits). The network estimate adds the RTP, UDP, IP and Ethernet headers of every packet, approximated here with a ×1.3 factor. The real value depends on packetization: at low latencies there are fewer samples per packet and the relative overhead grows.

Does the calculator work for AES67 as well as Dante?

Yes, as a design estimate: both carry uncompressed PCM over RTP/UDP/IP and the formulas are the same. Each protocol packetizes flows slightly differently and adds control and sync traffic (PTP), so treat the result as a network-sizing figure, not an exact measurement. Everything runs in your browser and no data is sent anywhere.

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 →