Finding the fault in an audio network

When an AoIP system clicks, drops out, loses sync or refuses to connect, I find out why. I work from packet captures and measurements, not from swapping parts until the symptom moves.

Faults I am often asked to find

A Dante network that keeps changing its clock leader. The audio glitches or mutes across many devices at once. I capture the clock traffic, show which devices are competing to lead and why, and say what to change so that the leader stays put.

A Dante network that Dante Controller says is healthy. The built-in diagnostics leave a good deal out. From the traffic itself I can show the priority markings actually in use, which grandmaster each audio and clock device is really synchronised to, precise packet latency and jitter, and exactly what is being sent: unicast or multicast, RTP or not, the packet time, the flow sizes and the channels in use.

Glitching that comes and goes with network load. The usual cause is priority: the stream’s packets carry the wrong DSCP marking, or the switch queues are not configured to honour it. I compare what is marked on the wire with what the switches are set to do.

A software device with heavy jitter, and a supplier who says it is within specification. I measure the jitter at the device’s own network port with hardware timestamps. That gives you figures the supplier cannot dismiss as a fault in your network.

Everyone blaming everyone else. When audio glitches, the sender, the switches, the timing and the receiver each get the blame. By capturing at more than one point, I show where the stream is still good and where it goes bad, which settles whose problem it is.

How I work

  1. You describe the symptom and send captures or configurations if you have them.
  2. I capture at the right points in the network, on site or remotely. On site I use a ProfiShark 1G+ tap referenced to GNSS, which timestamps packets to within tens of nanoseconds. An ordinary Wireshark capture with no time reference is accurate to a few milliseconds at best, and can be minutes out.
  3. I analyse the timing and protocol behaviour in Wireshark, with my own dissectors for AoIP stream and control protocols and my own Python analysis tools.
  4. You get a written explanation of the cause, the evidence, and the fix.

I wrote the analysis tools for this work. They measure PTP and stream timing in more detail than a general-purpose analyser shows.

Describe your fault