Best Way to Connect 2 Radio Stations to Send Live Programs to Each Other?

[chriscollins]

I have seen both WNTI and I agree with you.

We have two Barix that work great using PCM 44.1... They are on a Wireless Internet Link from the same provider who takes care of our studio needs. It is extremely reliable.

There is a Barix that another company that I deal with uses. They use the public internet and against my advice, they grabbed cable internet instead of getting a link from the same provider that hosts the receive end. They drop ALL the time at 96k AAC. They would be much better suited by a Comrex.

If you have control of the network, Barix can work great. If the network is public and totally out of your control, save your money.

 

[RolfTaylor]

Barix uses a proprietary "error camouflaging" algorithm in which it re-plays the last repeated chunk of audio if a block is missing. They are clearly NOT simply dropping the audio and go on - that's what others do. Extended listening tests have shown that the result sounds much better than if you simply drop the packets.
There are alternatives .. mainly, FEC (forward error correction) techniques. But they have a drawback too - they need more bandwidth, and they can only cover specific problems.

This alternative to a simple drop out is not at all proprietary and is very commonly included in many codecs. However, since a "lost packet" really means a "lost frame" the duration of the audio covered by a frame is rather large (incidentally AAC uses variable frame length) and the result is the "stutter" effect with which we have all heard before. six of one/half dozen of another as to whether this is better than a simple dropout of the same duration.

With APT-X there are no audio frames, so the dropout duration is purely due to packet size which is adjustable. So in that case you can reduce packet size to the point that dropout are invisible or some other compromise. This is a simple case of psychoacoustic error concealment, taking advantage of the fact the ear cannot perceive dropouts below a certain threshold.

In MPEG land there are a number of tools that can be employed but are not there as a default. The first is active psychcoacoustic error concealment. This is implemented by Telos and Comrex as part of their suite of tools (Telos ACT and Comrex BRIC). I recently became aware that Musicam Suprima's claim to have error concealment in all of their MPEG decoder algorithms.

FEC is not a particularly useful tool, particularly when delay is a criteria.