We Need A Standard For Standards

TL;dr Short history of audio networking. A brief comparison of two of the competing technologies AVB and Dante. A current, real world application of a large scale AVB based system addressing benefits and drawbacks. Busting the myth that AVB is really less expensive because it doesn’t require a license or proprietary parts. Not holding out much hope for better interoperability at least near term.

In his book Computer Networks creator of Minix Andy Tanenbaum professed “the nice thing about standards is there are so many of them to choose from”. Even 30 years since computers started getting interfaced to audio equipment we still can’t agree on standards for it. For the longest time we couldn’t even agree on which pin was hot on an XLR.

In the late 90s live audio started to shovel digital audio down a network instead of point to point. But we couldn't agree on a format let alone a networking scheme. It wasn’t sold on scalability or flexibility which were key advantages. The pitch was on using fewer, smaller cables which meant smaller conduit and reduced labor costs. Who cares about reduced frequency response, dynamic range and latency? Turns out a lot of people. Particularly on my side of the tracks. Close, no cigar. Yet.

For decades pro audio said if we get a standard to use for this processing and transport of audio everyone will flock to it. So everyone got to work. And created their own version of what the standard should be. Every AES Convention there were great prognostications about the upcoming standard. After the show everyone would return to Northridge, Buena Park, Salt Lake, Berkeley, Costa Mesa and Kidderminster, among other places to work on their own brew.

Somewhere along the last ten years we reached a tipping point in the adaptation of networked audio when advancements in technology steamrolled the standards process. Networked audio is no longer some far away tech fever dream. It is now the norm. The mass adoption was driven by the markets need for an application based solution not so much a solution that was derived from a standard. So most decided on a proprietary solution. Dante.

The good news at the time was there was also an open standard for audio networking though not widely adopted compared to Dante. It was the Audio Video Bridging Standard IEEE 802.1. From a technical standpoint it’s pretty sexy. Deterministic transport with stream reservations, accurate timing, bounded latency, traffic shaping that makes sure the most important packets like the clock get through first and the rest of the bits get there to be able to assemble them to turn them into sound. What’s not to love?

The org promoting the adaptation of AVB standards, Avnu, along with the members pitched (and still pitch) the open source technology as a lower cost solution to audio networking. Specifically this would be the lower cost of acquiring the base technology as it wouldn’t need to be licensed and could be chipset agnostic. Sounds good. Look how open source software, Linux in particular but also Apache, Bind, Sendmail, Postfix and others made the internet possible.

Dante is no slouch either though most of the magic is hidden under the hood protected by either an NDA or API. Not so much an issue as it works well. In fact that’s the selling point. It’s a solution not a standard. But like any proprietary IP you have to license it. After using both I say both work well. Both in the best of breed audio technology category. Varsity stuff.

Aside from technology implementation the differences were time to market, interoperability and market penetration. AVB had a couple of missteps early on. First it was later to market or rather implementers didn’t deliver turnkey solutions as quickly as their Dante counterparts did. In some cases though companies released both Dante and AVB devices. Because of this delay the market penetration aspect came into play. Can’t ship what you don’t have and can’t deploy what isn’t available.

To make matters worse initially there was no interoperability spec with which to integrate the various implementations of AVB. Everyone rolled their own. As a result many devices were islands in their own ecosystems even though they AVB plumbing. That was sorted out when they released the Milan version of AVB with an interoperability spec. Though Milan deployment is still hamstrung by a relative lack of devices. They’ve since posted a primer on the AVnu site that gives an under the hood overview of interoperability in the technology.

During the pandemic supply chain crunch Dante was hit hard. One of the longest lead times in the business. On a show I was building there were a couple of devices that needed to be switched to AVB devices. The project also had three other AVB nets. One Milan net and two that didn’t have interoperability. To get the interoperability between the nets the design called for conversions to analog between all the systems. That can get expensive. Some of it was particular gear choices while some was a lack of interoperability between the same manufacturer.

For example the playback rig and editing DAWs were converted from MADI to analog with Andiamos, the proprietary Avid AVB was output as analog at the stage box to be input to Dmitri and Q-Sys though we were able to keep some outs AES into some Lab Gruppen amps. The Dmitri Gnet was sent up to the catwalks for distribution to the arrays, broken out as analog then input analog into a device that formed the Milan AVB net for array distribution. It works well and sounds great. Sounding great meaning there are no acoustic signatures due to transport and the finished system sounded good as well. It has a robust feature set with high fault tolerance as everything was deployed in an online redundant topology.

Granted the scale and complexity of this particular design is an edge case. Only a handful of live shows in the world have that scope and scale. It does illustrate issues when choosing technology to deploy. Primarily interoperability which drives networking cost as well as the cost of the gear to roundtrip those analog connections when required. We need intermediate protocol converters when design dictates a couple of different schemes.

There are some converter solutions but lack of capacity rule those out for many apps. Particularly when they are powered by a wall wart and have no ability for hot failover. The Prosonus is a good example on where to start particularly at the price point but needs a bit of tweaking to make it to the bigs. The Direct Out Prodigy line is promising but as of yet has no AVB card and it’s pricey due to the scale and flexibility of the device. What we need is something like the Prosonus with 32x32 or 64x64 in a rack mount with dual power supplies. No routing, eq or other features just a straight one to one converter both ways for a couple/few grand street price.

Let’s focus on smaller use cases that are the bulk of live events. Gigs most people do. A drawback particularly for smaller gigs are the cost of the switches. AVB is a protocol that switches must specifically recognize so not any switch will not work. The design of the protocol is the thing that contributes to AVB being a technically robust solution. That means higher priced switches with most charging extra to license AVB capability in the switch. That’s where the cost comes into play.

Let’s say you have enough gear to do small corporates to an arena show. Switches are now as common as patch panels were in the olden days (because they are now patch panels). You may need one per rack or close to it. With Dante the switch requirements are not as strict as they are with Milan AVB so a switch for a Dante net is quite a bit less expensive. For $400-500 I can get a suitable switch that’s easy to deploy and widely available.

A single AVB certified switch for the same port count is roughly twice the price. Plus a couple hun for the AVB license. The total switch cost is at least twice or more to implement a Milan system vs Dante. If your networking design needs an advanced configuration or high availability you’ll be spending money on full featured switches anyway so it’s a wash. Gear with AVB interfaces generally aren’t less expensive than comparable Dante gear.

There is also a technical/training hurdle but given the direction AoIP tech is headed being able to understand and configure complex switching technology will be a given. For some it's already required. VLANs, QoS, understanding clock, network topology and failover mechanisms will be required. For smaller operators that’s going to be an issue at least near term. Vendor training is helping with that part.

While there is switch gear in the market that is relatively easy to configure it doesn’t give you the basics of design or how it should work. It's all unicorns and rainbows until it shits the bed. This isn’t specific to AVB or Dante (or even networking) but generally in understanding system configuration. Plug and play is cool until it’s plug and not playing.

In a perfect world these two protocols would have better interoperability. I can see why one camp or the other doesn’t want to build bridge technologies. This applies equally to both the Dante and AVB camps. They want to peddle their own. Understandable and in some cases you may not need to build outside either ecosystem.

But the siloing doesn’t serve the end users or the designers well. This isn’t a protocol or standards issue as much as it is a marketing and sales issue. This isn’t to promote one technology over the other. It’s to recognize that even now there are still significant barriers to interoperability. Until we can converge or implement parallel schemes of all AoIP audio interface nirvana will still be a pipe dream.

To answer the question at the top, no, AVB hasn’t lost the protocol wars. Yet.

Thanks for reading A Barking Dog! Subscribe for free to receive new posts and support my work.