Technology: Out of the Ether
Technology: Out of the Ether
Phil Ward aims right between the goalposts as two of the world’s leading audio networking gurus, one from each protocol, go head-to-head
Deciding between Dante and AVB was not, until recently, a straightforward choice: like deciding between shop-bought beer and home brew. Since the launch of AVB Milan as an application layer solution, the arguments have now entered a more level playing field in the same ballpark.
Jeff Rocha, global product management director, L-Acoustics
Why is there confusion about whether we should compare AVB and Dante?
I would say that, from my perspective, the biggest issue is that people historically were comparing AVB to Dante – or any other readily available network protocols, Dante being the one with the most traction – as if they were the same concept. But, in reality, you have to compare AVB with standard Ethernet protocol. So, if you think of Ethernet as being the network layer, Dante is then really an application layer that’s built on top of that.
Dante is a way of ensuring that all of the devices that are using it are interoperable on the Dante network. The challenge from our perspective is that Dante is ready-made for Ethernet: users actually see a benefit in being able to use a regular Ethernet switch. But, in fact, that switch has Ethernet weaknesses, and that’s what companies like L-Acoustics, d&b audiotechnik and Meyer Sound were concerned about – because Ethernet was not truly meant for real-time audio transport.
This meant that things like timing and synchronicity on the network, and also the guarantee of having bandwidth availability, were not really there. There are some workarounds that have been developed: Precision Time Protocol; DiffServ – things to try and prioritise data on an Ethernet network and make sure that real-time audio can happen. The reality is that if you have an Ethernet-based network and you’re trying to learn real-time audio and control over the same cable, and you end up with a lot of other data on the network, it can actually cause audio dropouts. Basically, anything built on Ethernet has some risks.
How does AVB overcome those risks?
AVB, by comparison, has things like timing and bandwidth reservation baked into the hardware. It’s the reason why you need an AVB switch. None of the foundational weaknesses of Ethernet are present in AVB. You could say that AVB improves the pipe through which the content flows – audio and video. The problem, though, is that it’s essentially a standards-based, IEEE protocol that gives a lot of flexibility in the implementation process. So manufacturers can make choices about stream format, how they deal with clocking, do they want to do redundancy and how are they going to do it … and essentially, for years, manufacturers that chose AVB as the solid foundation for their network implementation kind of built their own application layer on top of AVB and would make their own decisions about formatting, clocking and so on.
What that meant was you could put devices from different manufacturers into AVB, and essentially those devices were not speaking the same language – precisely because those same manufacturers had made their own decisions. That’s what Milan is all about. In many ways, Milan is something that could be more fairly contrasted with Dante, because it is the application layer solution that sits on top of Ethernet, and it is stronger.
That’s what gives the industry the level of interoperability it requires, and what our customers need so that when they plug in different devices, they just work. And they have all the assurances, with AVB as the underlying technology, that things like timing and guarantee of bandwidth for audio are taken care of – so you don’t have to worry about audio dropouts.
Do you still have to carry out your own implementation of the protocol?
No, that’s the beauty of it. From the user’s standpoint, Milan is even more plug-and-play than something like Dante, precisely because of the limitations of Ethernet. Due to those limitations, there’s still work that has to be done on the Dante side concerned with quality of service and other things. There is always some setup and configuration associated with any Ethernet-based network to make sure the hardware is configured correctly. With AVB, as long as you’re using Avnu-certified AVB switches and, alternately, Milan-certified AVB endpoints, you literally just plug them together and they work.
From a manufacturers’ standpoint, the Dante path was in many ways easy, I would say, because as long as you pay for Dante hardware, Audinate can help with implementation. There’s an incremental cost per unit but, from a development standpoint, it was simpler – that’s why it’s been attractive to manufacturers. AVB was harder – but not anymore.
Joshua Rush, vice president of marketing and product management, Audinate
How should we consider Dante as a commercial option?
The right way to think about Dante is a complete application stack – in fact, a solution as opposed to a protocol. To be honest, we view the protocols that actually connect the streams between devices as the most basic level of connectivity. In the development of Dante over the past 10 years, there were things above and beyond that level that were purposely done to make it easy to use – plug-and-play, discoverability, the ability to rename devices and having tools like Dante Controller so you have a third-party controller that’s maintained by one independent company. Dante Controller is important because it’s the software that you use to actually make the routes and monitor network health for latency, clock sync and so on.
Aidan Williams’ vision when he founded the company was to recognise that if it wasn’t drop-dead simple, and if it didn’t work on people’s existing equipment, it would be very difficult to get wide adoption. It’s important to think about it like that because, at that protocol level, we can choose to include many different options to create an end-to-end solution that people can buy into. We listen to our customers to figure out which ones to incorporate.
AES67 is a good example, because there were lots of requests for adding that in, especially from the broadcast sector. It’s one of the layers of the stack, meaning it’s not either one or the other – you can have AES67 within Dante and still get all the other benefits of the solution stack. The most recent one is SMPTE ST 2110, announced at NAB last year, behind which there’s a lot of momentum in the Alliance for IP Media Solutions. With all these other things we bring to the table, it’s better to think of Dante as an audio networking application rather than a protocol. It’s much more than that.
So, it’s not quite so finite and hermetic as commonly believed?
Frankly, we have a pretty big engineering team over in Australia and the expansion of the application is what we spend the majority of the time working on – additional things for the stack to make it more valuable and more usable. Dante Domain Manager is a great example of that: we were getting all these other requests for things like security, user management and dashboards so we developed a whole other layer on top of that stack.
Milan is positioned around interoperability, but that was how Dante was conceived from the very beginning. The news around Milan is basically validating that this basic requirement is something users really want. The challenge is that it’s still built on AVB and requires those specific switches, while one of the founding principles of Dante is that it needs to work on existing hardware. There was a lot of engineering work done in the early days to make that possible. Also, with AVB there’s still no central group developing the control software and making it consistent across the manufacturing sector. Again, Dante Controller is really powerful in being able to provide that window into the Dante ecosystem, and we spend a lot of time, energy and investment in continuing its evolution as a viable tool throughout the whole industry.
Your products do seem to have a consistent goal of universality …
Right – Dante AVIO Adaptors are a good example of that. We continue to get more and more Dante-enabled products all the time. There’s around 1,700 now. But sure, there are millions of products out there already in place that people might want to include on their network, so they need a simple way of connecting all of them. You want to be able to enjoy the experience that the software offers across all your gear – not just some of it.