Audinate's Dante (and SDVOE?), AVB/TSN and AES67: An Audio Networking Update After Infocomm 2018
June 30, 2019: Latest update here .
It's become an annual tradition-since 2009 I've been writing about networking from a live sound perspective after catching up from Infocomm (and this year after storm chasing). You can see last year's entry here, and in this overly jargon-filled post I'm assuming you're familiar with the A/V networking landscape; if not, much of the terminology I use here was defined in previous write ups.
This year, I would describe the state of the A/V networking world to be pretty much the same as last year, with some interesting developments on the horizon. The lion's share of the products on at the show running audio over ethernet are doing so with Audinate's Dante; a few (important) companies are running AVB.
Update July 6, 2018: Roland Hemming's annual audio networking survey data has been released, it's an interesting read and supports my anecdotal observations above.
Audinate's Dante
I unfortunately missed the Audinate networking world day-long event this year, so I might have missed some new products. But seeing new Dante product at Infocomm is such a common occurrence now that I didn't even really bother to try and document it. The AVNU (AVB/TSN) alliance did not have a booth as far as I know, but Audinate was there with a Dante booth and there were numerous "Dante spoken here" signs throughout the hall.
One of the trends I saw is that there are now quite a few small Dante I/O boxes on the market now, including the original ones from Amphenol, from Audinate itself and also NeutrikAll these smaller boxes that I've seen only deal with line input and don't have head amp control or phantom power, so if you want to use a mic input you need something like a Yamaha Rio. Yamaha showed their new V2 series of boxes but I completely forgot to get over there to look. To control the head amp/phantom on these you still need to use a modern Yamaha console or their R Remote software. Update 12:11pm: See comment from Uwe Weissbach below--apparently the new Rio's can be controlled from the front panel, which I think is great.
One cool new Dante product I saw was a USB DANTE Interface from RME that supports redundant Dante operation. This is a great a benefit over the Audinate Virtual Sound Card, and also offers more flexible clock options (I'll be looking into these for our haunted house to increase our redundancy). They also are making an AVB interface.
AVB/TSN
Cisco's Catalyst AVB switches were announced in 2016 (see my writeup here) and are now finally AVNU certified (and Cisco has a section on their website devoted to AVB). This is a long overdue development for AVB in large installations where the users can have the IT support to manage these high-level switches. Personally, though, for most typical live sound applications I recommend simpler switches; I personally use the Cisco small-business switches and that's also what Yamaha typically recommends. I asked the Cisco rep if AVB was coming to that line of switches but he didn't know.
Presonus also had an AVB switch on display, but, like the MOTU AVB switch which came out a few years ago, it's not AVNU certified. The Presonus sales rep on the stand didn't know what AVNU certification was (here's a writeup and photos of my visit to the AVNU Testing lab in 2013), so it doesn't seem to be a priority for them and they don't have any products on the certification website in any case. In 2015, a rep from MOTU said on the theatre sound mailing list in response to a thread I started, "AVnu certification is a priority for us but our interfaces and switch are not certified nor have we applied for certification yet. There were a couple of corner-case AVB features that kept us from starting that process when we released our first AVB interfaces last year. Once those are ironed out and solid, we plan to get certified." That was three years ago, and they are still not listed on the AVNU site, so it seems that certification is no longer a priority for MOTU either. So what's the value of the certification, if these well known companies are out there selling AVB product that is not certified? Since last year's Infocomm, in addition to the Cisco switch, there's only four new products on the AVNU certification list: the L-Acoustics LA4X amp/controller and P1 Measurement Platform/AVB Processor; and the Control4 S3-24P PoE switch, which describes itself as a "leader in the smart home market" (and I don't think they were at Infocomm as far as I could tell).
Connecting Dante and AVB and AES67
There seemed to be more connecting and interfacing boxes available, and AES67 functionality was advertised all over the floor (and I'm still looking for some non-Dante stuff that speaks AES67 so I can learn it--please contact me if you have any gear I could borrow!). Point Source Audio had this on their stand, the AuviTran AVBx3 Audio Toolbox, with both AVB and Dante card. Presonus was rumored to have a similar product, but the sales rep there said it wasn't ready in time for the show.
To Watch For the Future: AVB/TSN and Milan
The biggest A/V networking news at Infocomm was the announcement of "Milan", a new effort to help with AVB inter-operability. They had a roll out at Infocomm, but I was not invited--I only found out about it at all because I was hanging out talking to friends on the d&b stand. My friend and audio networking expert Roland Hemming, who I missed at Infocomm, did make it into the meeting and has written his insights which you should read here. "It was a nice presentation from the Avnu team", he said in his write up, "but gave the impression that the rest of us are crawling around in the dirt, barely able to connect anything presumably due to our lack of opposable thumbs." This is kind of the tone I too felt reading through the AVNU Milan web page here, and the whitepaper which you can download here if you register.
What is Milan?
The AVNU Milan website describes it this way:
AVB is an open standard that each manufacturer can use in their own implementation, but device interoperability isn’t guaranteed without certification. Avnu Alliance compliance testing and certification is ideal for network infrastructure switches and ensures interoperability at the network layer, but doesn’t outline specification requirements for the application layer such as media formats, media clocking, and etc. It doesn’t assure interoperability amongst Pro AV end devices. Milan does.
Effective inter-operability was a goal of AVB from the beginning. The first thing I wrote on AVB in 2009 had a quote from a (now gone) roll-out article written by one of the key developers of AVB saying a key goal of the effort was to, "...enable the construction of highly interoperable Ethernet networks capable of streaming audio and video with perfect QoS." So the very existence of this Milan effort for me points up the shortcomings of AVB, and one of the reasons I think AVB hasn't gained broader acceptance: nothing in the existing standard offers the user anything like the plug and play patching convenience we already have today with Dante (of course I've written a lot about that already). Even in Milan, from what I understand, the way the patching will done--which is what the users actually interact with--is still left up to individual manufacturers (or maybe a third party?) to handle. For competitive reasons there's not really much incentive for clean, unified plug and play multi-manufacturer signal routing that is consistent for the user, and I doubt it will happen any time soon in AVB systems, if ever.
What Will Milan Offer?
In the section titled benefits for "AV System End Users" (who are listed fourth after Manufacturers, AV Managers, and IT Managers), the Milan white paper gives us the following:
* Milan fulfills expectations for real plug-and-play net-work setup and functionality. Network structures don’t require setup or complicated switch configuration tasks.
* Networks as signal and control transport structures becomes easy, fast to set up and reliable. Users can concentrate on their creative tasks.
Neither of these things is true today in AVB systems, where real, practical multi-manufacturer inter-operability exists only in limited ways. But every time I fire up Dante Controller I'm always impressed at how fast it discovers everything in a truly plug and play way. Most Dante networks require little or no switch configuration, and I would say already allows users to concentrate on their creative tasks.
The whitepaper goes on to lay out a list of promises, and while the word "Dante" doesn't appear in the document it's clearly targeted at some perceived version of the Dante world, with references to not depending on a "single company", etc. And there some inferences in the paper that, if interpreted as veiled references to Dante, are not accurate (like they imply that you need to configure switches and QoS for Dante, which you only have to worry about for the largest of systems (think airport size systems).
In terms of technical details, the white paper offers the following four things:
Media Clocking Specification
Stream Format Specification
Redundancy Specification
AVDECC Specification for Endpoints
This is pretty low level, basic stuff; Roland has more detail in his writeup. And of course none of this changes the fact that to run AVB you still need to buy AVB-capable switches (certified or not). In terms of actual benefits for the user, the most interesting thing is the AVDECC implementation (which I had to look up). Regarding this, the whitepaper says, "... Milan defines a profile for professional audio devices with a small subset of the standard, and tries to remove all ambiguities from this subset in order to achieve basic inter-operability at the Control layer."
"Control" of the discovered devices is promised, and, being the controlgeek blog, this is something that caught my eye. As we've seen many times, real, multi-manufacturer control that is successful in the market is something that is rare in a competitive industry, and I've documented this in my book since the 1990s and here on the blog (I have examples under the heading "limitations of standards" extracted here). But specifically for the audio market, we can look back to the development in the late 1990s of AES24 and its eventual market failure and withdrawal in 2004. AES24's legacy lives on in OCA/AES70, but for similar reasons that's only achieved limited success in the market, as I've detailed starting in 2012. Again, OCA and AES24 were developed by very smart and capable people (and friends of mine), but market adoption has been limited for the usual competitive reasons: many audio manufacturers put EQ in their products, but it's very difficult to get them all to agree as to how that EQ should be controlled. And it was amplifier manufacturers, back in the days where you would buy them separately from speakers, who all offered incompatible, non interoperable control solutions and did not support the initial standardization effort. Sound like a familiar situation? Easy exchanging of digital audio streams, on the other hand, is something that's in everyone's interest. (And ironically, Audinate is actually in a position where it could dictate some basic control functionality, like at least head amp gain and phantom power status).
Who Developed Milan and Why?
This part of the whitepaper really felt condescending to me, and seems to have been written by a bunch of very smart people who haven't really done their market research to see what people are actually doing today in the field (or the text was fluffed up by a very competitive marketing person):
Milan is the result of 18 months of close collaboration amongst direct competitors including AudioScience, Avid, Biamp, d&b audiotechnik, L-Acoustics, Luminex and Meyer Sound. Milan was created by the technical experts designing the systems and driving product roadmaps to impress upon other manufacturers the importance of this technical transition for the future of their business.
Market leaders decided long ago that AVB is a technically superior network technology that guarantees deterministic delivery of audio, video and data, and offers a sustainable standard technology that is not limited by one company’s vision and its future development and support decisions for its technology.
...
Today, major manufacturers in the Pro AV space have taken the lead with the first tangible solution to promise deterministic, reliable and future-proof delivery of networked media.
...
The Milan initiative is a long-term approach to bringing about change across the Pro Av market, and product certification will guarantee fool-proof interoperability of deterministic networked Pro AV devices.
This all seems to reflect the viewpoint of many of my manufacturer friends who are not on the Dante bandwagon. They understandably don't want to base their products around a core technology from another company (Audinate), and many of them seem to think that Audinate will eventually get bought up and could stop development of Dante (which is what happened to Cobranet), or go in another direction or something. And they think AVB will be there waiting to take over. These are a lot of brilliant people but I think they are wrong.
Where Will it Go?
At my school, we bought a Yamaha CL-5 mixing console. I picked that board because it's widely used in the NYC event market, many of our graduates will encounter them in the field, they are well made and affordable. Given the reality of budgets, we will likely be using this console for the next 10-15 years. That means we're running Dante for the next 10-15 years, even if Yamaha and Audinate both went out of business. And many rental shops around here and touring sound companies are in the same boat. And with so much Dante product in the world (and the rumor that Yamaha owns a stake in Audinate) if Dante development stopped, it's likely a consortium of manufacturers would continue it anyway.
But in the end this Milan effort seems to be led primarily a group of high-end loudspeaker manufacturers (all of which I'm a fan of), and I'm not really sure how--other than guaranteeing that there will be a group to continue some kind of development--Milan will really benefit users. If you're buying d&b speakers, you're going to have to buy a d&b controller/amp anyway. If you're running all self-powered Meyer speakers, you could use anyone's speaker controller but why would you when Meyer makes such great controllers already optimized for their systems that's integrated with their modelling software? And L'Acoustics has been moving in recent years towards their multichannel, control-intensive L-ISA system, which is dependent on their own DSP. So while I applaud and wish well any sort of forward movement towards any sort of inter-operability, and I hope Milan succeeds, I don't think it's the game changer they seem to be advertising.
And I don't think I'm alone in thinking this way. I was emailing with a friend who works on major Broadway productions about this new effort, and he already knew about it and said, "It seems to me, as an end-user, unless you can get DiGiCo, Yamaha, Studer, and the console big-boys on board, any discussion of audio transport is missing a crucial component. If they can get DiGiCo to come out with a Milan card, that could be cool." But that wouldn't mean he no longer needs Dante. "Right now, DMI-Dante [interface cards are] how we interface with our audio networks, and it works seamlessly at the other endpoints (speakers, monitoring, other consoles, computers, and remotes)."
And Roland Hemming summarized thusly, "Having met with most of the Milan creators, I think, or at least I hope, they know this will not be the audio networking protocol to take over the world. Milan offers AVB features that frankly should have been there in the first place. It's a welcome addition to help AVB offer a robust, high-performance local audio network."
Two Networks for the Future
It seems to me (again, in the live sound world) we will remain for the foreseeable future a two-network world (plus MADI, etc of course), and while this is not ideal, that's the way these things always seem to work out. On many systems, Dante will connect the console, microphones, stage boxes, wireless mic receivers, recording rigs, in-ear monitors, measurement systems and associated gear and often the loudspeakers, with companies like Martin offering direct Dante to their boxes. But if the speaker system is Meyer, d&b or L'Acoustics, then these systems will be running AVB, probably on a completely separate network with separate AVB switches, and the two networks will be connected with some kind of rudimentary, small channel-count interface, a converter box, or--if we're really lucky--AES67 based network interchange.
But while Audinate has embraced AES67 and incorporated it into Dante, companies like Meyer have not embraced AES67 and don't seem to have any interest in doing so. I talked to my friends at Meyer about a huge and very cool AVB-based system they built for Metallica, and asked them how they got audio into their speaker system from the console. The answer? AES3, the two-channel point to point digital audio over XLR connector standard first developed in 1985. This is better than analog, of course, but we should be doing better than that in the 21st century. But in reality, this is an OK (but limiting for the future) solution, since in a concert system typically only a handful of the many channels of the overall system are sent over from the mixer to the speaker system, and this console/speaker system dividing line also reflects a common division of labor anyway, with the (human) mixer responsible for most of the stuff up to the console outputs and then the system tech handling the speaker systems and alignment. This is good enough.
And, as I've written for many years, "good enough" is where technology typically settles, until there is a compelling reason to move forward. Milan, at least as currently outlined, is a useful, incremental, and overdue improvement to AVB, but not compelling (or complete) enough to replace Dante in the market any time soon--certainly not within the time horizon of someone who has to buy or spec a system today. For the future, who knows, but of course, Audinate is not sitting still:
SDVOE Over Dante?
The world of video over Ethernet is still a bit like the wild west. One really interesting proof of concept on the show floor was Software Defined Video Over Ethernet (SDVOE) implemented by Audinate into Dante. They were doing presentations and had a working proof of concept system; I saw it patch video right through Dante controller, which is very cool.
At least for an end user like me, seeing things like this operate is when it really seems real to me. We've been seeing SDVOE development boards and so on for years, but when you see Dante controller patch video, even in prototype form, then it seems like a real thing. We'll see what comes in the next year.
More photos from Infocomm here.