SVG Europe Audio welcomed sound specialists to its latest live online discussion with the DMF Audio Forum, which took place on 24 June.
The event, which was sponsored by Appear and Telos Alliance, saw an expansive panel tackling the topic of the Dynamic Media Facility (DMF) reference architecture and its media exchange layer (MXL) for live sports audio environments.
Guests heard from Andy Rayner, CTO at Appear, John Schur, founder of Minnetonka Audio and president of TV Solutions at Telos Alliance, Phil Myers, Lawo CTO, Paul Markham, TV, radio and media technologist and SVG Europe contributor, plus Ian Wagdin, Appear, VP of Technology and Innovation, Tom Knowles, Solid State Logic, Director of Product Management, Rob Oldfield, Salsa Sound, Co-founder and CEO, and Henry Goodman, Calrec, Director of Product Management. The session was chaired by SVG Europe’s Roger Charlesworth.
Originally conceived by the European Broadcasting Union (EBU), the DMF reference architecture provides a blueprint for building flexible, scalable and vendor-agnostic media infrastructure and reframes the broadcast facility as a cluster of interchangeable, software-defined media functions running on general-purpose compute – on-prem, in a data centre, or in the cloud.
During the session, Raynor said on DMF and MXL that, “the most exciting bit is it’s a whole new way of thinking about bringing software applications together”.
He continued: “The ‘dynamic’ media facility was really born out of the fact that if you look at most people doing live production, a large proportion of their physical devices that they’ve traditionally used to fulfil the workflow for live production – be that the audio chain or the video chain – are redundant for probably 85% of the time, apart from 24/7 op stuff.
“People have built facilities for the last 90 years to [cater for] the peak workload they require, so they’ve built facilities to handle the maximum number of events on a Saturday afternoon when everything’s happening on a really hot summer’s day, and most of that is then redundant.
Watch the on demand recording of the DMF Forum here
“The whole point of a dynamic media facility is actually saying, ‘let’s build, dynamically, the things we want, when we want them’,” Raynor continued. “For many, many broadcasters and live production people, that’s actually a tiny fraction, timewise, of when it’s needed. So that’s the concept of the dynamic element.
“One of the key principles of the out-working of DMF and specifically the out-working of MXL within that is saying, the way we have done linear connectivity between devices in the real world, be that AES67, SMPTE 2110, and before that, SDI, and analogue audio interfaces, etc, before all of those components, once we’re within compute it’s much, much more effective and lower latency to hand the data that’s been finished in one process downstream to the next one in a very efficient, memory-mapped way, without having to copy it in loads of places, without having to serialise it then de-serialise it into the next application.” He concluded on the MXL aspect of DMF that, “this is all about a low-weight latency, highly efficient way of actually pushing the data.”
Raynor added on the benefits of MXL for audio processes: “You’re not beholden to any latency that’s related to any of the other mediaflows. [For] your audio flows, all you’ve got to do is get the other audio flows that you want to mix or process with your flow together, and do the work on those, so [MXL is] absolutely instrumental. That’s a premise that we’ve had in AES67, and -30 of 2110, but this takes it to other level as well, because of the asynchronous nature of that.”
Later in the conversation, Oldfield provided examples of where he sees MXL having huge benefits for the live sports audio industry: “We worked on a project looking at globalisation of content and we were taking in audio doing a speech to text, and then using that to trigger automated sign language and also translations and live closed captions and this was across multiple vendors; it wasn’t just us doing this.”
The project used SRT as it was a common streaming protocol supported by all the members of the team working on the project, he noted: “Having to send everything around – we were doing it using SRTs – was just massively adding the latency in, which isn’t necessary when a lot of these things could coexist on the same compute. There’s no reason why they have to be SRT; that was just the most convenient that we happened to all support.
“But being able to do that with MXL would be a complete gamechanger; we’d probably be looking at knocking seconds off our latency, not milliseconds. So that’s clearly a huge advantage.”
Oldfield continued: “Other aspects I’m liking with implementing MXL is things like when we do audio quality control; being able to basically just dive in and inspect certain audio assets and then write results to memory without having to daisy-chain multiple processes, writing metadata that can be picked up from somewhere else; this is huge. [MXL] makes all of that sort of setup so much easier.
“We’re working with a client at the moment where we’re having to do quite a lot of automation in the cloud, where we’re automating the configuration of various SRT streams, and this is another great example of a system that would be hugely streamlined using an MXL approach.
“So I think this [technology] is going to be quite a big deal, [especially] in the future as containerised applications become more and more a thing in the audio processing space. I’s [also] allowing these multi vendor approaches to be so much more efficient; everybody can do the thing that they do well, and it not be a complete nightmare in terms of cobbling everything together,” Oldfield noted.
