[00:00:00] Speaker A: Foreign.
Today on Broadcast to post, we're talking about a big shift in broadcast facility design. You know, for years, many studios and control rooms and newsrooms were built around fixed infrastructure and fairly rigid workflows. For example, if a newsroom wants to add a new replay system or support a second production control room, a fixed design may require re cabling, reconfiguring multiple devices, or even taking part of the room offline. And that approach can make upgrades slower and less flexible over time. But a dynamic media facility, or dmf, points towards a more adaptable model, a model that is built to evolve as technology and production needs change. So to help us unpack what that means for the future of broadcast, joined by Philip Myers, CTO at LAVO and Vice Chair of the MXL Council for dmf. I love those acronyms and my really smart coworker Steve Dupay from the Key Code Media Broadcast team.
Phil and Steve, thank you for joining us today. And let's jump right in. We have a lot of stuff to cover. For folks who are joining us today who are just hearing these terms for the first time, how would you define dynamic media facility, also called DMF and the Media Exchange layer or mxl, in kind of plain language? And how is this different from how current media facilities have been architected and deployed?
[00:01:43] Speaker B: So yeah, thank you, Michael. So I think the best way to look at this is if you look back across time. We've always built broadcast facilities, whether it be radio, TV or audio, in a very fixed way. So you design it to your maximum capacity, make sure that you've got every use case, every production type covered, and then in a lot of cases you build duplicates of it for that peak load when you might have elections or you might have concurrent studios in use. And so what that's meant in the past is we've always built things to the extreme. And one of the downsides of that is you've had to think well in advance of what your CAPEX expenditure was going to look like and what equipment you were going to put into that. So by the very nature of that, we need to make the media facilities dynamic, which means that we put into a foundational layer of compute and network infrastructure, but then we can overlay on top of that the productions that we need to do on a day to day basis. So whether that be a podcast, whether that be a radio studio, that on some occasions you want to add video into it to do visual radio, whether you want to have an audio control room, or whether you Want to have a full master control room.
You want to be able to spin up what you need for that production and when you finish that at the end of it, spin it back down. Give that resource to somebody else. More importantly, be a little bit more greener on the planet. Give back all that power and cooling and everything else that comes with it.
DMF is really about living that infrastructure and make it as dynamic as possible for the customer to give the best tools to their clients.
One key part of that is something called MXL Media Exchange layer. And this is to allow us to move the audio video essence with inside that compute layer in the most efficient way in an asynchronous manner. So that means that we can move that data as quick as possible between the media functions and remove all the transport layers that go in between. And by the very nature of that, remove things like audio video sync, adjustment latency from end to end and really allow us to get the maximum utilization of the compute by not having to put all of the additional inputs and outputs around all the media functions.
[00:03:57] Speaker C: Those are excellent definitions, Phil. One of the things that kind of strikes me is that most devices that are in facilities today, because they're built on what's known as an FPGA basis, are very specific in their purpose. So they're very limited as to what functionality they can do. Even though we can cram a lot of stuff. The idea of a dynamic media facility is it uses the newer technology that's found today in IT centric compute devices and allows you to change on the fly almost if needed.
What you have, how you do it, you can basically spin up a channel, if you will, spin up a processing block.
That's not really possible to do in an FPGA based architecture. So now we have the flexibility to change based on the needs and uses of the day and we can buy extra compute extra capability without having to worry about being able to use it in the future to do additional processes that breaks us away from the limitations of a fixed facility. And it can actually lower the CAPEX expenditure at the, at the front end by allowing you to get what you need and then grow as you need.
On the, on the MXL part of it, the, the concept there is that we have all these processes that are disparate that, you know, different people have different applications and we want to also link those together. So hopefully in the future so that we can use the best of breed if you across numerous manufacturers of particular processes and functional blocks.
[00:05:31] Speaker A: And what would you say is the driving Force behind this obviously changes in technology. Okay, because we can do it. Should we? Or is it because the expectations of modern facilities are just so drastically different at this point than they were 10 years ago that we had to find a technological solution for these new expectations.
[00:05:50] Speaker B: You have to look at other parts of the production chain. So if you look at the non life part of production, so play out edits, media, asset management.
They already made the transition into generic compute networking many years ago and that was really down to a couple of things. One, the network interfaces were sufficient to carry the data, and second of all, there's less of a latency requirement in that space. So as fast as possible is good. We've had to wait for the evolution of the sort of cloud titans, the AI boom to get the larger interfaces is the first thing because we move, you know, uncompressed data. It's quite big, especially if you're doing a 12, a UHD, 12 gig production and we have to move it at the lowest lanes. If you go to what Steve was saying earlier, if you're buying bespoke pieces of equipment for every part of your production, you're putting a lot of cost into your production, into your business. So every bespoke piece of equipment requires service level agreements, it requires in house knowledge, it requires also spares and everything else that comes with it. And what we see with customers now is they want to put one hardware layer in, they want to put the compute layer in, they want to put the network layer in and they want to converge non live with live. And really what's driving that with clients right now is the customers I speak to on a daily basis, they're asked to do more content at the same high quality for less money. And so they really need to get the most out of their infrastructure and maximize the operational efficiency as much as possible.
[00:07:20] Speaker C: That's exactly right, Phil.
The fact of the matter is that the IT world has increased the speed of data transfer and processing so much more than what the broadcast domain has that now we can move into their domain and take advantage of their economies of scale. The speed with which we could process signals basically in real time using FPGAs is now being matched by what you can do on a CPU GPU side. The other part of that is the transport layers are now incredibly fast. We used to lead the industry in serial communication link speeds, you know, getting up to 3 gigabits per second.
Now we're talking 100 gig, 200 gig, 400 gig, 800 gig is common in data centers, leveraging those kinds of speeds to move these large blocks of data in and out of the compute asynchronously makes it possible to do real time processing similar to what we do in FPGAs, but on a compute platform that ultimately is lower cost, gives us greater flexibility, and creates the true dynamic nature that we're looking for on these facilities.
[00:08:30] Speaker B: It's really, really sort of lots of content now being moved around in lots of different ways. And I think as a content creator you, you want to be able to consume that content, manipulate it and repurpose it. And I think the infrastructure has to facilitate that. So you know, what we see with clients today using sort of DMF capable infrastructure, if they have a multi viewer that's typically acquired to be a 2110 multiviewer and they need one input on it to be JPEG access, for example, I don't really want to have to go put it through a signal converter to then move it again into the system just to be able to put it on the multiviewer. What I want to be able to do is shut down the container that's running the input for number one on my multiviewer, reorchestrate another container that's JPEG assist capable, and just make it work.
[00:09:22] Speaker C: That's such an excellent point, Phil, because in traditional facilities you are either using the router to get all these disparate resources patched in, or, or you're physically going back into the engine room and repatching stuff to get the functionality that you need for the processing chain. With this dynamic media approach, we're now able to have the engineer sit next to the operator, to the creative talent, and they can have a discussion about what it is that they need to do, how this thing needs to look, and the engineer sitting at his computer can now instantiate those workflows quite simply, spin up what he needs, create the workflows without having to leave that spot. The speed with which that can increase the creative capability to adapt to whatever production quality, production value that they're trying to achieve is absolutely phenomenal and will level the playing field, if you will, between those that can afford really high end bespoke systems and those that have kind of been limping along with the band aid patched together stuff.
[00:10:29] Speaker A: If we kind of flip the coin instead of looking at zeros and ones and look at it more from a dollars and cents perspective. A lot of broadcasters have heard big technology promises before. So how do you respond, Phil, to the concern that DMF design may not actually reduce costs, but simply move the cost from a more fixed broadcast hardware into new layers of software, which of course requires some specialized staff and of course a more generalist IT staff to get us over this hump and keep running.
[00:11:00] Speaker B: Yeah, so I think, I think you got to look at it in one way. So the one thing I always say to customers, you know how to build a TV station, right? We've done it many, many times over many, many years. When you look at this new technology, you've got to, it's not really an apples for apples comparison because you're unlocking a whole different thing together. So the first thing I would, I would always say is you need to understand what your utilization looks like. So just because you have H studios today, you need to go do the analysis and say, okay, what does my total cost of ownership look like over those H studios? And you might find actually over a given time only four of them are inactive at the same time. So the first thing is then you understand what utilization is. You then can understand what are my common media functions, Multiviewering, audio mixing and so on. So do I need the maximum capacity of multiviewers in every studio right now? So in a lot of cases you put 16 pips on every multi viewer. But actually in reality at any given time you might have a quad split or a 9 way or 116. So you can I distribute the licensing for the inputs across all of my studios rather than license them up for every single one. The third thing that I think you have to look at it is can I get better utilization of my assets. So a great example of that is If I need 1024 channel audio mixer for one production on a Saturday, but then during the week I need four more distributed 256 channel mixes. I don't want to have to buy all of that. I want to be able to say, can I take the 1024 and split it into four 256 ones? Because that's again a great way to get better utilization. It's reducing down your costs. The next thing I would look at is do I need to have it all on? So what's my cost to run my facility? So we, you know, I walk around a lot of customers facilities and all this, all the galleries are on, they're all on. And a lot of customers will tell you, hey, I don't turn them off in case they don't come back on. That's the first thing. But if you look, when you look at the total cost of ownership over 5, 7 years. You've also got to factor in your energy costs, which globally is going up and up every year. You've also got your H Vac, you're getting your air conditioning and clearing out the heat from your facility. That also drives your energy costs. And with a lot of customers now, they're also looking at their carbon footprint and what that, what that costs, the last cost, and I always say is that is the biggest cost to any technology is people. So if you would follow what I said before, lots of bespoke pieces of equipment, lots of SLAs that have to be renewed every single year, lots of knowledge that has to be transferred across your organization that all incurs an ongoing cost.
So one thing there and what we see with customers that have made the transition into IP into 2110 is they're able to leverage the teams, the wider teams they on in their business, the network teams. And so as we go forward, the dynamic media facility is no different. It's a prerequisite for doing dynamic media facility is you have to have made the IT transition. So not only can you utilize the wider team in your organization, I would also say to you, when we look at dynamic media facilities being containerized over IP running on standard IT commodity technology, there's a much wider community of engineers available to us right now and working in other industries with the specialist skills we need. DevOps, quality assurance operators, engineers across those disciplines. Does that mean we don't, we don't need audio video engineers? No, we still need them because they are very focused on the operational application layer. But let's let the infrastructure be managed by the same teams that are running that IP IT infrastructure for other departments in the organization, whether IT be non live or even just the commodity IT infrastructure.
[00:14:56] Speaker A: While there is this technology that can kind of revolutionize how things are done. Seems to me there almost needs to be an audit, if I can use that phrase, to determine what the TCO is. So do you find there needs to be kind of an exploratory time where you're going with the client and reviewing the actual usage of the tools that are currently there?
[00:15:17] Speaker C: Absolutely. And one of the key factors that we add into our formula for what we call total cost of ownership is what kind of training your staff is going to need, where's your staff at today? In fact, treating the staff as that critical component that they are and what their skill level is, how much additional training is needed if you need to add an additional IT resource in order to manage these systems that all factors into that total cost of ownership. And what we find is that while there are significant savings in some of the physical infrastructure and how that's integrated, that is oftentimes offset and exceeded by the cost of additional staff or training needed for the staff. But when we look at it over a five year period, the staff capability and what you're able to leverage starts to bring that back into, back into parity or even to significant savings. One of the key areas Phil mentioned there was the reduction in power and cooling.
That starts to play in there a whole lot more as you go over time. The other thing that I think is really interesting that Phil mentioned there is you've got to have staff that understands how these things work and taking it people and training them how to be video guys and audio guys. Seems like, you know, maybe that's, that's really tough. But one of the other cool things that happens with this, this new dynamic approach using, using these new tools is that there's monitoring and self analysis capabilities that are coming as well that enable even, I don't want to say amateurs, but you know, people that, that maybe don't have the high end production skill level that, that somebody in the industry for 30 years would have. But there are enough things there that can help guide them as to how the signal should behave or what things are going on. Correcting for errors in lip sync, colorimetry, all those kinds of things to help the operator, that junior engineer, be successful in getting those signals operational. Lastly, I'll point out that in the terms of design, one of the things that always happens is there's a little bit of miscommunication sometimes between what the ultimate customer wants to have and what the engineer designed the facility has envisioned. And going to a dynamic approach on this compute based architecture allows us to over build a little bit, if you will, and naturally build into the architecture the capability to adjust to those changes without having a huge cost overrun and without having to add additional physical infrastructure. That part of it is really huge on the integration side. It makes it so that we can be more efficient, more effective and in our designs. And that gives the customer ultimately the ability to scale up to new technologies, new capabilities as their demands increase at their production area.
[00:18:15] Speaker B: Really one of the big benefits of moving to the dynamic media facility using containerized infrastructure. A lot of the monitoring, the observability is baked in from the start. And so what we see right now with clients is, is a lot of our clients are building out observability screens using things like Grafana and having really true end to end observability of things like the servers, the power supplies, pcp, how many containers are running, the health of those containers, everything all the way up towards the signals and what's the latency of the signal across the infrastructure. And this I honestly believe brings these types of infrastructure to life. So if you think about historical SDI based infrastructure, we'd have to put a really highly proprietary controller monitoring system across top of that with device drivers for every single bit of proprietary product under it. I know now if I have servers running containers, I get a lot of metrics from the servers. If it goes through an IP network, I get a lot of metrics from the IP network as well. So I can build up a picture of what's happening. And the great things from using tools like Prometheus with Grafana is I can roll that back across the whole infrastructure, back five minutes, back one hour, back one day, back 24 hours. So if I'm doing the same production every week on a Friday, I can look at, okay, what's the behavior? What's is the behavior every week the same across the compute and across the network. So it allows me to be a lot more proactive to the infrastructure in monitoring rather than reactive and only dealing with it when you have maybe an issue with audio, video on air. But also another key aspect of that, it can all be done through a web browser. So the engineers can work from anywhere. So the day of having to have the engineer sat right next to the bit of equipment in the engineering room, honestly they can now dial into it remotely, they can be sent an alarm when there's a problem. It really allows people to do their job really a lot better and a lot more efficiently than maybe it would have been in the past where you run around with a fabrics and a BNC and you're kind of trying to find in the jack field where things are going wrong in the infrastructure.
[00:20:30] Speaker C: Yeah, I think that's one of the beautiful things about it is that flexibility to operate from anywhere and with the level of monitoring that is now available, the quality of service capabilities that are present in the compute portion as well as in the network, it makes it so that again, what I see happening, and this has kind of been a long term vision, is that the operator can sit next to the engineer and they can discuss or the producer and they can discuss what resources it is that they need for a show.
And with a little bit of self analyzation and automation, rather than just writing on a piece of paper, start Putting that into the system and have it created the workflows for them. And then you're just using the monitoring tools to troubleshoot, to make sure that it's working radically. Speeding up the time from concept to delivery on a complete workflow in a production facility, that's huge because now you can spend more time as a creative person figuring out and performing that creative aspect, increasing the value of the production.
[00:21:39] Speaker A: Steve, I wanted to circle back about something you spoke about, which I don't want to gloss over, because it was very important when we started doing the migration from baseband to 2110. And of course, it required a lot of upskilling of engineers who had been working with baseband for decades. So what have we learned during that upskilling of engineers to learning baseband or to learn 21.10 from baseband? And how can we kind of apply this to the DMF technology shift?
[00:22:07] Speaker C: So the biggest change was, is that with an SDI signal that was so well vetted and so precise, and it had to be, because there's no recovery on it, you drop bytes and bits, you lose the signal. Right. Because that was so robust, we had to teach engineers, and we continue to do that, that you have to be a little bit more careful, a lot more careful. Let's be honest with how things are put together. The timing, making sure the PTP clocks are operating correctly, making sure that all of the file structures are correct for all the signal paths and associations in the nmos. So we had to take them from that.
Hey, I got the video signal anywhere, and I can see what's going on by just grabbing a BNC cable. Say, hey, we got to pay more attention to what's going on with that data path. So they've.
I'll say it this way, there's no broadcast engineer that I found that can't learn fairly quickly all the IT skills necessary to make it work. Some of it's kind of deep, but they can learn it if they choose.
I've also found that there's a lot of the IT guys, particularly if they're into gaming, that already have kind of a feel for how this video and audio stuff works. And it's not that tough for them to be trained in that as well. In fact, the new generation of operators, broadcast engineers, I see, are going to probably come from the gaming domain.
These kids that are able to build their own computers, that are building their own home networks, that are doing all these kinds of crazy things to create their own AI data center from a Bunch of old components. These are the kind of young people that are going to be able to take broadcast facilities forward. So it's nothing to really be scared of, it's just something new to learn. And for some of us it's pretty exciting to have some new technology to learn and develop some skills on.
[00:24:11] Speaker A: So Phil, what are some real world customer stories where a dynamic media facility approach creates a better outcome than a more traditional fixed facility model?
So what are the strongest early use cases for dmf?
[00:24:26] Speaker B: Yeah, so we have a large US OB company that's using our home apps technology extensively on really huge scale. So they've got 192UHD multiviewers, over 3,000 pips assigned in there. They can spin that up and down any size from 16 pips to nine to fours. And then when they reduce the multi users they can use it for other things. So whether it's UDXs, audio shufflers and so on. And that's really, I look at that use case where you're doing really tier one sports and saying first thing is latency is really important. Second thing, scale.
I would say in the US market the size of the trucks are huge compared to the European market. And so they're almost mini facilities on wheels. And so the only way to get that all in and make it reprogrammable on the fly is really to look at this technology. That's one use case. The second one is we have, we have this global sporting event going on right now in the world. We have a lot of infrastructure there from sort of venue side all the way through to the ibc and again being able to spin it up at the edge, spin it up in the core and take down what you don't need and put up what you do need. And so that use case is slightly different because you come do a global event, you want to build it quickly, do the event, collapse it and go do something else with it. And so in that use case if we would go back sort of 10, 12, 15 years ago, we'd send a lot of people because a lot of wiring for signals. Now we plug the compute up to the network, we set a DevOps engineer, he says, okay, I want to get this up and running. I want 20 multi viewers, some UDXs, some multi mixes and it's up and running very, very quickly. And that allows companies that are providing host broadcasting services from event to event to bring it up, use it, tear it down and move it as efficiently as possible.
And I guess the third use case is you Know, sort of we have a large US based news organization using our infrastructure as well. And you know, they have lots of lot of departments. And whether it's an audio, whether it's an mcr, whether it's again in a gallery, they have to change as their production changes, breaking news, elections, so on. So we've had people using this technology at scale and it was designed from day one to at scale because if you can do the large things, the smaller things are a lot easier to do.
One thing that we've always said, audio over IP is way more complex than video over ip.
And so if you can do audio properly over ip, I'm pretty sure you can do video over ip.
[00:27:14] Speaker C: That's exactly right, Phil. And there's a way to kind of break that down from a business aspect of it.
If you look at a truck that is designed just to do football or basketball, you may have it on, you may have it booked between 32 and 88 times a year and then that equipment sits idle. So years and years ago they started looking at how do we add a little bit of extra stuff to maybe go to a news event or a corporate event. And you kind of break them down into four general areas. And trying to put all the gear to support all those things, make it, make it possible for the truck to do all those things in a bespoke manner would just overload the truck.
With a dynamic media facility approach to that. Now I can quickly change it on the fly. And instead of just using the truck 88 times a year or 130 times a year, now I'm running the thing 180, 215.
Think of all the different things you can do just by moving the truck from point to point to point. And I'm reducing the number of people that are needed on the truck to support it. So I'm able to get better utilization of the resources. I'm able to lower my overall cost because I don't have to have as many people repatching stuff and figuring things out. And I'm able to increase the production value of whatever it is I'm doing because I have the resources there that are quickly and easily available to configure on these, on these new platforms. So truck companies are a key example of how that increased utilization can mean greater profitability.
It can also apply to rental facilities, even, even news production facilities that then lease their, their facilities during downtime to other, other productions that that need to be produced.
So I think if you're starting to get this whole picture that being able to quickly change and adapt to whatever the requirement is is the key benefit. That is exactly it.
[00:29:14] Speaker B: Yeah. And I think I would just add two other things. Sorry, Michael. One is, you know, when we talk about OB technologies, typically you would connect SDI up to the rear of the truck. Well, the nice thing with this concept is you can build stage boxes and connect at exactly where the point is and just bring back a couple of hundred gig fibers to the truck. So whether that's for capacity or whether it's redundancy, the ability to just set up and go is really a lot, a lot easier. Which means that you might not need the trucks out there for many, many days beforehand doing the prep work. The second side of it is, you know, we start seeing now more customers do remote production. And so you might have a truck, but I want to bring all the signals back. If I can just say to it, okay, now I need that output to be in JPEG XS or in srt. I'm better off to spin it up on the infrastructure and send it on its merry way rather than have to wire in all this extra technology to do it.
[00:30:09] Speaker C: The last part that I want to emphasize here is that these same technologies that can be basically on a private data center, in your truck, in your facility, et cetera, can also be instantiated in the cloud. So now you have the ability to leverage all three kinds of approaches to get the maximum benefit from the technology.
[00:30:30] Speaker A: So both of you have brought up many different reasons why moving towards a DMF topology is beneficial. And I think what folks can kind of get wrapped around the axle on is like, where do you start? What is the kind of order of operations of things you should be looking at, things you should be thinking about first, if they want to move towards this model without ripping out everything they have. So can you kind of start with where people should start to look internally?
[00:30:56] Speaker B: Yeah. So I would always say find a good partner, find somebody who can help you can advise you, you know, and I think if one of the things we've learned with the IP transition is consultancy is not a bad thing. You know, find somebody that's trusted, can help you can understand what you're trying to achieve. Some customers want to get the benefit, but they're already in a 24x7 infrastructure, so they need a plan. They need to say, how can I expand it out without it impacting or reducing my ability to keep my current facility? If it's a brand new build, even better, I would say always engage the first thing is always engage, try and understand what you're trying to achieve, understand the business benefits you're trying to achieve. That's really critical because on one side there's an engineering aspect, on the other side there's probably a CFO that wants to know what it's going to cost me and what are the benefits and the savings from it.
And I would say always don't look at it as it's hard, it's difficult, it's scary. Look at it a slightly different way.
The people coming into our industry now, it's not uncommon for them to build, pop up a studio, build up a YouTube channel, plug cameras into an IP system, transport from a software based vision or audio mixer, move it through the Internet to some sort of content distribution network and get it out. So this technology is already being used on a daily basis to produce audio, video content.
What we're doing with it in the dynamic media facility in customers is we're just using it at a higher quality and higher bit rates, but with the same real sort of principles of how singles are moving around.
And I think human nature is always to take one step back and say this all looks a bit difficult, but if we had done this in the past, we wouldn't have gone from analog to digital and we wouldn't have gone from HD to uh, we wouldn't have done SDR to hdr, we wouldn't have done stereo to immersive audio.
Really find the people, understand what you're trying to achieve, embrace change. That's really an important factor of this because there's change in the technology and there's change and maybe some of the roles in the organization change to being sort of more hybrid or being more focused on certain things. And you know, I would always say go speak to people who have done it. And so one prerequisite I always say to people is you have to have experience in IP networks. These are servers, off the shelf servers. They connect to the IP network.
You have to have a baseline understanding of that and from there I think you'll be in a pretty good shape to move this to where you want to get to.
[00:33:34] Speaker C: Absolutely agree, Phil.
I think that's getting with a partner who understands, talking with vendors that have solutions that, that help you to reduce your cost, lower your risk, improve your efficiency are all of the factors that you need to consider. Phil mentioned, instead of running all your audio lines, all these mic cables, all these video cables back into a core, is there a way that I can create video islands and then just run a handful of fiber cables back, much easier to distribute. And all of a sudden, when you look at those kinds of technologies, video switcher that is more compact, CPU based, that can change on the fly. Like we've talked about, it becomes dynamic, the world changes. And once you've got that kind of operation going, your efficiency and your savings in labor, your effectiveness in setting things up radically improves. And the same thing can apply in a production studio where you want to get to that next level. You want to add HGR or something else in a bespoke, you're going to have to replace a bunch of gear. But if you start on some core pieces that are dynamic in nature that you know more of this COTS based approach, then you're able to start expanding that out and gaining that benefit and not having to rip out all that infrastructure every time the GM says, hey, if we only could do this kind of thing.
[00:34:58] Speaker A: Phil and Steve, thank you so much for joining us today.
What I'd like to do is kind of transition to some live question and answer from our audience. So could you two stick around for a few more minutes?
[00:35:07] Speaker C: Absolutely.
[00:35:08] Speaker A: Excellent. We'll be, we'll be right back. Thanks for watching. Broadcast to post don't forget to follow Keycode Media on social and contact us about your
[email protected].
