Easy-to-understand and meaningful audio loudness measurement has proven elusive over the history of recorded and transmitted audio. John Kean, Senior Technologist at NPR Labs, has been researching the problem and finding solutions. John joins Chris Tobin and Kirk Harnack for an update on this aspect of audio recording and transmission.
Watch the Video!
Read the Transcript!
Kirk: This Week in Radio Tech, episode 214, is brought to you by the Telos Z/IP ONE IP Audio Codec, low-latency, high-quality stereo audio over private networks or the public Internet. The Telos Z/IP ONE is the best way to hear from there. Easy to understand and meaningful audio loudness measurement has proven elusive over the history of recorded and transmitted audio. John Kean, senior technologist at NPR Labs, has been researching the problem and finding solutions. John joins Chris Tobin and me for an update on this aspect of audio recording and transmission.
Hey, welcome in to This Week in Radio Tech. I'm Kirk Harnack and delighted that you're here. This is the show where, you know, we talk about broadcast engineering, Internet streaming, anything audio of a nature not necessarily pro-audio, but audio that we get out to the masses of people, and how we handle that audio and the metadata that goes along with that. Just how we try to keep those systems sounding good and reliable.
We've been doing the show for, gee, over four years now. This is our 214th episode, so glad that you're here. Remember, you can partake of this show live. We're on the air, on the interwebs, Thursdays at 2:00 p.m. Eastern Time, which I believe is about 1800 UTC. We also have the show available – you can listen to the GFQ Network, just live audio streaming.
You can also download the audio version of the show from the website, thisweekinradiotech.com, or from GFQNetwork.com, and you can subscribe. That's what I'd really like to get you to do. Subscribe to the podcast, that way it'll show up in whatever your favorite aggregator is. I use Beyond Pod on my Android devices. I also use iTunes. So you can have that show downloading and appearing automatically for you. You don't have to be in constant Internet range. You can just download it, watch it, or listen to it, and most of all enjoy it. So I'm glad you're here.
Now, let's get on with the show. Our co-host, Chris Tobin, is here, the best-dressed engineer in radio, although maybe not today. He doesn't qualify. Chris, where are you?
Chris: Well, today I'm in a communications room in one of three large skyscrapers here in New York City that house the broadcaster transmitters for both FM and television. This particular room has two TV stations, UHF and digital, on either side of me. That's the noise you hear, the blower motors. Then behind me is an array of two-way radio communications prepared for a couple of the local authorities here in town.Off to the side, I'm not sure if you can see it here, but this is a microwave linking between several buildings. In our lens, it would be SDLs for various back halls for two-way radio repeaters to communicate and have city-wide coverage. So yes, I'm in the high top of a building that has a large mast on top and having a good time.
Kirk: Good deal. We'll I'm sure break a couple times or set aside part of the show to get a bit more of a tour of where you're at now. Let's go ahead and bring our guest in, though. I'm delighted to have this gentleman on. I met this guy in the taxi line at the NAB show. I guess it was the 2013 NAB. I got real interested in what he had to say and what he was doing in the industry. So let's bring in John Kean, senior technologist at the NPR labs. John, welcome in. Glad you're here.
John: Well thank you, Kirk. Nice to be here today.
Kirk: You're going to be talking with us about . . . I'm sorry about the slight delay here. We're going to be talking about some of the things that you're doing at NPR Labs, and a lot of that has to do with loudness. I'll tell you what, before we jump into that, why don't you just bring us a little bit of what NPR Labs does and tell us about your role at NPR Labs, then we'll jump into this notion about measuring loudness.
John: Well, happy to. NPR Labs is a not-for-profit research and development center here at National Public Radio. We are a small, independently-funded operation, so we operate much like a consulting engineering firm. We just happen to consult both to internal NPR departments and to external customers. They could be public radio members. We've had members from commercial radio. We've had television, and we've been doing work quite a bit lately for the National Radio Systems Committee. We try to keep busy and try to keep ourselves funded through projects, and they're always interesting projects I think.
Kirk: Yeah, we had your colleague, Rich Rarey, on a few episodes ago, and Rich is pretty interesting. He's a colorful character. You are working though on a pretty specific area, and that has to do with loudness and measuring loudness. Would you kind of give us the "this is a football" take or speech about loudness? What are we talking about when we talk about loudness? How do we define that?
John: Well, among the projects that we do, and we do everything from radio frequency engineering and signal propagation to audio engineering, I discovered about two years ago when I was asked by NPR Digital Media to choose the best audio codec for Internet streaming for public radio. In the course of that study, I realized that we had an even more challenging problem to solve, and that was the mismatch in loudness from stream-to-stream and from content-to-content. And in doing that research, it led me to discover that a loudness metering standard had already been developed in Europe and was really the solution to our problem.
The reason that loudness measurement is important for audio production and distribution is that it solves the mismatch in loudness from program-to-program or between any content. When you solve that loudness mismatch, a lot of other things fall into place. You don't need as much audio processing, and it pleases the listener more as well. So those are the reasons that I found it to be valuable. I'd love to talk more about how it works, though, and how we use it here at NPR.
Kirk: So I'm responsible for a few radio station streams myself, and you know, I always thought I had a bit of an ear for audio processing. I hope so; I've worked for Frank Fody who does audio processing, and you know, Omnia and the competitors too, Orban and others, do audio processing that supposedly is intended for being encoded. So the problem is it doesn't have any clipping, that's the first no-no, and things like that.
But here's my point. I always thought it was enough to use a reasonable audio processor. I'm talking about streaming, here. Use a reasonable audio processor. In my case, I tend to use audio processors intended for streaming. Then hit the encoder. Now, I realize that the metering may make a judgment difference in how I'm hitting typically a Fraunhofer-branded encoder suite, but hit the encoder at a good, consistent level following that audio processing.
I've always been reasonably pleased with what I'm hearing back on my own streams. Of course, being a radio station, at the stations that I own, we're pretty tuned to having consistent levels from source to source to source, so the product that I end up getting out of my own stations I'm pretty happy with. Tell me, am I doing something wrong, or should I be concerned about more than I am concerned about in that scenario?
John: Interesting question. Well, it's purely a matter of opinion as to whether audio processing is needed or what type of processing. Clearly there is a value in having a safety limiter to avoid hitting full-scale and causing audible clipping, but what I've found in the process of using the loudness metering here and running our own test stream out of NPR Labs which comes from audio straight from NPR News Production, straight from the Neumann microphones with no audio processing.In fact, it is that very often you don't need additional processing, even in order to get reasonable matching of levels. And when you do that, at least this is - now, this is my opinion talking here. You don't need the audio processing, and sometimes it actually sounds more real and has more impact to the listener.
So audio processing is valuable, certainly, for broadcast because we have the FCC modulation rules to deal with; we have a noisy medium and at least with streaming we have a 16-bit linear digital audio system. So we don't have a dynamic range issue there. There is plenty of dynamic range. We don't have to process to protect for that, other than maybe the safety limiter. But in my view, you can have less processing and have a greater impact, a more effective experience for the listener. Sure, go ahead.
Kirk: I'm sorry, you have a great point about the 16-bit. Even though most streaming isn't using PCM linear, I do want to return to that subject because I've got some questions about what the effective dynamic range ends up being. Chris Tobin, you've done a lot of streaming work yourself. What has been your experience regarding loudness? At least out in the field, what ends up working and maybe what ends up being the biggest mistake that you're finding people who are streaming make with regard to the loudness and consistency?
Chris: Well, the first thing, before we started the show John and I were talking about the loudness meters and how to read them and understand the difference between VU, dB and full scale. So I would say the streaming experiences I've had with folks, one, they don't understand what loudness is. Reading the meters is almost impossible to understand.
And the other thing is, as we all kid about this, what is loudness?
It's relative, in some respects. I have a speaker. I have a knob. I turn it up and down. The meter on my computer, the sound part says the levels look good. I'm streaming. I have somebody dial up the website. They listen to the audio stream and go "Hey, it's really too low," yet it looks good on my encoder.
So the experience I've ran into is just education, people understanding how to use the stuff and the references as we were just talking about. It's going to be hard, you know? As far as quality of streams, pre-process the audio properly, don't hit the encoder too hard, and you should be in great shape. After that, as John pointed out, understanding levels and how to read a peak, average, integrated, and all that stuff . . .
Kirk: Chris, maybe your experience is similar to mine, but the biggest irritation that I have had, and it seems to be getting better in the industry and I haven't been irritated by it so much anymore, but the biggest irritation I've had with regard to loudness and streaming is when these insertion ads are done and they're doing it by I guess you call it stream-splicing. The ad is inserted at some place that's not the radio station.
Now, I know the people - I've met with some of the people that are doing this, and I realize that they get the concepts of equal loudness, but I don't think they have the tools to make that happen at this point when you're just splicing in a stream of some other content over what has already been produced. Has that been your experience, too, that that's the biggest shift in level in irritation? Chris?
Chris: It's not the technology. You're dealing with, I'll say, the industry who has . . . the industry has decided to split the responsibilities of what we consider the broadcast chain, the broadcast revenue stream. I've worked in several operations where we were streaming start-to-finish our product, inserted commercials for the stream for our product, and every level was where it should be. John would be very proud of it; he would've enjoyed it. His experience would've been the same as if listening to the radio.
Then, one day, the corporation decided anything relating to the Internet was this group of people. Everything considered broadcast will remain with the other group of people. The group of people with Internet have no idea how things work in the realm that we're discussing, and that's why 90 percent of the splicing that takes place never works right as far as levels, because they're not referencing the same thing the broadcasters are doing, nor do they want to learn it or understand it. They believe they know better because they've been empowered by those who believe they know better.
I know this is probably a nasty thing to say, the approach of the biting the hand that feeds you concept, but that is the reality of it and that's why you and I and others get aggregated by these level differences when insertion takes place. It's similar to what cable TV experiences as well.
Kirk: So John, anyway, so there's a take from a couple of broadcast engineers. I guess we have the biggest issue when things are out of our control, and it turns out they get kind of messy when they're out of our control. But maybe that's not what you're really talking about. I realize that's kind of an exception, and it's looking for trouble to externally splice a stream and insert other content and then return you to the stream, but I guess most of your thoughts about this are at the content creator's point-of-view, and from the point-of-view of the people who are actually getting the audio into the stream encoder. Would that be fair to say?
John: Yeah, I think that's fair to say. I don't want to get wrapped around the axle about audio processing. There is a value, and it's a very personal choice for broadcasters as well as streamers, but really the point that I'm making is loudness metering, starting at the production point, is the right way to go because it solves the matching between content produced by different people, by different content organizations, and it makes it more compatible. When it's handed off to a streamer or a broadcaster, it's easier to distribute and it tends to please the listeners more.
What we find as a side effect is there is less audio processing needed, and that, I think, is somewhat of a benefit in my opinion. But in any case, it is the right way to approach this. We are now doing that here at National Public Radio. All of the technical divisions, it seems, have decided to get on board with my recommendations. That includes audio engineering. All of the control rooms are being converted to loudness-based metering, including our bureaus in time. Our satellite distribution division and our digital media departments are all looking to use loudness-based metering because they realize it's going to resolve the inconsistencies that they get as feedback from our member stations and from our listeners in the public.
Kirk: I just thought of a whole slew of questions here. I'll preface this by saying it seems to me that one of the biggest problems that we've had over the decades with loudness from one content producer, typically I mean a disc jockey, but from one content producer to another, is when I started in radio in 1977 I suppose, you'd go into a commercial station, a rock station, and the VU meters on the old audio console are just slammed to the right and pinned. If they go below zero VU, there must be a problem.
At the same time, I got started in public radio. You go into the public radio station, and I started with an RCA console, ten pods or so. If the meters, the VU meters, were above minus 20, then you weren't watching them carefully enough. I swear, that's just the way it was. I don't know if it's still that way or not. I think it's a bit less so.
But my point is it seems you're talking about loudness metering, and I agree completely, we're on the same page. That's where things need to go as long as - and I assume you're also taking, in your vision of perfection here, you're also looking at a way that warns you if peaks are too high. But what I'm interested in is some kind of metering that is so dead nuts simple for an operator to interpret that he doesn't have to make a value judgment. Is minus 20 the right place, because the public radio station manager told me what's where it's supposed to be? Or the engineer at the AM rock station told me that the signal gets out farther if we modulate more.
So I'm looking for a meter that my daughter could walk in the door, look at that meter, and tell you if the level is being run right or not. Is that kind of what's in your mind as well? Because that sure would be exciting.
John: Yes, I'd say so, because loudness meters generally have a fairly wide scale range. In fact, we have a picture that we can show later, but the idea with most of these meters is they have enough available display range so that you don't have something buried at the bottom of the scale where you can't see that there is audio, or that it's off the top of the scale. It is going to be sort of visible within the range of the bar graph and easy to view.
Also, the ballistics of the ITU BS.1770 standard, which is the one that I'm talking about, has a ballistics so to speak, which has a rise time and a fallback, which is very natural and easy for the eye to follow. I would say it's a meter that is easy to work with and that even your daughter would find easy to pick up and use.
Kirk: Is this a good time to look at some of these pictures so that we have in our minds what you're talking about?
John: Sure, absolutely.
Kirk: Andrew, what pictures shall we have a look at first here? Whichever one . . . here we go. What is this we're looking at?
John: Well, let's see. If we back up to the one which is the blue and red line graph, that might be a good start. There you go. That one is maybe something that'll help some of the technical folks in the audience today to see the difference between a loudness meter and a peak reading meter.In this graph, there are two line tracings. The red one is the peak reading meter, which is what we typically use for all mixing consoles and distribution equipment. We're looking at the signal peak level. Now that has a value in telling us whether there is a risk of reaching overload and creating audible distortion. No argument there, that's a useful thing.
The other trace which is on this chart is in blue, and that's below. That's the loudness meter, the ITU loudness meter. What it's doing is following a characteristic which resembles the response of our own hearing mechanism to the loudness of sound.
Now when you compare these two charts together, these two graph lines together on the chart, you'll notice that there are times in the red, and I'm looking over at my copy of the chart on my screen, that the peaks reach full scale. But if you look down directly below, using the black double- headed arrow, you'll notice that at that moment, the loudness was not very high. That was for the very highest electrical signal peak. This is in about ten minutes worth of All Things Considered coming right from the studio here at NPR.
But if you look a little over to the right, you'll notice that the highest blue peak on that chart was the loudest moment in about ten minutes of audio, but the signal peak was almost 20 dB lower. So you can see right there that there is a poor correlation between signal peak and what our ears tell us is a sense of loudness. If we try to follow that red scale and use that for setting levels or for mixing, it's going to lead to these loudness mismatches. That was one of the things we discovered early on in this research.
Kirk: It seems there is a little technical problem getting that graphic up right now. If we can get it popped up in post-production, then it'll be on the copy that people are seeing who watch the playback of the show. But what you're talking about, John, is exactly what as an engineer I know to be true, and as a former disc jockey and board operator, I know to be true as well, that peak reading and the loudness that we perceive in our ear often don't correlate.Just as . . . and this is why audio processing, of course, is done for modulation technique, whether it's AM or FM modulation technique, is a transmitter . . . there we go, there's the graphic. A transmitter responds to peak voltage, or the peak of audio, and our ears respond to RMS or maybe a much more complicated algorithm than that. The notion of a broadcast transmitter is we try, because it's a bit of a noisy medium, AM or FM, we try to increase the average and decrease the peak so they are more well-matched to each other. Now John, now that we have the graphic back up, can you go ahead and tell us again what we're looking at real quickly there? If Andrew can pop that up. There we go.
John: Sure. So, you see the longest black double-headed arrow, and that's the one pointing to the highest signal peak in this ten minutes of sample program material which is mostly speech from All Things Considered, produced here at NPR. Looking down at the bottom of that black arrow, you notice where the loudness was running at that time.Now, looking over to the right, you see the shortest black arrow. That happens to be the moment when the loudest sound occurred in those ten minutes. You can see that because it's the highest blue peak. But when you look straight up to the top of the black arrow, look where the signal peak was. That was nowhere near the loudest, or sorry, the highest in terms of signal level. So that's just kind of a graphic indication of this poor correlation between what peak reading meters indicate and what we hear.
Kirk: You know, my eye wants to tell me there's probably a 70 or 80 percent correlation here between loudness and peak. I mean if you look at that graph, there's general agreement, but there are certainly enough places where there is a lack of agreement that okay, we can't just measure peaks and depend upon that, and we can't just measure loudness and depend upon that, because the equipment or the electronic ear responds to peaks, and our ears respond to a loudness algorithm.
Kirk: So we need a way to measure that gives us I guess both pieces of information, so we don't go beyond the peak because the electronic equipment that follows can't handle that. We'll get horrible distortion then. But we also want to maintain the loudness in an appropriate manner. Is that the basic goal we're looking at here?
John: Exactly. And a good example of this involves the mix of content which has been pre-processed. Most popular music, as you know, has quite a bit of audio processing, both dynamic gain control and peak limiting, so that the signal peaks are leveled out quite . . .
Kirk: You are the master of understatements. They are pretty processed.
John: Yeah, I'm trying to be tactful about commercial music production, but that is kind of the state of our industry. Now, if you were to mix content on a peak reading meter according to those signal peaks with say live speech, you start to then really show these mismatches where speech will have peaks, but they're going to be very brief, very high, with low average loudness by comparison with the processed music.The loudness meter is going to behave more like our ears tell us, and what we're going to start doing if we use this in production is have an easier time mixing, because we're going to see things on the indicator, the loudness meter, that match what we're actually hearing. We're not going to have this kind of mental battle going on between what I'm hearing and what I'm seeing.
John: So that will actually make the job easier.
Kirk: Gotcha. Wow, so, well, can we look at the graphic? Did you have a graphic of a meter? A metering device that might help us?
John: Yeah, there's a . . . I'm going to switch over to that chart here on my screen as well. Okay, if Andrew is able to bring this up, I'll show an example of a nice loudness meter. It looks very much like a real meter that you would see in the panel of a piece of equipment, but it's actually a Windows PC program which is free. You can download it on the Internet, if you look up K-Meter. It's available both for Windows and for Linux machines and it gives you a good idea of what a loudness-based meter might look like in general.
You notice the bar graph in green is the loudness indication, and right at this moment it is peaking very close to a target loudness which is marked as zero on the scale. That's, again, a kind of natural place to put that marking. In addition, there is a peak reading indicator which is the little floating red bar that you see above it. That will give you the indication of whether the program material is getting close to that digital full scale which actually occurs at the very top of that bar in this calibration, right at plus 20.
So you actually then have two meters in one. You've got the loudness indication in the solid bar, which will change color, by the way. You'll notice if it rose a little more, one more bar, it would go from green to yellow, indicating that you're starting to get into loudness range which may need some consideration. It could be the program material is supposed to be loud at that moment. Maybe there's a sound effect, or maybe there's a crescendo in music, or maybe someone is raising their voice, and that is a perfectly natural thing. We don't have to always lower things to the target level.
But on average, we want things to float around that target level. If they go much above the target level, they'll actually go into the red range as indicated here. But that little floating bar is always available there to tell us about the electrical or signal peaks, and so we've got a two-in-one display here.
Kirk: Now it looks like that meter topped out at plus 20 with reference to the zero. Would that be a level that would be then calibrated to zero dB full-scale? The top digital amount?
John: Yes, and in fact, you'll notice that there are some little buttons on the side of the display there.
Kirk: Ah, yeah.
John: That top one is lighted in green, and that gives you a plus 20 at the top of the display and zero at the target level point. However, if you go down, if you were to click on the button labeled "Normal," what it does is it puts zero dB at the top of the scale which may be what you're more familiar with, and minus 20 shows up right where we would have the top of the green bar.
John: So there are different ways of representing this, and it's perfectly a matter of choice for the user. I should mention that the European Broadcasting Union developed a standard called EBU R128 around this ITU metering scheme, and they, after a lot of research into different program material, decided that their target average would be minus 23. So it's perfectly appropriate to use minus 23, and that's what we probably will be using here at NPR. We're in the process of testing that right now.
Kirk: So that's minus 23 from what a peak would be at full scale if you were doing this based on peaks, which of course for the reasons we've just been discussing you wouldn't do that.
Kirk: I noticed that meter had other selections. There was K minus 20, K minus 14, K minus 12. I suppose minus 14 and minus 12 are other standards for dB from zero dB full scale that one might operate at?
John: They are, and I haven't talked with Martin Zuther who was the developer of this software, a really nice fellow over in Germany. He has put this software up on the Internet for free. I guess what he was trying to represent there were other scale ranges that might be familiar to people, especially if they were using analog equipment. But with digital equipment, because we know that digital has a very hard ceiling, it goes abruptly in clipping and it becomes very audible very quickly, you need about 23 dB between the average loudness and your digital full scale.
So the way the EBU standard defines this target is minus 23 LUFS, and the LU is loudness units, meaning that you're using loudness meter, but one LU is equivalent to one decibel in ratio. So we're 23 dB below the level in which a signal peak would reach the clipping point with a one kilohertz tone. That's the way they define it, and that's the way we're doing it here at NPR. That's the way I'm using it, in fact, with my private test stream from NPR labs, and it's working out very well.
Kirk: I noticed that meter also had buttons for the 1770 standard and RMS. RMS, of course, is what we've been taught about as engineers for decades now. What behavior changes with the meter, can you describe it, if you switch between these two standards for measuring loudness?
John: Well, the ballistics don't change markedly I would say, but I guess the purpose for having it there is if you're familiar with using an RMS metering system, or even a VU meter, it would be a little bit more familiar. It's really there more for reference, but the ITU-R meter is the one that's recommended. But the ballistics remain roughly similar, and the scalings also remain somewhat similar.
One of the differences, of course, with a loudness meter, is that it uses a frequency rating, because we know, thinking back to our old Fletcher-Munson curves that we might have studied in school, the ear has a different response to frequencies across the audio range. So there is a frequency waiting network in the loudness meter. There is also a specific rise or retract time for the ITU loudness meter, and so it will respond differently under program conditions than an RMS meter. They may track approximately the same, but they're definitely going to change once we put on different kinds of program material: different kinds of voice, different kinds of music.
Kirk: Before we take a break, I want to ask one more question about these meter ballistics. For people like me who for the last 34 or 35 years have been looking at analog VU meters at radio stations, what am I going to... I think my question is really this. How terrible, if it is terrible, or how good if it actually is pretty good, is the good old VU meter ballistic?
The big honking iron moving vein meter that I'm used to seeing on a big old Autogram console or something like that that's just got this nice VU meter? I always felt like my ear and my eye had good correlation looking at a VU meter. Now I realize we weren't considering headroom and digital electronics down the way. We were overdriving the inputs of tubes a lot of times. But what do you have to say about VU meter ballistics?
John: Well, it's funny you ask that because I did a paper for the NAB Broadcast Engineering Conference this year, which is in the proceedings, and on the first page of that article I mentioned that ironically we had less trouble matching the content of audio material in the days when we had VU meters because VU meters have a 300 millisecond integration time which isn't far from the way our ears behave. Our ears have a similar lag in how long it takes for a sound to persist before we judge its loudness. The VU meter just happens to do that in a reasonable way.One thing it doesn't do is consider the frequency weighting of our ears, so a VU meter isn't a perfect indicator of loudness, but it's certainly better than a peak reading meter, and that's the irony is that we replace the VU meter with a more modern and supposedly more accurate indicator which got us into this problem with loudness mismatching from content-to-content. So it is kind of a fun irony to talk about and think about the VU meter.
Kirk: You are watching This Week in Radio Tech. It's our weekly podcast about radio technology. Today we're talking audio loudness measurement with John Kean, a senior technologist at NPR Labs, so we're glad to have you along with us.Chris Tobin is with us, too. Chris is in a high-rise on the roof of some place in New York. It remains a secret. Hey, Chris, there you are. Chris, when we come back from the break, can you give us a little tour before we resume our loudness conversation since you're in such an interesting place?
Chris: Sure, I can do that. That's not a problem.
Kirk: Okay. Well, I'm sure there are state secrets that are there so we don't want to give any of those away. All right, Chris Tobin in an undisclosed skyscraper in Manhattan, our co-host on This Week in Radio Tech.
Our show is being brought to you by my friends, and thankfully my employer, Telos, and the Telos Z/IP ONE. I just happen to have one right here. The Z/IP ONE is an IP audio codec. You know, for years and years, gee, it's been 20 years now since Steve Church, the founder of Telos, brought out the original Zephyr.
That was the ISDN codec that just revolutionized so many things in the broadcast industry. We didn't have to any more arrange for expensive satellite time; we didn't have to suffer with long installation times of equalized program lines. We could do a remote broadcast from almost anywhere, including other countries, and just have a call-in over ISDN lines. Yeah, at first ISDN was hard to find, and we used switched 56 lines. Those were possible and at first available, then ISDN lines.
Well, you know, ISDN is going away. We get confirmation of this over and over again at Telos, and we also find that where they're not going away, they're getting awful expensive. I talked to a voiceover artist just a few weeks ago who told me his ISDN line went from $90 per month to $350 per month for the same ISDN line. No change; same one he's had in for years. The phone companies don't want to be in that business, at least not on the individual ISDN BRI business. They're still selling PRI circuits and will for some time, but if you don't want to buy a whole PRI for $500 to $1,000 a month, then you want a BRI and those are getting hard to get.
So what's the solution? Well, believe it or not, it's the Internet, or private WAN networks. But there are problems with using the public Internet for reliable, low-latency, high-quality media transmission, and you know what those are. You know, packets drop out. Not every packet is guaranteed to make it from one place to another. Packets arrive out of order. Packets arrive with a great deal of jitter. You know, ISDN had virtually no jitter, and yet the Internet, we're going to get jitter on every single packet. There's 50 or 60 milliseconds of jitter going on right now in my connection with the GFQ Network.
So the Internet is not a perfect place, and a lot of times the problems we find are actually in the last mile or in your own home or office network, and so we've got whitepapers on the Telos website telling you how to mitigate those problems. And one way to mitigate those is with this guy right here, the Telos Z/IP ONE. It was designed from the very get-go to deal with the vagaries of the public Internet: packets not arriving on time; packets missing altogether; packets having a great deal of jitter from one to the next; and packets arriving out of order.
Let me go through a couple little things here on the Z/IP ONE. By the way, there's now almost 2,000 of these out in the field at radio stations, at voiceover artists' homes. Okay, I've got three of them at my office in Nashville. Actually, the place I'm at right now, I'm at the University of Florida in Gainesville where they're just now putting together an emergency network of all the public radio stations in Florida, and they're all connected for emergency communications with these right here, Telos Z/IP ONEs.
We've been testing them from here to Orlando, here to Miami, and just all over the state they're building these out with redundant networks and also redundant backup through, believe it or not, NPR. NPR Satellite Services is providing redundancy to the public Internet infrastructure here in Florida.
So on the front panel, there's headphone control. You can actually switch between transmit audio and receive audio, so you can listen with headphones to what you're sending, and listen to what you're receiving. There's an OLED display here on the front panel, it's bright and easy to read; a control knob here; and buttons here for dialing or inputting text if you need to do that. Of course, the whole thing you can get into with a web GUI and do all the programming and dialing from the web GUI.
On the back, well, we have audio inputs and outputs. Let's see, here we go, I'll turn this toward the camera, that'd be better, not toward the microphone. Audio in/audio out analog. There's also Livewire connection. Oh, there it is, the Livewire connection to hook to a Livewire network, and also another Ethernet port for non-Livewire, for hooking it up for the streaming and for doing your web browsing.
There's also contact closures on the back, so you can input closures that show up at the other end of your connection, and receive closures as well. These closures can also be used to indicate the quality of the data that you're receiving. You can set it to where if you miss just one packet, it'll give you a closure. You can set a threshold of packet misses and it will give you a closure based on that threshold. It'll also give you a contact closure for receive lock to show that you are locked to the incoming stream.
In addition to all that that I've just described, there's a whole bevvy of internal features that make communicating over the public Internet an easy thing to do. We can look at graphs of the rather poor performance of packets coming into a unit, coming into one of these Z/IP ONEs, and we do buffering. We also do agile connection technology where two Z/IPs will talk to each other over that Internet connection, and they will say "Hey, my buffer is just fine, receiving just fine. Everything is good; keep doing what you're doing."
But if packets are coming in out of order, too many of them, we're getting low on the buffer, one unit will actually tell the other unit "Hey, slow down. Let's wratch it down to a lower bitrate and see if that improves things. At the same time, I'm going to increase my buffer time so I have a better chance of catching all your packets."
Way too many things to talk about right now in a short ad, but check it out on the web, and there are videos on the website as well. So if you will, go to telos-systems.com. I've got a couple videos on there describing how this works, how to make it work with Luci Live software so you can go out in the field with your cell phone and do a high-quality remote broadcast that way, too. So check it out on the web. It's the Telos Z/IP ONE at telos-systems.com. Thanks very much for supporting This Week in Radio Tech.
All right, back at it now. Hey, we've been talking to John Kean, the chief senior technologist at NPR Labs. On this show, we have a kind of rare opportunity. Where in the world is Chris Tobin? Chris, if we could take just a moment away from John's time and have you give us a little tour of wherever you're at? Tell us about it.
Chris: Okay. Just so we know, it's very loud here. I'm in a building, one of the many buildings here in New York City that host broadcasters, public safety folks, CAVI [SP] services, radio gear, large antenna structures and tower structures on the rooftop. So I'm in one of the rooms working with a couple of clients, and the noise you hear in the background is actually one TV transmitter to that side, a 75KVA UPS to that side, and several dozen six-foot-high, three-foot-wide two-way radio cabinets with a lot of blower motors in it for keeping things cool.
Behind me, as I turn, is a combining system for those of you in the RF world, ham operators and amateur operators. That's a TX/RX combiner. It takes multiple transmitters, combines them together, and then one antenna is used for all those transmitters. To this side is the actual transmitters themselves. I had to blackout which agency uses them, but they're standard Lambo [SP] [radios]. The microwave linking between buildings is done on that side there, and those microwave units are radio and Ethernet, so they actually have a microwave signal. RF comes out of it and converts it to an Ethernet signal, into a Cisco switch, and into the network of the various people using them. That's part of it.
So now you have all these antennas and structures and things that sit on top of a large tower that you drive by. Here's how they come into the building. I can't zoom in, so this is the best I can offer you.
Kirk: That's good.
Chris: That's what they call a wall mart penetration, and those cables are the transmission lines up to the tower structure above me, and they come in through those insulated holes. Just down at the bottom is a grounding panel, so the transmission lines are grounded so if there is any lightning strike or anything of that sort, hopefully the lightning just gets arrested at that copper panel.
So now we take a look over this way, and now we're looking at where there is a combiner system for a UHF TV transmission system. That's not a Dalek for those of you who are Doctor Who fans. Then off to this side here, in the center of the screen, you'll see the 75KVA UPS. And just above it is part of the air conditioning system that keeps the room at a comfortable temperature of approximately 68 degrees Fahrenheit. The humidity level is about 30 percent to 35 percent. That's part of it.
Then there's the compressed - or the nitrogen that we use for transmission lines. Remember, transmission lines are an RF signal. It's electricity. Moisture and vapor do not mix, so nitrogen is pumped into the lines, and that's what they call a manifold. Those various valves and meters that you see are feeding nitrogen to the various transmission lines like the one right above us. That is two of the TV transmission lines going up to the rooftop. There they go through the wall.
Let's see if I can get some light on it. Oh, that's the best I can do there. So that's what you have on this floor. Then in the back of the room is some spare transmission line, six inches in diameter. So let me put this back to here so I can show you one more thing. Yes, the sound in the background is the lowers coming on and offline.
So what does one do when you're measuring signals? In audio, we have XLR connectors like on the bottom of this microphone. Then in the RF world, the transmission world, we have transmission lines. But still, the test equipment uses a small end connector like on the end of this device, this mechanism I have here, but then the transmission line itself, as you can see, is six inches wide.
Kirk: Holy cow.
Chris: That's a biggie. So for those of us that have late night work here at the transmitter site and you have nothing to do waiting for people, you can adopt a nice little party hat. A little heavy, but it's fun to walk around with.
Kirk: [Laughs] Wow. I guess it's frequency-dependent, but what kind of power level is used with transmission line that's that big?
Chris: Oh, this? 17,000 watts? 25,000? Upwards of more. It depends.
Chris: Then it just gets reduced down for the test equipment. That's the idea behind what you see.
Kirk: I see.
Chris: This reduction is taking place.
Kirk: Hey, those UHF . . .
Chris: You'd put an end connector on here.
Kirk: Yeah, it looks kind of ridiculous, but that's what you've got to do.
Chris: That's just some of the things we're doing at this place today.
Kirk: Wow. Those TV transmitters, you said there were two UHF TV transmitters in that room. Is that right?
Chris: Yes, yes. Digital TV, digital UHF. They're low-power TV here in New York City.
Kirk: So on the ceiling there, I saw some plumbing for those. What was that, coaxial transmission line or was it wave guide?
Chris: Well, it's actually transmission line and wave guides. Wave guide when the signals are being combined from the two transmitters, so they come together in the wave guide, then they're converted to the round coaxial - not coaxial, the RF plumbing. So yes, wave guide gets converted then sent off.
Kirk: To me, wave guide . . . I mean I'm not in the TV world for transmission. Wave guide is amazing. Explain how it works? Well, it's like coax, but there's no center conductor.
Chris: Yeah, pretty much. Wave guide, it's what's inside. It bounces inside. So let me take this little guy for demonstration purposes. So if you can think of wave guide in a rectangular form, this is circular, but wave guide operates with bouncing the signals on the edges inside this cavity if you will. You can tune it by literally putting little plungers, or screws if you like, into the path. That will tune the wave guide to what you need it to do. Or you can just take a hammer and bend it, and that'll do it as well, but it's usually not recommended for a normal operation.
Kirk: Well, and that's one difference. Coax is not tuned. It works up to a maximum operating frequency that has to do with the size of the physical components and such, but wave guide is tuned. Different TV channels use different sizes of wave guide.
Chris: That is correct. As a matter of fact, the wave guide . . . no, I don't have the smaller stuff here. The wave guide for seven gigahertz or two gigahertz is this wide, if you can see my fingers.
Chris: Then for say channel 34 or channel 30, which is much higher, the wave guide is this wide. That's all tuned, you're absolutely right, unlike coax which is based on power. There are certain frequency restrictions, but it's a little more broader than wave guide which is more specific.
Kirk: Sure, sure. Well Chris, thanks for the tour. I appreciate that.
Chris: Any time. It beats sitting in the studio or sitting at home or in an office, so why not?
Kirk: Now you're in a noisy place there, and that's a great segue back into our conversation about loudness with John Kean. John, welcome back in. Thanks for holding on for us. We've got about ten minutes or so left in the show. You mentioned earlier, and I thought hey, let's get back to that subject, that with AM and with FM transmission, we have a somewhat restricted signal-to-noise ratio.Even if you don't want to do audio processing for the purpose of just being loud and aggressive-sounding, you may want to do audio processing to get your symphony concert from being an 80 dB signal-to-noise ratio of loudest to softest down to a 20 or 30 dB ratio so that when you transmit it over this noisier medium you can hear everything on the other end. You're not cranking your radio up just to hear hiss and hear the violins over the hiss.
But then we also think we need to do audio processing so we can help the listener in their listening environment. Granted, AM and FM, that signal-to- noise is going to be fairly consistent; it's the technology. But a listener's listening environment may vary from a quiet room like I'm in, or like you're in, to a place where Chris is or a moving car. So what are your thoughts about loudness and maintenance thereof for the benefit of maintaining a good listener experience that may not be quite true to the original?
John: Well, we might have the opportunity to answer that question later this summer. The Consumer Electronics Association represents a large number of manufacturers of consumer electronic devices, and they're interested in this issue of loudness matching from content-to-content, particularly when it means that their customers come back to them and blame their hardware.So the CEA has asked me to develop and carry out a study on loudness range requirements for consumers. In other words, as you said, if you're in a quiet environment, how much loudness range would you prefer in order to hear all of a particular musical passage? Or a talk program. It might be much wider than if you moved out onto the city sidewalk and you were using little ear buds in your ears. You might have a much nosier audio environment for background, and therefore you might prefer less loudness range.
And so what we're going to be doing in the audio lab here at NPR Labs, which is a large about 18 by 23 foot NC-20 noise rated room, extremely quiet, it's one of those rooms where when you walk in you hear the blood pumping in your ears.
John: Well, we're going to be doing testing with a quadraphonic array of loudspeakers reproducing different sound environments at calibrated sound pressure levels, and we'll be having people listen both on loudspeaker and with some selected ear buds to audio material of different loudness range, and we're going to be asking those listeners to tell us whether the loudness range is too great, and they would prefer less, which is a good guide as to how much audio processing is needed. So we'll actually be putting that report out, and it'll be a public study later this year, to answer that question.
Kirk: John, I've always thought that one ideal way to solve this problem of listeners being in all kinds of different environments, and radio stations and other content creators and transmission people squeezing their audio for perhaps that worst possible case. You know, someone in a garage working on a car wants to listen to a talk show and some music, and there are all kinds of noises going on, and so he's got to crank that radio up and the radio station is probably processed pretty heavily.
Wouldn't it be great if - and maybe this is too much for a consumer to do, I don't know, but wouldn't it be great if there were automatic processing that could take place in the receiving device? I know we've got a little bit of that in our technological society right now. Many car radios, you can adjust to adjust their loudness based on the speed of the car. And of course there are noise quietening headphones which provide a nice environment right there at your ear.
But I always thought it'd be pretty interesting if receivers, radios and now computer programs and play out systems, would have a very simple, very predictable algorithm that would not colorize - it would not color the texture of the audio at all, but would simply give it less dynamic range or more dynamic range in a very predictable way such that those who are treating the audio from the transmission point would be aware that that's out there?
And audio processing at the point of transmission would have some awareness of whatever the algorithm is in the consumer's device, so as not to fight with that. Then let the consumer, or maybe even let the radio automatically, monitor the loudness level in the area and adjust its process accordingly. The actual loudness output of the radio wouldn't change, but the dynamic range would change. That's my utopian ideal of how it all ought to work. Any agreement to that?
John: I think that's a fantastic idea, and one that I think you ought to rush out and patent because that's something you definitely could take to the bank. And, in fact, the data from our study which I'm going to be doing with Dr. Ellen Sheffield at Towson University who is a psychiatrist, or psychologist rather, is going to be designing the listener testing process for this with me. What we will be able to identify, I think, in the study is to give you the data on how much range you really would prefer given a particular SPL noise environment.And as you know, there are already devices out there, there are tablets and smart phones with microphones built right in, so it would just be a matter of writing the software to measure the SPL of the noise and make a decision on how much to restrict the loudness range according to what our listener tests indicate. So I hope that we'll be able to give you the data, and you can make a product.
Kirk: I've got to believe that somebody else has already thought of this. In fact, I thought that Dolby, with regard to television audio - okay, it may not measure the noise in the listener's environment, but it can provide a consistent listening environment, and it is possible to have controls at the user end.
John: You're right, yeah. Television had to face this several years ago with the complaints about loud commercials, and of course as we all know, Congress passed the CALM Act which mandated devices to control unruly loudness, especially from commercials that didn't align with the program content. They're very aware of this. Of course, they have no way of knowing what the noise level is for listeners under different environments, so I think we have the opportunity here to develop some original research which might help develop an improved product, which would actively determine are you listening in a quiet living room? Or are you listening in your car with cabin noise? Or are you out on the city street? Then helping do an adjustment for the listener, so they don't have to be doing that gain-righting process, and it certainly avoid this problem of a one size fits all solution where we wind up processing for what we think is the worst-case environment at the expense of maybe the listeners who could hear it with full range. I think we'll be able to make an advance with this data.
Kirk: John, considering we just have another minute or so left, what final thought would you like to leave us with? What would you like the listeners and the engineers who are taking in this podcast, what would you like to leave them with with regard to loudness measurement?
John: Well, probably folks are curious about these meters. I mentioned the K Meter which they can Google on the Internet and look up and they can download.Also, I've written a loudness measurement tool which is available on the NPR Labs website, which is NPRLabs.org. Just go to our tools section, and that's a unique little software which will allow you to decode Internet streams. It provides a high-fidelity player. It will give you an integrated loudness measurement, so if you're curious about how does the loudness of my stream match to the EBU-recommended standard. It'll be able to measure that. It also measures audio files of a variety of codecs, and it does it at about 30 times the playing speed so you can assess the loudness range, an ambi average loudness of an audio file, a one-hour file, in about two minutes.
Kirk: Oh, wow.
John: So you can download that for free.
Kirk: That is very cool. Very cool. John, this has been fascinating. I'm sorry that we're out of time, but I'd love to ask you to come back again sometime when you have maybe some results from your studies about those listening environments. That'd be very cool to get an update on.
John: Well thanks, I would love to. It should be an interesting result, and I'll look forward to feedback from your viewers and listeners.
Kirk: All right. Our guest has been John Kean, the senior technologist at NPR Labs. Chris Tobin has been with us as well with a quick tour of an undisclosed transmitter site on top of a skyscraper in Manhattan. Chris, you still with us? Did Chris drop off?
Chris: I'm still here, yes. It's noisy, so I'm trying to keep the microphone muted when I can.
Kirk: I'm sorry you haven't been on. There's this natural dichotomy between where you are with all that noise, and talking to John Kean.
Chris: Well yeah, I originally was not planning on being here today. I got a call for sort of "Not an emergency, but we need some help with a few things. Are you available? And if you are, that would be great." Sort of saying "Wink, wink, really, can you get here as fast as you can?" So that's why I'm here today, otherwise I was trying not to be in a noisy place for John's conversations.
Kirk: Although actually that mic does a great job. Your voice is very clear above the noise. Hey, if folks want to be in touch with you, Chris, you are an IP solutionist. I think you have all the business you can stand, but if somebody wants to outbid your other customers, where do they reach you?
Chris: [inaudible 01:02:46], that's all. You'll reach me that way.
Kirk: Where's that?
Chris: By the way, John, if you're down the hall from Steve Dinsmore, tell him I said hi.
John: I'd be happy to, Chris. I'll definitely run down the hall and tell him.
Kirk: All right. Chris, thanks a lot. I appreciate you being here, and wherever you are on short notice. John Kean, thanks again for being here from NPR Labs, and also say hello to your colleague Rich Rarey. He was a blast to have on the show as well a few weeks ago.
John: Will do. My pleasure, Kirk.
Kirk: All right, our show's been brought to you by my friends at Telos, and the Telos Z/IP ONE. Go see it on the web. Here's one right here. We've just put 30 of these in throughout Florida, working great. On the web at telos- systems.com, and look for codecs, the IP codec, the Z/IP ONE. Well, coming up on shows in the next few weeks, we've got some exciting guests for you. You'll just have to stay tuned to see who they all are. We're working on a couple of really cool ones.So that does it for this show. Thanks to Andrew Zarian at the GFQ Network for producing the show. Be sure you check out all of the shows that are on the GFQ Network. If you like this show, I know you'll like What the Tech, another GFQ show, and some other shows as well. You might like Mat Men. Check it all out at gfqnetwork.com. That's it and we'll see you next week on This Week in Radio Tech. Bye-bye, everybody.