This morning at KDX we are doing an interesting study. Operating only on our AM frequency with FM and streaming turned off, we are broadcasting a string of 3 programs and looking under the hood at the audio formats of those programs.
The first two shows are from PCJ Media including "The Happy Station Show" and "Media Network Plus", both in MP3 stereo with a sample rate of 44.1 kHz, but their bit-rates are set differently: the former is at 128 kbps the latter at 256 kbps.
The third of our test shows is also MP3 but its sample rate is at the unusual 16 kHz with bit rate 96 kbps, the program being "World of Radio with Glenn Hauser". The low sample rate gives the program the sound of a shortwave broadcast and Glenn's program is mainly about shortwave, but we don't know if he does this intentionally.
We have done many other comparisons in the past weeks and can tell you that virtually all radio programs distributed as audiofiles come in the MP3 format. I think only once in 11-years did someone send shows in the WAV format.
Although many stations stream in AAC we have never seen an audiofile distributed in AAC.
The most common two sample rates of distributed programs are 44.1 kHz and 48 kHz.
Of all the programs run by KDX some are produced in stereo, others in mono. The bit rates are all over the map, there is no standard: 24 kbps all the way up to 320 kbps.
It would be interesting to be told why the many producers choose the settings they use.
I work with a lot of audio producers (mostly for podcast, some radio). A lot of them actually don't make their choices, the software does it for them.
One fellow had a very old bit of software he had not updated, which would only encode at a low bit rate. Others just let the default setting in Audacity take care of the chore. It never occurred to them to change the settings.
Many compress their files so tiny as not to eat up whatever precious space is left on their 80 GB hard drives. Others keep them small for the benefit of the receiving party (who doesn't want to wait 20 minutes as their DSL pulls down a large 320kbps MP3 file), while others have a data cap on their ISP plan and keep the rate low so they don't pay extra.
And frankly, some just have bad ears and dont care.
What is really a headache is a file encoded with Variable Rate Encoding instead of Constant Rate. Let a file like that be imported into your playback software and you'll get stutters, chirping, and other nasty things as your computer tries to decode on the fly.
When a show airs on AM, an MPEG file with a freq response of maybe 11 kHz (22k sample rate) isn't at all bad. Vinyl records were just as limited, if not worse. Its when the encoding rate is so low that the audio is filled with gargle and drifting. Then its crap.
That's About It
I can bet that your take on the situation is right on the nose, BillyBurg.
Probably few users of audio software have the comprehensive education needed in several related fields: the audio spectrum and digital storage technology, to simplify.
Understanding broadcast specifications is useful, knowledge of human auditary physiology, musical theory, also useful. Room or outdoor acoustics, perhaps the list goes on.
Interesting what you said about VBR (Variable Bit Rate). I have never been any more than curious about it, thinking that its description seems attractive, but most producers use CBR (Constant Bit Rate) and now we know why.
When I released regular programs for use by other stations a few station owners demanded at least 128 kbps and felt that anything less is low quality in their opinion, but for streaming and AM radio and compact file size you are right... lower bit rates can do the job very nicely.
In the early day when digital audio was getting introduced on computers, I had programs and hardware on Amiga Computers that did a fair job of capturing and editing 8-bit sound, except that I found that 8-bit audio gives me a headache. I have never seen a scientific explanation why that happened.
At work we download weekly dozens and dozens of syndicated programs for broadcast, either on one of two 100,000 watt FM stations, or a 5,000 AM station. All these programs arrive as mp3 files. Most at 128, a few higher. These are anything from simply two minute news programs to three hour music shows. They all sound fine and never has a listener complained that the fidelity was poor.
The Oompah Hour is distributed at 44.1/128 in stereo and is carried on 27 stations, most FM and no one has complained about audio quality. Of course, all the music is from records, some scratchy so you know it's authentic.
TIB
Distributing by Stream
I'm kind of bending the original subject of this thread, which dealt with the technical specs of digital audiofiles...
Today we're looking at streaming, having just scrolled through many of the 14,146 stations presently available at the Icecast Directory where KDX can also be found. I like to learn what other stations are doing and how they are doing it.
As always the majority of them are sending pop music by automated playlist, it is very rare to hear a live voice on any internet stream.
A surprising number of stations are highly distorted because of over-driven levels, many have excellent sound as evidence that digital sound can be very good, and there are some silent streams where the operator may have fallen asleep.
Icecast delivers many audio formats including MP3, Ogg Vorbis, Opus, WebM, AAC & AAC+. Since our last survey there is a distinct increase in the use of AAC.
Very few news stations are found most of which are licensed AM FM stations carrying public radio or one of the corporate news networks.
KDX appears to be the only independent news talk radio station streaming on the Icecast Directory.
CB, I've always thought a useful tutorial video would be one where a Part 15 operator 'chased' levels all the way from the front end of the process right to the transmitter and showed the viewer what it all meant.
First, show the original audio, and whether or not it needs normalizing and/or peak limiting before being ingested into the on-air computer. Especially in talk programming, one peaky lip-smack or 'tsk' will throw off normalization as the process will scale all levels against that peak. Decide on a proper average level for automation import -- I use -14 dBFS which, on my gear, just tickles the "zero" below the red zone on my analog meter.
If it arrives too loud, normalize it down. But if it comes in loud AND distorted, do not air it.
Second, match mixer levels going in and coming out. Make sure all levels are in the same ballpark. That means computer output, telephone/Skype levels, mics and CD/MP3 players. A wise man once said, "the 'gazouta' has to be the same as the 'gazinta'."
Lastly, what is coming out of the airchain processor (comp/limiter)? Is it overdriving the transmitter? Back it off if it is.
At any point along the way, if any audio is not reproducing properly, reject it, replace it and find out where it is sucking and why. It's better to go with a good-sounding Best-Of replay of a production than to air a new but lousy sounding program. If it's on your station, you have every right to demand an air-quality file. Otherwise, there are plenty of show providers out there who would love to be on your stick.
As another in the commercial broadcast industry, I can confirm all content sent to our full powers arrives in MP3. Almost always 128kbps 44.1 khz, however due to compatibility issues we tend to convert most of it to WAV. This has no impact on audio quality, the losses from mp3 are aleady there but WAV atleast wont make it any worse. Our playback software has issues with mp3.
A Lot Packed Into Digital Audio
YES Billy Burg, that would be a great video - showing in clear detail how to tailor the level during a production process.
Only the other day I snapped my fingers while making a point on the microphone, and the "snap" put a peak on the audio that dominated the entire attempt to normalize the audio.
And a whole other dimension opens up regarding the technique of producing a decent video for YouTube. We could talk at book length about how many poor videos get posted.
Turning to the point made by Mighty 1650, I do a twist on what you described as far as converting MP3's into WAV format for special reasons: I happen to master my productions in WAV, and when using sound clips from other people's MP3 programs I convert them to WAV so they will be "preserved".
And now for a dose of reality.
In the "real world" no one takes time to tweak and produce what they're putting on the air. In the real world no one has time for that.
After nearly 44 years on the air in commercial radio I have never worked at a station where any serious amount of time was spent tweaking levels after the fact. Of course for many of those years things were recorded live to tape and diddling levels afterwards, diddling with eq and compression and all that after the fact didn't happen. How about real radio stations where a good share of the programming is LIVE?
The whole key, the secret to not sounding crappy, is to ensure whatever you're putting on the air is not distorted. That's it. Distortion in, distortion out. There's no 'after the fact" fix for that to fix it, except doing it over.
Your station, any station, whole have SOME sort of processing before that audio gets to the transmitter. I can't say that I've ever been anywhere where there wasn't some processing, that at the very least kept the levels under control. I presently am chief engineer for three stations, two 100,000 watt FM's, once "adult contemporary" ( whatever the heck that is these days) and one 5,000 watt AM that's oldies (early 1950's to mid 1970's) with a heavy dose of news and sports (probably 40% of a typical day is live people talking and live sports, either national, state or local). We also carry some sports on the FM's. I have worked for two of these stations for about 30 years, and the last we just bought two years ago. So I've listened to them for thousands of hours. Audio sources are everything from mp3's, CD's, records, Tapes, sports feeds from national networks to local high school games. Live remotes either via computer of Marti, weather reports from meteorologists who wife their reports from various qualities of home studios. School news programs recorded by kids with cheap computer mics, to agency ads that have eq and levels all over the place.
All our audio, on the air, sounds perfect. Because the processors handle this. Low levels, hot levels, tinny eq, whatever. Now, at these three stations each has a recent and expensive Orban processor. BUT, here at my Part 15, the processing that is included in my Procaster does nearly as good of a job, at least as far as controlling the levels and most of the overall sound. Naturally the Orban's have a ton of other features and tweaks offering far more control -- such as not bringing up the crowd noise in the background of a baseball game when the announcer is quite for a few seconds in a break in the action, and keeping things under control when the team wins the championship and the announcers and crowd go nuts.
Real, live radio has varying levels and emotions. Of course the operator can ride the gain, but no matter how perfect they are they're gonna mess up. I've beat into the heads of everyone on the air it's ALWAYS better to come in a bit quiet that to blow the needles off the meters. The processor easily takes care of quiet programming bringing it to where it needs to be, and it can bring hot levels down to where they should be, but if you're pumping in distortion, all you'll get is distortion -- at the right levels.
Get some processing and forget all that other BS. People always like to make it more complicated than it needs to be.
Now, obviously you don't want to be so low that the processor is bringing up a bunch of noise with the wanted audio, but in today's world of digital recording there's no tape hiss, etc. You've gotta be REAL low before a processor can't make it right. Or at least right enough that a listener can't tell the difference.
In real radio with all the producers, commercials from various sources, interviews, news clips, and all the varying audio sources you'd need a staff of people whose job is only to eq, tweak levels, normalize, and screw with audio to "get it right" before it went on.
You do not need a $7,000 Orban to get your on air levels and basic eq under control and sound better than if you didn't have processing and tried to do it all yourself.
Sure, in the olden days you were dealing with tape and well used machines that would have tape, hiss, hum, and lord only knows what else, plus tube electronics, etc that in the abuse world of commercial radio would create noise and riding the gain tight was important. But with digital recording, no tape, silent solid state electronics, etc. as long as you're not distorting, you're good.
Now, I know there are those who will preach that if you ride the gain, produce the interviews with an hour spent in post production, etc.. things will be BETTER. Maybe they will. But better enough that a listener will notice the difference? No. And these are things that will likely go on the air once. The news changes everyday. Those commercials only run so many times a day and for only so many days. That interview only airs once, etc. It's different if you're producing a record that's hopefully going to be played on the air for decades, and the bass track is distorted, or the eq is way off, but that's different that the living, breathing thing that is broadcast radio.
Use common sense and basic decent techniques and you're fine.
TIB
In Defense of Audio Art
Tim, are we saying that all chefs are alike?
Certainly your "real world" point of view comes from the perspective of the wholesale requirements of full time radio stations, but things are different at the point of production of distributed material such as studio produced commercials or podcasts.
The example of sports announcers being balanced against loud crowd noise is the kind of ratio I most recently considered in fine-tuning the post-production compression of my voice microphone: unlike the common approach of gating background noise out of the studio sound, I prefer to "hear the room" in my own Blare OnAir shows, and a slight nudge with the compressor does the exact amount of room "lift" while allowing my voice to dominate.
Big city studios are known to put in hours producing a 30-second commercial for distribution to stations.
Other program sources put their effort into bad jobs, like whoever does the MP3s of the non-commercial version of the daily Thom Hartman Show, which are crushed in compression so severe that the sound has little "presence", and I spend a few minutes running the shows through an EQ slope that restores a broadcast quality.
Looking into how the Hartman Show is done I've found that it is done as a TV show on Free Speech TV, with Thom wearing a lapel mic on his shirt, which does not give the same sound as a studio microphone in front of the mouth.
Some of the best audio of any programs I air is heard on the daily RT News that has employed a slight reverb, well composed musical themes and bridges, and distinct voice readings, and an elastic compressive action that contributes to the "dimension". It is a very entertaining newscast with appeal even to people who don't like news.
At the radio station level not all stations sound like they have a Tim working for them. Some radio engineers know their RF but have no sense for audio.
Some of the worst ever audio I've ever heard comes from "big city studios". Mostly they seem to spend a lot of time making things awful. The worst are those we receive for car dealers. Clients obviously tell them to make their commercials as loud as possible. These are a solid wall of super compressed maximum level eq'ed to hell blobs of sound with no dynamic range whatsoever. Believe me, what comes from big city studios and agencies is, for the most part, crap. And I'm trying to put all that politely.
Other big city studio techniques include out of phase stereo effects, stereo ads that don't translate well to a mono AM station (and there's never a mono mix version available), and my favorite, the ads that have two people having a conversation. In stereo. With one person in each channel. So a listener has to be within reasonable range of both speakers to understand the conversation. This is rarely the case. Even in a car. Especially in a car. When one voice is in the passenger side of the front seat and the passengers legs are partly blocking the speakers on that side, the driver only hears his side of the conversation well. It's the stupidest idea ever. It probably sounds cool in the studio and impresses the client but is impractical in the real world. Horrible sibilance on vocal peaks is another common problem. Ugh.
As some of you may know, I also provide a voice-over and commercial production service. I produce roughly 5,000 scripts per year. 99% of these want NO processing at all in the final product. Now, I'm sure in the case of straight voice work they may be mixing it with other production elements and adding their own processing. But even in completed commercials I do that are ready to air zero processing is done in the studio. Remember, the air-chain of the station is where the processing happens. It doesn't need to be double-processed. This is part of the problem with the car dealer ads. They process the hell out of them then we pump them through our air chain.
The few who do want some compression, or maybe some reverb, etc are those getting things produced for a podcast or internet show and aren't using any traditional broadcast style processing.
In 44 years on the air I've never worked for a station that had any processing in the production studios. It's just not necessary. Unless of course you're applying some sort of special effects like echo, etc.
I'm sure there are big city studios that do wonderful work, but they're not the ones churning out car dealership or political ads by the hundreds everyday for every market in the country.
TIB