Do you know anything about streaming radio levels?

[Unregistered]

Hello. I hope that you are well and enjoying life in general.

I apologize for crashing a forum that is intended for professionals, but I have a question that only you can answer. This: Is there some kind of standard for audio levels in streaming radio shows? If so, why are they so crazy wacky?

The NPR levels are usually strong, except once in a while they'll stream a show at about -20db (as displayed on my soundcard). Then another show will just crash in at full level. The Glenn Beck podcasts on Soundcloud are totally unleveled. From moment to moment the levels are either inaudible, mumbling off mic, or some recording blowing the speaker right out of the cabinet. I have to compress every single show myself if I want to listen to it.

I mean, don't web-streaming stations have VU meters? Shouldn't they?
I think they should.

 

[Bill_981]

I've been in radio for nearly 50 years, starting in high school as a jock, and trained by a real engineer to respect the VU meter and attempt to maintain consistent audio levels. I also learned how to "duck" the music when I talked, so as not to be drowned out by it.

The people running boards today use the meter bridge as a copy stand for liner notes, etc. When they can see the meters, the seem to feel the pointer must remain solidly attached to the right-hand end of the scale. If they need to make themselves heard over the music, they just turn the mic pot up more. In the analog board days, they could probably get away with it, as most consoles had 11-15 dB of headroom once the +3 point was reached on the meter. Actual level riding isn't as necessary as it once was because of the modern audio processors used today. On the other hand, if they try to push the levels of today's digital boards beyond full scale (0 dBfs), some pretty miserable distortion is created. Once the digital word is all 1's, that's all there is.

That said, podcasts are usually recorded directly off the board, edited quickly on Adobe Audition or something similar, and posted quickly. Similarly, many radio station streams may have some light processing in front of the encoder, (some have none), and if the input level drops out of the processing window, then it'll be lower than it should be. The larger stations/groups should have processing that's similar to what's used on the air, but tailored specifically to low bit rate audio to make it sound good. I have an Optimod 8400 in the HD mode feeding a professional sound card in our streaming server, and it sounds very consistent from element to element. Our streaming service covers our commercials with music during our breaks, but I've adjusted the Optimod output to match their level.

 

[TheBigA]

I agree with a lot of the above, but I'll add this: I've also done work at major recording studios where many of the records you hear on the radio are done, and real time recording levels are all over the place. It seems that recording engineers are taught mainly to keep it out of the red...the rest can be fixed in post. And since most recordings today are assembled from multiple performances in post production, the actual levels never seem to matter. It just pushes the real work of recording level uniformity down the chain, often to the mastering engineer, whose job it is to set the final level for the CD. Now with CDs going away, there are some in the music industry who are trying to dodge the mastering phase. That means it's possible that the standards we've come to know for levels may disappear completely.

What does this mean for streaming and podcasts? Good luck!

 

[Goat Rodeo Cowboy]

I worked in radio during the analog "we use glowing little tubes" era. Then I left the business to chase other dreams and sometimes be pursued by other nightmares. (Those are stories for another day.)

Through the years I dipped my hand into a little bit of audio here and there. Set up a sound system for the wedding of the employer I was working for. Considerable dinking around with church sound systems. Then sometime around 2005 I got thrown into house-of-worship sound big time and took on the responsibility of editing, mastering and distributing CDs of the worship events. I had a lot to learn very quickly. There were no tubes anywhere in the building! Everything was solid state. And there was no analog recording. The original master was direct to CD.... so to coin a phrase for the occasion.... I was "baptized" into digital audio whether I was ready or not.

The tell us "The Internet" is your friend... with Google as your guide. As dumb and uninformed as I was, I could quickly see that a lot of people who thought they were experts were highly misinformed. People who were writing essays on the internet to explain how digital sound works obviously had no business turning the computer on. Some of these people, as uninformed as they are, are today producing podcasts. Some of them are running 'streaming' operations. Some of them may have jobs working in the music industry.

Within the last month I came to realize that a publisher of audio-books that I am working with was publishing specifications on volume levels that were driving me nuts. So I asked for clarification. I NEVER got a straight answer. Finally out of frustration I created a miniature seven-minute long pretend audio book with each chapter named after the numeric value of an RMS audio value. The best answer I was able to get was: "If you will set you levels to be like the second chapter in your recording, you will basically be meeting our specification."

It was 13 to 18 dB lower than my understand of their standard! Why they wouldn't have a direct conversation on the subject, I still don't know. From that experience I can tell you that the various sound editing software applications (DAW) do not all meter/measure/display/identify audio levels the same way. I downloaded trial versions of several major editing programs and that is when I knew why I was struggling to meet the standard... and the recordings were sounding so strained, so tense, so bent-out-of-shape.

I went into a retail electronics store one day and asked for a CD read and write drive replacement for my computer and told them what my problem was and why I needed a replacement, and what I expected of the replacement. I thought they were going to ask me to leave the store because I was in violation of their "We don't serve stupid customers" rule.

I suspect the guy who wanted to throw me out probably publishes one of the blogs that has volume levels that puzzle you.

The radio stations today have some really wonderful devices to get the volume right. But they are expensive. Way beyond the budget of many people who stream or blog. But if you have a small briefcase full of cash, Mr. Orban and some of his competitors can fix you up with an electronic box that will give you PERFECT level, every day... all day long.

 

[Unregistered]

Should it be a surprise that Glenn Beck would drive the needle as far to the right as it will go?

 

[TheBigA]

The radio stations today have some really wonderful devices to get the volume right. But they are expensive. Way beyond the budget of many people who stream or blog. But if you have a small briefcase full of cash, Mr. Orban and some of his competitors can fix you up with an electronic box that will give you PERFECT level, every day... all day long.

I don't know...all of my audio editing software came with numerous "dynamic processing" plug ins and "hard limiting" plug ins. And to make matters worse, I didn't pay full price for this software. Plus there is in-line audio processing and limiting hardware one can buy for about $300 that will do the job just fine. Unfortunately, even free software or cheap hardware doesn't matter when the people involved don't recognize the problem in the first place.

 

[SirRoxalot]

"Guest", if you'd like less wringing of hands, and more substantive answers to your question, scroll down to the "Podcasting" forum and/or "Engineering" or "Engineering Tips" forums and repost your question. There are some very smart people who have learned to address the problem successfully, and will offer your concrete answers to your questions. Particularly in the Engineering forums. They'll make specific software recommendations, and can help you find audio harmony.

 

[Goat Rodeo Cowboy]

I don't know...all of my audio editing software came with numerous "dynamic processing" plug ins and "hard limiting" plug ins. And to make matters worse, I didn't pay full price for this software. Plus there is in-line audio processing and limiting hardware one can buy for about $300 that will do the job just fine. Unfortunately, even free software or cheap hardware doesn't matter when the people involved don't recognize the problem in the first place.

Good points. When I composed my post, I was thinking primarily about the signal of a radio station or on-line steaming delivering an around-the-clock series of audio segments and components.

I think of editing 'software' having the ability to level out the sound of the one program segment at a time... the one currently in the editor. The problem comes when a radio station collects audio from various recordings and network feeds, each of them coming in as a separate module... and maybe each using a different standard as to what is "proper gain".

The second problem with audio is that it is a little bit like "Who is in charge of the kitchen?" in a restaurant. Adjusting the parameters of audio delivery is like recipes for cooking. Every cook has their own 'artistic values' for seasoning to taste. Someone has to be The Dictator when it comes to establishing the settings on the software or the hardware.

I personally find setting the software for gain control programs to be something close to a "Voodoo Religion" in actual practice. Our original poster has observed that the Voodoo Religion folks are in need of a creative Shaman.

 

[Robert Bass]

Let's be clear about spmething here.

In the digital domain, the volume scale is dBfs, which stands for dB full scale. I notice the OP quoted one program audio level was showing -20 dB. I'm going to assume his leveling system is using dBfs, so what he's seeing is -20 dBfs. Ideally in streaming an unprocessed audio signal with full dynamic range, you want the level to be as loud as possible without exceeding -3 dBfs. Ideally, RMS levels should be around -15 to -12 dBfs, with peaks no higher than around -3 dBfs. Processed audio streaming levels would depend on how much processing was used. If the audio is heavily processed, you can get away with lower PEAKS so long as the RMS levels are between -18 to -12 dBfs.

R

 

[Goat Rodeo Cowboy]

Ideally in streaming an unprocessed audio signal with full dynamic range, you want the level to be as loud as possible without exceeding -3 dBfs. Ideally, RMS levels should be around -15 to -12 dBfs, with peaks no higher than around -3 dBfs. Processed audio streaming levels would depend on how much processing was used. If the audio is heavily processed, you can get away with lower PEAKS so long as the RMS levels are between -18 to -12 dBfs.

R

I recently significantly embarrassed myself... and at the same time carried out a "Battle Royale" with an audio book publisher on these topics.

It seems to be conventional wisdom that audio must never exceed -3 dBfs, but that seems to be "folk lore" which was generated in the 1990s and keeps rising to the to the top. Why -3? Why not -1? Why not -0.1? (Based not on what you can see on screen as your software paints of picture of the wave form, but from the statistics screen where your software tells you right down to the 'gnats eyebrow' what your highest peak is in digital values?)***see footnote

But the real witches brew in discussing digital audio is trying to wrap our brains around RMS values. I learned the hard way that Adobe Audition has TWO different calibrations for RMS: TOTAL RMS and AVERAGE RMS. If my memory is correct, what Twisted Wave calls AVERAGE RMS is what Adobe calls TOTAL RMS. I downloaded several software programs in TRIAL version and found out that various programs will rate the same audio file with differing RMS values. SO!!!! WHO is the Gospel on this? If I decide that RMS -20 is the ideal programming level, which software package is the correct package to use to measure that value? And would I use the Square Wave value as zero, or choose the Sine Wave value as zero?

Here is an experiment that will dazzle and amaze you. Record some spoken word content at a slow pace... leaving some longer pauses than you might usually do. Be dramatic about your delivery. The "produce and process" as you normally would. EQ. Compression. Whatever. When you are happy, ask your software to give the RMS value of that audio. Write it down!

Now take a copy of that file. Do absolutely NOTHING to change the actual audio content, except highlight the gaps between words, the gaps between paragraphs, the gaps between the syllables. Remove the silence. Make it sound much like one of those government required disclaimers at the end of a commercial where everything runs together. (Your software program may offer a "silence remover" feature that will do this for you.

Now you will have two audio files where the audio peaks are identical in value. But when you read the RMS value of each file, there will be a significant difference.

You can't broadcast or stream Heavy Rock Music and interviews paced the way Terri Gross does on NPR at the same RMS value and have listeners saying: "Boy, you really know how to set your hardware and software don't you!!!"

P.S. I now produce my book narrations with peaks of -3dBfs and shoot for an RMS of -20 but have difficulty keeping it under -17.5 RMS. With few exceptions, Book Narration fits my description above of 'spoken word at a slow pace'. Compress it down to -12 to -15 RMS (TOTAL RMS in Audition) and it begins to sound like a commercial for a WWF event.

***Footnote on 'Gnat's Eyebrows": Here in Georgia we have expertise on the anatomy of gnats.

 

[Robert Bass]

-3 peak is a good headroom safety margin. Obviously an occasional peak exceeding -3 may slip through here and there. But if you used -0, you risk peaks exceeding the clipping point because you have no way of seeing it coming. If you know you are occasionally seeing anything higher than -3, you are better aware that you are approaching the clipping point.

R

 

[SirRoxalot]

Goat, you have to understand what RMS means - Root Mean Square. It's a calculation generally used to allow DC formulas to be applied to AC current, which typically has a fixed frequency and fixed amplitude.

"The root mean square (abbreviated RMS or rms), also known as the quadratic mean, is a statistical measure of the magnitude of a varying quantity. It is especially useful when variates are positive and negative, e.g., sinusoids."

Audio has neither fixed frequency, nor fixed amplitude. When you try to apply define the mathmatic mean, it WILL be affected when you remove silent periods. Think of it in terms of amplitude values. Your original file registers 8,6,4,0,0, 2, 4,6,8,7,4,0,0,2,4,8,4,0. The mean is not exactly an average, but it is affected if you change the values included in the array - especially if you remove a lot of zeroes.

The calculation of RMS is very complex, and will depend on the sampling rate and number of bits in each sample. If you want to play with the math, check this out:

http://en.wikipedia.org/wiki/Root_mean_square

One advantage of the digital domain is that once you convert to digital, further processing SHOULDN'T add noise. That doesn't mean that it won't add distortion, however. Compression and limiting certainly add distortion. Limiting in particular can change the shape of a wave form from a curve to a square, which WILL change the sound.

Too many people use plug-ins to erase dynamic range, turning wave forms into brick-like density. The plug-in that most people should be using is "normalize", generally at -3 dB. This leaves headroom for overshoots. It may not squash the dynamic range the way that compression does, but that's a good thing. Most streamers add additional compression in the final stages, and adding compression to a compressed file really messes up the audio.

If you're responsible for the final product, you want to add compression at the final stage of the delivery process. Companding (compression of loud material and expansion of soft material) should only be done ONCE, and I'd suggest a lighter touch than used by many radio stations these days. Most people put a "brick wall" limiter at the end of the chain to prevent overvoltages from damaging downstream gear and/or exceeding the bandwidth of the delivery pipe.

 

[TheBigA]

If you're responsible for the final product, you want to add compression at the final stage of the delivery process.

All this is wonderful, but in the world of streaming and podcasts, there are no standards. It's the wild west. it's radio without the FCC. It's radio without engineers.

With regards to the examples in the OP, I'd ask was he listening to a stream from the local station, or the national service. There's a difference between Glenn Beck and The Blaze in terms of standards. Same with NPR Washington and NPR local.

 

[Goat Rodeo Cowboy]

Goat, you have to understand what RMS means - Root Mean Square. It's a calculation generally used to allow DC formulas to be applied to AC current, which typically has a fixed frequency and fixed amplitude.

"The root mean square (abbreviated RMS or rms), also known as the quadratic mean, is a statistical measure of the magnitude of a varying quantity. It is especially useful when variates are positive and negative, e.g., sinusoids."

Yes, I have considerable knowledge about 'RMS' than what I included in my post. This thread appears to be providing useful information for folks who are relatively new to the concept of keeping audio levels consistent and somewhat under control.

I am aware the RMS and Quadratic 'technollogies' are also used in statistics and various technical calculations to help focus data toward its primary direction and trend, rather than let everything from political surveys, audience studies and industrial failure-point-calculations from being badly skewed by statistical points that are sometimes called 'outliers'. Now that we have all done our due-dillegence in 'marking our turf' in areas of expertise, let's try to focus on the curiosities and need of the original poster.

Some audio process devices, both hardware and software, take the cheap way out and do not do the heavy lifting needed to avoid distortion and artifacts. Depending on how one chooses to define distortion in the audio world, it is possible to do significant compression and limiting (*** see footnote) of audio in digitized audio without distortion. At that point we get into a debate: Are artifacts tht creep into audio processing a form of distortion or are artifacts a unique and seperate breed of contamination?

I use Adobe Audition. I won't attempt to debate whether it is the best, as good as, or inferior to other audio editing software. It is "what I learned on" and "what I grew up with". But as audio software goes in the broadcasting, streaming and podcast world, it ain't shabby! If I choose the wrong feature available in Audition and use it too aggressively, I can easily produce an audio product that has the beauty and attraction of spoiled and rotting chopped liver. But, it I choose the correct features, use them in the correct order, and use them with restraint, I can get beautiful sound out that has integrity!

So part of our best answers for the original poster probably should include: There is NO software or hardware that is guaranteed to product great audio output on its own. Whether you are using a piece of broadcast audio hardware you paid $7,000 to purchase, or a piece of $59.95 audio software for a computer, the understanding and skill of the person doing the install and doing the process will determine whether you audio stream sounds like it has fidelity or sounds like a cheap bicycle trying to change gear ratios as the chain jumps from sprocket to sprocket.

(***footnote) I was sitting in radio station control rooms back when the only automatic gain variation was done by a tube-type 'PEAK limiter'. About the time GATES introduced the STA-LEVEL which was designed to adjust overall sound rather than just the short peaks, the brother-in-law of our station owner was working for a company developing one of the new 'overall gain controllers' and he brought a prototype in and we installed it. (No brand name was given at the time. I wish I could go back and learn who he was working for and if it was the prototype of something that has become a household name in the industry today.) All of that to say... back in the stone age, I cut my teeth on terminology: a limiter worked on instantaneous PEAKS and we called that limiting. A device that analyzed audio over a longer time span and slowly dealt with program content gain was (at that time) called a 'compressor'. Fast forward to today and read the instruction manual for today's digital audio software and read about limiting, compressing and expansion, and you realize that it is a handicap to have my old fashioned language definitions. I am not sure the copy-writers who produce instruction manuals today have any concept of what might distinguish these various words from one another.

 

[SirRoxalot]

Goat, I was simply trying to explain why you got the different RMS results after you removed the silences. You seemed confused about that in an earlier post.

 

[Goat Rodeo Cowboy]

Goat, I was simply trying to explain why you got the different RMS results after you removed the silences. You seemed confused about that in an earlier post.

It was a great contribution to the conversation. I wasn't confused. I was attempting to share with people the fact that RMS readings are not as "finite" as some other audio measurements. AND that you cannot always trust some of the measurement tools that will tell you what your RMS value is.... whether it is right or not.

As one person who posted said, this conversation might have been a bit more appropriate if it were in the ENGINEERING forum, but I suspect some of the people who want to do streaming and other content preparation may never set foot in the Engineering form because the conversation there gets too technical, and does not contain these self-correction side-bar comments that we ventured into here.

The learning process for "newbies" in the field of audio has always been a bit of craziness. The internet has made it possible to reach out and find so much information not readily available to us 30, 40 and 50 years ago. Unfortunately, the Internet is not very good at self-regulating. We used to assume a good hard-cover book meant some one had invested a lot of good money in publishing the book and that they had "vetted" the content. (Wasn't always true!) But today, the bad information, the folklore, the misguided "knowlege" gets a free ride. When you are new, when you are the student, how do you know which is which? Conversations like this let the readers make up their own minds who to trust, and who to ignore.

Then, there are people like me out here.... that people have no idea what to do with.

 

[Unregistered]

All this is wonderful, but in the world of streaming and podcasts, there are no standards. It's the wild west. it's radio without the FCC. It's radio without engineers.

With regards to the examples in the OP, I'd ask was he listening to a stream from the local station, or the national service. There's a difference between Glenn Beck and The Blaze in terms of standards. Same with NPR Washington and NPR local.

Although you'd expect professional organizations to be doing better than they are.

The Beck shows come from Soundcloud, and their levels are hysterical. The NPR streams come from NPR. If the same meter shows one NPR show hitting the ceiling and another show peaking at -20, and the content is similar, it's a pretty good presumption that one is pretty quiet compared with the other. Especially when my wireless headphones won't latch on to the quieter show.