Using a Loudness Meter

[rcbuse]

The first 10 minutes of this video talk about distortion introduced by peak normalization. The audio examples he plays are quite amazing.

[rcbuse]

I should also say I'm a huge fan of R128. The k-weighted loudness measurement, the loudness range LRA, and the oversampled True Peak measurement are fantastic.

[selig]

The first 10 minutes of this video talk about distortion introduced by peak normalization. The audio examples he plays are quite amazing.

Should be noted in the first examples he's creating a worst case scenario, where the sine wave is a multiple of the sample rate. For one, that's highly unlikely to happen unless you use technical tones such as 100 Hz, 1 kHz, or 10 kHz. So if you don't use a tone that's a multiple of the sample rate, you will eventually have samples that represent the actual analog value and thus show the true peak level. But more importantly, unless your music consists of test tones, you'll not likely have this issue with actual music waveforms. But unfortunately he fails to mention how much headroom is typically required to avoid inter-sample peaks. He DOES go on to suggest the headroom required for mp3 conversion at different rates, fwiw, and if you release your music as mp3s this should definitely be noted.

So how much headroom SHOULD you leave for CD quality masters? No one agrees 100%. Some say 3 dB to absolutely avoid inter-sample peaks, but others say that's only necessary for extremely peak limited material.

I've long used around a full dB of headroom just to be sure, but I don't use much peak limiting when I "self-master" (typically 2-4 dB gain reduction max). In Reason I also use Ozone's Inter Sample Limit feature just to be sure, since you can't predict what playback system will be used.

I'm curious what others have experienced, and what "best practices" they've found.

[8cros]

I since 2011 experimenting with R 128 for music.

*I can give an example of a broadcast application in the mix: your mixer channels, and bus groups. This is the program audio.

*In addition the principle of moments and short measurement - it's magic. This applies to balance the sounds of different length.

I'm talking about the experience and practice and not the video.

And the most important thing for me. This ability to work with vocals.

[8cros]

Of course it hurts me to hear that it was not for the music, and for broadcasting.
It squeezes my heart.
What's the difference? What Florian not a producer? I hate the producers, they produce a mass indistinguishable from the background music.

I'm not interested in the fight between the broadcasters and musicians. That still was not enough. I've seen of engineers mastering. A lot of. And all of them as machines.
Most Reason users, give them a head start in mastering and commutation strategy.

[8cros]

to everyone i think it is my fault why we went off the topic because i was asking about how to use/read the volume meters in SSL for mixdown (i did not know how they measureed). fwiw selig has helped me significantly to understand how to read the meters and given me his view on suggested peak level which i am experiment with now but sometimes hard with kids who always seem to steal away my music time!

You multi account? (I'm kidding of course)

Selig, you refute everything. This is something new? No. So many people do. And you can even to disprove that the earth is round. Even the fact that your leveler is something useful.
Please, go along with us and not against us.
If you have questions, try to find a solution and we will help you.

[normen]

Should be noted in the first examples he's creating a worst case scenario, where the sine wave is a multiple of the sample rate. For one, that's highly unlikely to happen unless you use technical tones such as 100 Hz, 1 kHz, or 10 kHz. So if you don't use a tone that's a multiple of the sample rate, you will eventually have samples that represent the actual analog value and thus show the true peak level. But more importantly, unless your music consists of test tones, you'll not likely have this issue with actual music waveforms. But unfortunately he fails to mention how much headroom is typically required to avoid inter-sample peaks. He DOES go on to suggest the headroom required for mp3 conversion at different rates, fwiw, and if you release your music as mp3s this should definitely be noted.

So how much headroom SHOULD you leave for CD quality masters? No one agrees 100%. Some say 3 dB to absolutely avoid inter-sample peaks, but others say that's only necessary for extremely peak limited material.

I've long used around a full dB of headroom just to be sure, but I don't use much peak limiting when I "self-master" (typically 2-4 dB gain reduction max). In Reason I also use Ozone's Inter Sample Limit feature just to be sure, since you can't predict what playback system will be used.

I'm curious what others have experienced, and what "best practices" they've found.

Yeah, its back down to "just make your music sound good" - if it sounds better with more harmonics from limiting then thats a good thing. But its important to *know* what technically happens - and thats best seen with test tones. And to be fair, that is the core of the keynote so its obvious he'll mention these things. He goes into how limiting is detrimental for the audio quality - not how it can be used as a means to get some sound.

As for inter-sample peaks - by definition the loudest inter-sample peaks occur at high frequencies, those that approach nyquist. And by definition hard limiting causes high frequency content to be added. As for when inter-sample peaks occur when mastering - theres no need to guess, you can measure (or rather calculate) that with a plugin and go right to the edge before you send out your mix for the glass master. Its a wee bit more complicated for digital compression because you don't know which algorithm will *decode* your material - they might behave differently and create (inter-sample) peaks at different levels.

Anyway, obviously you're only trying to add information to the video, not refute any of it. Bottom line is still that no producer needs to care about R128 while mixing - he only needs to know that he doesn't have to compromise audio quality anymore when using a limiter because he's in some kind of "race".

[8cros]

I can add that laudness, pushed me to use isotope often.
Not only on the master.
I do the gate G8 method on each track, by using the isotope. And I can see the difference.
I started doing less rigid compression.

[selig]

to everyone i think it is my fault why we went off the topic because i was asking about how to use/read the volume meters in SSL for mixdown (i did not know how they measureed). fwiw selig has helped me significantly to understand how to read the meters and given me his view on suggested peak level which i am experiment with now but sometimes hard with kids who always seem to steal away my music time!

You multi account? (I'm kidding of course)

Selig, you refute everything. This is something new? No. So many people do. And you can even to disprove that the earth is round. Even the fact that your leveler is something useful.
Please, go along with us and not against us.
If you have questions, try to find a solution and we will help you.

8cros, you simply don't speak the truth. I don't know what your point is any more. I'm beginning to think you don't even have a point, you just want to refute what I say just to be contrary.
I'm not "against" you, and there is no "us" for me to be against. Can't you see that sometimes some of us agree, sometimes we don't? There is no "group think", though I've been accused of going WITH group think, and now accused of going AGAINST group think. I'll never be "right", so I instead simply share what has worked for me - dispute that if you must, but what's your point?


Sent from my iPad using Tapatalk

[normen]

I can add that laudness, pushed me to use isotope often.
Not only on the master.
I do the gate G8 method on each track, by using the isotope. And I can see the difference.
I started doing less rigid compression.

Why do you only start to hear something when you read a meter? You should hear how good or bad your mix is without that.

All the good mastering engineers didn't push their masters to extremes because they didn't hear how it was bad for the audio quality, it was because their clients said they wanted it to sound as loud as X.

[selig]

Should be noted in the first examples he's creating a worst case scenario, where the sine wave is a multiple of the sample rate. For one, that's highly unlikely to happen unless you use technical tones such as 100 Hz, 1 kHz, or 10 kHz. So if you don't use a tone that's a multiple of the sample rate, you will eventually have samples that represent the actual analog value and thus show the true peak level. But more importantly, unless your music consists of test tones, you'll not likely have this issue with actual music waveforms. But unfortunately he fails to mention how much headroom is typically required to avoid inter-sample peaks. He DOES go on to suggest the headroom required for mp3 conversion at different rates, fwiw, and if you release your music as mp3s this should definitely be noted.

So how much headroom SHOULD you leave for CD quality masters? No one agrees 100%. Some say 3 dB to absolutely avoid inter-sample peaks, but others say that's only necessary for extremely peak limited material.

I've long used around a full dB of headroom just to be sure, but I don't use much peak limiting when I "self-master" (typically 2-4 dB gain reduction max). In Reason I also use Ozone's Inter Sample Limit feature just to be sure, since you can't predict what playback system will be used.

I'm curious what others have experienced, and what "best practices" they've found.

Yeah, its back down to "just make your music sound good" - if it sounds better with more harmonics from limiting then thats a good thing. But its important to *know* what technically happens - and thats best seen with test tones. And to be fair, that is the core of the keynote so its obvious he'll mention these things. He goes into how limiting is detrimental for the audio quality - not how it can be used as a means to get some sound.

As for inter-sample peaks - by definition the loudest inter-sample peaks occur at high frequencies, those that approach nyquist. And by definition hard limiting causes high frequency content to be added. As for when inter-sample peaks occur when mastering - theres no need to guess, you can measure (or rather calculate) that with a plugin and go right to the edge before you send out your mix for the glass master. Its a wee bit more complicated for digital compression because you don't know which algorithm will *decode* your material - they might behave differently and create (inter-sample) peaks at different levels.

Anyway, obviously you're only trying to add information to the video, not refute any of it. Bottom line is still that no producer needs to care about R128 while mixing - he only needs to know that he doesn't have to compromise audio quality anymore when using a limiter because he's in some kind of "race".

Agree 100% - my mention of the tones was to point out they are worst case, not that there's ANYTHING wrong with knowing what worst case represents (I do that all the time myself). I've been more often accused of being too technical than not, but I'm a musician at heart, learning engineering only as a means to an end - my music.


Sent from my iPad using Tapatalk

[8cros]

I can add that laudness, pushed me to use isotope often.
Not only on the master.
I do the gate G8 method on each track, by using the isotope. And I can see the difference.
I started doing less rigid compression.

Why do you only start to hear something when you read a meter? You should hear how good or bad your mix is without that.

All the good mastering engineers didn't push their masters to extremes because they didn't hear how it was bad for the audio quality, it was because their clients said they wanted it to sound as loud as X.

I am ready to make an example for this purpose.
But not now. I'm going to limit its presence on the forum. Because of the constant wrangling I keep getting warnings.
And my personal example will be criticized.
I do not think this is extreme. gate method retains all the audio information intact. Therefore, on the master we have less summed peak.

[normen]

I am ready to make an example for this purpose.
But not now. I'm going to limit its presence on the forum. Because of the constant wrangling I keep getting warnings.
And my personal example will be criticized.
I do not think this is extreme. gate method retains all the audio information intact. Therefore, on the master we have less summed peak.

Well if you're dishing out (very) personal critique you should be able to stomach factual critique, right? The actual level of the master bus in a digital system is pretty much irrelevant, whats relevant is the level of the *output*, after the master bus. Also I didn't say anything about you doing something extreme, I said that the extremely limited (almost destroyed) masters that came out since the 90s (from Oasis to Metallica) are NOT because the mastering engineer didn't hear what was going on - they did - they just proceeded to do what their clients told them to.

Theres no need to do anything sophisticated here, just back down the limiting. With the wonders of R128 your track will still sound as loud as the others in the playlist - even if you decided to make your master peak at -10dB for some reason - the BROADCAST instance will handle that for you now. It simply doesn't matter anymore.

[8cros]

I am ready to make an example for this purpose.
But not now. I'm going to limit its presence on the forum. Because of the constant wrangling I keep getting warnings.
And my personal example will be criticized.
I do not think this is extreme. gate method retains all the audio information intact. Therefore, on the master we have less summed peak.

Well if you're dishing out (very) personal critique you should be able to stomach factual critique, right? The actual level of the master bus in a digital system is pretty much irrelevant, whats relevant is the level of the *output*, after the master bus. Also I didn't say anything about you doing something extreme, I said that the extremely limited (almost destroyed) masters that came out since the 90s (from Oasis to Metallica) are NOT because the mastering engineer didn't hear what was going on - they did - they just proceeded to do what their clients told them to.

Theres no need to do anything sophisticated here, just back down the limiting. With the wonders of R128 your track will still sound as loud as the others in the playlist - even if you decided to make your master peak at -10dB for some reason - the BROADCAST instance will handle that for you now. It simply doesn't matter anymore.

Yes, you are absolutely right. We are in audio nirvana.
But to make the G8 to master even the isotope. It is extremely difficult.
Zen

[selig]

…I said that the extremely limited (almost destroyed) masters that came out since the 90s (from Oasis to Metallica) are NOT because the mastering engineer didn't hear what was going on - they did - they just proceeded to do what their clients told them to.

Had the opportunity in the early 90s to have a very interesting conversation. I was in nirvana, aka, a mastering suite in NYC with Bob Ludwig, mastering a project I had co-produced and engineered (but not mixed). At the same time, my brother was chief engineer at Z100 in NYC, probably the #1 radio station in the US at the time. I gingerly mentioned the subject of "loudness" to Bob, which was already a hot topic in 1991-2 when I was there. Seems there was already a lot of "finger pointing", with broadcast engineers saying to not master so hot as their processing was designed to work on music that wasn't as "hot", and mastering engineers talking about how much processing radio was doing to their mixes. At the time, there was no "push" to master this project loud, we just made it sound as good as it could (and Bob did little to the project overall). I DO remember regretting bringing up the subject - Bob's first response was a deep sigh, indicating this was a subject he had giving a lot of thought to in the past.

But in the end Bob did agree that it was NOT the mastering engineers pushing things hard and louder, it was artists, producers, and label heads doing so. And all the while you had the broadcast engineers saying the "louder" you made your mix, the WORSE it was going to sound on the air. Go figure, it was more like some sort of addiction. Some artists even said things like "I don't want the listener to have to turn up the volume when they play my record", to which I would respond, "why not, that's what volume knobs are literally designed for". These conversations went on all through the 90s, to the dismay of mix and mastering (and broadcast) engineers alike. By the 2000s, sadly, there were no more conversations, just a blank stare and acceptance your mix was going to be crushed to death for no good reason.

It is for these reasons I applaud broadcasters adopting a standard that can end the madness (aka, the "loudness wars"), allowing engineers/producers/musicians to go back to mixing for musical/emotional effect, which CAN include wanting the mix to be "loud" (understanding the consequences), if that's what you're in to. Or not.

[8cros]

But Reason that does not, for 6 years. It's not on alihoopa
I am using live samples and crest factor of 10 decibels -1 dB peak.
Next I hear electronic tracks. On alihoopa. They are close and they are louder 3-5 decibels. And I cry.
Our Challenge is also not regulated. Oh yes, our Challenge have an authoritarian system assessments.

[Galaxy]

Thx guys! You've made a lot of sense out of all this! The take away here is R128 -23 is a broadcast standard for loudness normalization, great!

I think it was Normen who mentioned intersample peaks initially, which I think is a very important consideration when finalizing your music. Selig, you mentioned a loose rule of -1 peak with ISL on ozone. Older convential wisdom suggested -0.1 to -0.3 peaks. Later we learn, because of intersample peaks upon reconstruction waveforms on shitty DA converters, we need to limit to lower digital level, correct? Forgive me for all the questions, I just find this subject fascinating and put a lot of pressure on myself to "get it right".

I watched an izotope ozone video, when version 7 advanced came out, that mentioned something about even going lower, to say -1.5, because intersample peaks can still pop up on crappy codecs, etc etc.

Here's something to consider, and if any of you have experience or further knowledge about this, please do share your comments, opinions and knowledge. Apple TV airplay? Anyone use it? Ever heard your own music on it? Try cranking up your iOS device's volume all the way and use your surround sound tv system volume as your playback monitor level control, as you should. My experience has been tons of distortion, yeah sure I can turn my iOS device down a few notches to reduce/eliminate some of the playback distortion, but from some of my tests, izotope has got it right. -1.5 seems like a safe level to limit to avoid the above mentioned issues.

I have other comments about how some music just works at louder crest factors than others and vise versa, but this post is about limiting ceilings.

[Galaxy]

Btw, it's worth mentioning. Selig, your stories/info about NYC and your brother are very interesting and impressive You're lucky to have someone that close to you with this much knowledge and experience when it comes to broadcasting

[normen]

Thx guys! You've made a lot of sense out of all this! The take away here is R128 -23 is a broadcast standard for loudness normalization, great!

I think it was Normen who mentioned intersample peaks initially, which I think is a very important consideration when finalizing your music. Selig, you mentioned a loose rule of -1 peak with ISL on ozone. Older convential wisdom suggested -0.1 to -0.3 peaks. Later we learn, because of intersample peaks upon reconstruction waveforms on shitty DA converters, we need to limit to lower digital level, correct? Forgive me for all the questions, I just find this subject fascinating and put a lot of pressure on myself to "get it right".

I watched an izotope ozone video, when version 7 advanced came out, that mentioned something about even going lower, to say -1.5, because intersample peaks can still pop up on crappy codecs, etc etc.

Here's something to consider, and if any of you have experience or further knowledge about this, please do share your comments, opinions and knowledge. Apple TV airplay? Anyone use it? Ever heard your own music on it? Try cranking up your iOS device's volume all the way and use your surround sound tv system volume as your playback monitor level control, as you should. My experience has been tons of distortion, yeah sure I can turn my iOS device down a few notches to reduce/eliminate some of the playback distortion, but from some of my tests, izotope has got it right. -1.5 seems like a safe level to limit to avoid the above mentioned issues.

I have other comments about how some music just works at louder crest factors than others and vise versa, but this post is about limiting ceilings.

About inter-sample-peaks (and btw Selig mentioned them first): Its not so much about shitty converters, its just a matter of fact that they occur with digital audio:



The problem is the analog circuitry after the D/A converter. It might or might not be able to handle these peaks.

As for "caring" about them - if you enable ISP limiting in Ozone it will automatically limit the audio to a level where upon reconstruction the inter sample peaks will be below 0dBFS. So if you enable that and Ozone doesn't limit you are safe. You can go all the way to 0dBFS. This is for PCM audio (WAV, AIFF). But (as said) you can't be sure how the wave looks after compressing (to mp3, aac etc.) and then decompressing the audio again. Depending on the encoder and decoder there could be new inter sample peaks above 0dBFS coming up. From my experience these differences should not be above 1 or 2 dB or so though (much less than ISPs can be "naturally" anyway), so those 1.5dB should be fine, yes.

For Apple products they all use the same codecs and Apple provides a neat plugin that allows you to check your audio for ISPs before and after compression/decompression. Personally I didn't experience what you describe with distortion, going from either HDMI in an AppleTV or the mini jack in the AirPorts but theres a lot of devices behind that chain so it could be anywhere down that line..

And yeah, although this topic is about loudness per se most of the discussion is about the actual theoretical limits - loudness is a vast topic by itself and depends on arrangement as much as processing.

[selig]

The problem is the analog circuitry after the D/A converter. It might or might not be able to handle these peaks.

Help me understand something I alluded to earlier…
In that pic we see only one half cycle of the wave. Is it not logical to assume that the next time around (or eventually) a sample point may fall at the peaks, and thus the peaks would be read accurately (assuming the highest peak is reported)?
As I mentioned previously, IF the frequency of the sine is a multiple of the sample rate, and IF the samples don't happy to land on the waveform peaks, this example will be true for all cycles. But otherwise, the sample points will fall at a different point each cycle, eventually falling on the highest peak - or am I missing something here?

[normen]

Help me understand something I alluded to earlier…
In that pic we see only one half cycle of the wave. Is it not logical to assume that the next time around (or eventually) a sample point may fall at the peaks, and thus the peaks would be read accurately (assuming the highest peak is reported)?
As I mentioned previously, IF the frequency of the sine is a multiple of the sample rate, and IF the samples don't happy to land on the waveform peaks, this example will be true for all cycles. But otherwise, the sample points will fall at a different point each cycle, eventually falling on the highest peak - or am I missing something here?

Sure, thats true. Actually a limiter could start to "oscillate" from such an effect - or you could get oscillating digital distortion/clipping. Thing is that actual program material is WAY more chaotic than this Even in a simple arrangement of sine waves the high frequency wave might "ride" on a low frequency one.

[avasopht]

Some artists even said things like "I don't want the listener to have to turn up the volume when they play my record", to which I would respond, "why not, that's what volume knobs are literally designed for"

Well you have two types of scenarios where turning up the volume is simply not an option:
1. People listening from portable battery powered players (particularly when outside or on a noisy train).
2. Home systems that are not that powerful such as televisions, low wattage music systems and laptop / desktop speakers.

[avasopht]

Also I think there are two sides to the need for loudness. It's not just about having the drum loud, it's also (at least in the style of music I do) about the character of loudness and aggression.

For that, there's no meter (although I'm sure such a characteristic could be measured).

[selig]

Some artists even said things like "I don't want the listener to have to turn up the volume when they play my record", to which I would respond, "why not, that's what volume knobs are literally designed for"

Well you have two types of scenarios where turning up the volume is simply not an option:
1. People listening from portable battery powered players (particularly when outside or on a noisy train).
2. Home systems that are not that powerful such as televisions, low wattage music systems and laptop / desktop speakers.

True to a point - you're assuming the volume is maxed already, and there's no further headroom left.

One other thing you're overlooking is the tradeoff - by brick wall limiting a mix you don't just "get" loudness. You also get artifacts. So it's not a free lunch - you don't just choose louder or softer, you also choose more distortion, less distortion etc.

You're also overlooking the other case volume knobs are designed for - turning DOWN the freaking loud tracks/commercials. You can ALWAYS turn the volume down, even in the cases you mention.

[selig]

Help me understand something I alluded to earlier…
In that pic we see only one half cycle of the wave. Is it not logical to assume that the next time around (or eventually) a sample point may fall at the peaks, and thus the peaks would be read accurately (assuming the highest peak is reported)?
As I mentioned previously, IF the frequency of the sine is a multiple of the sample rate, and IF the samples don't happy to land on the waveform peaks, this example will be true for all cycles. But otherwise, the sample points will fall at a different point each cycle, eventually falling on the highest peak - or am I missing something here?

Sure, thats true. Actually a limiter could start to "oscillate" from such an effect - or you could get oscillating digital distortion/clipping. Thing is that actual program material is WAY more chaotic than this Even in a simple arrangement of sine waves the high frequency wave might "ride" on a low frequency one.

Perhaps that is addressed with the "inter-sample limit" feature in Ozone?