Final Output Level

[huejahfink]

Here's my thoughts on the matter.

Are you pulling in lossless files to Logic to look at or lossy ones?
MP3 or AAC may very well have originally been mastered to a lower value than zero but due to the inaccuracies of the lossy codec, will probably peak higher when they come out the other end. Just something worth considering since if you are pulling in lossy files to analyse, it could be the case that the original WAV master did have a bit of a safety ceiling before it was converted by the shops systems.

That said, you are correct - lots of commercially available CDs and lossless downloads are hitting 0dB full scale. Whether or not this is important seems to be a matter of taste.

If you want to be cautious, leave a slight ceiling and check your output with an upsampling meter that will display potential inter-sample-peaks. These are peaks that do not show on the digital file itself, but may appear when the data is converted into an analogue signal. Decent gear will have a bit of room to handle such peaks gracefully and output a smooth signal, where as less thoughtfully designed or cheaper options may not - causing additional distortion.

Ultimately, I think it has more to do with mastering style than anything else. If you are going for a "tasteful" conservative master, where the peaks are not hitting fullscale the whole time, then it could very well pay to have a bit of a ceiling such as you mention - since no-one is going to miss that final half a dB in such a cases. Although conversely, audio that lightly taps 0dB every once in a while would probably sound great even on lesser equipment.

If the audio is clipped or aggressively brickwalled -- perhaps it doesn't matter so much, since the audio is already 'distorted'. Conversely again - one could argue that leaving a safety ceiling would be even more important in such a case, in order to keep any distortion strictly to that rendered in the file, and not encourage further distortion downstream.

I apologise if my post does not give you a specific answer, but I'm not sure that there really is one - and hope instead that it provides some food for thought.

[Mcramblett]

Thank you! Your answer was thorough enough that it answered many other questions, I appreciate it

[huejahfink]

You're welcome!

[Chris Bauer]

It is regarded as good practice to set the limiter at somewhere between -0.3 and -0.1 dB. In all likelihood, you will not hear the difference between these levels. The important thing is that your masters do not create digital clipping. This is a form of distortion that limits a signal once it exceeds the threshold of 0 dB.

[MAzevedo]

I've had a number of clients who checked refs on a system that blinked the 'over' light any time a sample hit full scale. Leaving a few 10ths of a dB (I eventually settled on 0.3) of headroom meant never delivering material that someone further down the chain would think of as clipped, and leaves a bit of wiggle room for intersample peaks and overshoots from the near-inevitable lossy encoding on the consumer end.

[sanddigger1]

I'm checking the level after mp3 conversion and leave all the headroom needed to avoid any peak hitting 0. Often it is 0.5 but the last one was 1 dB. To this day, noone complained .

[Mcramblett]

Looking back through my 'practice' masters, I was setting my limiter at an output of 0.0 and from my 24-bit version I dithered and made it a 16-bit .wav, then listened to it again and still had no clipping or distortion. I then made that .wav an AAC or a .mp3 through either logic or iTunes, and I never realized that once I converted it that it was causing distortion in those formats. But when I set the limiter output to -0.3, the results were excellent for an AAC. Perhaps I need more for an .mp3?

[fastlanestoner]

It could also be a discrepancy in your metering vs. the ME's metering.

As for level, I set the ceiling to .3dB bc of the logarithmic nature of decibels.

[Mcramblett]

So I should be fine at -0.3 if I know its going to an mp3? The site CDbaby distributes the audio files to sites like itunes and reverbnation, and they suggest a 16-bit .wav. So considering that the file will become either an AAC or mp3, would I be safe to set the .wav file I upload to -0.3?

[sanddigger1]

Originally Posted by Mcramblett

Looking back through my 'practice' masters, I was setting my limiter at an output of 0.0 and from my 24-bit version I dithered and made it a 16-bit .wav, then listened to it again and still had no clipping or distortion. I then made that .wav an AAC or a .mp3 through either logic or iTunes, and I never realized that once I converted it that it was causing distortion in those formats. But when I set the limiter output to -0.3, the results were excellent for an AAC. Perhaps I need more for an .mp3?

converting 24bit to 16 is very unlikely to change the peak level, no resampling takes place , contrary to sampling rate conversion or lossy format conversion.
Not all lossy formats are equal : ogg or aac handle peaks much better than mp3. (In addition not all mp3 encoders work the same, unlike ogg or aac encoders). I noticed that ogg was rarely more than 0.3 dB higher than the original .wav while mp3 was often adding up to nearly 1 dB (and sometimes more !)...

Originally Posted by Mcramblett

So I should be fine at -0.3 if I know its going to an mp3? The site CDbaby distributes the audio files to sites like itunes and reverbnation, and they suggest a 16-bit .wav. So considering that the file will become either an AAC or mp3, would I be safe to set the .wav file I upload to -0.3?

You'll not be totally safe but, in my experience, most of the intersample peaks are 0.1 to 0.3, higher peaks being very rarely more than a handful per song.
0.3 dB of headroom will likely handle most of the peaks, specially if your mix was'nt clipped or heavily limited. So, if you dont want to "lose level" and are not a maniac (I'm one), -0.3 to -0.5 should be good.

[Mcramblett]

Awesome, thanks for the help!

[Rossd25]

-.3 works fine.

You would be surprised by how many commercial CDs contain intersample overs (most). Ozone has a intersample prevention option, but to me it's better to drop the ceiling to -.3 and call it.

Hardly matters.

[Filthrill]

I'm just sayin...load a few tracks of the new Rihanna CD (or Chris Brown or anything in that vein) into your DAW. I use Pro Tools. Mind u that these files are not AACs or MP3s. They are from the WAV master, yes? Can we agree up to this point? Let's continue. IF her mastering engineer set his limiter between -0.5 to -0.1 the songs would not be almost constantly clipping into the red. Is this correct?

So, can we all agree that these mastering guys did in fact set their limiters (the ones that actually used limiters) to 0.0 or very close to it? (((I say that last part since the Sonnox limiter always the ceiling to be set by very tiny increments, smaller ones than other limiters allow. Sorry to be obvious but I've found that a lot of people don't know this)))

[Filthrill]

Sorry guys, I don't have any patience. Anybody who would like to respond to my previous post is still welcome to do so.

Let me ask ya Slutz an easier question tho:

Are any of u setting different ceilings for your waves vs. your MP3? Or r u just going w/ a lower/safer same ceiling for both formats.

Keep in mind that I bounce my MP3s directly from my Pro Tools mastering session via their MP3 Export Option. I do once bounce as a wave & another separate bounce as a High Quality MP3 & always both w/ same ceiling.

Any response appreciated.

[stinkyfingers]

TB EBULoudness - ISP/ "True Peak" meter
for .wav = 0dBTP and mp3 = -1dBTP is what i use for levels...
i use Elephant which is very hard to get anything by when set up "right"...
almost like trying to get by an, um....elephant...

[Babaluma]

I use the same output level for .wavs and .mp3s, but I run my MP3's through Replay Gain with the "Max No Clip Gain" setting after creating them, to avoid ISPs. Means my .mp3s might be a little quieter than some, but they won't be clipping any cheap consumer DACs down the line.

You could do it the other way though, like you suggest, and just run the ceiling of your limiter at -1.5dB or something, for the .mp3 versions.

[Filthrill]

Hey Slutz, good lengthy post today:

After loading a shload of random store-bought CDs last night, both newer & not-so-newer, I came to the conclusion that some MEs are abusive w/ the output ceiling & some are sensible. I spent about 4 hours of time doing this experiment to see what the hell is going on. Man, they were all different. For instance, Kanye West's last CD is clipping past 0.0 off the charts. Where's the sensible ceiling? I dunno. This is from the CD! Katy Perry's Teenage Dream was hitting -0.1. Okay, kool. Of course I loaded in her MP3 & it was clipping past 0.0 at certain points just like others on here have predicted. Audioslave's Revelations CD was clipping quite a bit too. Kelly Clarkson's All I Ever Wanted was right at 0.0 & never clipped into the red tho. Eminem's Recovery was all over the place. Some songs were at -2.1 & some were at 0.0 w/ clipping. Some were at -0.1. There was no real continuity. Tons more I'm not mentioning but nowadays it looks like -0.1 is considered conservative & I didn't find many below this. 0.0 is very common, some with clipping, some w/o.

Honestly I've been setting mine to 0.0. I'm going to stick w/ -0.1 from now on & just bounce MP3 the same. I vow to new go past this! I can just pull down the ceiling to that on my Waves L2 (don't worry it's just for dithering) last in chain. I can live w/ MP3 hitting 0.0 after the conversion & possible even clipping occasionally. I'll run some tests & listen.

[Babaluma]

The Kyari Pamyu Pamyu J-Pop single "Pon Pon Pon" last year had around a million actual overs, and who knows how many ISPs...

[jrhager84]

I've had most success with -.3 dB. Dunno why. I'm anxious to hear how terrible I am for doing it this way!

Sent from my SAMSUNG-SGH-I777 using Tapatalk

[stinkyfingers]

Originally Posted by Babaluma

I use the same output level for .wavs and .mp3s, but I run my MP3's through Replay Gain with the "Max No Clip Gain" setting after creating them, to avoid ISPs...

for best results, you really need to eliminate the ISPs before the mp3 conversion process...

[Babaluma]

Originally Posted by stinkyfingers

for best results, you really need to eliminate the ISPs before the mp3 conversion process...

I do, I use Elephant, Sonoris Meter and RME DigiCheck to make sure there are no ISPs when I do the final limiting and dither down to the final lossless file format. But one of the side effects of creating lossy files from lossless files is the creation of new ISPs. Running MP3s through ReplayGain ensures no ISPs in the lossy version too. Really, this is all very basic stuff and can easily be tested by oneself.

[stinkyfingers]

^ah ok...i understand now...this is in addition to precautions already taken...
i never really checked an mp3 after for ISP, i'll have a look...curious now...
most of the time i see about +0.8dB max for ISP (if any), so i use -1dB for mp3 conversion and "assume" the mp3 is all set...
but...i can't remember the last time i actually made an mp3 for real world use and not just for testing trivial things because i'm bored...(i can't stand to listen to them and i wouldn't subject anyone else to it either...)
but but...sometimes (rarely) people request them from me, so...i try and at least protect the general public from the dreaded ISPs in mp3s...maybe i try harder now...or not...

[Babaluma]

Yep, check a few mp3s that you've converted from lossless files with no ISPs, some will still be fine, but others will have mysteriously grown them.

MP3Gain will automatically "turn down" the MP3s in increments of -1.5dB until no ISPs occur. If you want to do an album and keep the track to track balance you can use the "Max No Clip Gain For Album" settings, which will turn all the tracks down by the same amount (even if some of them are not clipping). I can't remember the last time someone actually requested them though. It

Of course, you can do it the other way round too and just use a limiter ceiling of -3dB or something. I think I'll do more of this if I ever need to follow the new EBU R128 standard.

Agree with you on MP3s generally being pretty crappy, but I have sometimes found it hard to pick 320 CBR in a blind A/B test with the original, with some sources.

[Filthrill]

Originally Posted by Babaluma

I do, I use Elephant, Sonoris Meter and RME DigiCheck to make sure there are no ISPs when I do the final limiting and dither down to the final lossless file format. But one of the side effects of creating lossy files from lossless files is the creation of new ISPs. Running MP3s through ReplayGain ensures no ISPs in the lossy version too. Really, this is all very basic stuff and can easily be tested by oneself.

Looks like I can check for overs w/ my Waves Dorrough Meter plug-in. Anyone see an issue on relying on this plug-in for this? I'm on Mac & am a lonely Pro Tools HD user.

[Filthrill]

Originally Posted by Filthrill

Looks like I can check for overs w/ my Waves Dorrough Meter plug-in. Anyone see an issue on relying on this plug-in for this? I'm on Mac & am a lonely Pro Tools HD user.

I'm quoting myself now. Wow!!! So I mastered an Electronic-type Rock song yesterday that got pretty loud at the end & I used the Waves Dorrough Meter plug-in & set it to "OVERS" but never saw anything indicating any overs & I'd say for sure that there had to be some. Anyone have any experience w/ the Dorrough plug-in for this purpose?

[stinkyfingers]

is Dorrough Meter an ISP/"true peak" meter, even...?
try the ToneBoosters meter if you haven't already, it works good for ISP...
or iZotope RX is nice to view the waveform...

EDIT: the Dorrough meter is not an ISP meter...

[Filthrill]

Originally Posted by stinkyfingers

is Dorrough Meter an ISP/"true peak" meter, even...?
try the ToneBoosters meter if you haven't already, it works good for ISP...
or iZotope RX is nice to view the waveform...

The ToneBoosters stuff looks really nice. I like Izotope products too. I see that according to it's description the Dorrough displays overs. I'll do more research. User error possibly? Hmmm....

[Babaluma]

There are actual overs, and there are ISPs (which don't look like overs, but create overs when the digital data is converted back to analogue), they are different things, and each can be measured in different ways. I admit it can be confusing. Only some meters (usually labelled ISP, OS or "true peak") claim to measure ISPs, whereas all should measure regular overs.

[jrhager84]

Originally Posted by Filthrill

The ToneBoosters stuff looks really nice. I like Izotope products too. I see that according to it's description the Dorrough displays overs. I'll do more research. User error possibly? Hmmm....

I hate that TB have no AAX/RTAS options... They look really cool!