… meaning I don’t want to stir up a value discussion on how loud they SHOULD be…
but I’ve got a question about how loud they actually are. I looked at the audio file of a CD today, Keesha’s “We Are Who We Are”. I did some stats on it, and 0.0 dB had been hit. Then looking at the waveforms, I could see numerous parts where for a couple or three samples, it was pegged at 0 dB.
I was just wondering why it didn’t sound all grotty and distorted like mine do when I hit 0 dB. Is it because a few samples in a row hear and there aren’t really audible? Was that engineer “sloppy”, or just brilliantly living on the edge of distortion?
Thanks for any thoughts on this TECHNICAL question
Brilliantly living on the edge is a good descriptor. And usually some very expensive compressor/limiters. The ability to turn out a product like this is the art and science of a mastering engineer.
It is very hard to accomplish. It is a matter (for us mixing amateurs) For me personally. I get each track as close to 0db as possible through compression and EQ. Each time the combination of tracks I’m working on peg the 2bus (master bus) too heavily, I start bringing them down. All the while, I’m adding more tracks into the mix. (With 40 or so tracks, your eyes will start to bug out of your head with this method, but I’ve found it works well).
When I’m finished, the 2bus it peaking over 0db frequently, but not crammed at the top. My setup for the master bus is Multiband compressor and either a loudness maximizer or brickwall (type) limiter. Since I use compression on almost all the tracks, I use the multiband only in the low-mids to thin that particular boxy frequency range out even more. Then it goes through the limiter / loudness maximizer.
It takes some work, but with this method, you can get the master output a hairs breath under distorting and still have a touch of dynamic in the song. The meter at the 2bus should waiver just under 0db, but could peg there more than 60 / 70% of the time on the loudest passages of the song.
It is truly a razors edge there. It takes a lot of really keen listening. If you get your mix sounding good in general before any mastering, check for any distortion on a good set of headphones when you attempt to set this kind of limiting.
Be aware that you can also get intersample peaks which don’t always show up on daw meters. I don’t know all the technical stuff but it has something to do with waveform sample points and sudden spikes on the waveform in between the sampling points. This could be the problem with your own hot mixes. There are plugin meters available.
Good point and something I forgot to mention. Typically I’ll use the Loudness Maximizer in Ozone 4 which has the option to prevent intersample peaks. It seems to work pretty well, but I have to listen reaaaallllly carefully to make sure I’m hanging just below that razors edge.
Having just had an album mastered by a pro, he wanted final mixes that peaked at around -6dB to -4dB which gave him the margin to be able to mess with EQ, mid-side balance and overall loudness without worrying about any nasties that might come from accidental peaking. He then wanted some reference tracks from albums I liked and took that as a guide for the final loudness of the album. And event then, he pointed out that one of my references was a bit too loud for his taste.
I trusted his professional judgement. I’m glad I did, I have a master I love which isn’t the last word in loud, but doesn’t sound squished at all.
I think the thing we forget is that 24 bits give us an outrageous amount of headroom and even mixing with peaks down at -12dB still leaves the noise floor down at practical inaudibility. Particularly comparing to the days of analogue tape.
Hi - I went and did some more stuff to this track to make it louder (again, just as an exercise, I’m not saying here that everything needs to be louder). It’s just a stereo piano track, starts out kind of slow and low, then speeds up and gets some more volume.
What I did is look at the tracks’ overall waveform, for spikey things that stuck out significantly higher than what was around it, right up at 0dB, liking sort of like stray hairs sticking up from a porcupine. I zoomed in to sample level on the audio editor, and (gasp?) redrew the samples so that the peak was less, more like its neighbors. I went and did that throughout the track, renormalized, and got an extra 4dB out of the track. I was afraid messing with the actual samples would change the track entirely, maybe to something like Lincoln’s 1st Inaugural Address, but in reality my golden ears (ha!) couldn’t tell any difference.
Is that “allowed”, or more to the point, does anyone do that manually? It sure was tedious doing it manually. Should that best be done by a limiter? If using a limiter, it seems like it would be too easy to be too heavy-handed, and shave off a bit too much. Do some limiters allow you to know just how much you’re shaving off?
Is that what the professional engineers/masterers do to make tracks louder, or …
Thanks for reading this and any thoughts/responses -
You did it the “hard” way, but also the way that will maintain the most of the musical quality. And yes it is tedious. I’ll do that only on rare occasions, especially if there is an error (a glitch due to some sort of malfunction like an electrical noise, etc…). In that case, if a redraw doesn’t work, I’ll just zoom to the sample level and cut it. A few samples cut out of the track won’t appreciably change it’s position… and if it does I can always line it up again.
…it would be too easy to be too heavy-handed, and shave off a bit too much. Do some limiters allow you to know just how much you’re shaving off?
Just bypass the limiter, you’ll see how much the levels peak over 0 db. That’s the best way to tell.
Is that what the professional engineers/masterers do to make tracks louder, or …
I think you’ll get a different answer no matter who you talk to. You can do a kind of pseudo-limiting using compression. Using an EQ to notch out unfriendlies and “honky” noises will get you a couple extra db, and these things always help because then you can use the limiter “less” (for a lack of a better way to put it).
IMHO, the best quality sound you can possibly get comes before any editing on the 2bus (master). If you can get the mix rich sounding with pretty even response before adding any plugs to the 2bus then chances are you’ll only need to squash the highest peaks.
A few days I redrew a ridiculous peak [somebody had bumped into a mike at a choir concert I was recording]. I’ve not done that through a whole track, but as a rule, I do stuff by hand rather than with plugins. I’d never employ me in a sound-engineering company or professional studio. I’m glad you posted this thread, and I’m glad of the other responses. Helps me learn.
Keep in mind that when your recordings hit 0db you may, in fact, actually be hitting well above 0db. With digital, once you get to 0dbfs…thats it. If your signal was actually trying to get to +3then you’re gonna hear it (the more severely you turn those waveforms into square waves the more funky harmonics you start generating). A mastered CD in which the signal just’s touches 0dbfs (but does’nt exceed it) wont have the "grotty"ness problem. It’ll just be loud.
Markone mentioned that pro mastering engineers prefer mixes that are a bit down in level (max levels of -3 and many prefer even lower). Bob Katz (mastering engineer extraordinaire) says the same in his “Mastering Audio” book.
Of course, if your producing something thats not going to go to the mastering engineer, then you really have little choice now-a-days but to work with it to try and maximize levels. But there’s a difference between maximizing level and hyper compressing. I see nothing wrong with maximizing levels. Hyper compression on the other hand, doesn’t sound good to me. It’s very fatiguing to listen to for any length of time.
A reasonable argument can be made about the relative levels of tracks on a particular CD. If one imagines the CD as a holistic work, thats listened to from start to finish, then the relative level of the tracks IS an important consideration. The soft ballad that follows right after the major rock anthem aught to be quieter. Now-a-days though, my impression is that much music is really more single tunes as opposed to whole package. Sgt. Peppers wouldn’t be the same if every track was maximised.
If I’m doing something that’s not going to go to an ME, I’ll generally drop a waves maximizer plun on the stereo bus, set the max level to -0.2 and start squishing till I get as much level as possible without getting a hypercompressed sound (unless thats what I’m looking for).
Like a lot of us, I got caught up a bit in the Loudness Wars but I’m way past that now. If somebody wants my track to be loud they can turn the effin’ volume up.
The problem using plugins to achieve/increase apparent loudness is that they
invariably destroy the impact of drums
add nasty distortion if overdone
So anymore I do what Bredo referred to: I just hit the red and let the convertors chop off any overs. Sounds MUCH better than limiting or loudness maximizers.
And please don’t tell me I must be “doing it wrong” with regard to plugin limiters/maximizers… there are any number of “pro’s” who have forsworn using plugins for this purpose, because they suck
I recall reading this same thing in Bob Katz’ book but I think I remember a caveat something like…“newer convertors perform much(?) better”…(caveat non emptor on this quote!!!)
The dilemma in all this is that louder always sounds better (not “IS” just “SOUNDS”). So when you hand your CD to someone and they pop it in right after listening to the radio or some commercial CD, if yours doesn’t pop at something close to the same volume, it’ll sound weaker than what they were just hearing. I hear this practically all the time from small bands that I record. If I give them a mix that’s peaking at about -3 to -4, without fail, they’ll go listen to it in their car’s and come back asking why it’s so quiet (“I had to turn the car stereo way up to really hear it”). One note on this…I have my monitoring system calibrated to use Bob Katz’ “K20” monitoring system but I find that I prefer to record and mix with the volume far below that (83 db SPL just wears me out to fast).
This is so common, that I’ve started having discussions with bands early on in the recording process to educate them about this and I tell them straight out, if you intend this to be a serious release then I strongly reccomend that you take the mixes to a mastering engineer. I then give them a general explanation of what the mastering engineer will do for them.
But I also tell them that, if this is intended to be a CD that you sell at your gig’s or use for general promotion…you may not need to spend the bucks on an ME. Understand the tradeoff, but it may be a worthwhile one. Bottom line, if you’re going to take the mixes to an ME, dont sweat the level of the mixes. The ME will want the extra headroom to work with and they’ll get it hot for you. If they’re not going to take it to an ME then I’ll throw a maximizer on the stereo bus and we’ll squeeze it a bit till it sounds commercial to them.
Just to return to the OP’s core question…
A question for you would be this…ignore for a moment that your recording’s sound distorted when you hit 0db. With your’s and others both hitting 0 db, do your recordings still sound SOFTER than others? In other words, are you wondering how to get your recordings to “SOUND” as loud as commercial recordings WITHOUT distorting?
The answer to your questions are, yes they still sound softer, and yes I want them to sound louder without distorting - for the record, not because I think that’s always the right thing to do, but because I want to LEARN how to do it. I believe I know why they sound softer - I checked out the average RMS of my tracks and they are much lower than commercial tracks. No surprise about that, as I play with a lot of dynamics, but what opened my eyes is that even at the loudest portion of my tracks (normalized to -0.01 dB), my average RMS is significantly lower than that of a commercial track’s loudest portion.
This can be seen visually - there is no “white” on their track waveforms at the loudest portions, where at my loudest portions there is LOTS of white.
I guess they are compressing more heavily, to make the “white” go away?
I will try that, just to see if that is what happens.
It’s the rms or average level that gives you the loudness not the peak level. You can get a mix that sounds loud that peaks at -1dbfs or even lower.
Getting a tune to compete with the loudness of commercial releases means making some compromises in sound quality and dynamics especially if you do it your self with plugins. If your converters are good enough clipping them by a couple of db and recording back into your daw can help, but only if your converters clip nicely.
Exactly, Mr Dave – there seems to be some confusion in this thread between peak values and rms values.
I don’t pay attention to either anymore. Over time I’ve arrived at a consistent monitor level that’s comfortable but loud enough to RAWK, and I mix against that pre-set level. Sure, they’re not as loud as the typical commercial track, but they’re full enough. I hear pretty bad distortion on many latter-day cuts that have been squashed too much.
Well said Doug. you’ve learned the easier way. You’ve tride the option and ignored the social rules.
Bollocks to the nonsense about intersample peaks. And bollocks to clipping being evil.
Chopping off the overs just distorts at those points where the overs are chopped off. Using limiting or loudness maximizers. And the whole pansy track is distorted.
You have far more technical knowledge than but I have to say that I have experienced intersample peaks causing distortion. I had a mixdown that kept distorting on my laptop onboard soundcard but played fine on my main setup. The peak meters in cubase did not clip. In fact maximum output was -0.3dbfs. I inserted an intersample peak meter in cubase and voila, there was the problem. Admittedly that this is the only time I have experienced this.
Regarding loudness, I find I can normally get a quite loud mix with the use of compression rather than limiting and it sounds much better to my ears than using a limiter to smash the mix. Heavy handed limiting always seems to change the mix. A little soft clipping can sometime work better than a limiter too.
There is no “perfect” way.
Everyone has their own opinions as to what is the best way.
To be realistic, the initial point of degradation of a sound recording starts at the air in front of the microphone and ends at the ear. Virtually every piece of hardware AND software in the signal path is going to add some kind of degradation to the original signal. The last best hope of the recording is it’s “shape” once it reaches the ears.
No actually. Physically, the bit’s may be stored right next to each other and you can visualize them being transmitted in a continuous stream where each one is “right next to each other” but you need to remember that, where digital audio is concerned, there is a time domain that applies to that stream of bit’s.
If you’re sampling at 44,100 samples per sec, thats 44.1 samples per millisec. That means that there is a gap of 22.675 microseconds between each sample.
Intersample peaks which may exceed 0 dbfs aren’t theoretical. The output stage of a D/A converter is, essentially, an amplifier. When the input signal to an amp changes, ideally, the output will change at exactly the same rate and will always reflect the exact same amplification ratio. This is not possible though. There is always some delay in the response of the amp (sometimes called slew rate) and practically every amp will overshoot the target output by some amount.
Here’s where intersample peaks can happen. Imagine that your D/A converter get’s a sample that tells it to provide an output voltage level of .9 v and lets say that 1 v represents 0 dbfs, 22 microseconds later another sample comes along that wants the converter to give an output voltage of 1 v. the converter output ramps up to that but overshoots slightly before settling down to 1v. This overshoot is an example of an intersample spike in which the output of the converter is driven beyond 0dbfs even though it may never have received a sample that actually asked it to do that.
For me personally. I get each track as close to 0db as possible through compression and EQ. Each time the combination of tracks I’m working on peg the 2bus (master bus) too heavily, I start bringing them down. All the while, I’m adding more tracks into the mix. (With 40 or so tracks, your eyes will start to bug out of your head with this method, but I’ve found it works well).
I’ll do that only on rare occasions, especially if there is an error (a glitch due to some sort of malfunction like an electrical noise, etc…). In that case, if a redraw doesn’t work, I’ll just zoom to the sample level and cut it. A few samples cut out of the track won’t appreciably change it’s position… and if it does I can always line it up again.