What is the Ideal dB Level?

[HakimCallier]

The poll was meant to be an underhanded pitch to new music producers.

 

[moses]

wow, great post 73*

 

[73*]

wow, great post 73*

Thanks,

But I know there is tons of stuff I left out, hopefully you want to fill in some of the gaps? After all you do have a deeper understanding of audio than me.

 

[HakimCallier]

wow, great post 73*

I want to really sit down and read it. It looks like a good post.

 

[Emmapeel9]

Cool post 73

You mentioned some rough RMS guidlines in the post; that is -12rmsfs to -20rmsfs. To me these look more like RMS after the mastering engineer has finished rather than what you would send to the ME.

I've got Moses table (Sticky!) which has the rough numbers as -18rmsfs to -30rmsfs for the mixing stage.

Can there be a rough guidline for RMS to go along with recommended volume level in the poll?

EP

 

[Claborn]

-12 to -20 rms is a good range to send to the ME, that just means he doesnt have to work that hard. when he's done the range will be more like -9 to -10 or even -7 to -8 (if he likes the loudness war)

i'm a visual learner so i took some snapshots of what 73 is talkin about.

this is a track i did in it's final mix form. i put my mastering on it and it goes on sample cds and for artists to listen to. they like loud things. it's mainly just to give my tracks that professional sound when they hear it, like a preview of what you will get after mastering of the final product. All my mixing and master effects are done in Reason but statistics read using Cool Edit Pro (a.k.a Adobe Audition) where you can load up your bounced track and hit analyze to see what kind of numbers you're doing. I mainly look at the Total RMS Power rating. -8.15db RMS is plenty loud for just listening to it, my dynamics are still good as you can see my wave isn't a brick wall. Peak volume is -.18db and no clipping samples so my limiter is doing it's job nailing right under 0db and not letting anything through.

after the artist selects the track and goes to record, that's when i go back, take off the maximizer/limiter and send them this (if tracking out is not an option). Without my mastering effects my track is now -13.11 RMS and you can see how much smaller and more dynamic my wave looks. My peak amps is right at 0 and you can see i have some clipping samples so i'd go back and turn down the master fader a few DB's, as i have no limiter now to catch those. After those tweaks, this is what should be sent to either the mixing engineer to add vocals or the mastering engineer, there's room for any processing that needs to be applied. hope this helps.

 

[moses]

yeah, but even your "clean" file is clipped. -13 or -14dB is way to high during mixing. i would turn everything down by 10dB again.

-14dB is what you should have AFTER mastering. -8dB is ridiculously squashed - this is a boring sounding brick wall.

 

[Claborn]

Not sure if you skimmed my post or not Moses.. i was actually looking forward to seeing a reply from you but, i did say that my clean file was clipped and that i go back and turn it down. and that my -8db file wasn't used for anything other than making beat cds and listening sessions for artists.

And i'm not saying i even like these numbers, if it were up to me music would be released at -25db with no auto tune... but this is what this industry asks of me so i give it to them.

 

[73*]

Claborn, One thing to keep in mind is that the "suggested RMS levels" are for Forte/Loud passages, and they are not supposed to be the average level throughout. Moses is correct in that what you've posted up looks like they are both overly limited (but looks don't mean anything in audio).

In my original post I really should have stressed the irrelevance of the exact numbers durring actual mixing or mastering because of all the unique factors that come into play. The unique spectrum of each song, the instrumentation, arrangement and the style all have their influences on what is appropriate level wise. Durring mastering the order of the songs paly a large roll in this also, but obviously Bob Katz has done a much better job writing about this than I (see Mastering Audio).

The underlying danger with a discussion like this is that individuals that may not have the mind for critical thought and analysis are going to walk away from this with another "mix by numbers" guide to go along with the EQ frequency guides. This ultimately leads to more confusion and frustration because if you mix solely by numbers you'll end up doing things that your intuition (the number one thing that you should rely on) tells you isn't right. I'm sure we've all seen a number of posts from others wondering why even after reading the sticky threads at the top and following thier guides still don't have that "professional" sound, or don't understand why things don't sound the way their inner ear wants them to.


All these numbers are good for is establishing that very important base line. The "Reference Point", that you can return to time and again. Once establishing this reference point you're free to trust your intuition and mix with your soul all the while resting assured you aren't ruining your audio unintentionally. The reason the suggested levels are where they are is to provide enough headroom for most styles of music (not sure how to address Power noise or other experimental forms of music). If these levels are too close to the 0dBFS cieling you risk limiting (double meaning intended) what you can do with a song.

 

[Jet City]

a couple of things here.... Why is there +18dbspl on the list (moses you beat me to the punch on that)? If we are going to have a serious discussion with the intent to help others and hopefully all of us coming away from this with a deeper understanding of the topic at hand, i feel it'd be best not to include any "red herrings" (ie:+18dbspl & +4dbfs*).

Another thing that i feel is important to include in this discussion is the rms level of the audio. The rms level is far more important than the peak level. Peak level doesn't tell you anything of value, and does not relate much to what we hear coming out of our speakers.

After all the root of this discussion is loudness and the goal here is to hopefully establish some sort of guide lines that we can use in order not to tie the hands of the mastering engineer. As i'm sure most of us are aware the subjective loudness of a song is heavily related to the average rms level of that song. Now is it going to be very useful for us to start making generalizations of where the rms level of every song should be? I don't think so, and most of us can agree that if every song is different how are we to reasonably create a specific target that we all should shoot for? It just can’t be done, at least not without the sacrifice of the art form that we are dealing with. Sure, this is a mixture of art and science but the science must always be subservient to art or all we end up with is a soulless collection of sounds. So along with a guide line (or range of guide lines) we are going to need to exercise some critical analysis with respect to the values we get from our meters, and know when to simply ignore them for artistic purposes.

What's needed here is deeper understanding of what the db levels (as measured by a daw, both peak and rms) mean and what conclusions we can draw from the values our meters are giving us. It’s important that we understand these things, as simply putting a limiter on the 2bus and forgetting about the levels altogether is certainly no solution, and eventually leads to more severe problems than the one it’s a supposed “solution” for. Luckily for us, there has already been a lot of thought put into this very subject by some very intelligent minds. One of those people that need to be mentioned is bob katz. His suggested calibrated monitoring system (the k-system) allows us to establish a baseline for loudness and gives us a real-world tool to evaluate just where we are loudness wise, based on what we hear through our speakers. Now there is a touch of irony in this, as once most people calibrate their monitoring chain to this system they find themselves ignoring the meters altogether, yet they never end up with mixes or masters that are lacking in the headroom that a given song requires (ie: Unintentionally clipped and squashed and full of unwanted distortion).

peak level: It seems pretty self explanatory but it’s worth reviewing here for the sake of some of the less experienced. What a ppm meter (peak program meter) gives you is the instantaneous highest level of a given sample (the majority of us are dealing with digital, so let’s assume it’s a file playing in your daw of choice). It is really just the level of the sample in relation to 0dbfs (typically expressed “–x” db; where “x” is the level below 0dbfs), now the reason that this on its own tells us little to nothing about what we are hearing is because our ears and brain do not relate a high peak level necessarily to a high loudness level, there is also a frequency dependant sensitivity that our ears have (see fletcher munson curves). Remember this; all it takes is one sample to reach 0dbfs (or our target level of -6dbfs) and we would have satisfied our guide line that we are discussing here. If dealing with a 44.1 khz sampling rate, that one lonely sample is only going to be 1/44100th of a second long and the rest of the file could be dreadfully too low. So mixing a song and making sure you’re hitting -6dbfs isn’t a good goal in and of itself. Really any specific target for peak level is a bit lacking in a solid logic or philosophy. At most there should be a rule of thumb that your material doesn’t exceed 0dbfs at any point in time (assuming we’re talking about fixed point systems). Now a very broad rule of thumb isn’t what we want here so there has to be more to this. If we hope to do anything useful, this is where we want to start looking at rms levels.

rms level (root mean square): This is more of an average level and relates a lot closer to what we perceive as loudness. In the digital domain it’s very similar but not exactly the same as a traditional vu meter. This is the level that a lot of the great engineers will warn you to watch if you’re trying to mix with headroom. Although it’s worth mentioning again that our perception of loudness is frequency dependant (again see fletcher munson curves) but for this discussion that can be somewhat ignored as we are dealing with the metering of the 2bus (master output) of our daw that will have a relatively wide bandwidth signal passing through it most of the time. Depending on the material, you will see recommendations that your peak rms level during loud passages should be sitting anywhere between -20dbfs rms to -12dbfs rms. These are great guidelines to use and they help you end up with a mix that still has a fairly healthy level of dynamics, and not just another victim of the loudness war. However these guidelines still aren’t enough on their own and to follow them blindly is not a good idea. So perhaps we can get better results if we tie the two together, and call it something like “crest factor”. no, that’s not my original idea; i’m just trying to be clever…

crest factor: The term “crest factor” generally refers to the difference between the “peak level” and the “rms level”. For example, say a section of a song has a peak level of “-7.2dbfs” and an rms level of “-18db rms” you’d end up with a crest factor of 10.8. Very simple math! But what does this tells us? How do we interpret these numbers? Well, in the simplest sense it can help to think of things this way: Program material that has a large crest factor is typically going to be described as being dynamic, and the transients (you know good things like drums) will have a strong impact in the mix. Now as you can imagine program material that has a low crest factor will often sound compressed but this isn’t always the case, as music that doesn’t have transients generally will have a low crest factor despite possibly being very dynamic (classical, ambient etc). Now, if you’re dealing with your typical pop/rock/hiphop/rnb/dance music and it has a really low crest factor you can pretty much guarantee you’re dealing with some brick walled noise. Too low of a crest factor on most material and you end up with a lifeless mix that is fatiguing to listen to.

Hopefully i haven’t lost everybody yet, because now is the stage that we get to put these numbers to practical use. By practical use i really just mean that you want to keep an eye on their relative levels. If you find that your rms levels are getting too close to 0dbfs rms (low crest factor) you’re probably killing your mix and tying the hands of the me. If you have too high of crest factor you might be mixing a song too weak for a particular genre. But that’s the beauty of it. It’s up to you as the mix engineer to interpret what the song needs. It’s all about balance, when mixing we want to retain the punch of the drums and give focus to things like vocals, lead lines or what have you. It’s not about trying to squeeze every single decibel out of every bit we have.

What do you do when you’re trying to retain dynamics and headroom but the sound from your speakers is too low? Simple turn them up. That is the basis for the k-system. Having your monitors turned up to a calibrated level that is loud enough to allow the quiet parts to be heard and the loud parts to have impact. It’s a reference point, a very solid one that allows you to trust your ears. Once accustomed to it, you’ll know how loud something is going to be, because you have a reference of how loud things sound. The important thing about a reference is that you can go back to it anytime you need, and that it stays the same (that is why it’s calibrated). It will allow you to mix on intuition and get repeatable consistent results. Seriously read the article linked and you can search gearsluts mastering forum for a few extra help sessions bob katz has given on it (the op started the latest one).

This still feels incomplete but i’ve wasted nearly my entire work day on this. But i may review and add more once i’m home…




*while i included +4dbfs on the list of “red herrings”, it’s worth noting that even though at first glance most people will be inclined to disregard it as a level that does not exist on a digital storage medium these levels can occur when excessive levels are being reconstructed by a dac see this (http://www.tcelectronic.com/media/nielsen_lund_2000_0dbfs_le.pdf) paper. Here is a link to a free plug-in meter that is designed specifically to detect inter-sample peaks. http://www.solid-state-logic.com/music/x-ism/index.asp

*preeeeeeach*

 

[moses]

Not sure if you skimmed my post or not Moses.. i was actually looking forward to seeing a reply from you but, i did say that my clean file was clipped and that i go back and turn it down. and that my -8db file wasn't used for anything other than making beat cds and listening sessions for artists.

Oh, in this case, i misunderstood your post, your "i'm a visual learner so i took some snapshots of what 73 is talkin about." fooled me.

Anyway, there's perhaps an important fact about digital audio storage that might help you to get a better "feeling". nearly all modern audio apps work with floating point numbers (usually 32bit or 64 bit). This kind of number representation has several advantages. Let's have a rough look at these number representations, there are two main methods to represent numbers in digital systems (but there a more of course, but these two are used in audio dsp).

I'll try to explain them in a musicians context, this is not an accurate technical explanation.

Integer numbers have a fixed range. an 8 bit number for example is ranging from 0 to 255 (256 steps in total). the number is directly stored as an array of bits (the more bits, the more steps you have). this has two important effects when storing and especially when processing audio signals. the first is, you have a fixed max value (255 in this case) - any value above gets clipped (this also means it has a limited headroom). the second is, the precision will suffer if your whole signal (track) does not use the whole range. say, your signal has a max peak at 120, than you'll only have 120 steps of precision left (which is basically a precision of 7 bits).

Floating Point Numbers are stored in multiple parts. one part is a simple integer number - it contains the normalised value. another part basically stores the position of the decimal point (the are more parts i won't mention here). and here's the nice stuff: 1) floating point numbers don't have a fixed range - they don't clip, in other words, they have an unlimited headroom. 2) no matter how big or small the number is, it always uses it's full precision range.

now, back to what all that means for the modern floating point audio apps:

a) they cannot clip internally (but you should avoid hitting the zero, many plugins don't work properly when overdriven).

b) no matter how low your level is, the precision is maintained.

but take care, most audio files are stored as integers (but you can also use fp files) and AD/DA converters only accept integer signals! so, don't overgeneralize this. i'm only talking about closed floating point systems. but in any way, 24 bit numbers have many steps (256 times more than our ears can percieve!). So, even a 24 bit integer number is REALLY huge.

and finally, don't forget that any analogue device (even your monitoring amp) has it's own sweet spot around 0dBVU (which is -18dBFS or -12dBFS on most converters). everything above will be more or less distorted.

the conclusion is, you should work with generously low signals between recording and mastering, because it doesn't harm and avoids any kind of unitentional squashing and distortion that happens when constantly working at the limit. the K-System helps a lot to keep the levels nice and healthy. working at the limit just restricts you - it's not "hot" at all.

hope this helped.

 

[Claborn]

oh ok Moses good stuff, nice post. and 73 you're right, it does look pretty beefed, but it's a bass heavy club track and thats just how it came out. that's why i always do my statistics last and mix everything by ear first... looks aren't everything lol.

 

[HakimCallier]

-14dB is what you should have AFTER mastering. -8dB is ridiculously squashed - this is a boring sounding brick wall.

More and more I am realizing the reality behind mastering. And though the numbers are correct, there is still something missing. Mixing and mastering cannot be numbers alone, although I thought this way for many years. I am beginning to use my ears more and more. And although I am a convinced K-System user, I will admit that I have pushed the K-System to areas that I doubt Bob would approve, +9 in K-14 for example. But it made good musical/business sense IMHO.

 

[Claborn]

To go along with HakimCallier.. in all reality, it's not even up to the engineer what DB's are ideal for audio mixdown. The client has the final say so, like i said, if it were up to me music would be released at -25 lol. I've done mixdowns where the guys in the studio are telling me to "smash that hoe, i want it to sound full!! make it loooud!!" so i'm cranking compression and limiters and when they are finally satisfied, i check my rms meters and 90% of time my reaction is.."umm... whoa".

 

[milkybeatz]

great thread im coming out of this feeling like i just leeft the most information class of my musical life, haha. and i know not to mix by numbers and such but im having a little trouble unerstanding the crest factor 73 mentioned, your example was 10.2 but you mention low and high factors being overly compressed or lifeless. at the current time i cant experiment to hear what they would sound like(im sitting at starbucks) lol but im just trying to see what numbers were talking as being too high or low.

 

[Drethe5th]

If its for a mix straight out the DAW then using a limiter and cranking things to 0db is fine.

If its for mastering, the file you send should be -6db.

Read any blog or post in a magazine or online forum about mastering and they all say the same thing. Mastering Engineers hate when people send in tracks that are cranked to 0db. Especialy if it was blindly done using heavy amounts of compression. It leaves them no room to properly compress and raise levels in the way that they prefer. So in other words there was no point in sending them the files to be mastered in the first place because they couldnt do what they truly wanted to do with the material.

Its like lining up your own hair cut, and then going to a barber shop and asking them to give you a haircut just like the guys picture on the wall. Youve already started in a certain direction good or bad. In most cases its bad because if you were that good you wouldnt be asking him to cut your hair to begin with, youd do it yourself. Sure they will accept the work and charge for it, but it is pointless. Theres not much if anything that they can add to the situation and that is the basic reason for getting a master done, so that it sounds better then the "naked" mix by it's self

 

[Noyzi]

If its for a mix straight out the DAW then using a limiter and cranking things to 0db is fine.

If its for mastering, the file you send should be -6db.

Read any blog or post in a magazine or online forum about mastering and they all say the same thing. Mastering Engineers hate when people send in tracks that are cranked to 0db. Especialy if it was blindly done using heavy amounts of compression. It leaves them no room to properly compress and raise levels in the way that they prefer. So in other words there was no point in sending them the files to be mastered in the first place because they couldnt do what they truly wanted to do with the material.

Its like lining up your own hair cut, and then going to a barber shop and asking them to give you a haircut just like the guys picture on the wall. Youve already started in a certain direction good or bad. In most cases its bad because if you were that good you wouldnt be asking him to cut your hair to begin with, youd do it yourself. Sure they will accept the work and charge for it, but it is pointless. Theres not much if anything that they can add to the situation and that is the basic reason for getting a master done, so that it sounds better then the "naked" mix by it's self

I remember sending a track out to be done and the ME sent it back asking us to drop the level to -6db and remove compression cause we didn't know what we were doing. Needless to say that we kept going to him for the rest of the project and was happy with the outcome after his advise.
BTW... I learned a few good things in this thread... my eyes hurt now but you guys really know ur shyt. thx a mill.

 

[LeodPenrose]

below zero? There are a lot of numbers under zero on ppm! Which of those numbers is preferred?

-12dbfs = 0 db

 

[HakimCallier]

I feel it'd be best not to include any "red herrings" (IE:+18dBSPL & +4dBFS*).

Moses already commented on this point but it has been noted.

After all the root of this discussion is loudness and the goal here is to hopefully establish some sort of guide lines that we can use in order not to tie the hands of the Mastering Engineer.

Actually the root of the discussion is NOT loudness. I am asking what level the mixing engineer should find peak value upon a final mix, in preparation for the next stage of audio production. That's all.

However, just as in the process of audio production, the loudness factor is an important element of the final product. I know that the loudness issue is weighing heavy with you and others at Gearslutz right now which is why this element may be at the forefront of your thinking rather than at the end like mine.

I think your information on RMS and "loudness" in general is excellent and worthy of its own thread.

Oh and by the way the SSL X-ISM is an excellent FREE meter. Its a life saver when you are so into the art of your mix that you start to loose track of the finer technical details like word length.

http://www.solid-state-logic.com/music/X-ISM/index.asp

 

[Emmapeel9]

Does this one get auto-bumped because it is a poll? Got 5 stars too. Worthy thread off course.

When do we find out if we have won or not? Yeah! 18 people going to disneyland. (If right)

EP