KVR Mix Workshop - Week 9: Mastering

How to do this, that and the other. Share, learn, teach. How did X do that? How can I sound like Y?
RELATED
PRODUCTS

Post

(haha, ok so mastering isn't actually a part of mixing. It is, however, usually the next step...)

Week 9: Mastering

Original Final Version
Cordelia's version
JCM's version
Poonna's version
Shanecgriffo's version
Shogger's version
My original unmastered mixdown of Flesh And Bone (warning: 75MB uncompressed WAV file: 32 bit float, 44.1kHz)

(note that Cordelia used her own mix, whereas everyone else used my mixdown).

Yes, this is extremely long. If you think reading this is a lot of work, spare a thought for how much work it was to write it! I wrote this for all of you!

So here we are, at the final week of the KVR Mix workshop!

My mastering process follows these stages:
1) Initial setup
2) Tone correction
3) Dynamics and headroom correction
4) Bouncing and encoding

But first, a few words about the concept of mastering...

As you're no doubt aware, there has recently been quite some interest in mastering on KVR lately. There's a lot of noise about, and I'm not sure there is a clear understanding of what mastering is and what to expect from it. Some people think mastering is inserting a limiter on their 2-bus while their mixing (and composing!). Some people think mastering is a mysterious process that adds magic mojo to their music. Some people think mastering is a process for crushing the dynamics out of their songs. Some people think the best mastering engineers never use compressors or limiters (or any digital equipment). There's a lot of misinformation around, and a lot of misunderstanding.

I am not a professional mastering engineer, and I'm not going to speak for professional mastering engineers. I have, however, been mastering my own music for over ten years and I've developed a process based on my experience in analytic listening, DSP algorithm design, psychoacoustics, and almost two decades of studying music. This is a description of my best process at this point in time, but I expect it to improve in the future. In addition, I'm only going to discuss mastering for internet distribution. Mastering for CD is a slightly different process which I'm not going to discuss here. Likewise I'm not going to address mastering for other formats, such as radio, vinyl, television or film.

My approach is that mastering is not about making a song sound "good", it's about making it sound "right".

Let me say that again. Mastering is not about making a song sound "good".

The character (the "mojo") of the song should be addressed in preproduction, recording and mixing. If you wanted it fat, you should have used fat sounds. If you wanted it shiny, you should have used shiny sounds. If you wanted it to sound awesome, you should have written an awesome song. Mixing is where sounds are shaped. I realise this might be a little frightening to those who were waiting for mastering to add the magic juju. I actually find my process to be good for beating laziness. I simply don't have the option of settling for a less-than-ideal mix on the basis that it will be fixed in mastering. This way I work harder to get The Sound at the mix.

Mastering is the process of preparing the mixdown for distribution. Nothing more, nothing less.

The mixdown sounds great in the studio, and it's a 32 bit uncompressed wave file. We want a master that sounds great everywhere (especially computer speakers and iPods), in a data-compressed format that's suitable for internet distribution. This is what the mastering process is for. My mastering motto is "First, do no harm". This means that I first assume that I will not change the sound. I then only apply the minimum required processing to achieve what I need to do.

1) Initial setup
First, you'll need to find two or three commercial songs that sound similar to Flesh And Bone. Key characteristics of reference tracks are:
- Female lead vocals
- A guitar-strong mix (but not guitar-heavy!)
- Pop production values (lead vocal and snare are prominent, high overall average level)
- Commercial budget and backing (not an independent or low budget release)

Essentially you need to find tracks that share a similar sound and production values. This is to make relevant comparisons quick and relevant. It's faster to accurately compare the tone and dynamic behaviour of two tracks if the overall production values are similar. It also makes the mastering more relevant, as the song is most likely to be played alongside similar songs (people who enjoy Flesh And Bone probably also enjoy similar-sounding music). Usually I go through my music library until I find about ten songs that I think could be relevant. I then cut that list down to the three best (most relevant ) songs. These songs must also be from different albums and different artists.

Once I've chosen the three reference tracks, I bring them into a new project in my sequencer, alongside my own mixdown. Each song has its own track, and each is a single audio file beginning at 0:00. I always have one track soloed at any one time. This way I can switch between songs by simply changing the current track (using the up and down arrow keys).

Next the tracks need to be level-matched. You MUST do this by ear. I usually find that my mixdown is about 12dB below the commercial reference tracks. Remember: this is not RMS or peak, but average level as you hear it. Keep the mixdown level at unity gain (fader at 0dBFS) and reduce the level of the reference tracks. This way you don't have to change your monitor gain - you can keep monitoring (for now) at the same level you mixed at.

By this stage, you should have four tracks in your sequencer - your mixdown and three reference tracks. One should be soloed, and when you change between the tracks they should all be at the same volume level.

2) Tone correction
At this point you might be hearing some pretty obvious tonal differences between the mixdown and the commercial reference tracks. Personally, my mixdowns are usually too strong in the lower mids and too weak in the low bass and the highs. Other common problems people might have is too much lower bass or upper mids, or perhaps not enough highs.

To correct the tonal balance, you'll need a very flexible EQ with as clean a sound as possible. This is likely to be different to your mix EQ, which (at least in my case) has fewer controls and more "funk". My mix EQ is designed for quick operation and character sound. A mastering EQ, on the other hand, is something that you'll spend a lot of time tweaking so it needs to have full and precise control. It's also being used exclusively as a corrective tool - the last thing you want when correcting the tone of your mixdown is for the EQ to be introducing additional changes (character) to the sound that you don't have full control over. For this reason, avoid any plugins that add harmonics or saturation (or any other non-linear behaviour). You'll need an EQ plugin with at least five fully-parametric bands (with controls for frequency, gain and bandwidth), and you should know how to fine-tune gains to within a tenth of a decibel (0.1dB).

If you don't have a suitable EQ, grab the free version of Electri-Q (posihfopit edition).

The process of correcting the tonal balance is iterative. Start with broad boosts or dips to get your sound in the general ballpark of where the commercial references are. Keep checking against the references. Gradually get more and more precise with your adjustments, adding more bands as necessary. You might find it useful to doublecheck your work on headphones or other speakers, as they'll give you a different perspective and allow you to focus your attention on different parts of the frequency spectrum. You might also find that you have to change the overall volume of your mix to keep it the same as before, to keep the comparison even.

I am aware there are automated tools that can do this. I prefer to do this manually instead of using automated tools because adjusting the tonal balance depends on the overall perceived tone of the song. This is different to doing a blind (deaf?) mathematical comparison because there are important factors that the current mathematical models do not take into consideration. This goes beyond the psychoacoustics of phenomena such as harmonics and masking (although they are important). I'm talking about human judgement - taking into consideration the stylistic differences between the mixdown and the references. For example, one reference may have a thinner kick drum and a deeper melodic bass than your mixdown. Based on a "pure" frequency analysis your mixdown might be too strong in the lower mids where the bass is. Listening to it though, you might decide that this is where the character of the bass lies, and that it is a major component in the mix. Hence, you could decide not to suppress it, even though doing so would bring it in line with the reference. Remember also that I'm not talking about radical changes to the sound. I'm talking about very subtle adjustments - these types of "human" adjustments are in the range or +/-3dB. Not something that's going to radically skew the balance of the song, and not something that's going to make the difference between a good and a bad end result.

You'll know you're ready to move to the next stage when you can quickly flip between the commercial reference tracks and your mixdown and they sound like they're all on the CD or radio station. Your mixdown doesn't need to sound EXACTLY like the references, but the difference between your mixdown and the references should be smaller than (or at least the same as) the differences among the references themselves.

3) Dynamics and headroom correction
The next step is addressing the dynamics.

Remember my motto? First, do no harm. Don't add a compressor just because it sounds like a cool idea. Don't bust out your multiband juju just because it's what everyone else is doing. Remember - don't change the sound unless it's absolutely necessary.

First though, a quick side-discussion about levels.

Levels ain't levels. There are three main types of levels that I talk about: Peak, RMS, and average.

Peak levels are the highest digital values that are in the waveform as it exists in the computer (or other digital equipment). While peak levels aren't directly related to how we hear the sound, they are crucial for correct gain staging in digital gear.

RMS is a way of measuring audio levels that more closely reflects how we perceive sounds. The thing to keep in mind though, is that RMS level are still momentary - variable from moment to moment.

Average level is the overall volume level that we hear. You can think of it as a longer-time version of RMS. For example, if a song has a loud kick drum that is "poking out" of the mix, the RMS level is jump each time the kick sounds. The average level, however, will stay (relatively more) steady.

Here's why they're important. The average level of the mixdown needs to be at the same level (distance from 0dBFS) as the reference tracks. The difference between the peak level and the average level needs to be controlled (usually reduced) in order to bring the average level to a certain position relative to 0dBFS. Widely-varying RMS levels will make this difficult, because if the RMS level is varying a lot the average level will be significantly below the highest RMS level. This is a problem because there's a physical limit to how close you can bring the peak level to the RMS level before you get too much distortion. This in turn limits the degree to which you can reduce the headroom (reducing the peak level without reducing the average level).

You might need to read that a couple of times. I know I did! I'll give you a moment...

Now that you've taken all that in, you'll understand the constraints in reducing the headroom requirements of the mixdown. The more you can reduce the headroom, the more freedom you have in setting the "loudness" of the sound compared to the rest of the world.

You'll also understand that while we have tools to help us reduce the headroom, we might not need to use them to their full potential. That's right. First, do no harm. If you don't need a limiter, don't use one. If you don't need a compressor, don't use one. If you don't need multiband compressors or dynamic EQs, don't use them. Understanding levels will help you decide if you need these tools or not. But what if you need them? Approach these tools in the order of least damaging to most damaging...

Let's assume you brought down the levels of the reference tracks by 12dB, and they're all at the same average level as your mixdown (which is still at unity gain).

First, look at the peak levels of the mixdown. In a perfect world, the peaks of your mixdown would be no higher than -12dBFS. This means you could raise the gain by 12dB and the song would be at commercial levels. Of course, in today's world or loudness wars this is extremely unlikely!

The first tool to try is a limiter, or more preferably, a clipper. Used sparingly, these tools are the most transparent of all dynamics-shapers. Remember - transparency is key. First, do no harm. The problem with most digital limiters, however, is that have an integrated threshold and input boost control, which means that as you apply more limiting, the audio level increases. This typical method of "pushing up" the level of the mix makes it difficult because you're changing two things at once - the peak/average ratio and the overall level as you hear it. Not only does this complicate the hearing process, but it also makes it easy to ignore audio damage because the more damage you do, the louder it gets (and as you know, humans tend to perceive louder music as "better").

To get around this, you should insert two gain plugins, so your chain looks a little like this:

gain1->limiter->gain2

Set the limiter to limit at 0dBfs. When you switch it on, you should hear no effect. Then slowly increase the level of the first gain and simultaneously reduce the level of the second gain by the same amount. This effectively allows you to reduce the threshold of your limiter while keeping the average level unchanged. Do this 1dB at a time, and listen carefully to what you're doing to your sound. Be sure to do this while listening to the loudest part of the song - this is where headroom is most critical.

If you get down to -12dB (or whatever your references are at) without reducing the overall level of your mixdown as you hear it and without doing too much audible damage to your sound, then you don't need to do any further dynamics processing! Skip to the next step! This is usually the case with most of my masters.

If, however, you couldn't get to the target level without damaging the sound, then you'll need to use some more sophisticated tools. Which tools you use will depend on the nature of the audio and the type of damage that the limiter or clipper is doing.

If you can get close to your target (within 3dB), then you might want to try using a different limiter or more aggressive clipper. There is some variation in how much different limiters and clippers can reduce headroom while retaining transparency. Different tools will excel on different styles of music and types of sounds too. Usually smoother acoustic or ambient electronic music will benefit from cleaner digital limiting, whereas rock or other music with deliberate distortion (such as this song!) will benefit from more aggressive clipping.

If you can't even get close, this kind of trouble will occur when the RMS levels are widely varying. Listen carefully to what elements in the mix are causing the wide variations. If it's just a single sound (such as the kick or snare drum), you'll get your best result by going back the mix and reducing the level of that sound. Typically it doesn't take much (usually less than 6dB) to make enough of a difference without disrupting the balance of the mix. You might have to revisit the master EQ though!

If it's several elements that are fluctuating a lot, you might need to employ a compressor. Start with the fastest attack and short to medium release. If your compressor has an RMS detection mode, use it. Regarding threshold and ratio, there are two approaches you can use. One way is to use a high ratio and high threshold so that only the moments with high RMS values are compressed. The other way is to use a low ratio and low threshold so that the overall ebb and flow of RMS levels keeps its momentum but the lowest and highest levels are gently brought closer together. Which approach you use will depend on the nature of the audio and what you need to do to it.

Just a quick note - make sure your limiter or clipper isn't active while you're adjusting the compressor! Otherwise things get EXTREMELY complex, and nigh on uncontrollable!

If the RMS levels are mostly stable but with just a few spikes (and it'll be these RMS spikes preventing you from reducing your headroom sufficiently) then you'll be best served by the high-ratio high-threshold approach. This will keep the average level unchanged, and only bring down the volume during those moments that the RMS levels are spiking. Set the ratio and threshold at their highest (there should be no gain reduction yet). Then gradually lower the threshold while watching your meters (both peak and RMS, as well as the compressors gain reduction meter) until you've given yourself the room you need. Typically the amount of gain reduction will translate very closely to the amount of additional limiting or clipping you need. Your attack and release times will depend on the compressor you're using. For a dedicated mastering compressor or a compressor detecting RMS values instead of peak, the fastest attack time will probably be suitable. If you're using a general-purpose compressor though, you might find a fast attack time causes some distortion or undue modification of the sound. In this case, use a much slower attack time than normal - start at about 50ms, and go out from there. The same goes for release time. Start with a fast release for a dedicated mastering compressor and a medium-to-slow release for a general purpose compressor. If you're finding that you can hear the gain reduction, you'll need to increase the release time until it sounds transparent (but is still fast enough to reduce the level of the RMS spikes without changing the average level). If the compressor still sounds like it's "clamping down" on the signal, try reducing the ratio. Keep in mind though that this means you'll be getting less gain reduction, and so you might need to lower the threshold further, which then might bring you dangerously close to reducing the average level too. Be careful and be conservative.

If the RMS levels are widely fluctuating, you might be better off with the low-ratio low-threshold approach. Start with the ratio at 1:1 (no gain reduction at any threshold!) and the threshold at the lowest level possible. Set the attack and release time to fast or medium settings for now (the advice in the previous paragraph applies here too). Now here's the tricky part - gradually increase the ratio while simultaneously compensating for the change in volume. Depending on the compressor you're using, the volume will either increase or decrease as you increase the ratio. Unlike the high-ratio high-threshold approach, the compressor will be working all the time. It'll also be more difficult to tell how much compression is enough. Because of the nature of what you're doing as well, you'll also need relatively more audible compression to reduce the headroom requirement of the track. To get a sense of the amount of headroom you're reducing, look at the difference between the highest and lowest levels of gain reduction on the compressor.

Either way, don't apply more compression than necessary, even if it sounds cool! Remember - you're trying to preserve the sound of the mixdown as much as possible! When you think you've applied enough gain reduction, add the limiter or clipper from earlier and try to get to your target level. Make sure the compressor is out of the gain loop though! Your chain so far should look something like this: EQ -> Compressor -> +Gain -> Limiter/Clipper -> -Gain. For a particularly troublesome mix, this could require an iterative process - repeatedly switching between adjusting the compressor and adjusting the limiter. Also keep an ear out for the tone as well - applying heavy gain reduction will also change the tonal balance of the song.

You know you're ready to move to the next stage when you the second gain plugin is reducing the same amount of gain as the channel faders for the reference tracks. If you've got this far, congratulations - the most difficult part is over!

4) Bouncing and encoding
Turn down your monitors by the same amount as the second gain plugin. Everything should suddenly get quieter.

Deactivate the second gain plugin. Your mixdown should return to the same volume it was at all this time, but the peaks should be hitting 0dBFS now, and should be at commercial levels.

Insert your favourite dither at the very end of the processing chain. The VERY end. Make sure it's set to 16 bits.

Set your locators or region for the section to be rendered. The start of the section is easy - it should simply be the moment of the first sound of the song. The end of the section is a little more difficult. You'll have to use your judgement to choose an appropriate length of time. In this world of zero-gap playback, you'll have to assume that the moment the end of the section is reached, the listener's music player starts playing whatever is next. You also might need to do a quick fade (make sure the gain stage you're fading is before the dither!) to make sure you don't chop off any reverb tails.

Just a quick side note - This particular song is relatively easy to set the bounce section because there's a definite end to the song. That is, it doesn't end by fading out. When I have a song that I want to end by fading out, I render the mix without the fade - I just have the last section continue at full volume. I then perform the fade in mastering. This is partly because the nature of the fade depends on the next song in a multi-song package, but also because dynamics processing can sometimes damage the linearity of the fade (making it sound like the fade starts slow and then suddenly drops off). For this reason it's better to apply the fade after any final dynamics processing.

Once you've set the bounce section boundaries and applied the fade, bounce the file to a 16 bit 44.1kHz stereo uncompressed WAV or AIFF file.

Take a moment to relax and breathe deeply.

Next, encode the uncompressed file to a lossy compression format. I recommend MP3 CBR (constant bitrate, as opposed to variable birate) for maximum compatibility. Variable bitrate (VBR) has the potential for high quality at similar average bitrates, but in my experience there are still (in 2008!) some players that have trouble with VBR. This is the same reason I choose MP3 over AAC or WMV - better compatibility.

The actual bitrate you choose is a matter of contention. For a distribution format, some people prefer 256kbps. Some won't settle for less than 320kbps. Others will demand nothing less than a lossless format such as FLAC.

In my personal listening tests, I've found 192kbps to be sufficiently indistinguishable from the original. That's not to say there's no difference - but I've determined (to my taste) that the difference is not noticeable under regular listening conditions. Even in the studio, I'd only tell it the audio was compressed if I were A/Bing it with the original (and listening EXTREMELY intently!). A couple of years ago I was using 128kbps, but I started to notice some subtle artefacts in casual listening. Since moving to 192kbps I haven't noticed any artefacts in casual listening. I've also not received any credible complaints about the quality of my encoding.

Once you've encoded the file, you need to tag it. (I provided specific advice to participants for tagging this project)

And now... CONGRATULATIONS! You have a master!

<- Week 8

-Kim.
Last edited by Kim Lajoie on Wed Nov 05, 2008 12:08 am, edited 2 times in total.

Post

We finally made it! Woo! :party:

I'd like to extend a special thanks to everyone who participated - whether you stuck through the whole thing, joined for a few weeks, or just jumped in at the end. No matter what your level of participation, I appreciate it. It is you, the participants, that made this workshop possible!

-Kim.

Post

hey thanks Kim,
the workshop really helped fine tune my EQ skills during the mastering stage, its made my current mastering project(for my own band) easier and better sounding!

Post

Thank you Kim.
This has been a wonderful experience. I learned more in these nine weeks than I had in years before this, just struggling on my own. You've taught me a completely new way of listening, my approach to mixing will never be the same. I hope (I believe) I have learned the important concepts you have tried to teach us.
You have been so generous with your time, sharing your talents and giving excellent feedback and advice.
I'm sorry to see the workshop end, but I'm happy with the results, proud of my progress, and ready to apply all of his new information to my own mixes.

It's been fantastic fun as well. :)

Thank you sincerely!

Post

Kim, i wonder, is the info from the other weeks (mixing workshop)
available as well? i wouldn't mind being able to read through it as well, seeing as i only came onboard in the last week. thanks if so.

Post

shanecgriffo wrote:Kim, i wonder, is the info from the other weeks (mixing workshop)
available as well? i wouldn't mind being able to read through it as well, seeing as i only came onboard in the last week. thanks if so.
At the bottom of each week's post there is a link the the previous and next week's post. I'll also soon put together an index and link to it from my website.

-Kim.

Post

cool, might wait for the index. :wink:
thanks

Post

Hey Kim,

I want to give you a huge thank you! :tu:

This was so great of you putting all this together and working for many hours to get this done. Big thanx for your advice, attitude, patience and friendliness during all this.
For time restrictions I didn't participate the mixing part though I often had an eye on it and doing one or two sessions but never sent them in. When I find time I'll do the other mixing parts to learn from your written advice. I wished I had something that good when I was at the audio engineering school back then.

So big love again, what you did is very precious. :love: :band: 8)

Shogger
What?

Post

Much appreciation to Kim. This workshop was great practice for me and a quite enjoyable piece to mix. The work you put in was obviously tremendous. My thanks for all the time and effort you put into this.

Post

As with everybody, thank you so much for your time and effort conducting this workshop. I joined in only at the last week because I missed the news for the first week, but only this one week has already helped me a lot. Thanks again! :)
Peace, my friends. I'm not seeking arguments here. ;)

Post

I know I haven't really got involved with this but I've followed it throughout and I have to say you are a truly great man, Kim. Thank you, sincerely.
I need a soundcloud ¬¬

Post

(note that Cordelia used his own mix, whereas everyone else used my mixdown)
not that it matters, but i always thought cordelia was of the female persuasion.

it's a shame i couldn't take part in this, but all the same it's awesome to offer your help so generously. :)

Post

fabi wrote:
(note that Cordelia used his own mix, whereas everyone else used my mixdown)
not that it matters, but i always thought cordelia was of the female persuasion.
Me too, but then again I thought Kim may have been that way too. As for me I was suspected to be a boy with a fetish for techno... for the record i'm neither. :hihi:
I need a soundcloud ¬¬

Post

fabi wrote:
(note that Cordelia used his own mix, whereas everyone else used my mixdown)
not that it matters, but i always thought cordelia was of the female persuasion.
She is.

By the way Kim, have you posted your comments on these various mastered versions? They are fairly similar overall, and all are equally loud. It would be interesting to see what your analysis is. :)
We escape the trap of our own subjectivity by
perceiving neither black nor white but shades of grey

Post

Caysi wrote:
fabi wrote:
(note that Cordelia used his own mix, whereas everyone else used my mixdown)
not that it matters, but i always thought cordelia was of the female persuasion.
Me too, but then again I thought Kim may have been that way too. As for me I was suspected to be a boy with a fetish for techno... for the record i'm neither. :hihi:
Ha. I didn't notice the "his" in Kim's post.
Yes, I'm of the female persuasion.

So, what became of Yonyz' and ngarjuna's final masters?
Did I screw up? Was I supposed to master your final mix, Kim, instead of my own?

I had more trouble with the mastering of my mix than I did with any other stage of this process. Now I know I'll definitely hire a professional mastering engineer to master my mixes if I need to.
I think I'd just gotten too familiar with my mix and couldn't hear it anymore. I also found out that my monitors have limitations I need to address, although using headphones to hear the lowest frequencies helped.

Now I'm going to take the time to listen to all the masters, and hopefully learn a bit more.

Post Reply

Return to “Production Techniques”