One Mastering Engineer's Guide to Mastering

White_Noise JPI

New member
Hi guys, I started out self mastering my music about 4 years ago. I didn't know much when I started, but I knew that most mastering services I was being targeted by on Soundcloud and such were not doing much special and I decided to try and figure out the dark art for myself. These days, I suppose you could call me the "resident mastering engineer" for a netlabel that I love, and my work is called - by pro DJs and radio stations - commercial quality. I'm always experimenting with my mastering chain, and of course no two songs are exactly the same. For instance, right now I'm looking into clipping VSTs for the end of my signal chain as I hear that many of the top mastering houses are going that direction and just letting things clip in their A-D converters (though I'm fully in the digital domain). Anyways what follows is a step-by-step plugin by plugin breakdown of what I do to take a solid mix and turn it into a mastered track ready for distribution along with explanations of what everything is supposed to do and why I think it should work that way. I hope this helps some of you out!

My mastering chain is a lot of layers, each doing a little at a time. In that way, you can make big changes if you need to, but it can still come out relatively subtle. There’s no hiding lopping 3db off everything above 2k, but you can make it not sound like the whole mix went through a LPF and still make big moves. First, I use Airwindows bitshift gain to put the song into the right loudness range on the way in. This is only for when I get something where I was left 2db of headroom or max peaks of -24 db. Bitshift gain can change the level of the audio with no noise, very simple calculations, and next to no latency. Unfortunately, it can only do this in increments of exactly 6 db, so you only use it to take a mix that would otherwise not work at all and make it work. I didn’t have to use this on any tracks for the past few albums, but you want it around when you need it.

Once that’s done, we have a little more housekeeping to do before we get to the good stuff. Subsonic frequencies and the lowest audible frequencies really cannot be played properly by any but a handful of systems in a given area, and odds are the most of us do not have access to such a system. So, I took a tip from Deadmau5 when he said that he cuts everything below 40, even 45 hz in the mastering process, and he uses a free vst called Engineer’s Filter to do it. Those low frequencies can build up a lot of energy that no one can hear, but will make saturation and limiters misbehave. We want rid of them and I use Engineer’s filter with an order 20 HP Papalous filter to do it. You can examine some of the other models available, but the goal is the same: to cut the low frequencies quickly WITHOUT introducing phase issues in the filter response. You only have one octave to work with, so a gentle 6 db/octave highpass isn’t going to do much for you. But, even izotope’s finest brickwalls cause some phasing issues. I’ve deemed that the equivalent of a 96 db/oct filter with the particular phase response offered by Engineer’s Filter’s models is the right compromise for me. Just keep it well away from the bottom of the kick to be sure.

With those adjustments, we’re ready to begin taking the track from where it is to where I feel it needs to be given the context. The first thing I do given any mastering job, is listen to the whole project beginning to end and take notes on each track about what needs to change, any special elements I want to highlight, and any elements I want to try and leave alone. After that, I leave it a day, look over my notes and, if they still make sense the next day, I begin with fresh ears, usually with the quietest track first. I can work up to two hours in an evening, usually, before I start to make incorrect choices due to hearing fatigue, which is fine since that’s about as much time as I often have anyways. If I have time to work in the morning it’d only be an hour, then I’d have all day at work to recover, then I could still get in 2 hours at night. I focus in by looking at the loudest part of the wave print and start playback just before that, looping that section of the song.

The first processing that happens (after gainstaging and sub-filtering from earlier) is a broad EQ adjustment that sets up the track for subsequent processing. I use Accustica Red EQ for this as it sounds good, has a reasonable ability to adjust the balance of the mix, and I can get something I like out of it very quickly. Accustica doesn’t distribute this one anymore, so I’d investigate (and have set up) TDR Slick EQ as an alternative. With some tweaking you can get a very similar response to Red EQ, though I’m not as fast in this one so I still don’t really use it. But for the day that Red EQ breaks or I lose that hard drive, Slick EQ is my alternative (and is probably the technically superior product). In any case, this is only slightly going to affect the end product, but it’s really setting up what we want to drive into the next stage of processing.

And that next stage is a mastering tape preset I made for myself in Satin. I use the crosstalk here to give a suggestion of physical circuits touching the audio, where -70 db of crosstalk is considered very good hifi gear and anything lesser is only going to be more ( I use -70 by default). The general idea is that it’s a very clean tape machine. I do have the opportunity here to reduce tape speed, change pre-emphasis, bias, or pre-post companders to alter the response. My default settings are the highest fidelity while still impacting the sound and I can degrade various aspects from there. Satin is great for this because you can see in real time what the tape machine’s response is to various controls. So, even though I’m no tape expert, I can get the response I want very reliably. When not doing anything special, Satin’s job is to slightly attenuate the extreme frequencies and help bring the mix together between crosstalk and tape compression. Airwindows has some good free alternatives with ToTape and Iron Oxide, which even have some tricks that Satin doesn’t (though no real-time response graphs).

Next, and this is a relatively new part, I use the free TDR Kotelnikov to take off just the harshest elements. I’m only looking for 1-2 db of gain reduction from an already very clean plugin. I slightly modify the Tight Mastering preset. I play around with the detection filter to avoid pumping, that is not the goal of this kind of compression. Instead, I listen to the delta channel and I want to take out any of the harder percussion that tape hasn’t softened (and isn’t in the bass). I use no makeup gain or increased output. Instead, I bring the dry mix volume up to around -25 db. I’m not sure of the technical description for what this does, but to me it feels like it takes any space in the mix and fills it in with related musical content. The effect seems to be that it very transparently raises the level some in the quieter parts of the track without killing dynamics – I think.

After that, the song is usually in the ballpark of what I think it should sound like, but there still needs to be more precise work done to dial in how all the instruments, or various frequency bands, are working against each other, and for that I use Ozone 8. I don’t know of any single plugin that can replace Ozone in my workflow (I know how all its modules work too well and am too used to the Izotope sound from using Alloy 2 and now Neutron 3 for my mixing), but I will attempt to offer free alternatives to the individual modules so that you can try my ideas for yourself if you want.

First, I go into Ozone’s EQ and run another highpass filter (flat, 48db/oct) at around 30-35 hz, just in case Satin added any rumble. I then serve the song as needed, using a combination of broad bands to shift the tone of the track and narrow ones to push certain instruments in or out, working from the bottom up mostly. One trick I like and use in a lot of songs is a tight boost of 1 db or so anywhere from 90 to 130 hz. This is where I hear the upper part of bass drum and I like to emphasize that a bit because it’s the bottom range of what most systems can play clearly. Knowing how bass drums work though, most are based on some sort of fast frequency sweep, so keep in mind that anything you do is going to affect the kick’s frequency as well as it’s attack and release. Fortunately, the subtle boost I like seems to just make the attack of the drum pop out a bit more. That’s my only special trick. Tmain reason this is being EQ’d now is no longer to set up for subsequent processing, but to try and get the song exactly the way I feel it should sound. There’s still a few more modules of Ozone to go through, so if those don’t respond the way I want I will go back and change this EQ to get it right. The best free alternative off the top of my head would be TDR Nova.

Next is the Exciter module, and this has a subtle, but as far as I know unique use by me. I make use of each of the 4 available bands with the Triode character. Tubes are fast (faster than transistors) so you can get a very dynamic sound out of them and not sacrifice the clarity of the music too much because everything stays nice and tight in the time domain. So this ends up behaving more like a very transparent, broad, dynamic EQ the way I use it. I only use the bottom band at all if I feel the subs are sparse enough that they need extra weight (after the rest of the processing, this is not often), and I use the mid band more or less as you’d expect, bringing the drive level to between 1 and 2, leaving the mix maxed out, and seeing some harmonics come in. But for the upper two bands, it’s a different story. For these, I bring in the drive to about .3 at most, and this has the effect of eating up a bunch of energy without actually turning the level down. I don’t know if this is a processing artifact, or if it’s down to the way that this saturation really behaves, but it just takes all the edge right out of the upper frequencies and makes it quite comfortable without any shrillness as we push the mix louder later. I’ve started to realize this is heavy handed just with the Open House release, so I turn the mix down on these bands now to get some of that energy back. I used to use the tape mode for this, but I found that it wasn’t fast enough and turned to tubes instead, though tape can have a similar effect, albeit not as transparent. I’ve already mentioned Airwindows’ free tape emulators, they also have console saturation plugins and distortions on offer. Check Spiral 2 for some of their best work.

After that, things are a bit more flexible, but my default is to go into Ozone’s Imager. Again, I make use of all the bands here and I rarely ever widen things out. More often I narrow songs and try to get them just a bit tighter, more condensed, and less jarring in the move from mono to stereo. This is where I check things like mono response, channel flipping, phase inversion, and such to make sure that everything responds as I’d expect a normal song to. If all goes well, this is the last stage of processing where I have any creative input. Ozone offers the Imager for free, but it’s only a single band. I don’t have experience with any other multiband stereo wideners, free or paid, but you could run separate instances on a few mixer channels and filter the input on each to get a similar result.

The next stage is the Maximizer, and I use IIRC 4, set to Modern, with a character level between 1 and 2 and true peak sampling. I adjust the threshold until I’m hitting the kick reliably, but just a little and look at Youlean Loudness Meter to see where the loudness stands. When I think it’s right, I unloop the loud section and play the song start to finish. I make sure nothing sounds off in the sections I haven’t been working on, make tweaks to the limiter to get the track to around -11 LUFS in Youlean (and other parts of Ozone to make the whole song sound good), and make sure I don’t hear any obvious distortion or changes in tone made by the limiter. Because IIRC 4 is a multiband limiter, it’s cleaner and more transparent than the equivalent broadband limiter in ideal conditions, but it can make significant changes to the audio going into it if the song does something unexpected or if the limiter has to be pushed and take off 7 db of gain. If this is the case, multiple instances of limiter will be in order with eq or dynamic eq in between to compensate for changes the limiter could make. There are lots of great limiters out there to match whatever your tastes, I like loudmax and Limiter No. 6, but look around if you don’t.

Once the loudness is there, I set final headroom of 1.5 db (to leave room for lossy rendering) with loudmax, which also catches any peaks that Ozone let slip through (and is also free). I just leave the threshold at 0 and turn the output down. After that, Airwindows makes another appearance with Not Just Another Dither, which is the best dither in the world IMO. As I understand it, it uses a combination of noise shaping (which most dithers have) and gating (which most don’t) and the noise shape and gating is based on the music you put into it (which nothing else I’m aware of does). The net result is that you get the noise you need to avoid quantization error but you hide that noise in the music itself, so the noise floor isn’t the lowest on paper but it’s inaudible in practice (Airwindows has much more detailed videos talking about this if you want to know how it works, but I’m just very happy it does). Then that goes through Youlean for my monitoring and out of the DAW.
 
Last edited:
Back
Top