Urban music is often a springboard for new producers and engineers, and 21 Savage's chart-topping album was no exception.
Atlanta trap is problematic for those who believe(d) that hit songs necessarily involve a verse-chorus structure and a catchy hook. Many trap hits lack any discernible song structure or hook, instead getting by on vibe and the flow of the rapping. Melodic and harmonic ingredients tend to be pared down to a couple of keyboard parts and/or a sample, with the 808 replacing the bass, hard-hitting kicks, a fairly thin-sounding snare and prominent hi-hats. As a result there's often very little happening in the mid-range.
The trend has also created opportunities for new producers and mix engineers, and as a result, the Inside Track series has recently featured more and more hitherto unknown mixers. 21 Savage's I Am > I Was ticks all the aforementioned boxes, from the musical style to the fact that 12 of the 16 tracks on the deluxe version of the album were mixed by a relatively unknown mixer. I Am > I Was also features the usual huge posse of producers, from the well-known (Boi-1da, Metro Boomin, and the seemingly ubiquitous Louis Bell) to the obscure, plus a parade of all-star guest artists including J Cole, Childish Gambino, Post Malone, Travis Scott and more.
The full name of the unknown mixer in question is Chanbuny Maddox Chhim, aka Maddmix, who works from a studio in his house in Atlanta. "My parents were from Cambodia, and lived there during the time Pol Pot was in power," he explains. "They were fortunate to have survived and to escape the Khmer Rouge regime. They had sponsors from Richmond, Virginia, so they settled there for a bit, and then made their way to South Carolina, which is where I was born and raised, and they still live. Nobody in the family was musically inclined, but I played trumpet and French horn as a teenager, and then tried to make beats, using Fruity Loops, and recording into Cool Edit Pro and Sonar. Local rappers came to me to record their songs, and I noticed that I was better at the more technical side of making music. So in 2010, when I was 20, I went to Full Sail University in Orlando, Florida, where I spent two years doing the Recording Arts programme.
"After finishing Full Sail I moved to Los Angeles, and tried to find an internship. It was really hard to get into a studio, but just as I applied at Larrabee Studios they needed a new runner. So for one and a half years I was getting food, coffee, cleaning toilets, all the fun stuff you do as an intern! Then Jaycen Joshua's main assistant left, and I got a chance to work closely with Jaycen, which I did for three and a half years. I left Larrabee in the summer of 2017 to become freelance and worked for a year out of Fab Factory Recording Studio. Because I was working remotely more and more, I decided towards the end of 2018 to move to Atlanta, because it's closer to my family."
Atlanta is the urban music centre of the world, and as the new kid on the block of a sizeable music scene, Maddox could easily have struggled after arriving. Luckily, though, his previous experience had given him a head start. "Because I had been handling Jaycen's professional emails and client relationships, I had kept close relationships with his clients, especially managers and A&R people. After arriving in Atlanta, Ericka Coulter of Epic contacted me, and said there was a project that urgently needed to be mixed. I mixed one song, and they loved it, so I went to the studio where 21 was working to meet him and his team. Everything was cool, and I offered to work in the studio there as well, but they said: 'The mixes you did at home were pretty good!' So I mixed the rest of the project in my studio over two weeks, late November to early December."
Maddox worked on the mixes alone, while keeping in close contact with Epic's A&R people, Jennifer Goicoechea and Ericka Coulter, and 21's engineer, Mac Attkisson. "Mac recorded 21 at a studio in Atlanta called Tree Sound, and he put all the files together, and gave them to me on a hard drive, so there was no chance of the tracks leaking, something which 21 was very concerned about. Mac would hand me the vocals with the instrumentals, the latter as individual track outs. After that, all I really had to do was organise my workflow, and start mixing. I did about a mix a day, and would get feedback from Jennifer and Ericka, and they would play the tracks for 21 to get his notes, which were also sent to me.
"I usually prepped the session the day before I mixed it, and the actual mixing would then take me four to six hours, depending on the song. Prepping and organisation were very important. I wanted the sessions organised to suit my workflow, to be efficient and be able to knock out these mixes as fast as possible. Nowadays people want things really fast! That means organisation is a key thing in general.
"The first thing I do when I open the session is to go through all the tracks and listen to see what's what, and I will then organise the session according to how I like it, which is the vocals at the top, then the drums and then the music. Once I have done that I'll import my mixing template, which includes aux effect and group tracks, and master tracks. These go above the vocals, so I end up with my mix bus and master tracks at the top. I'll also do the colour-coding and sending audio tracks to aux subgroups and so on.
"I usually get a rough mix, or reference track, from the producer or artist, and when I open the session I will A/B that against the reference mix, to get an idea of what they want, but also to make sure there's nothing missing in the session, and to compare levels. Once I've done all that, I start mixing, and then it's a matter of beating the reference. That's my job. Of course I take on board what I'm asked to do, and with 21 there were remarks like, 'We like it like this, so keep it like this,' but also, 'Do your thing and make it better!' That was a fine line to walk. These guys will usually have listened to the demo for months, so I tend to keep the production — ie. the music and drums — as close to the reference as possible."
"The main foundation of hip-hop is drums, so with every mix I did for 21, I made sure that the drums are hitting hard and sounding big and full, and then everything else found a place around that. This meant that I started with mixing the drums, got them to a good place, and after that I brought in the music, and finally the vocals. If at the end of that process anything needed more massaging, I would go over the entire mix again and tweak things here and there. Doing it in this order didn't normally lead to problems, because these tracks usually don't have a lot of mid-range, which makes it easy to mix in the vocals. In fact, they're pretty sparse in general, giving me a lot of room to do what was needed.
"Getting the drums and the 808 in particular to sit right is a matter of gain staging. I use the same approach for every mix that I do, with pretty much the same levels in every mix — I'm talking drum levels, 808 levels, kick levels, music levels, vocal levels — and that translates pretty well across all mix sessions. To get to those levels I sometimes have to clip-gain tracks up or down. Many of the producers work in Fruity Loops and put a limiter on everything, and their tracks come in at zero, so are really loud. If you start mixing from that point, you'll be overloading the mix bus, so that requires a lot of clip gaining down.
"My gain staging process is kind of hard to explain, as I learned it from observing Jaycen, but I usually have auxes called 'All Vocals', 'All Music' and 'All Drums', and I look at those and watch the levels and make sure they're not overloading. My 'All Drums' usually is the loudest, and is almost hitting red in Pro Tools. My 'All Music' and 'All Vocals' aux tracks usually are halfway up the meter. The meters in Pro Tools 12 and 2018 are a bit different, I think there's more headroom in those newer versions. But if the levels are too high, I pull the relevant tracks down with clip gain. I then check the mix bus and make sure it's all good.
"The biggest issue is the 808, because that usually is what overloads the mix bus. When there's too much low-end information taking up all the space, there's not enough headroom for music and vocals. So I tend to start mixing the drums with setting the level of the 808 and the kick, and once you have that right, the rest of the mix is really easy. I often duplicate the 808 track, and I smash the copy with a compressor and/or distort it, to get some harmonics, like around 100Hz, so you can also hear the 808 on non-hi-fi systems. I'll then blend the duplicate track in with the original track."
"The heart of my system is a MacBook Pro with Pro Tools," says Chanbuny Maddox Chhim, "which is plugged into my Apogee Symphony I/O, and from there the signal goes to my SSL Sigma SuperAnalogue mixer, and then to my Grace m905 monitor controller, which feeds my monitors, the ATC SCM25As and Yamaha NS10s. I also have the KRK VXT4s, which I don't use very much. My SSL Sigma gives me that analogue sound. Running the signal through analogue circuitry makes a subtle difference. It adds a little bit of width and openness to my mixes, and my low end is definitely more round and full. When I was at Larrabee, I was around both working on consoles and working off your laptop, and if you can find a plug-in, or in my case some outboard, that can recreate what an SSL is doing, you can mix off a laptop.
"I know the ATCs from working with Jaycen, who was listening to them the entire time I was with him. When I was at Fab Factory I was on the ATCs all the time, because I was in a treated room, which sounded great, plus I could play as loud as I wanted! However, although I have treated my room in Atlanta with some acoustic panels, it still does not sound too great, and I have not quite gotten used to it yet. So when I am in here I start mixing on my headphones, the open-backed Sennheiser HD 800S, and then I'll flick to the ATCs, and then the NS10s, and then I go back to the headphones.
"I spend probably 60 percent of my time mixing on my headphones, mainly because I can judge the low end better on them. I dial that in with the headphones on. The thing with low end is that many people don't have access to hi-fi speakers, which means that I also have to reference on things that people actually listen on. So in addition to my headphones, I'll also listen to really crappy ear pods, and I do a lot of referencing on my laptop speakers. If I can hear some bass on the laptop and the ear pods I know I'm in the right ballpark! But at the moment I am trying to decide whether I want to work in an outside studio again, or build a studio in my house."
- Written by Shayaa Joseph, Jermaine Cole, Dacoury Natche, Anthony White & Shelia Young.
- Produced by DJ Dahi & J White.
Maddox illustrates his mix approach with a detailed look at 'A Lot', the lead track and lead single from the album. The song samples the 1971 cover, by the band East Of Underground, of the 1970 song 'I Love You For All Seasons', originally performed by the Fuzz and written by Sheila Young. The rest of the music of 'A Lot' was co-written and produced by DJ Dhali and J White. Surprisingly, the mix session for the sparse-sounding track contains 78 tracks, but the first 39 of these are Maddox's template tracks, including effects and subgroup auxes and mix print tracks. Track 40, 'Lead All', is a master aux track to which all vocals in the track are sent, and 41, 'MadVoxComp', is its double, with parallel compression and limiting. Below that are six 21 Savage vocal tracks (42-47, purple), three J Cole vocal tracks (48-50, dark blue), four vocal ad lib tracks (51-54, pink), 12 drum and percussion tracks (55-66), five vocal sample tracks (67-71) and seven synth tracks (72-78). Each of the groups mentioned here feeds its own aux track, and in many cases there is an additional Master Fader track controlling the bus level entering these auxes.
Tracks 15-39 are aux effect tracks and Master Faders from Maddox's template, with reverbs, delays and also distortion; of these, only tracks 15-26 are active in the 'A Lot' session. All aux effect tracks go to the 'All EFX M' Master Fader and then to the 'All EFX M' aux (track 14). Tracks 6, 8, 10 and 12 (red) are the main group aux tracks Maddox describes in his account of gain staging, above, each with its accompanying Master Fader; these are all sent to an SSL Sigma summing mixer, the output of which comes back up on 'Mix Bus' (6), and is then printed on track 5. The top five tracks are mix print tracks (blue), with tracks 1-4 consisting of, says Maddox, "rough mixes and references for the arrangements of the J Cole version".
A high resolution Pro Tools Edit window screenshot is available to download as a ZIP file (see right-hand Media sidebar).
Although there are relatively few plug-ins on the drums and hardly any on the music, Maddox uses dozens on the vocals. 21 Savage's five vocal aux tracks, for example, all pass through a signal chain comprising Antares Auto-Tune, UAD 1176LN E, Waves Renaissance Vox, FabFilter Pro-DS, Waves C6, another Pro-DS, FabFilter Pro-Q2 and another C6, with only the latter one or two plug-ins dropping out on a couple of tracks.
All these tracks go the '21 Lead' aux (42), which has nine plug-ins, including compressors, de-essers, and four(!) more instances of Pro-Q2, as well as four sends. Both 21 Savage and J Cole's aux group vocal tracks then go to a vocal aux called 'MadVoxComp', and from there to the 'All Vocals' aux, each of which hosts two plug-ins; so in total, every 21 Savage vocal track passes through 19 to 21 insert plug-ins, as well as being sent to up to seven auxes. The situation with J Cole is even more elaborate, though complicated by the fact that some of the sends came with the vocal session that Maddox received from Cole's engineer.
"I guess my thought process behind my mixing is kind of messy!" laughs Maddox. "The thing is that I will start with adding compression and EQ on a vocal, and then if I hear something else that needs correcting, I will just add another compressor or EQ. I never go back and take plug-ins off or reset them, because I don't want to go back on what I already have, because I might mess it up. So I just keep going forward and adding plug-ins until I get the sound I want. That works for me. Some people like mixing with very few plug-ins, but I will mix with whatever I need to get what I want.
"The plug-ins I have on these inserts I use frequently. I guess you could call them my clean-up plug-ins. The 1176LN worked on 21's vocals in this session. I may also use the UAD Tube-Tech CL 1B or the Waves RCompressor. In this session the 1176 made 21's vocals really full and in your face. Next is the RVox, which is also is a good plug-in for bringing the vocals up front and making them really full. It is very subtle, but it makes a big difference. If I feel like the vocal does not sit up front enough I put that on, compress it a little bit, and it usually fixes the problem for me. The 1176 has a similar function, but I use it more for actual compression and the RVox more for tone and to make it sound bigger.
"Next is the Pro-DS, which as a de-esser obviously helps with the sibilance. Some vocals have more sibilance than others, and sometimes I use just one, but in this case I felt I needed two. I love using multiband compressors, and the C6 is here for some general compression, but there always are some frequencies that need adjusting, but that I don't want to take out with an EQ, because you need those frequencies. When I use a multiband, I find the frequency that bugs me, and set a threshold, so the frequency is only dipped when it needs to be dipped. If you look at my C6 plug-ins, they often address harsh frequencies that a de-esser can't take out. I'll have a really tight notch on the C6, and then just compress that frequency a couple of dB."
Of the '21 Lead' aux, Maddox says: "There are four instances of the Pro-Q2 on the signal chain, which is again an example of me working cumulatively. I now use the Pro-Q3 a lot, which is a dynamic EQ, which can do pretty much the same thing as the C6. Then insert 4 is the McDSP MC404 multiband compressor, which I use pretty often, and then I have the Eiosis Air EQ, the C6, the Kush Clariphonic EQ and the SSL G EQ. I use different EQs for different purposes. The Q2 has unlimited bands, so I use that to search and dip frequencies I don't like. I really like the high end on the Air, which opens up the vocal and makes it sound really good. I also often do some scooping in the lower mid-range with the Air, because it does a really good job of that. I use the Clariphonic mainly for the Clarity knob, which widens the vocal just by having that on. I really like that on vocals and use that in almost every session.
"The sends go to several aux effect tracks with delays and reverbs, but for this project they told me that they wanted the keep the vocals pretty dry, so I didn't use much reverb. Mostly just a small room reverb really subtly in the background to give it some space. The main reverbs that I have in my template are the Slate Digital VerbSuite Classics. They have great emulations of popular reverbs like the Bricasti and so on, that sound really good. I also use the UAD Lexicon 480 a lot. Those are my go-to reverbs. I occasionally use the Waves RVerb and TrueVerb. My main delays are the Soundtoys EchoBoy and the Waves H-Delay. For distortion I use the Soundtoys Decapitator and Devil-Loc, sometimes the Dada Life Sausage Fattener, which is fun, and the distortion pedal in the Waves GTR3 Stomps plug-in.
"The session was originally a 21 Savage track, and they later decided to add J Cole. His engineer sent me J Cole's vocal session with reverbs and delays already set, so I pretty much imported his vocal auxes, and just matched that to the mix that I had already going. However, the inserts are all mine, and similar to what I used on 21. There also are four Pro-Q2s here, plus two instances of the C6 multiband, and the Clariphonic EQ. The 'Slap' is a send to one of my aux effect tracks, just with a 30-millisecond EchoBoy delay, to widen the vocals a bit. I stripped all the other sends down, to match J Cole's vocals with a track that was more on the dry side, and actually disabled all the sends apart from the 'Slap' on the J Cole aux group track, 'Leads'. Finally, both '21 Leads' and 'Leads' go to the 'Lead All'  and parallel 'MadVoxComp'  tracks, and the latter has the Waves CLA-76 and L1, both for more presence and volume."
With regards to the drums, Maddox says: "Track 57 [orange] is the 808, and track 58 the kick, and track 59 the kick duplicate for parallel compression, coming from the Waves SSL Compressor, and it also has the Waves PuigTec EQP-1A. The 808 track has the Waves RCompressor and Pro-Q2. The rest of the drum tracks don't have any plug-ins. As I mentioned, I tend to not do very much to the backing track. The producers do a lot of the EQ and compression; it's part of production nowadays. When they are sending me the files, everything sounds pretty good. I may just EQ something here and there to fit what they do in the mix that I am doing, but there won't be a lot of riding levels, and for panning I also tend to follow the production as it came to me. Incidentally, I mentioned gain staging earlier on, and you can see on these tracks what kind of levels I am after.
"All drums go to the 'C Drums Master [Fader]'  and the 'C Drums' aux , and on the latter I have five plug-ins: the UAD API 2500, Waves NLS Channel, Slate Virtual Mix Rack, UAD Neve Preamp and Plugin Alliance Black Box Analog Design. This signal chain is fairly common on my drum aux parallel. The 2500 compressor just adds some glue. I don't like to compress the drums too much, because otherwise you end up flattening transients. The NLS and Virtual Mix Rack both add harmonics, and most of the heavy lifting is done by the Black Box, with the Saturation knob — it makes the drums sound round and full.
"There are no plug-ins at all on the individual music tracks, which all go to the 'Synth All' aux [and its associated Master Fader], and for some reason I did the processing on the master, using the Waves Doubler, the Brainworx bx_shredspread, and the UAD Tube-Tech CL1B compressor. The Brainworx gets the music to spread a little, to make space for the vocals."
As explained earlier, the 'All Music', 'All Drums' and and 'All Vox Master' subgroups were each routed through Maddox's SSL Sigma and back into Pro Tools on the 'Mix Bus' track. Each of the three aux tracks has a few plug-ins, and the 'Mix Bus' has a fairly involved signal chain.
"'All Vox' has the Avid Lo‑Fi and Waves RCompressor.I take the Sample Size down to 15 [bits] with the Lo‑Fi, which is a trick I learned from Jaycen. It works like a form of compression, and makes the vocals a bit more up-front and louder. The RCompressor is set to a ratio of 2.56 and is there to keep the vocals in place. 'All Drums' has the Acustica Diamond, in a special edition made for Dave Pensado. I felt that the drums needed a little more 100Hz, so I boosted with that plug‑in. I like the way it works on the low end. I also used the Lo‑Fi on here, on the same 15-bit Sample Size setting as with the vocals.
"I use the signal chain on the mix bus in most sessions. It starts with the iZotope Ozone 7, using the EQ, Exciter, Dynamics and the Imager for the top end. I kind of transferred these settings from Ozone 4, and then tweaked them, because the two plug-in versions don't sound the same. After that it's the Clariphonics EQ, Brainworx bx_digital V3 Modus EQ, Stillwell Audio Event Horizon and FabFilter Pro‑L. The Clariphonic EQ opens the overall mix up a little, and the Brainworx has a spreader for more stereo width. Sometimes I automate the Brainworx, for example making the choruses wider if that needs to be more of a moment. Pensado put me onto the Event Horizon. I use the Soft Clip button for gain. Finally, the Pro‑L is there to keep the signal in check. I sent all tracks to Colin Leonard at SING Mastering in Atlanta for mastering."