I completed a song mix-down a few weeks ago, and it was slightly out of the norm. For one thing, R&B Pop is not my typical genre, though the instrumentation and general formula are very similar to the Electronic Dance I’m usually generating. The real challenge here was the very quick deadline. Who can resist a good challenge? In hindsight, it was an interesting episode of setting priorities and calling on experience. Here’s how I succeeded and where I may have failed.
“I’m A Wreck” was produced by JP “The Beat Digga” Nebres (Pinay, DnH) and performer Sway Peñala (American Idol top 24, season 5) for Sway’s debut album, My Story (Amazon, iTunes), released on 18 May 2010. The instrumentation, the recording, and the musical arrangements were done. My job was to take all of those pieces and make them sound like a song.
We are all using Apple’s Logic Pro now, so it was a fairly simple process of making sure all of the song’s assets were collected in its session folder, zipping it up, and shooting it across the Internet. I’ve been an evangelist for Dropbox with everyone I know, so transmitting the 250 MB session was an easy drag and drop.
I fired up Logic and took my first look at the song. Left to right, it was a typical 4 minutes or so. Top to bottom, there were nearly 40 tracks of pitched instruments, drums and vocals. Things had already been massaged into a “working” mix. It’s nearly impossible to work on the production of a song without working towards a mix at the same time. So I had the basic idea to work from. I listened to that a good 5 or 6 times to get the feel of the song. In most Pop music, the vocal is the top priority. It’s what most people relate to, it typically carries the most memorable melody or “hook” and it’s probably why the song exists. But I grew up on good electronic music, so I love my drums and bass as well. Hello killer bass line in today’s song, we’re going to have some fun.
Since we were all using the same basic software, we didn’t have to print the musical instruments as audio tracks before moving the song around. So a lot of the original software instrument and effects plug-ins were still up and running, ready to be tweaked. (Logic is the foundation, but everyone’s rig is further customized with a large list of add-ons. I’m currently running around 150 plug-ins.) This is excellent, because it can convey how the producer wanted that specific sound or vocal to be presented. I can also trace back to the raw source if there’s a problem or if I think I can achieve the same result in a better way. The first problem arose, however, when I discovered two tracks using plug-ins that I didn’t own. One was not being used anymore, fortunately, but the other was playing a sampled flute riff throughout most of the song. Dropbox to the rescue again. While JP and I were on the phone, he bounced that single track down to audio and dropped it in our shared folder. This wayward track had been recorded with the same starting point as the rest of the song, so within seconds I was able to drag it into the song, and boom, there it was, without a worry about timing or tempo. The 21st century is good.
Now I had a complete mix sitting in front of me. What first? I like to start with the fun stuff. Drums! But it’s not easy poking around 40 tracks of simultaneous sounds that you hadn’t seen until an hour ago. Step one is organization. The tracks were already grouped in a semi-logical way. Pitched instruments like the bass synthesizer, guitars, piano and that flute were in the top lanes. The vocals followed, with two lead tracks and twelve tracks of backing harmonies in various spots. The drums were having a party at the bottom of the page. They were mostly samples playing from Native Instruments Battery 3 and Logic’s EXS24, with a few Ultrabeat sounds salting things up.
Totally aside: My personal color coding mantra starts with cyan on drums, mostly in tribute to my usage of Battery for many years. If something stands out or is being played by an unusual instrument, I’ll make it stand out visually as well, hence the bright red for the Ultrabeat track. I use bright blue for the vocals, a little darker on the lead tracks. This is probably because I started using Logic in 2001 when working with Sharyn Maceren, and blue was her color at the time. Musical instruments get the rest of the palette, with pastels and browns going to vintage styles, bright colors on synths, and usually a solid black, white or gray for the bassline(s). The goal should be whatever feels right and gives you the quickest clues without thinking about it again. Alternately, if a rainbow running down your screen is going to ruin your mood, forget everything I just said.
Drum recon revealed no less than three layered kicks, two snares, a clap which had to interact with the snares, two different hi-hat cymbal tracks, and various percussion bits and transition effects. I often start with the kick. It moves speakers, it’s fun, what else can I say? Layering multiple percussion like a kick drum can be tricky. You can’t just slap them all together and expect them to agree. If they all share a similar peak bass frequency, you’ll now overload on that and lose definition in other frequencies because you have to drop the overall level. Step one is figuring out what each layered sound is going to contribute. One kick supplies the bass boom around 60 Hz (but may still need a notch lowered at around 200 Hz to soften it up). One kick will supply the punch and character anywhere from 750 Hz up. Logic’s built-in Channel EQ actually sounds very good now, and its UI easily lets you get very specific. If for any reason it’s not sounding good, my old flame, Waves Renaissance Equalizer is often my next choice. TriTone Digital HydraTone is my favorite EQ for audio quality, but it doesn’t do precision well. Any of the sounds may be improved with a little intentional distortion or compression. That’s where my favorite medicine for kick drums comes in: CamelPhat (or it’s free cousin CamelCrusher).
Chances are, layered drums will have tiny timing differences as well. When combined, they may obviously flam if their attacks are too far apart, or may cause low-mid frequency boosts or drops if they are close but not perfect. To fix this, you can adjust the timing on the tracks overall. Since we still had the plug-ins up and running, I dove into Battery 3 where I could see the waveform display of each sound for a little extra assistance. The final task to handle here was that one of the three stacked kicks was used by itself in certain breakdown sections. So it had to work well solo while also playing nicely with others. I spent a good hour getting this all to happen with a combination of careful timing adjustments and EQ. The final output of all three tracks was combined on a common output (bussed to an Aux channel), where I applied a final dose of CamelPhat flavor.
The snares all had different jobs to perform as well. “SnareKap” was an old school sound, fairly harsh. This was best used for its character, so I used CamelCrusher to compress and lengthen the decay with a bit of Audio Damage Fluid to give it a great stereo image, which also tends to reduce harshness. The second snare was beefy, with a clap leading into the beginning. I rolled some bass off because it wasn’t necessary and lopped some high-end to leave room for the first snare. Because the clap sound preceded the snare hit, this obviously brought up timing worries again, but the tracks were sequenced accounting for timing differences and not hard quantized. Very minor adjustments were necessary, if any. On the third channel, the lone clap sample was solid, but boring. So I dropped some compression on it with a delayed-reaction attack so that it clamped down late and molded a “pop” into the beginning to give it some aggression. Logic’s new Compressor, which I believe first appeared in version 8, is excellent here. Now it sounded snappy and fit well in the stack, but right before each chorus and during a break in verse 2, the clap is front and center by itself. So it also got a soak in Fluid to be a little more interesting stereophonically.
The rest of the drums fit into place pretty easily once the kick and snare were set. The primary hi-hat was a TR-808 sample panned left. Since the real thing is synthesized and not a sample of an actual cymbal, I always take the high-end on 808 ‘hats very seriously. Samples rarely reproduce them well. Here’s where TriTone Digital HydraTone shines. Its “Avln 37” mode is awesome on treble, boosting it in the most pleasant way and, if necessary, it also seems to magically repair bad stuff. Again, the UI is not very precise for surgical control, but for dipping a wide range of frequencies through the magic, it’s fine. The other ‘hat was more of a tik-tok midrange time-keeping sound. I also sent it through HydraTone to clean it up, but not so much to make the treble louder.
Next up was the bass line. Because it needs to interact with the low-end in the kick drum, frequencies are even more important to worry about. The bass line and the kick must sound good on their own, but they also need to sound good if they overlap. With our style of music, this often means that the kick is king and gets to delve into 60 Hz and below. It is usually possible to high-pass filter or low-shelf a bass line at around 180 Hz without any obvious loss of “bass” feel. In this case, the original sound didn’t have much in this range anyway, and the producers had boosted from 88 Hz down. I simply backed off on this a little to make room for the kick. Another EQ boost was happening at 1800 Hz for the nose of the bass to come through the mix.”https://www.glenngutierrez.com/wp-content/uploads/2010/05/Sway-ImAWreck-BassClip.mp3″
The other instruments were a simple matter of cleaning up each sound one by one. I rely on experience here to give me an idea of where each piece will fit into a mix, frequency and image wise. Otherwise, you’ll just spend more time going back and forth to see what works. The only sound which went counter-intuitive here was the grand piano. A sound like that might normally sparkle with clarity, but since many other sounds had that job, including the lead vocal in overlapping spots, this piano ended up living in a more midrange frequency range. However, the piano plays one of the primary melodies during the chorus between vocal lines, so it had to be easily audible as well. A small boost in the upper midrange achieved that. Hopefully, you didn’t notice any of this until I brought it up. I also like to roll off the low-end on most instruments which don’t need it, to reserve that space for the kick and bass line. That brings us to the vocals!”https://www.glenngutierrez.com/wp-content/uploads/2010/05/Sway-ImAWreck-InstClip.mp3″
The background vocals were already nicely labeled, imaged and level matched. Few adjustments were necessary. I slapped Logic’s simple pitch correction plug-in on these channels not to fix any glaring mistakes—Sway is an excellent vocalist, as Simon Cowell and company will tell you—but to keep the held harmony notes very precisely matched. This prevents audio beats from being generated, causing an almost imperceptible but very unpleasant midrange “acoustic roughness” which is harder to control and mix. Similar to the musical instruments, I also roll the bass off of all vocals as high up as possible without affecting the beefiness of the singer. This can be anywhere from 200 Hz to 800 Hz, depending on the vocalist and the range of each vocal track.
Some de-essing was necessary on the background vocals. When they are singing the same words as the lead vocal, at the same time, sibilance (S and T sounds) can overlap and become unruly. Psychologically, the backgrounds don’t need that sound at all and the listener won’t notice. But you don’t want to just lop off the high-end because the “breath” and “air” is also part of the magic of mixed harmonies. If there are some egregiously bad areas, you can go into the audio file itself and manually reduce loud sibilance. It can be surprisingly easy to spot, isolate and gain-change by -5 db or so. On this particular song, there was only one of these on the lead vocal. I can’t even remember where it was now, which is exactly what you want. Ideally, though, you’d want a device which listens for S and T sounds, reduces the volume of just those sounds quickly, then leaves everything else alone. This can be done by side-chaining a specially EQ’d (to enhance those frequencies) duplicate channel of the vocal into a compressor. It will then over-react to just those sounds, often in the 7-9 KHz range, and duck the volume of the original track when they occur. Luckily, smart people long ago combined all of this into one device. In software, you can also do it all manually or call up a dedicated de-esser plug-in. The problem here is that many just don’t work very well, and that includes Logic’s built-in De-Esser. One of my favorite alternatives was the Waves Renaissance De-Esser, but it wasn’t included in the Waves bundle I ended up with on my current rig. Luckily, I stumbled on an interesting substitute with the KeyToSound Dynamic EQ. It’s a 4-band parametric EQ with compressor-like controls on each band, and a vintage-style smooth quality. I basically use one of the four bands, set it to react to the offending frequency, and it sounds great. The bad news? KeyToSound became Koblo, and they went out of business soon after. So no updates and no fresh installs will be possible, but I’ll keep using it until I cannot (still working fine in OS X 10.6.3). Minor thing, de-essing, funny how finding a good solution is such a big deal.
The final step to a nice vocal is compression. We aren’t trying to smash the voice and end up sounding like a radio jock, but a tastefully done control of dynamics makes everything sit just right, keeping the vocal in front without blasting the mix apart. It also makes sure nothing ever dips too low to be intelligible. My weapon of choice here is Waves Renaissance Vox. It is a specialized compressor designed just for vocals. Actually, it’s probably a series of more than one gentle compressors, but that’s their secret sauce. Renaissance Vox goes on the background vocals after ganging them together, and on the combined pair of lead vocals. (I recently tested Voxengo’s Voxformer as an alternative to Ren Vox and it does a very nice job as well, with more control. It even has de-essing!)
You’ll notice some edited vocal tracks at the end of the song. The first is a little snippet which drops down from the lead on track 19 into “woo” on track 20. This was to give Sway’s ad-lib yell a little extra effect action. I could have automated some effects on the original track to come in just at that spot, but I resist using automation if I don’t have to because it has a way of locking down options and it can get messy if you automate several things on the same channel. Automation was also important when track and channel counts were finite. They aren’t anymore. In this case, it was very easy to snip out one word, copy it to a fresh channel, and customize what happens to it with a 100% wet effects mix. This leaves the original intact and consistent, only the effects are heard from the track with the copied word, and it’s easy to see what I did when I load this song up a year later.
The backgrounds which appear to be muted after bar 86 were actually bounced down to a separate sub-mix because they were competing for dynamic space in the backgrounds bus. Since the lead vocal was vamping out by this time, we were left with two powers fighting for the space of one, and the result was lower volume of both because of automatic compression. Again, automating a level increase or compression threshold change might have fixed it, but each vocal bit was eating into the other, so one was simply moved outside the box. Now you can hear the very cool “I’m a wreck, I’m a wreck” part clearly, alongside the “I’m crazy, I’m crazy, I’m going insane” at lower level with a wider image, and neither is competing for level.
While all of the math is going on above, the flow of the song is also in your hands. Part A leads to part B. Sound X counters the other. Between vocal lines in the chorus, the kick does that “pah-pah.. pah” which slaps the rhythm in your face, but you love it. This bar pushes while that bar pulls. The bass is a little hot in verse 2 but it’s awesome and you can still hear the vocal.. since you turned the bass down 3 db. Does the build-up actually build up sonically as well as musically? Does the outro actually wrap up with that “sense” that things are coming to an end? These are all part of the emotion of a mix as well. After I sent the final mix back to JP, he wasn’t feeling the long intro any more, so I whipped up four alternates. The last one was the anti-intro, with a cold start at the first verse vocal and minimal percussion (much easier to do now in the days of digital). That’s the mix on the album now.”https://www.glenngutierrez.com/wp-content/uploads/2010/05/Sway-ImAWreck-30Clip.mp3″
And that’s basically a mix which was done in a day. Am I happy with the results? Very much so. Is it perfect? Hardly. “You never finish art, it is only abandoned.” (Leonardo DaVinci, not George Lucas.) First of all, it’s a bad idea to complete any mix in one session. As you listen to your song over and over, especially with the mix out of whack, you can burn your ears. Most people notice this when they come home from a rock concert. Hopefully, you aren’t mixing at that volume the entire time, but you’re also going to be spending a lot more than 2 hours doing it. A good night’s sleep will usually get you back on track. I was being extremely careful to avoid the burn while doing this mix, since I knew the extra day was not an option. I still could have used the extra time to test the mix on more systems, and maybe back off a little more bottom end from the bass line which chugs a bit on some speakers now that it’s been mastered. But it does lend a certain beefy feel to the song. I also have a self-serving theory that every mix needs a little something slightly “wrong” to add character. I was sure that the mid-range open hi-hat in “Do You Miss Me” (the Dreamhouse/Radio Mix) was way too loud after it went to manufacturing, but it became a unique identifier for that song.
I also did the artwork for the album release. But that’s a whole other story.