Vocoders carry the unfortunate and undeserved reputation as the cheesy robot-voice one-trick pony, but in fact they are versatile spectral processors, woefully underused and misunderstood. In this article, we won’t dive too deep into the countless uses of Ableton’s Vocoder, but we will take a look at how its noise source can be used as a creative layer in drum sound design.
Softer, Brighter Hats
One of the great properties of the vocoder is that it inherently functions as an envelope follower. An envelope follower “follows” the path of the ever-varying input signal amplitude, and then makes that path contour available to control other parameters (in this case the amplitude of our carrier signal). This allows you to easily create a parallel layer where you choose the harmonic/timbral content while the dynamics simply match the input signals dynamics - matching dynamic layers with very little work at all. Hi-hats are a particularly good candidate to layer with noise because they're mostly just rhythmic white noise themselves. I'll add some hi-hats to my set to get to started. For this demonstration, I'm using a loop from my Hi-Hat Pack which features recordings of acoustic hats played by various drummers.
As you can hear in the example (click the play button in the bottom left of the gif), real hi-hats typically have a lot more midrange than electronic samples. I want to remove some of that frequency range but also create a softening effect by layering white noise, something I wouldn’t be able to do with an EQ adjustment alone. The vocoder will impart this softening timbral effect all while perfectly preserving the dynamics and groove of the acoustic drum loop - “playing” the white noise only when the hats are playing. I'll go ahead and add Live's stock vocoder and make sure the carrier (the timbre of the additional layer) is set to "noise". The XY pad below the carrier selector is great for changing the texture of the white noise, but I found only a minor tweak to this was needed in this case as I'm after the traditional white noise sound to blend with my hats.
Now it's just a matter of playing around with the settings on the device while listening at 100% wet. For hats, I'm looking for a brighter sound, so I know I'm going to immediately pull the formants knob up so the vocoder focuses more on the upper frequencies in the noise. The depth control attenuates or exaggerates the amplitude contour applied to the carrier, changing the dynamics from totally flat to just the peaks poking through. I know I'm going to want the vocoded sound to be tight with the original hat loop, so I crank the depth to 140% to better match the peaks of the hats' amplitude envelope. I'll also set the release to something close to the release on the hats I'm working with - a good starting place would be anywhere from 80 ms to 200 ms, but it will vary based on the original signal.
At this point, I'll switch my focus to the heart of the vocoder - the filter bank. Vocoders are essentially a bank of bandpass filters that detect amplitude differences at each band and dynamically copy this behavior to the filter bank affecting the carrier (white noise in our case). A lot of this will be trial and error, but for these hats I liked the sound of fewer bands and also made the filters wider (by increasing BW) and switched them to the grittier "retro" setting. It's also a good idea to play around with the range - this can make a dramatic difference in the unpleasant washy phase sound you might be hearing. Finally, I start drawing over the bands looking for the best balance. For hats, my focus is on attenuating the upper range that can be rather harsh.
Finally, I pull back on the Dry/Wet knob to get a good mix of the original signal back in. The vocoder noise is rather quiet, so something around 75% should do. Now that I'm hearing the final result, I'll usually go back through the settings and tweak them, especially matching the release time so the noise doesn't bleed past the hats.
And there you have it! The vocoded layer is much easier on the ears than the harsh, thick original loop and the softer sound will blend nicely into a mix.
Snares are another good candidate for noise layers because that's essentially the sound snare wire produce. Envelope-shaped white noise even takes the place of a snare in some classic drum machines, so adding a layer with the vocoder will help add some of this electronic feel to any snare sample. I'm starting with a simple loop that has its components isolated on individual tracks.
After placing the vocoder on the snare's track, I start adjusting the dials again. I think the noise could be just a little brighter, so I make a very minor clockwise adjustment to the formant dial. The real focus, however, is on the release, which I'll turn to around 500-600 ms to add a nice decay to the noise.
Now I adjust the filter bank to similar settings to the hi-hats. I've found that the lower and wider bands set to retro work best for the purer white noise sound I've been going for with these percussion examples. For the snares, I'll allow considerably lower frequencies through than I did for the hats to ensure the fundamental is captured by the vocoder.
After pulling back the Dry/Wet to add some of the original sound back in, this is the result. It has the feel of a vintage drum machine and adds some much needed brightness and decay to the snare sample. You can hear the difference in the above audio example as the snare switches between dry and vocoded each bar.
A common problem in a mix is that the kick becomes buried and muffled and indistinct. A kick is generally built of two parts - the bass fundamental and the click - and these issues are often the result of the click not doing its job very well. One could add more of click by manually adding a sample to each bass drum, but since we'll often be looking for a white noise-esque sound anyway, the vocoder will be a great tool for adding this automatically. Here's the loop I'm using, with a purposefully muffled and muddy kick.
If you try loading a vocoder on a drum like this, you'll notice it sounds pretty awful. This is because the heavy bass doesn't translate well to the white noise and by default the release time is quite long. I know that I want this synthesized click sound to be very tight, so I'll bring the release down to just 50 ms and raise the depth quite high as well right away. Then I'll make the usual small adjustment to formant and modify the noise source to be a little less bright.
The biggest strides are made in the filter bank, as usual. I'll set to fewer bands again, but the critical step is in adjusting the frequency range. The lower frequency should be set around 300-350hz so the bass is ignored by the vocoder. Because there is so little information in the mids and highs of a bass drum - and particularly in this muffled one - the vocoded click is quite quiet, so I boost it using the gain control as well. From there, it's the usual balancing by drawing across the filter bands.
The audio example shows the difference after the synthesized click has been added with the vocoder, as it switches between dry and wet each bar. The kick itself becomes more defined, and you can tell that this will cut through a dense mix better as well.