Tech Notes

Signal Flow

Overview

Trying to manipulate a vocal recording's significant level changes with a single processing stage leads to frustration: settings that sound great in some sections sound too subtle or extreme in others. So it's usually necessary to use multiple processing stages, with each stage handling part of the overall processing.

Voca's processing stages complement each other, each stage improving on the previous stage's output and improving the following stage's predictability, making dialling in a controlled and consistent vocal sound quick and easy.

Soften

Why Soften processing is applied towards the beginning despite appearing to be the final stage in the plugin interface:

Applying compression and saturation to a voice tends to level out the relative balance of consonants, sibilance, and vowels/body such that even a smooth sounding vocal can become slightly harsh or too sibilant. So it's helpful to think of Soften as a final polishing stage, a finishing touch dialled in after compression and saturation.

However, suppose a voice does have very apparent harshness and sibilance. In that case, it's beneficial for those characteristics to be smoothed out first, preventing compression from digging in too hard due only to the vocal's tone.

So, softening the spectrum of the voice before compression results in a smoother sound overall. Yet it's often easier to hear the right amount of softening after compression has evened out the voice's dynamics. Showing the Soften control on the right keeps you moving forward and reduces the need to jump back and forth between controls.

Optimise

In addition to changing the target level that Auto Input pulls your vocals towards, clicking Optimise adjusts the internal threshold of both compressors and the headroom of the Saturation stage.

This means that Optimising doesn't drastically change the amount of compression and saturation you may have already dialled in, and you're able to get the same amount of stability and focus even on vocals that were recorded too quietly or loudly - but without Voca unbalancing your mix by changing the overall volume for you.

The result of Optimising is that you'll get the most benefit from Auto Input without needing to think about whether or how much to adjust a target, input, or output level control.

What is a compressor?

Compression is a powerful tool for manipulating a sound's movement, groove, and character in a mix. Compression can shape its overall character and control its stability in front-back space by rapidly changing a sound's level over time, hence the relationship between transients and sustain.

Note

Compressors are commonly associated with reducing dynamic range, but, as we'll see, this is often not the case.

Typical compressor controls and how they work

Threshold: Set the level at which the compressor will begin changing the signal. The compressor reduces the output level as the input signal level rises above the threshold.
Ratio: Set how strongly the compressor reduces the output level.
Attack: Set how quickly the compressor reduces the output level while the signal level is increasing above the threshold.; The attack time is never instant since this would cause distortion. As illustrated above, this means that compression often exaggerates a signal's transients by letting the level continue to rise (or overshoot) during the attack phase before attenuating the sustain of the signal.; That is why it is inaccurate to say that compression reduces dynamic range. More often, it increases dynamic range by exaggerating transients relative to sustain.; If the attack is very fast, the signal can be delayed relative to the compressor's level change to prevent such overshoots and achieve more consistent peak level control. This delay is called Lookahead.
Release: Set how quickly the compressor adjusts the output level while the signal level is decreasing back down towards the threshold.

Auto Input vs Compression

Both Auto Input and Compression change the level of a signal over time. There are two key differences between them:

Auto Input changes the level slowly and without changing the character of the sound. Compression changes the level much more quickly and is designed to change the sound's character.
Auto input applies level attenuations and boosts to bring the signal closer to a target level. Compression applies only level attenuation and only when the level is higher than a certain threshold.

Note

Here we are describing "downwards compression". "Upwards compression" is also possible, though it is much less common. Auto Input can be considered a combination of upward and downward compression, though this isn't quite how it works internally. See our Oxford Dynamic EQ user guide for a quick introduction to upwards vs downwards compression.

Compression behaviour

Instead of re-using our Oxford Dynamics compressor and limiter or analogue-modelling any old classic compressors, we designed two new compressors. They take some inspiration from various FET and Opto-style compressors often used on vocals, including 1176-like and LA-2A-like units, the Elysia Alpha, and the Tube-Tech CL 1B.

Neither compressor is intended to behave exactly like any of those units. Voca is our twist on the series combination of a fast FET-style compressor and a slower Opto-style compressor, taking advantage of what's possible in the digital realm rather than being shackled to an exact emulation of old hardware.

Two compressors in series

Why compress twice? Isn't overcompressing bad? Why fast before slow?

Recall this paragraph from the beginning of the page:

Trying to manipulate a vocal recording's significant level changes with a single processing stage leads to frustration: settings that sound great in some sections sound too subtle or extreme in others. So it's usually necessary to use multiple processing stages, with each stage handling part of the overall processing.

For instance, when using a slow compressor like the LA-2A, a large amount of compression often sounds excellent, achieving the stability your vocals need. However, sudden level surges from consonants or high-pitched notes can cause a significant increase in gain reduction that takes a while to release. This can result in a temporary drop in volume as the input level drops before the compressor has released back to unity gain, defeating the purpose of using it to make the vocals sound more stable.

Suppose the slow compressor's release time is reduced to prevent this. In that case, background noise and breaths between words will be made louder as the compressor releases, adding an aggressive pumping movement which is often undesirable.

If, instead, sudden level surges are caught by a fast compressor stage first, whose release time is short enough to prevent an audible volume drop, applying the slow compressor next gives more consistent and predictable results.

When set up appropriately, multiple compression stages can give more stable vocal volume without creating aggressive pumping, even with substantial amounts of gain reduction.

Stabilise and Squish

Stabilise adjusts both compressor thresholds in the same direction: higher Stabilise settings mean lower thresholds and more compression.

Squish adjusts both compressor thresholds in opposite directions.

Low Squish settings result in a low slow threshold and high fast threshold, giving predominantly slow compression.
High Squish settings result in a high slow threshold and low fast threshold, giving predominantly fast compression.

Fast compressor

The attack time is less than 0.1ms, with lookahead (Recording Mode disabled) for consistently near-instant peak control.

The release time depends on how quickly the input signal level changes: faster when the level is changing quickly and slower when the level is changing slowly. The release time also tends to be longer after more extended periods of gain reduction, helping to prevent the compressor's fast gain reduction envelope from directly following the waveform when the input signal has significant low-frequency energy, preventing the compressor from adding distortion.

The ratio reduces slightly as gain reduction increases. This helps to smooth out sharp transients while avoiding over-compression of sustained signal levels or short-level surges that would cause vocals to get lost temporarily in the mix.

Slow compressor

The attack and release times depend on how quickly the input signal level changes: faster when the level is changing quickly and slower when the level is changing slowly. The release time also tends to be longer after more extended periods of gain reduction.

Like the fast compressor stage, the ratio reduces as gain reduction increases.

Notes about both compressors

We would need many more plots to visualise the compression behaviour

The attack, release, and ratio behaviour of both compression stages depends on the input signal's movement and the overall gain reduction. The plots above capture only a single snapshot of this behaviour for one signal and setting.

Considering the two compressors as opposites is tempting due to how the attack and ratio behaviour appear in the above plots. However, the differences between the compressors are more complex when using musical signals.

The best way to explore the behaviour of the compressors is to listen to various Stabilise and Squish settings on a wide range of audio, including vocals, percussion, and bass instruments with significant low-frequency energy and high-frequency transients.

Compression and filtering the level detection path

Both compressors filter the input signal before detecting its level so that the onset of compression is balanced across the spectrum perceptually. Without such filtering, compressors tend to be oversensitive to low-frequency signal components and sound overcompressed and unnaturally 'pumpy' before sufficiently controlling peak level from high-frequency components.

Compressors often have a simple high-pass filter in their level detection path, which defaults to disabled, to compensate for this. This means the compression is not well balanced by default, and if the filter's cutoff frequency is increased too high, low-frequency components don't trigger compression at all!

Voca's level detection filtering has been designed more carefully so that compression sounds balanced without needing adjustment.

Benefits of low-distortion compression

From the fast compressor section above:

The release time depends on how quickly the input signal level changes: faster when the level changes quickly or slower when the level changes slowly. The release time also tends to be longer after more extended periods of gain reduction, helping to prevent the compressor's fast gain reduction envelope from directly following the waveform when the input signal has significant low-frequency energy, preventing the compressor from adding distortion.

Here is the harmonic distortion added by Voca's fast compressor (foreground) compared with a popular 1176 emulation (background), for a slowly level-modulated 50Hz pure tone, with both compressors applying 10 dB of gain reduction with similar attack and release settings.

With a pure tone input, the resulting compressed signal sounds like it's been saturated. With a music signal, such as a deep voice or a bass guitar, the result is a less desirable buzzy distortion (which is more challenging to show in a static frequency analyser).

This distortion on low-frequency input is the norm with fast compressors. If you don't want such distortion, you either have to use a compressor that gives you control of the attack and release times and know how to use them, or you need to switch between different compressors for different sounds.

Since Voca's fast compressor doesn't add such distortion (except in Recording Mode), you can use it on any signal with the same clean results. And then dial in the amount of Saturation you want without having to modify your desired compressor settings.

Recording Mode, latency, aliasing

When Recording Mode is enabled

The fast compressor's lookahead is disabled, attack time increases slightly, and the release behaviour changes slightly, tending to be slightly slower.
The Saturation section's oversampling is disabled.

This results in zero input-output latency, slight distortion from the fast compressor with low-frequency input, and aliasing from the Saturation section. Overall Voca sounds very similar, though a little less clean.

This trade-off is worthwhile when recording vocals through Voca and unnecessary when mixing.

Aliasing is a kind of inharmonic distortion that occurs when frequency components are created which are higher than half the sampling frequency. Aliasing can sound like atonal bell-like distortion with test signals and very simple tonal signals. With most audio signals with more complex spectra, aliasing sounds more like modulated noise. It's the same phenomenon that can cause wheels to appear to freeze or rotate backwards when filmed and is responsible for the effect shown in this video.

The Wikipedia page about aliasing is a great place to get more detail.