Avoiding clipping: OASYS PCI gain structure
Contents
Clipping is audible, but meters look OK
Output volume in analog synths
Program gain structure - background information
Meters show
clipping
If the meters show clipping in a Mixer Channel, Send Bus, or Output Bus, check the following.
Fader gain should be 111 or lower
The faders on the Mixer Channels, Send Busses, and Output Busses all include 6dB of gain. This is handy for boosting low-level signals, but can also cause clipping with higher input levels.
So, if you are seeing clipping on the meters, make sure that the fader is set to 111 (0dB) or lower. As a shortcut, double-clicking the fader will set it to 111.
If this doesn't solve the problem, see Clipping in Effects, below.
Clipping is
audible, but meters look OK
If you can hear clipping, but the meters in the mixer don't show any overloads, try the tests under Clipping in Effects, below. If the Effects are not causing the clipping, try reducing the Patch volumes in the Program Edit window, as described below.
Program Edit window Patch volumes
The default Patch volume is 95, which in general provides a sufficient amount of headroom. Depending on the Patch algorithm and the number of voices being played, it may be possible to set the Patch volume higher, or it may be necessary to set the volume lower to avoid clipping.
For more information, see Program gain structure - background information.
Output gain in
analog synths
The analog synth Patches include a volume knob with +12 dB gain, so that you can get your signals as hot as possible. However, if filter resonance is high, settings above 0dB may cause clipping.
Clipping in
Effects
Temporarily bypass Effects
Many Effects can boost volume, and potentially cause clipping. This clipping may be internal to an individual Effect, or masked by lower gain in other Effects later in the chain, so the clipping may or may not show up on the mixer meters.
As a quick check to see if Effects are responsible for clipping, bypass all effects in the channel or bus.To do this, option-click on any of the Effects slot numbers.
If this eliminates the clipping, then re-enable the Effects by option-clicking again. Then, bypass individual Effects one by one until you find the one (or more) causing the clipping.
Finally, once you've identified the Effect(s) causing the clipping, see below for suggestions on avoiding clipping with particular effects types.
All Effects - reduce input level
In addition to the specific tips noted below, the single most effective way of eliminating clipping in any Effect is to turn down the input level.
EQs
Both cutting and boosting will increase the gain within the EQ itself, and can cause internal clipping even if the resulting output level is low. To avoid this, reduce the input level, or reduce the amount of cut or boost.
Compressors, Limiters, and Gates
These effects often include output gain. If they are causing clipping, reduce the output gain.
Resonant filters
High settings of resonance can cause clipping within the Effect. To avoid clipping, reduce the input level, or reduce the amount of resonance.
Effects with feedback, including delays
Feedback can increase gain, causing clipping. To avoid clipping, reduce the input level, or reduce the amount of feedback.
With mono-input effects set to "L+R" input, input gain should be -6dB or lower
Most mono-input effects can get their input from either the left channel only, or a sum of the left and right channels (usually labeled "L+R"). The L+R setting sums two signals together, which means that - without reducing the input level - it's possible to cause clipping.
Setting the Effect's input level to -6dB will ensure that the summed L+R signal doesn't clip at the input. (If the Effect has gain internally, you may need to reduce the input level even more.)
Program gain
structure - background information
Basically, all systems (analog and digital) will have some point of maximum gain, after which they begin to clip. Analog systems may do this gradually, but in digital systems, there's usually no middle ground. The irony is that, right up until that clip point, the louder the signal is, the greater the S/N ratio and thus the better the sound.
So, the trick is to get it as loud as possible before clipping.
In synthesizers, it becomes difficult to realize the optimum gain automatically. This is because we have more than a single signal (say that ten times fast) to worry about; instead, there may be 4, or 8, or 16, or more individual voices, each summed into a single channel. More complex, those voices each have their own varying levels. So, how do we deal with achieving an optimum gain structure?
The ultimate in conservatism would be to make sure that nothing clips, ever. This is pretty easy to achieve; you just assume that each of your voices is playing a full-scale sine wave, and scale the volumes back so that the summed result is just under clipping.
The result of this technique will be that your voices are pretty low-level, in which case S/N ratio suffers and overall output level is low.
Most of the time, though, the voices will be considerably softer than a full-scale sine wave. For instance, they may have a loud transient, and then a softer sustaining tone. In this case, each individual voice can be considerably louder before clipping than is assumed in the conservative gain-scaling scheme above. In the worst case, with a number of full-scale waves sounding at once, clipping *will* be able to occur, but this can be handled at the programming stage, by reducing internal volume levels.
This is the approach used on the OASYS PCI. Levels are set so that, in the general case, sounds won't clip - but kept hot enough to maintain a good output level, with low noise. The Korg sound designers take care of the rest; when they create a sound which causes clipping, they scale back the Patch levels in the Edit Program window.
Note that the optimal Patch levels will vary, depending on both the Patch algorithm being used, and the number of voices being used.
The result is that some human discretion is necessary, but that the overall output level is hotter and S/N is better.
Korg's QC department ensures that factory Programs can play the preset number of voices without any clipping. For your own sounds, you'll have to do your own QC!
This also means that if you increase the number of voices allocated to a Program, you may need to reduce the Patch volume(s) to avoid clipping.
Note that reducing the main volume by adjusting the mixer fader levels will *not* solve this issue, since the clipping is happening before the mixer.