MSpectralDynamics attack speed and distortion

[anton_s]

Hi Vojtech,

I would like to know whether it is possible to specify different attack speeds for different frequency ranges in MSpectralDynamics. When I use multiband compression, I usually try to use a slower attack speed for lower frequency bands, in order to avoid distortion of the waveform.

If it is not possible, could you perhaps recommend another strategy to avoid distortion?


Thanks!

Anton

[MeldaProduction]

It's not really possible, I was thinking about adding it, but it seems like an overkill actually. Basically one reason is that there is no distortion with MSpectralDynamics! It doesn't work like the classic dynamics processors. So, the only reason this would be useful is if you intentionally want a different response for different frequencies. It could probably be useful, but note that the threshold curve can actually do half of the work in that case.

[anton_s]

The MSpectralDynamics manual itself mentions the problem of distortion in numerous places. For example:

Slow modes usually produces more pumping, but less distortion. Conversely fast modes reduce pumping, but cause more distortion. The type of the distortion is different between modes, so you may actually profit from distortion caused by some modes, as the generated higher harmonics may enhance the audio. The default fast mode provides a good compromise between distortion and pumping.

and

If you need a very short attack, the attack stage may be too quick and cause distortions. In limiters this attack is often 0ms, in which case it becomes a clipper.

The manual recommends using a combination of the "Look-ahead" and "Peak hold" parameters to control distortion:

Look-ahead is commonly used in limiters along with very low (often 0ms) attack times to avoid distortion. With 0ms attack time the limiter is immediately following the input and when the level gets above 0dB, it turns it down to 0dB, so the attack stage is effectively being clipped. To avoid distortion produced by this effect, you can increase look-ahead and peak hold to the same value, say 1ms. As a result the attack stage occurs before it actually occurs, so the distortion is still present, but in much lower levels and usually is masked by the forthcoming transient.

- Is it possible to avoid distortion altogether by increasing the values of Look-ahead and Peak hold?
- Is there a minimum value for these parameters that will ensure that even the lowest frequency (20Hz) waveforms are not distorted?

[MeldaProduction]

Yeah well , the thing is, you might noticed the docs are sort of automatically generated and these parameters are just shared with MDynamics and other processors. Anyway there could be a sort of distortion, but in a different sense that you are used to - it technically cannot produce the classic higher-harmonic distortion like other processors. So I'd suggest: simply ignore it!