I’ve haven’t really used this before on the Signal Follower but a recent project was giving me some strange problems when trying to duck a percussion sound from another, I was getting a little click at the start instead of clean ducking. So I played with the lookahead and found it was exactly what I needed. So now I’m curious about it, the renoise manual is pretty vague about it. Are there any sure fire situations when you would need it? Or just as and when you feel like it? Is there a way to work out how much is needed? And what happens with very large values, the ducking would happen far too early, like before the target drum sample starts?
I wonder if it’s for fixing up plugin latency compensation perhaps? To be honest i try to avoid using the signal follower for ducking unless i work with vocals and other non one hit samples. I use a key triggered custom envelope because it gives you so much better visual representation and abilities to fine tune the ducking precisely and independantly from the source sample. If you feel this kind of ducking seems too static and lifeless you can use a signal follower to slightly move the offset of the custom LFO.
Lol, some lesson on relativity? Never used lookahead limiters before?
Lookahead will quasi look into the future. Using relativity - by delaying everything else. That’s why pdc must be active, you will essentially add latency for your whole project, but not for everything the signal follower is reading from. The manual seems clear enough for me on these points. Normally there would be some delay for the reactions of the signal follower, we’re in the digital, processing buffer latency world. But by reading input at normal latency, and having everything else delayed much more, you can have some headroom to process signals in advance.
This means, if you have two clicks and want to duck one to another with a signal follower. If both clicks happen at the very same time - normally the ducked one will still sound the attack before being cut, because the reaction of the signal follower has latency. But you can shift latency with the lookahead parameter, at the sweet spot the ducking will be perfect and no attack will come through from the second sound. This is what you observed, and is one of the purposes of the parameter. But you will add the same latency that the signal follower would be able to read “into the future” to your whole project by this as a tradeoff for this.
So, it may sound impossible at first, but lookahead will make the signal follower react in advance to peaks, by some configurable amount.
Edit, yes, other uses are possible. Sometimes the very large numbers can result in interesting effects when you experiment in a wider sense than just ducking stuff. It is a timing parameter - you can basically shift the envelope of the signal follower in time. Only ahead though, it seems, delaying might also be a nice use for it. But it will mess up realtime performance when using “live”, so regard it as a mixing/mastering thing, not some live action tool.
The LFO way is something I keep meaning to implement but haven’t got around to yet and instinctively grab a signal follower.
Never used lookahead limiters before?
No not really.
at the sweet spot the ducking will be perfect and no attack will come through from the second sound.
I understand that but is there really no way to find the sweet spot other than listening to it? How would I know when I need to use it at all and when not to? Surely for something like ducking a bass from a kick there will be a small period where both are playing at the same time by default. So I would always need to use lookahead?
The manual does state what it basically is but still after reading that, as well as your details description of it, I’m still somewhat in the dark about it. Why have a ms value at all if it is just a case of doing it by ear every time?
Doing it by ear is ALWAYS the best approach.
Here, I have found the sweet spot to sometimes be less than 1ms, which is a bit tricky because the value field accepts only whole ms unless you type in the values. And you might want to experiment with the low/high-cut filter, high frequencies tend to be detected faster, more precisely than the low ones.