DSP: possible to split audio to 2 volume levels? peaks and..

[mayo]

Hello

Is it possible to make something like double gate I mean to split signal to 2 parts

one from -100db up to -10db, and next one for peaks -10db to 0 db?

i can create gate which pass only signals above some treshold level, but need to separate all under this level

why? Because for mastering plugin, I want to process only peaks of signal with different algorithm than rest of signal.

If this is possible can you show some schematics please?

Thank you

[Spogg]

So you’d need a limiter for the lower level signal path to not go above -10dB say. Then the gate you made can go to separate processing to the limited signal, and at the end the two signals would be combined.
Is that the idea?

Cheers

Spogg

[mayo]

So you’d need a limiter for the lower level signal path to not go above -10dB say. Then the gate you made can go to separate processing to the limited signal, and at the end the two signals would be combined.
Is that the idea?

Cheers

Spogg

yes but problem with limiter is that when you cut signal -10db it creates lot of distortion, artifacts, harshness etc.
so Im thinking about some other methods how to split signal

[Spogg]

... yes but problem with limiter is that when you cut signal -10db it creates lot of distortion, artifacts, harshness etc.
so Im thinking about some other methods how to split signal

Yes that’s the problem with limiters. There’s a trade-off between a fast response and side-chain ripple.
I’m afraid I don’t know of an alternative solution…

Cheers

Spogg

[mayo]

... yes but problem with limiter is that when you cut signal -10db it creates lot of distortion, artifacts, harshness etc.
so Im thinking about some other methods how to split signal

Yes that’s the problem with limiters. There’s a trade-off between a fast response and side-chain ripple.
I’m afraid I don’t know of an alternative solution…

Cheers

Spogg

yes yes, thanks, maybe something like inverted gate? but I dont know if it is possible to code it...

[tulamide]

As a guy with a lot of imagination but no clue about Flowstone's DSP, I throw this in:

An audio signal is nothing else than a bunch of discrete samples, each representing part of the amplitude and therefore a dBFS value. As naive as I am, I would just check each sample, and if it is below the threshold (say, -10 dB), I'd send it to output 1, else to output 2.
Timing is still intact, but there are now two audio streams, that can be treated differently and later re-combined.

Stupid?

[ChrisHooker]

Not stupid at all!
Attached is an example through code.

The trouble comes in processing the two signals after this split. If their levels change at the threshold crossover point set in this first step, then recombining them will result in clicks where the level jumps between the two. Therefore, I'm not sure how useful this would actually be.

[Spogg]

@tulamide
The problem is that the system needs to respond to “loudness”, probably as an rms value, which needs a finite time to evaluate. At an individual sample level you would have a wave shaper or distortion module.

Cheers

Spogg

[ChrisHooker]

Perhaps the best approach would be to use the split signals as a side-chain, and process the main signal in two ways (run the processes in series of each other). When the lower signal is present (or just any signal at all, before the split) you have your 1st effect on the main signal. When the upper signal is present, an additional effect takes place on the main signal.

You can have these effects utilize RMS or peak referencing, however you want.
No recombination is necessary, since the split is only used for detector side-chaining, not for being processed itself.

[Spogg]

I was thinking about this all wrong!

If you have a loud signal it can’t be quiet at the same time (unless Schrödinger was involved).

So, you use an envelope follower of your choosing (peak/rms etc.) and the side chain value created is used to route the incoming signal through one effect channel or the other.
The routing (multiplexing in effect) should be “soft” to avoid sudden shifts. So at say -12dB channel A gets the whole signal and at say -6dB channel B gets the whole signal. In between those crossover levels the signal is shared, so at -9dB both channels get an equal amount of the signal.
At the end of the 2 processing chains the signal is combined.

I think that approach would work and be quite useful and potentially interesting.

Cheers

Spogg