Q: MSpectralDynamics - v1106 Detector Input modes

[DarkStar]

I have read the Info about this parameter but am still a bit confused.

I can understand that each mode calculates a different level for the detector. And, here is a time-graph of the Compressed output (and peak meter levels) for the 8 detector modes, for a fixed input and side chain signal (C3 sinewave)



But I cannot work out when to use each mode. These modes have got me stumped:

Input - SideChain subtracts the levels of individual frequencies in the side-chain from the main input. When used as a compressor, it lets you exclude frequencies present in the side-chain from being compressed. When used as a gate it can do the opposite - by reducing the level of these frequencies it prevents them from exceed the threshold, so it can be seen as one way to solve collisions.

SideChain - Input works the other way around. It can be used to attenuate frequencies present in the side-chain, but not those present in the main input, hence cleaning the input signal.

Multiply multiplies the levels of individual frequencies. This may be the best way to solve collisions, since this operation produces (processes?) frequencies present in both inputs at the same time.

Can someone shed some light on them?

For those three modes: what detector level would be calculated for Input = -12 dB and Sidechain = -18 dB? And for Input = -18 dB and Sidechain = -12 dB?

[vectorwarrior]

Yeah, I'd be curious about this too, having read through the info like you did. I dont use SpecDyn for mixing/ducking purposes so I doubt I'd be using it but I'd be interested regardless.

I didn't see a feature request for this, so I'm guessing this was something that would help out Melda's future tools, but perhaps it was an idea for a specific use case or was something in a tool like Ozone/Neutron/etc that they wanted to implement.

[MeldaProduction]

The idea here was to somehow use both signals for processing, mainly for avoiding collisions, but I think it could have many uses. The rest of it was just a playground . Try using a filtered noise with some slope, say +6dB/oct for one signal and -6dB/oct for the other, the results should be easy to see there.

[vectorwarrior]

Yeah, i was thinking about this this morning and it got me thinking of some interesting creative uses that I'm going to investigate when i get some time.

Which got me thinking... would this be useful to add to MMorph? As in, multiplication of A and B, etc.

[MeldaProduction]

It cannot be applied to MMorph, I'd already add it when I was developing it . It works very differently...

[DarkStar]

Playground ? Yes. But if how they work is not clear they are unlikely to be used, other than for experiments.

Anyway, can someone help me out with the "dB-arithmetic"?

For those three modes: what detector level would be calculated for Input = -12 dB and Sidechain = -18 dB? And for Input = -18 dB and Sidechain = -12 dB?

Input - SideChain
-12 - -18 = ? dB
-18 - -12 = ? dB

SideChain - Input
-18 - -12 = ? dB
-12 - -18 = ? dB

Multiply
-12 x -18 = ? dB
-18 x -12 = ? dB

[vectorwarrior]

I would presume it's

Input - SideChain
-12 - -18 = -6dB
-18 - -12 = 6dB

SideChain - Input
-18 - -12 = 6 dB
-12 - -18 = -6 dB

Multiply
Okay, stumped here... I can't imagine it's 12x18, which is what I was expecting, because the value becomes way too small.

[DarkStar]

^^^
Nah, dB levels aren't linear. After all:
-18dB + -18dB = -12dB
-18dB + -6dB = -4.2dB

[vectorwarrior]

Yeah, i did some testing with signals and realized my estimates were off but couldn't figure out the solution. Your numbers feel right to me, how did you calculate them?

For multiply, the signal is definitely MUCH quieter in the tests I did.

[DarkStar]

I put MNoiseGenerator on two tracks, using white noise, sent them both to a third track and looked at the meter levels. Alternatively use MOscillator.

[jmg8]

It would probably be a lot clearer if you make the two tracks have very different audio. Like Vojtech says, put low slope on one and high slope on the other. It should be easier to see how they interact with each other.

[MeldaProduction]

It would probably be a lot clearer if you make the two tracks have very different audio. Like Vojtech says, put low slope on one and high slope on the other. It should be easier to see how they interact with each other.

Exactly!

And stop doing math . I think, not sure, dB is NOT used there, it is processed in linear domain.

[DarkStar]

<< never mind>>