DSP Robotics Support

I have samples thats different volume levels.

And i got a rompler.

I wonder how can i make the samples same level inside the rompler. so its balanced. Normalized.

Rex told me i prob need some kind of ruby. Anyone thats good with this kind of things? That can help me out pls?

Why do you post the same topic twice in two different places? Anyway, just use the "Norm" float array module. It makes the array samples between -1 and 1 and normalizes different volume levels. Then you can send those float arrays to the stream and each should be more or less equal loud.

adamszabo wrote:...just use the "Norm" float array module. It makes the array samples between -1 and 1 and normalizes different volume levels. Then you can send those float arrays to the stream and each should be more or less equal loud.

Interesting!

I thought that was for single cycle samples. Can it handle a wav sample of a few seconds with tens of thousands of elements in the float array? If yes, would it take a long time to process? I never tried this.

Adam where is the other post? I must have missed that. Because i think the other post was about same volume for wave arrays. Not samples. mybe its the same i dont know. i tought it was bit different. not sure.

It seems to me that it is easier to do batch processing of samples in some editor. Why unnecessary processes in the scheme?

StereoSpace wrote:It seems to me that it is easier to do batch processing of samples in some editor. Why unnecessary processes in the scheme?

That's my thought as well. Why trying pseudo-realtime normalizing at the cost of high CPU load for a sample player? Just normalize them beforehand.

EDIT: Just in case there's a lack of understanding, normalizing means running through every single sample point and increasing its amplitude by a relative value, that's dependend on the highest peak difference to the norm value. So, if you want to normalize to 0 dBFS, you first have to go thorugh the whole sample and check each sample point to find out the peak value. The you calculate the increasing factor. You then run through the whole sample again and for every sample point apply the percentage increase of the factor.
That's not a task you'd want a plugin to do, that's meant to exist in a time sensitive environment like a DAW.

I know well of Johan's specific application.
The samples are most usually, if not all normalized.
Personally I can't be bothered batching and would normalize with the method Adam described.
The normalizing within FS happens at trigger rate, and I assume it is cpu costly,
but that's fine for the 2-6 sec samples.

I think what may too be applicable here is some compressor and limiter implementation imo

I can either batch all the banks in SoundForge or use FS normalizer

I checked out what Adam said and it works perfectly and fast! I tested in a wave player using a 5 second sample reduced in amplitude by -18dB and it was normalised instantly.

I’m impressed!

Many thanks Adam

Yes, thats exactly what I mean Spogg, I thought it would work. In a rompler you usually dont have long samples, you mostly have short ones or loop a sample for a few seconds, and this array normalization should do the trick as a quick fix, but I agree, each sample should be normalized by default.

Spogg wrote:and it was normalised instantly.

tulamide wrote:Why trying pseudo-realtime normalizing at the cost of high CPU load for a sample player?

Really, I don't get it. There we are, constantly crying for help towards Martin and others for ASM optimized code, just to save those 2 nanoseconds per second. And then we implement pseudo-realtime normalizing? Or are concerned that graphics could be using CPU? Sometimes I'm really confused about certain decisions.