Digital Disease ^ Limit the difference between two samples

Hello !

At the moment i took a closer look art the waveforms of “modern music” i wondered about the Waveform in some kind of digital audio. My English is not the very best - i’m native german - so please let me show a screenshot so you can see what i mean:
Bildschirmfoto vom 2018-05-04 indivisible before 3.jpg
This is a typical waveform. I want to decrease/ limit the level of difference between the samples to an useful scale or ratio. My Tweeter or mid range drivers are not a rail gun or something like that. It can not be useful to power them up cause i want to enjoy musik and not the settings of the limiter in my (or orhers) controller(s).

Is it possible to define the max difference between two samples ? I tried to manage this with several filters but cannot find one with any sense. In short sections i can use the “repair” funktionb in audacity, but it is only for very short sektions / limited to 128 samples. Annother possibility is to edit the waveform (the “draw” feature with the pencil) but that takes to much time doing this.

here another example:
Bildschirmfoto vom Digital Disease 2.jpg

A low-pass filter will reduce the distance between adjacent samples.

yes, a low-pass will reduce the distance between the sample points. But i wont reduce the distance of every single sample - only the silly ones have to be reduced.

in my opinion: therefore i need a dynamic component in the low pass filter.

Sounds like you want a “multi-band compressor”.

“Calf Studio Gear” includes a multi-band compressor, but some versions crash in Audacity.
“Zam-plugins” include a multi-band compressor, but I’ve not tried it myself.
There’s probably others.

Another way to do that without distortion is increase the number of dots.

Increase the sample rate (44100) to a higher number. A common video standard is 48000 and studio recordings use 96000.

The distance between dots should decrease because the accuracy of each sample goes up.


Hi Koz,
Your suggestion- inrease the sample rate - is funny. That will decrease the difference between two adjacent samples. But my intension ro decrease the difference between two samples is not because the furthermost sample will feel very lonley (maby too lonley) and might became sick. My intension is more physical. If you generate a short noise (e.g. 3 seconds white noise and in comparison pink noise) you can zoom in and see the difference in the waveform. I want too reduce the “like white noise” waveform sections so they became a bit more like pink noise.
The tecnical matter is, that there is often too much “energy” in the high section. How much electrical power is in white noise?
In professional audio systems the Amps will emit their full power @ 0 dBfs. as you can see in the screenshots (digital disease) , there is too much power in the high frequencies section. IMO no PA system has the ability to reproduce those sektions (like shown in the screenshots digital disease 1 or dd2) instead what you can hear is the limiter.
The white noise signal has 6 dB crest and to much power in high frequencies.
The pink noise signal has estimated 15 dB crest and is more likely a typical sound you can hear for a while without getting rid of it.

@ steve: The multiband compressor is able to reduce gain? . (increase crest?)

Like a standard compressor, a multi-band compressor reduces the gain when the level exceeds a specified threshold. The difference with a multi-band compressor is that specified frequency bands may be compressed differently. In you case, you would want low and mid range frequencies to not be compressed at all, and only high frequencies to be compressed.

In you case, you would want low and mid range frequencies to not be compressed at all, and only high frequencies to be compressed.

ok i’ll try the Multiband Comp and leave a comment after that.

Ok the Multiband Compressor works. Let me say, it doesn’t deliver the best result but it is the easiest way to reduce the “digital disease” effect so far.
Another wav is to safe the “infected” Audio File as .wav file, cut off the header and copy all data into a Calc Sheet. Then you find the “infected samples” via “conditional formatting” and define the max difference (dB) between two Samples. After that you’ve to copy back and replace an appropriate header. This is a bit more nerdy but delivers the best result in an acceptable lapse of time.

Thanks to all !