How does saturation increase preceived loudness?

[sambaji]

When I compare the original pure signal, in this case, white noise, with the same signal but with added saturation, set the same RMS (average) db level, the peak level of the saturated signal is higher? Is that how saturation increases the perceived loudness? I thought saturation compressed the peeks (lowered the dynamic range) in the signal, or is that just for transients, not the overall frequency response? My guess is that either for is that saturation mainly adds higher frequencies that are audible, increasing the peak levels, but do not alter the RMS as much as lower frequencies do. My understanding is that streaming services such as Spotify normalize their loudness output to -14 LUFS integrated, not the true peak db. Getting higher peaks with the same LUF, or RMS, means that the track sounds loader.

Thanks.

[Halonmusic]

Not an expert but i think saturation compresses the peaks and transients, (as you already mentioned) and smooths the signal a bit then adds gain as you push it higher. Saturation is intended as a subtle effect applied here and there but can of course be overdriven if you want to.

[Tj Shredder]

That is correct. Saturation does compress the sound like any analog tape did in the old days. Sound engineers back then loved to push it into the red...
A habit that got lost in the digital age, as red on a digital recording sounds bad - very bad...

[sambaji]

Whether it decreases or increases the true peak to RMS ratio or decreases it (increase peaks values, when RMS is gained match), seems to depend on the spectrogram of the incoming signal (source material) and the saturation plugin and its settings. Using the same settings on Softubes free Saturation Knob plugin in both cases (Saturation turned to about three quarters and "Neutral" selected), white noise, which contains more power in the upper frequencies than pink noise, resulted in a decreased peak level, which is what you would expect, whereas pink noise, which contains more power in the bass frequencies (ie.g., ) resulted in increased peak levels. In these two examples, the RMS was gain matched to 0UV, -18dbfs.

On the other hand, using Voxengo's free Tube Amp plugin compressed the peak levels of both white and pink noise, which according to the above replies and what I read online are the expected results. I think the moral of the story is that not all saturator plugins compress peak levels, regardless of the source material. Saturation plugins vary in their non-linearities, algorithms, and eq response curves. They can give contradictory results as in the two examples above.

[arkmabat]

as red on a digital recording sounds bad - very bad...

People would pay a pretty penny for a digital effect that did this like tape.
Edit: Oh, I'm an idiot. That's what this IS apparently?

[4damind]

The human ear is more sensitive to some frequencies than others.... when you add such frequencies, it seems louder, even though the level does not change and nothing is compressed.
Paul Frindle, the developer of the Sony Oxford Inflator, had written something about this on Gearslutz

[arkmabat]

Sociobiology paired with musical knowledge could prove to be very interesting. Even just the range of human hearing has its specific reasons grounded in evolution.

[fese]

Whether it decreases or increases the true peak to RMS ratio or decreases it (increase peaks values, when RMS is gained match), seems to depend on the spectrogram of the incoming signal (source material) and the saturation plugin and its settings. Using the same settings on Softubes free Saturation Knob plugin in both cases (Saturation turned to about three quarters and "Neutral" selected), white noise, which contains more power in the upper frequencies than pink noise, resulted in a decreased peak level, which is what you would expect, whereas pink noise, which contains more power in the bass frequencies (ie.g., ) resulted in increased peak levels. In these two examples, the RMS was gain matched to 0UV, -18dbfs.

On the other hand, using Voxengo's free Tube Amp plugin compressed the peak levels of both white and pink noise, which according to the above replies and what I read online are the expected results. I think the moral of the story is that not all saturator plugins compress peak levels, regardless of the source material. Saturation plugins vary in their non-linearities, algorithms, and eq response curves. They can give contradictory results as in the two examples above.

I really don't think white noise is a good signal to test saturation as it already contains all frequencies at equal intensity. There are no 'peak levels' and saturation has no effect here. I would use a pure sine wave.

You're right that saturation plugins differ very much in their approach, and if you don't know how they work internally, you cannot really compare them regarding level. Most of the time you drive the saturation stage with the input gain, and that is what can cause increased levels. Some of them have compensated outputs or other gain stages. This is most likely what you observe.

Saturation in general reduces the dynamic range of a signal as it cannot reproduce a signal above a certain threshold correctly, that's why we say that it compresses the sound.


Edit: I think this is a rather good article about it: https://www.sageaudio.com/blog/masterin ... tering.php

[Ah_Dziz]

They eat peaks and generate harmonics. This is by way of a transfer function that would ideally respond dynamically to the input level and frequency content. In software you have anything from simple waveshapers to complex models of analog circuitry. Any approach can sound nice depending on what you feed it. I would look at some analog circuit design for simple transistor based amplification up to differential amp based designs. This will give you plenty of insight into what people are trying to do with software. Generally the analog designs are trying to be as clean as possible so the bits of distortion you get in nice designs are generally pretty subtle. Try out a simple "drawable waveshaper" plugin and see what the different curves do to the signal both in level and frequency content. I would use a sine wave for more scientific readings and something like a nice drumloop for a more real life application test.

The short version is you are trading peak level for more high frequency content. At the extremes you will transform a sine wave into a square wave.

[arkmabat]

I'm going to test this out, thanks to this thread.
https://www.softube.com/saturationknob
"One Knob To Rule Them All."

[sambaji]

I really don't think white noise is a good signal to test saturation as it already contains all frequencies at equal intensity. There are no 'peak levels' and saturation has no effect here. I would use a pure sine wave.

You're right that saturation plugins differ very much in their approach, and if you don't know how they work internally, you cannot really compare them regarding level. Most of the time you drive the saturation stage with the input gain, and that is what can cause increased levels. Some of them have compensated outputs or other gain stages. This is most likely what you observe.

Saturation in general reduces the dynamic range of a signal as it cannot reproduce a signal above a certain threshold correctly, that's why we say that it compresses the sound.

Yes, I am also testing the plugins with static and sweeping waves, both frequency-wise and dynamically, as well. However, I am finding using both white and pink noise helpful. From my tests, it is possible to add saturation to white and pink noise and both hear and see it with a spectrogram. It's subtle, but it's there. Using Melds MMultianalyzer I am able to overlap both the pre-saturation and post-saturation signal together and visually compare them. For example, I am able to hear and see the EQ/frequency response curve of the 3 modes of the Softube Saturation Knob plugin. In my mind, this makes sense considering the white/pink noise is dynamic, non-static, frequencies are constantly being regenerated at different times, although the overall amplitudes remain the same. When frequencies of saturation overlap frequencies of the original signal, wouldn't you get the summation of both amplitudes and notice an increase in those areas?

[sambaji]

According to https://ledgernote.com/columns/mixing-m ... se-mixing/ , The RMS (root mean square, basically an average) of pink noise level is about 3 dB lower than the peak. The noise will have some volume variance but no more than plus-or-minus 3 dB.

[DaveClark]

Although the first step in studying how saturation affects perceived loudness is to make laboratory measurements of the signals involved in a particular saturation algorithm, a subsequent step is to relate these measurements to known pschoacoustic effects.