View Single Post
Old 06-12-2019, 10:40 PM   #36
Human being with feelings
Join Date: Dec 2012
Posts: 4,094


...shows what happens at any given signal level between silence and 0dbFS.


...looks a little weird. I lowered the threshold instead of input gain, but then I added an arbitray and incorrect amount of makeup gain to make it even visible in this picture. if it was your project, you could zoom in, but this was obviously a mistake.

If I do this...

You can pretty clearly see that near the zero crossing, it is not really a 1:1 correlation. The gain causes the output to get to and around the curvy part and then hit the limit really fast. That's not the display clipping off, it's ReaComp.*

OTOH, if I do this...

The linear portion of the "circuit" (near the zero crossing) does map 1:1 visually with the original input. It still hits the curve and limit just as fast.

Which do you prefer?

One could (assuming any of the gear we use was DC coupled, which it probably isn't) put in a series of DC voltages to the input and read the DC voltages that come out. It doesn't matter actually if the circuit multiplies (adds gain to) that DC voltage or the DC voltage is just bigger to begin with. By the time it reaches the actual non-linearity, that DC input will have a given level, and will produce some DC output. One could measure a number of different points between 0V and wherever the output stops getting bigger in order to make a graph. As the number of those points approaches infinity, it starts to just be easier and faster to run a triangle wave through it.

Edit/PS - If I use a sine wave instead...

...well, there's less gain so that I can make the point, but the input is already curvy, so how can I tell what part of the curve is the original and what part is ReaComp?

I definitely understand that you need to use a pure sine if you want to do the spectrum analysis route to see which harmonics are "generated", but if what you want is something like this...

...then the input should be triangle or saw.

*Actually, I did manage to crop out the very top of the picture of track 1's waveform, if I'd scrolled up just one more click... I'm not shooting it again.

Last edited by ashcat_lt; 06-12-2019 at 11:02 PM.
ashcat_lt is offline   Reply With Quote