So while wanting to test something else, I stumbled upon the differences event volume, pre gain or volume automation have on the rendered audio.
Test setup: new empty Cubase 14 project, 128samples buffer size, 100Hz sine wave audio file on a new track. Three tests, each producing a steep 6dB decrease in level, at the same position for respective event (1/4 of the sine cycle):
I repeated the experiment with a buffer setting of 512 samples, which didn’t make any difference (which is good in my book).
Both event volume and volume automation produce visible and audible distortion of the waveform, as opposed to the pre-gain automation, which is properly smoothed and is just a nice fade.
Whether the resulting distortion matters, is of course dependent on the material, and on how much you increase/decrease the level.
project file, for anyone who - like me - obviously has too much time on their hands cubase automation.zip (1.2 MB)
Yes, forgot to write that, of course the volume automation is a configurable setting now, but I set it to the highest possible, which is 32 samples (I wouldn’t know why anyone would want to go lower). I just set it to 8 samples, and the result is, as I would expect, even worse:
What good is the sample accurate volume automation if it leads to distortion? A proper fade might not be sample accurate, but leads to better audio quality overall.
The funny thing is, when you draw volume automation via the selection tool and then use the “scale vertically” handle, Cubase actually doesn’t draw a 90° line (which wouldn’t be possible anyway), but a slight angle:
to me that looks like the zone that is used for lowering the volume was not properly selected.
The “issue” looks like the start of the selected area was not a the 0-crossing, but in the middle of the raising sine wave.
So at THAT point Cubase lowered the volume, and - to me - that seem to cause that “kind of distortion”.
At least that one below looks extactly like that:
When you change the level, regardless of where, you should end up with a distorted signal. You are literally changing the waveform by changing the amplitude. If the input signal is a sinewave then the output will not be a pure sinewave that looks the same no matter how you go about it.
Increasing the precision of volume automation should mean that the level changes you ‘request’ of the automation system should be executed more precisely which means that if you have drawn an “instantaneous” drop of 6dB then the more precise the automation playback is the faster the attenuation should be. If it was infinitely fast you would see a purely vertical drop to the next value (6dB down).
The faster the amplitude change the “sharper” the visible “edge” in the waveform.
To me this all seems correct.
(I just had my second cup of coffee though so it may take a while to see why it wouldn’t be)
From a purely technical standpoint, you are of course correct. I don’t think though that this is a case where the mathematically correct solution is the best one, at least not for the quality of the audio. Which is why e.g. all good audio plugins (like TrackControl) always apply some form of smoothing for parameter changes, and Steinberg do it themselves e.g. in their new Volume plugin:
If you look at that render in a spectrogram, there are still some artifacts visible, but barely audible anymore.
(Of course, the lower the frequencies, the higher the likelyhood of shaping the wave…)
I freely admit that so far, this is nothing that I encountered in real life, or at least not that I have noticed or attributed to automation, most likely because recorded material has overtones anyway and you do need to hit a low enough frequency at the right point with a more drastic change …
Well, of course, I deliberately selected that point for testing purposes. In real life, you are not guaranteed to always hit a zero crossing with the range selection tool, so unless you really zoom in and control your start and end of range, that can easily happen. At least not in the automation lane, even with “snap to zero crossing” enabled, that only works on the audio event.
Whether or not it is the best solution is really debatable. I’m sure you can come up with plenty of cases where you really want the automation to be as precise as possible. Granted, we’re talking about 8 versus 32 samples or whatever it was, but still.
Also, I wouldn’t say that this is definitely “smoothing”, it could be as simple as having less precision for different parts of the automation system for performance reasons.
Ultimately I think it should be the user’s choice how much precision automation should have, not the DAW’s choice. And think about the alternatives you’re faced with here:
the system is smoothing processing and that limits maximum precision
the system is more precise and you can smooth it yourself by adjusting the automation curve’s angle / rate of change
I am not sure I find a reason where I want a a volume automation to be so abrupt that it may create audible distortion, as opposed to having a short fade… but as I wrote, an option to configure the degree of slant when using the handle would be nice.
The question of course remains why this is handled so inconsistently in Cubase. Four functions (event volume, automation for volume, pre-gain and the volume plugin) - four different results.
But then, we are used to inconsistencies in Cubase, aren’t we
