I’m trying to do my first serious mix of a very complicated electronic track. I’m starting with the bass which has all sorts of out of control moments in it. In a single measure, my bass line consists of 5-6 tracks. Each track has parts that were mixed down from multiple synth sources. My first task is to get everything about the same level and to do so I’m starting by making sure the peak levels for each part of each track are the same. I’m doing this because some transients/events/peaks on notes are much smaller (and sometimes quieter) than others.
My current process is to use the “Volume Handles” to move each wave to the same peak level using the Cubase meters. I select each part (usually one note) and loop it while moving the volume handle up/down to make them all (-3 db).
On the one track I’ve managed to do this on, it sounds great, and even though it’s peak and not RMS, I’m happy with it because it’s the first pass and I’ll go back and adjust them to sound right where needed in the next pass.
Here are my questions/thoughts:
From what I understand, my 44.1 kHz, 32 bit wave files are being mixed in a 32 (64?) bit domain where there’s massive amounts of headroom. When I increase the wave form of something to bring it up to -3 db, even if the waveform image is taller than the part’s box, it isn’t clipping right? (Assuming the original audio isn’t clipping.)
My understanding is that using Volume Handles is like using the fader or gain makeup on a channel - it isn’t re-rendering the file. My understanding is, that Normalizing (which is effectively what I’m doing by hand with the Volume Handles), does create a new audio file and therefore introduces artifacts while re-rendering the audio. Are these ASSumptions correct?
ASSuming that Volume Handles don’t introduce distortion or artifacts, I would like to find a way to automate my process. Is there a way to select a part or multiple parts and have the Volume Handles of each part set to the correct makeup gain to set them all at the same peak value?
I haven’t looked into macros yet. Is it possible to use the data collected from the Statistics function to set the Volume Handle amount? (I’m guessing it’s not.)
I wonder if he means that any time a file is manipulated, there is a small rounding error which introduces a small degree of distortion? I don’t know if that’s true anymore … I think I remember reading something like that long ago, with Cubase VST?
I’ve submitted a formal support ticket for this to see if they can help.
I’ve read in many books on mastering that normalizing adds artifacts. From what I recall, re-sampling is required simply because of the FFT algorithms required to read the data as levels, and then sum, and then convert back to data.
I’m also guessing that Volume Handles are part of the gain stage, and as numerical data in Cubase, can be considered in one calculation along with pre-fader and gain makeup values for each channel during mixdown.
I’d like to think 32-bit files make this a moot point, but I’m going to be doing a lot of this now, and I’d like to be sure.
It’s not “normalising” as in mastering where you try to get all the tracks to the same loudness
Cubase does peak normalising, no re sampling is needed no dithering, just a scan of the file for the peak level and a gain change to move the whole file up to the set maximum peak level.
I’m trying to accurately reproduce low sine waves so that large sounds systems can reproduce the tones. Artifacts will produce inefficient movement of the cone and reduce the impact of the vibration.
Maybe I got the part about FFT wrong, that may only be an issue for transforms in the pitch domain. Initially, I thought it was a simple gain change too.
Any errors are going to be so small that they are irrelevant.
If you are worried about any differences between a fade handle gain change and a normalize, do a null test.
Infact make a sine wave at -6dB (or something) copy it, normalise to 0db drop the gain by 6dB and see if it nulls against the original, repeat at -12dB, -18dB, -24dB, etc untill you get a difference!
They would have to be some large artifacts to interfere with what you’re trying to do!
And most large “sound systems” are high pass filtered anyway.
And then I exported them from the Audio Channels and the main Output Channels. They are saved as:
1 - No Change - Export Output Channels.wav
2 - No Change - Export Audio Channels.wav
I then did a series of Normalizing calls and saved them out in order:
3 - Normalize to 0db.wav
4 - Normalize to 0db to -3db.wav
5 - Normalize to 0db to -3db to 0db.wav
6 - Normalize to 0db to -3db to 0db to -3db.wav
7 - Normalize to 0db to -3db to 0db to -3db to 0db.wav
The result is that normalizing a file from a different source (RME) created a difference, but only when it was normalized. Once the file has been normalized, comparing 2 files at identical DBs proves that they are identical.
Created a mono track, put test gen on. 1Khz sine at -23dB rendered 4 bars 32fp mono.
Copy track. Normalised copied track to 0dB and back to -23dB, perfect cancellation. Repeat normalised of copied track another 5 times 0dB to -23dB, still perfect cancellation. Did 3 times with same multi-normalised track 0dB to -50dB a further 3 times…and back to -23dB. Perfect cancellation. Got fed up, gave up!
Took new copy of original, normalised to 0dB, used volume handle on original to increase gain by 23dB, result not always complete cancellation! It seem the volume handle gain setting is imprecise. Conclusion that no harm was done to the signal, incomplete cancellation is due to non matching gains.
The result is that normalizing a file from a different source (RME) created a difference, but only when it was normalized. Once the file has been normalized, comparing 2 files at identical DBs proves that they are identical.
You are probably comparing a 16 bit file against a 32float file, once you have processed it, it will be a 32bit float file. and if you then exported it as anything other that a 32float file there will be differences.
Try it with a file you didn’t make in Cubase (like the test file I used). It’s a MINOR issue, as illustrated by my results, but I’ve realized it doesn’t matter anyway!
And yes, the Volume Handle is not accurate, since the meter isn’t showing the floating point value.
BUT, despite all of this, I’ve discovered the main problem is that normalize is pre EQ and the Volume Handle is post EQ.
I EQ every channel, and so even with everything set to the same value pre, I have realized I have to do it by hand anyway.
Unless there’s a way to automate gain adjustment post-EQ, I’m going to have to do every slice by hand.
OK, back to work! It does sound really nice, so it’s been worth all the hard work anyway.
The point is that Cubase Normalise is the same as Volume handles as is the same as channel gain, and if no inserts or EQ used is the same as the Fader.
The goal is to get the same volume post EQ, despite where the gain change is done. The only way I can think of doing this is to mixdown the files with EQ and do it post - which I don’t want to commit to for obvious reasons.
If you are looking for a sound to be the same and level matched, you would be better off with your part or loop in MIDI, load the sample you want, Then trigger it by the non-level matched part. Quantize the velocity and have a fixed level sample.
In observation of your thread here, it appears you are going against what the music wants to be for the effort it is taking to achieve what you want is severely opposed by overthinking and lack of a cooperative work flow.
Music has a way of making itself come to be. If you become the muse, your path will become more clear and less effort will be needed.
