Hello, question from a beginner here…why is it that when I record a few tracks, each monitoring between -21 and -9 individually, and I put them together, it peaks out the main mixer on playback? I thought that if I wasn’t peaked out on individual tracks, they wouldn’t peak when layered, but I must have been mistaken. Any assistance would be greatly appreciated. Tony
When Audacity exports your final work it just adds up all the tracks. And yes, if you decided to play all twelve parts of an orchestra “safely”, you’re likely to seriously overload the final show. The only current work-around is select all the tracks and keep Amplifying to a negative number (top number) until the overload goes away.
I think you can play the show and watch the green flashing sound meter. Keep tuning until it doesn’t peak to red.
We’re discussing a graceful and semi-automatic way to deal with this. Everybody who does overdubbing runs into the same problem sooner or later. It should be a “push this button to set levels” and the problem goes away.
The tracks are all summed up when they are rendered in order to play.
The conservative calculation is (all tracks have the same gain in dB):
-6 dB = 2tracks
-9.54 dB = 3 tracks
-20 dB = 10 tracks
This is only true if all tracks have the same content.
For non-similar tracks, the allowed number of tracks before clipping can (max)be doubled.
You can control the peak of all tracks by temporarily creating a mixed track (Shift-Ctrl M, when all tracks are selected and unmuted).
Use “Amplify” with all tracks selected, including the new one. After ok, you can delete the mixed track.
Make another “mix and render” test track and you’ll see that the new track has 0 dB.
Ensure that the format is 32 bit float because this can handle gains greater than 0 dB and doesn’t hard-clip the peaks.
By the way, you can give some headroom when you amplify the tracks; increase the proposed number if it is negative, e.g. -9 to -10 dB this gives one dB head room.
In summary, the procedure for the semi-automatic way is:
-Ctrl A -->selects all tracks
- Ctrl-Shift M → Mix and render to a new track
- Ctrl A
- Effects → Amplify, Ok (or higher negative value)
- Delete mixed Track
And that illustrates why we want “push this button to set levels and the problem goes away” similar to Effect > Amplify > OK, but for a multi-track show.
Also note that if your default is 16-bit rather than 32 (like most of my stuff), this process will probably destroy your mix. There’s a very serious discussion thread about how to design that tool. Simplicity isn’t simple.
It looks to me still like a very easy effect to implement.
Lets call it “Master Amplify”.
The amplify does currently the following when applied to multiple tracks:
It searches for the highest peak and proposes a change of (0 dB minus peak).
If we have for example tracks with +3dB, -3 dB and -6 dB, it will propose amplify by -3 dB.
This gives 0 dB, -6 dB and -9dB for the tracks. This will still clip in total because the amplitudes (linear) are 1 + 0.5 + 0.3535 = 1.8538 linearly or 5.36 dB.
including a mix and render step (automatically done in 32 bit float) would return this peak of 5 dB or less (dependant on the content).
That’s the reduction the “Master Amplify” should propose.
Thus the new amplitudes would be:
-2 dB, -8 dB and -11 dB and adds perfectly up to 0 dB.
You can add head room by typing a negative value into the “New Peak” box.
In other words: the effect does a “Mix and Render”, measures the peak and reduces/amplifies the selected tracks by this amount, minus the head room. I can’t see anything complicated in this.
No, the actual process needn’t be complex, but I remember the discussion branched off into whether it was desirable to permanently change the original tracks or not. If you do, exporting would be simple, but you could be presented with really teensy “rendered” tracks and trying to edit them further would be very difficult. If you don’t, you have to remember to apply the tool each time you export and the possibility exists of delivering trash to a client because everyone is always in such a rush.
When I last touched the thread it was down to creating a particularly talented “Master Fader” (from the large mixing desk model) and proper management of that fader could solve this and possibly several other production problems as well.
I’m all for it. Particularly since the problem seems to have cropped up as a forum post about every week or so since I brought it up.
(by the way, sorry for the windows shortcuts above, the Mac uses probably Cmd instead of Ctrl)
Koz, do you see any destructive aspect for my method?
The tracks themselves are not re-rendered individually (these would destroy the envelope, pan and gain settings).
Furthermore, the amplification of the tracks hasn’t to be permanently, the found value could simply serve as a gain offset for all tracks.
It’s in principle possible to create a master gain slider for all tracks, however this wouldn’t work with the real “Rendered” peak over all tracks because each slider movement makes a mix down necessary and that would be far to slow.
The master gain had therefore to be relative (as it is in fact always the case).
However, I’d would vote for my proposed effect because it is absolute and works for all selected, non-muted tracks whereas the master gain slider would change the overall gain by an offset, regardless of selection, state or actual “rendered” peak.
What’s more, I can’t set the gain slider properly with regard to the playback monitor because it isn’t accessible for VI users.
Essentially that is the same as the “particularly talented Master Fader” that Koz mentioned.
Users would be able to add or subtract an offset for all gain controls by moving the master fader. This provides a quick, rough and ready way to control the overall volume of a multi-track project without affecting the relative levels of the individual tracks. Currently the only way to do this is to manually apply the offset to each gain slider, or to (destructively) amplify the project.
The “talented” part of this is the ability to set the offset automatically by scanning the project. The scanning is similar to Mix and Render but without changing the tracks. Audacity only needs to calculate the peak mix level, so writing to disk is not required.
While manually adjusting the Master Fader would be largely guesswork, on large projects the scanning would still be quite slow (even though it does not actually “render” a new track), whereas just changing the offset would be virtually instant.
As I see it, there are still two effects involved.
- The usual “Master Gain” slider.
It manipulates the gain offset for all tracks in the project, including muted and new ones.
A intelligent behaviour makes no sense in such a constellation. It can not keep track with all mute states gain increases amplifications etc.
It would drag down the overall performance.
- Gain Normalization
It calculates the mixed peak for all selected and unmuted tracks and manipulates the individual track gains (or does it destructively).
It is therefore a effect on command and not automatic.
Yes the “auto” setting is an effect on demand. It is only “automatic” in the sense that you set the desired maximum mixed peak level and the effect sets the gain offsets for all tracks, it doesn’t just tell you what offset to manually apply.
It could optionally have a “Replay Gain Calculation” check box that does the calculation in relation to perceived loudness.
The mix and render procedure above will still be needed for “Export Selection” if one wants to maximize only this portion over several tracks.
I’m currently trying to write a Autohotkey script for this.
@Robert - I don’t think you voted yet for a “master fader” or other solution. Do you want to vote for some method?
I’m not very fond of a general master gain slider. That’s a remainder of the analog domain.
However, I like the “Normalize Project” part of it.
OK - you said:
Normalize - It calculates the mixed peak for all selected and unmuted tracks and manipulates the individual track gains (or does it destructively)
I don’t like “destructive” myself, but is your vote for “destructive” or for manipulating the sliders of selected and unmuted tracks?
If the latter, why ignore unselected tracks which will be included by straight export?
Because I am not sure about this. A gain slider should set an offset for all tracks.
The normalization should be relevant for export, thus muted tracks had to be excluded. Then, there’s the possibility to export selected tracks into one file and the normalization should therefore be valuable for only those tracks.
Let’say we have:
- Guitar -12 dB Gain
- Bass -4 dB Gain
- Drums -5 dB Gain
- Vocals -3 dB Gain
We could now add an additional offset of -4 dB with the Master Gain. This would probably apply to all tracks.
A Project Normalization might correct all gains to -5 dB. All gets messy if we only want to export an instrumental version of the song.
If the normalization only works on the unmuted tracks 1 - 3, the whole balance will be destroyed. The solution is to calculate from 3 tracks but to apply the gain change to all 4 tracks.
The overall gain with the vocals track enabled will be to high, but another normalization will correct this while keeping the original distribution.
Without qualms, I vote therefore for a Master Gain Slider…
…that adds an offset to all tracks
…that has a normalize button next to it
…that calculates the peak from all unmuted tracks…
…and applies it again to all tracks.
… is reset to 0 after such a calculation which allows afterwards setting an explicit head room.
…discards the selection (because rather a rare case, workaround with muting or copying to new project)
…works non-destructive (because this could cause problems with different bit-depths)
I also vote for a check box in the export dialog or preferences “Normalize to Master Gain Setting”.
Is it up to the user to do another normalization? What happens if user wants wants to export instrumental and full (all tracks) versions of the project you mention? Or makes any other changes to mute or selection states after Normalize?
Should the Normalize if necessary be recalculated and performed on unmuted tracks at export time (if a user enables an option like you mention to Normalize to Master Gain level on Export?
Does this have a text input box where the user chooses the level to normalize to?
Do you mean it does Edit > Select > None? If so, why?