Sound Level to Plugin is wrong

Hi all,

Using Audacity 2.3.1 in MacOS Sierra.
The problem I have is as follows:

I load a song that I know has been exported at -23 LUFS.
When I play it in Audacity, checking the loudness level with the Youlean AU plugin, the audio is always around 6 LUFS too high.
Using other songs, same thing.
I also notice that the track, mixer or master volume controls have no effect, it’s as if the audio being sent to the plugin is pre-fader.

Any one else experience this?

Further to my previous post, it seems I’m not the only one.
Someone else having what sounds like exactly the same problem but using another loudness measuring plugin:

He was using an earlier version of Audacity, but it seems the problem remains in later versions.

The conclusion thus far, is that there seems to be no fix to this (judging by no replies from the other post).
Pity, as Audacity is a really easy piece of software to use and get to know, but cannot be used for any serious leveling work.

Did some more tests and what I believe is the problem with Audacity is drawn below using a simplified block diagram:

As can be seen, it looks like the track audio going to a plugin is pre track fader, “B”, all other audio editors and DAW’s are post fader as
shown in “C”.

As previously stated, this is a very simplified diagram, but I think it illustrates exactly the problems that I, and others, are experiencing.

In other words, you are saying that the Youlean AU plugin measures the playback level and not the level of the audio data.

Note also that there are different ways to measure loudness of mono tracks. Some plug-ins treat mono as a single channel through one speaker, while other plug-ins assume that mono plays through both speakers. The two methods differ by 3 LU.

Hi Steve,

What I’m saying is that any plugin will measure the audio data that is presented to it.
In Audacity’s case, it’s pre track fader and also seems to be affected in other ways that I don’t know.
So I guess yes, the plugin reads the playback level.

To give you an example, I import audio that has been checked and confirmed to be at -23 LUFS.
In Audacity, the readings on all the loudness plugins that I have tried, show too high a level, not just the Youlean one.

The same plugins show the correct levels on Reaper, Logic and even DaVinci Resolve.
On all the above software, the track fader controls the audio level being fed to the plugin, with 0dB making no difference.
The same is not happening with Audacity.

Yes, there will be a difference between mono and stereo.
I only use stereo though out my whole work flow and it’s always one stereo track, not two discrete monos panned left and right.

Looking at your illustrations again, I got that wrong way round. You are saying that the plug-in measures the level of the audio data, not the playback level. That’s normal for how plug-ins work in Audacity - the plug-in code processes the data in the track. Example: If you normalize a track to 0 dB, then the audio will peak at full track height (0 dBFS) irrespective of the track gain or track pan sliders.

I’ll have to leave this here as I don’t have the Youlean AU plugin.

The plugins are available as AU and VST.
Must also add that it’s not only the Youlean one that is reading too high.

Another thing that I have noticed is, which may be related, there is definitely a discrepancy between what Audacity+Nyquist measure WRT peak and RMS and what other plugins measure, not just the loudness plugins.

It’s as if the audio being sent to any AU or VST plugin has gain on it, even though the track slider is at 0dB.
I have tried applying a “gain” of -6dB to compensate, the plugin then shows the correct level within Audacity, but when exported and checked, sure enough, it’s too low.

I vaguely remember trying this on Windows as well (win 7 pro 64 bit) with the same plugins but using VST versions, and the same problems showed up.

I understand you don’t have the plugins so can’t answer, but can you confirm that my block diagram a few posts back, is indeed correct?

Yes. The plug-in acts on the track data (B), not on the post-fader playback stream (C).

Thank you Steve.

Some more tests that will hopefully help others, or even shed some light for the devs of Audacity, as to where the problem/s may lay.

There is definitely a discrepancy between the audio data fed to say Nyquist and external plugins.
In the example below, I took exactly the same song, imported it into Audacity and Reaper.

In Audacity, the spectrum and level analysis (Span, not the built-in one), shows clipping and a certain RMS and TP value.
However, in Reaper, those clippings are not there and the RMS and TP values are lower.

Interestingly, when wavestats.ny is run, the results (RMS and TP) agree closely with Reaper and not that of the audio in Audacity being fed to the same plugin.

If I also run “Find Clipping” in Audacity, there is nothing, further demonstrating the discrepancy.
Somehow, when audio from Audacity is fed to external plugins, AU or VST, there is some gain being applied.
As to where and why this is happening, I have no idea.

The freq. response is slightly different for each as I took the screenshot at different times, but the readings still hold true.

Just a thought - What is the sample rate of the audio that you are testing, the sample rate of the track, and the Project Rate?

In this case it was 44.1 KHz for all, in both Audacity and Reaper and the project rate was set to the same as the sample rate.
The problem also occurs when all is set at 48KHz, tried with another song at that sample rate and ensured that the project rate matched.

It struck me as odd that your “Audacity” illustration shows a significant level above 16 kHz, whereas the “Reaper” version shows virtually zero in that frequency range. Similarly the 20 Hz band is over -60 dB. It’s as if the “Audacity” version has an additional signal mixed in.

What happens if you run the analyzer on silence?

Hi Steve,

That is correct, the freq. response that you see, is not static like the built-in one in Audacity.
It updates periodically (settable) according to the audio content.

In both cases, I let the song play for at least 70% of it’s duration, as the RMS, TP and clippings are averaged or accumulated.

So that part that you see the extra 16KHz energy, was in that part of the song when the screenshot was taken.
Irrespective though, notice how there are no clippings with Reaper and this was confirmed with “Show Clipping” in Audacity.

What is telling is the dB scale difference between Audacity and Reaper.

To get a better understanding on how “Span” works, below link to one of many on Youtube.