Hi Tony.

My main point is that even when two differently-produced pieces of music are equally normalized, they will still sound like they are playing at radically different volumes. Normalization won't make all your songs sound like they are at the same volume if they are produced differently.

I support your main point: Normalization (if you refer to the one that increases the amplitude (almost) as much as possible without clipping) will not make sure that two songs sound as being the same volume if they aren't ment to (and produced to, to use your words). However, there are only a few possibilities to make two songs sound as loud as each other (to a certain extend):

1. Normalize in the classic (amplitude) way. This leads to songs with similar peak volume. only similar, not the same, because of the difference between amplitude and volume.


2. Normalize in the more modern (power) way. This leads to songs with the same peak volume, but that peak volume might be reached during 80% of the playtime for one song, but only for 1% of the other, leading to different 'experienced' volumes nontheless.


3. Use one of the above, but change the normalization factor dynamically during the play (also called compression, like the voladj-Kernelpatch does). This will probably lead to extremely similar experienced volumes, but might disturb the artistic experience during the play of a single song if there are prolonged intentionally (more) silent periods.

The whole thing is: Todays Pop music is almost always compressed (volume wise) and almost all Pop CDs are also normalized to approx. 98% of the maximum possible peak amplitude, so normalization and also compression (again, volume wise) won't give you any advantage in most cases. If you are in an extremely unfriendly (loud) environment, you probably do gain some davantage by using dynamic sound compression (like voladj), if you are in a friendly (silent) environment, you probably will loose a little of the intended artistic experience.
If you are listening to classic or older Rock/Pop music, normalization (no matter if amplitude or power wise) will probably help in both environments, while dynamic compression imposes the same (dis)advantages as with modern Pop.
So, when you do use normalization, my tips are:

• Be aware that, depending on your music collection, you won't gain much, but leaving CPU time aside, you don't loose anything either.
• Therefore, use normalization, but if you ask me, tell your software to only normalize (to approx. 98%) if the peak volume (resp. amplitude) is below 95%. Your would be able to tell the difference between a 95% and a 98% song anyway.

The reason why I won't recommend normalizing to 100% is that the encoding to MP3 and it's subsequent decoding sometimes increase the peak amplitude of a song to some extend, and using 98% normalization, you don't take the risk of that causing any clipping on the empeg's digital part.

Do we agree now?

cu,
sven

Do we agree now?

Yes. You're more or less saying the same things that I wrote here.

But I'm not sure what the difference is between numbered items 1 and 2 on your first list. If the resulting songs all have the same peak volumes in both versions, how are they different?

I understand your point about the difference between analyzing RMS volume versus analyzing peak volume, but if in the end, the files get normalized to the same amount (ie, no peak ever clips and all peaks fall in the same range), then the analysis method doesn't matter. Do you have an example of a piece of software that does #2 without resorting to dynamic compression?

Hi.

Right, _if_ they get normalized to the same amount. But actually, if they are analyzed differently, they will get normalized by different amounts. Also, RMS volume is still a little bit off the analysis I thought of. Do you know of the sone (as opposed to dBA) loudness indicator? Like dB(A), it weights different frequencies differently, so a full-peak 50Hz tone gets a different loudness indication than a 5000Hz tone, but sone is linear (dB is logarithmic), so a sone 2 loudness will actually be twice as loud as a sone 1 loudness. Now normalising due to the same peak amplitude becomes a whole different story as normalising to the same sone values.
Anyhow, I haven't yet tried this out, so I can't say how much different this would be. I know there is normalisation software out there that uses some sort of power based normalisation, but I don't know which one does. So for now, I will keep using software that does the simple peak amplitude normalisation.
I don't like dynamic compression all that much, because that also changes the relative volume of different parts of the same song, and I can't really influence that to a level I would like to take influence on it, so I skip over dynamic compression for now.

cu,
sven

So how do these other normalization routines cope with this pathological example? Can you give an example of software that handles it?

So, this is probably a stupid question, but would the VolAdj work for a system where the RioCar is plugged into the head unit (and thus equalizer and loudness don't work)?

Yes.

However, it will only fix the volume level on the MP3 files you play, not on anything you're playing on the head unit.

But it will work with any volume, loudness, or equalizer setting. Because it doesn't actually change the volume setting, it simply modifies the dynamics of the raw digital audio data.

Hi Tony.

It would probably amplify the classical part at the expense of some clipping, or attenuate the Pop part to a very low volume. At least if the "loud" part of the classical one was long enough. AFAIK, that type of software still avoids clipping if possible.
And no, like I said, I know how that type of software should work, and that there are some such programs out there, but don't know any specific package that uses this approach. I have a tendency to absorb interesting facts about different technical things but to forget about real world implementations of that. Sorry. If I stumble about some such software, I will say so in here.

cu,
sven