Sound normalisation

[Keith Knight]

Hello again

Having installed an app that Martin has devised to get round the Memory Full problem I have.  Would it not be possible for him to do an app which normalises the output stage of the B2 so that the mp3s and flacs and wavs all play at the same level please?  The interenet is full of programs that can be downloaded on a computer, maybe it could be added to a B2?

Best wishes 

Keith

[Iain]

I agree.  Had to create a "Loud" album to get round this.

Not ideal.

Iain 

[Iain]

Agree - would love this.

Volumes of, say, modern RnB tracks are so high (Mary J Blige, Amerie, etc) compared with my other stuff.

I even made an album called "Loud" to counter this problem, but that is far from ideal on such an expensive product.  (I have bought  3 of them)

On Tuesday, January 25, 2022 at 2:14:36 PM UTC keiths...@googlemail.com wrote:

[Mark Fishman]

What is an album called "Loud", and how does it address the problem you describe?

[Iain]

Hi, Mark,

Thanks for your interest.

Sorry. I used the wrong word - duh!

Well, Brennan backup fatigue was setting in after several days and experiments - should have called the items "Playlists"!. 

It fails to address the problem by almost 100%, because popular and favourite loud tracks one might wish to include in, say, "MumsFaves", clearly, will not be included in those Playlists.

For years, I have felt that the lack of normalisation on the Brennans is the biggest drawback of them.  I have 3 of 'em.

The software "Switch" with which I am currently changing a B2's FLAC files to WAV (for a JB7) can normalise the tracks at the same time, but I have had to buy that software - the freebie does not convert more than a part of the FLAC drive's contents.

Regards

Iain

[Rearwing]

This comes up on a regular basis and it is one of my few really irksome issues, if normalisation is wanted it has to be for me on an opt in basis. I have a collection of recorded music spanning several formats and many different versions of the same piece depending upon when it was recorded, when it was re-mastered and when it was digitised; the reason behind this is that I wish to hear what the artist recorded and how it was mixed at the time. 

Dynamic range for me is very important, if a particular piece was mixed and mastered to be “quiet” in order to be able to give “room” for the full dynamic range and even the studio/concert hall acoustics to be heard, then I want to hear that. When normalisation occurs, it strips out the nuances of the process and veers towards elevator music to my ears, and I really don’t want that.

So please, bear in mind those of us who want to hear the difference and don’t want to be normal(ised).

[Mark Fishman]

I believe that "normalize" (or -ise, to cover both sides of the pond) is the wrong word -- or rather, the right word for the wrong function. Here's the description of that function from NCH Switch's online help:

Normalize

To 'normalize' is to adjust the volume so that the loudest peak is equal to (or a percentage of) the maximum signal that can be used in digital audio. Usually you normalize files to 100% as the last stage in production to make it the loudest possible without distortion. Another reason to normalize is to have multiple tracks sound equally loud, or to have equal average loudness.

This description further confuses the issue by propagating a common misunderstanding about how we decide something is "loud" or "soft", a confusion that also requires us to disambiguate (I'm stealing Wikipedia's word) "compression".

Normalization does indeed do what that first sentence says: it adjusts the overall gain so that the PEAK level of all tracks is the same. For music of similar type, produced and mixed for similar purposes -- e.g., all acoustic jazz, or all heavy metal, or all string quartets, recorded and mastered at around the same calendar year -- since the peak-to-average ratio of the music will be similar, adjusting the peak level (like turning the Volume control on your amplifier) will make everything seem similarly loud or soft.

The problem is that our perception of loudness has little to do with peak level and almost everything to do with peak-to-average ratio. The so-called loudness wars that have occurred since the 1990s (and earlier) involve compressing the dynamic range of music to make the AVERAGE level much closer to the peak level without simply pushing everything into clipping. Heavily compressed (not file size, but dynamic range compression) music sounds LOUDER than natural dynamics at the same peak level, because the average level is higher.

IF you are playing tracks from many different albums, mixed and released at different times (when artistic or commercial considerations were different), normalization doesn't cut it, but it's easy to do. Much harder, and more necessary, is gain analysis to determine the AVERAGE level rather than the peak. The ReplayGain algorithms were developed for this purpose. What you will find, in fact, is that to avoid clipping and still make all tracks have the same apparent loudness, you almost always need to make recent pop and rock releases QUIETER, because you can't make older releases (or proper acoustic music -- folk, jazz, classical) louder without forcing them into clipping.

Rearwing has it right: this should be optional, and preferably reversible. If you ever play an album straight through, you will want the dynamic relationships between tracks to be what the artists intended.

Cheers -- m.

[Mike W]

I haven't done any serious work on this but I can't find a system which does this dynamically as tracks are playing. There's a number of apps and software which will enable you to do level adjustments and then save the results.

Maybe this is seen as not desirable on higher end systems or I haven't looked far enough for systems such as the Brennan which do have this capability.

Mike

[Mark Fishman]

There are two ways to use ReplayGain: one is to place a tag in the file that spcifies an amount of playback gain to be applied at the time of playback. That's reversible. The other is to recalculate all the sample values, thus "baking in" the changes.

To make use of the tag method, the playback software has to read the tag and turn the gain up or down in software, as the tag indicates. There are many programs that do that sort of thing, e.g., foobar2000. Brennan software would have to look for, read, and honor the tags, but IMHO asking users to use some other software to do the initial (one-time) analysis and apply the tags would not be an enormous burden.

If you really want on-the-fly dynamic compression, though, perhaps that is best done in hardware. Something like this, placed between the B2 and your amplifier:

https://www.amazon.com/dbx-266xs-Professional-Compressor-Processor/dp/B004NDFRVC

[Daniel Taylor]

This is an interesting discussion.  Personally, if I wanted normalization, I'd want it in the form of an automatic volume control that made the peaks from each album (not song) the same max level.  Some songs on an album are intended to be a lower volume than the others and I'd not change that.  However, such automatic volume control would not make even the loudest songs on an album seem to have the same volume because of the compression used on some albums and not on others.  So, not only the volume would need to be adjusted, but the compression too.  Compression can be added, but can it be removed?  Do we really want all our music to sound like it's in the loudness wars competition?  Personally, I would not want a normalization scheme that involved compression.  The dynamics of a song are a significant part of its soul.

For many years, we've experienced the variable loudness problem when playing CDs in succession.  If the next CD has a noticeably different volume level, we adjust the volume control.  Now, because all the music is on the same device, we expect the device to adjust the volume automatically.  I don't think that's a realistic or reasonable expectation.  No matter what algorithm we use, the device cannot know what volume level will sound best to each of us.