Dear guys,

Most of the problem we had is, unlike any other professionals, we don't have a body to set us a standards. I will take web professionals for example. They have The World Wide Web Consortium (W3C) — an international community where Member organizations, a full-time staff, and the public, work together to develop Web standards.

I think at some level, it's good to have some sort of standards and guidelines for audio engineers to build upon. Maybe take my words and spread it to everyone until it reaches Friedemann Tischmeyer, Roger Nichols, Ian Shepard, Trent Reznor, and to their peers alike. Maybe from there, maybe if my words reached them, maybe there will be a well established committee for us, for us to get a guidelines or standards to deal with such matters.

Best Regards,
Futurewine ACDC.

 

Music industry does have standards [K-standard, LU], but musicians don't obey them and producers don't obey them. Generally out of ignorance. People who are knowledgeable just roll their eyes, keep their mouth shut and do what's asked, because "the client is always right". Then we talk amongst each other and laugh at them how stupid people are to keep ruining their own work in the name of loudness which is not loudness, just ruins audio in a bad way. It's completely ridiculous.

A part of the problem is that people simply don't understand digital and even less analogue audio. They don't understand metering, what it means. They think all audio meters are the same and dBFS and dBU are the same, too. No idea what peak, true peak, RMS metering is, either. You, as a musician don't have to understand all these "technicalities", really. I give you that. But it would be enough to, for example, just put a VU-meter [Klanghelm, TBProAudio...] on the master and make sure your audio doesn't exceed 0dB on it. That's simple, isn't it? Forget digital peak meters. By using VU meter or a LUFS meter [ToneBoosters] and gain staging your tracks proprly you will never have to worry about levels ever again.

Thanks to loudness equalisation algos that puts all the tracks at the same RMS level this will slowly change, when "loudness warriors" realise how limp, soft and just poor their songs now sound in comparison to other, more dynamic and properly mastered tracks. Tracks that have been mastered using VU, K, or LU standards will sound great, on the other hand.

 

Actually, @mozee is right, there is no such thing "in real world" - that's why we call the PCM decibels dBFS.
Also note that 32-bit floating point has 24bit resolution plus 8bit worth of floating axis.

Regarding integer numbers, it's actually all the way from 32,767 to -32,768. The encoder won't put out 32,767 all the time, that would actually be a DC output offset by full scale voltage of the DAC/Amp. At worst case scenario, it'll put out square alternating between 32,767 and -32,768
It appears you have a lot of info to share, but maybe you can say it in more informative and less condescending way.

 

Last edited: Dec 11, 2017

Thank you.

 

http://dr.loudness-war.info/album/list/year/desc/1

I admit, it still looks pretty bad

 

Non existant DR is the main reason why i mostly gravitate to older releases.

Been listening to the Dennis Edwards- Dont look any further, remaster and sounds really good, well done.

 

Absolutely because it has five beats going on. If you haven't done so as of yet, take time out and check out the whole album...great stuff.

 

I'll have to respectfully disagree.
If your DAW shows 0dbFS on its master (rendering) channel, while rendering a file @ 16-bit depth ("CD quality"), one of the following is true:

1. The number 65,536 or, as you've correctly pointed out, 32,767 when the same number is interpreted as 16 bit signed integer (nice catch! see what happens when I gloss over things ) will appear in the resultant file.

2. Your DAW does some thinking for you -- normalizes your render if the result exceeds 65,536 (or, as you've correctly pointed out, 32,767 when the same number is interpreted as 16 bit signed integer -- can't be too careful, as I have been reminded just now, things like this must always be spelled out).

3. Your DAW is coded by children and/or rodents, or is broken. The "or" is inclusive, "and/or."

So noted.

Yes, the same 1111 1111 1111 1111 can be 32,767 or 65,536, never ceases to amaze.

A lesser man than I might point out that clipping is, indeed, DC voltage while it's happening (a succession of identical sample values, 65536 65536 65536 ... 65536 or, as you correctly pointed out, 32,767 when treated as a signed integer), and that pointing out that an audio signal is, indeed, an alternating current in this context is the frickin' pinnacle of pedantry, but not I. Ohhhh no. I learned my lesson!

I got so much love to give, and no one wants to take it :'(

 

Clipping, even the most harsh one, is still an alternating signal, otherwise it'd be outputted as silence in most capacitor coupled amps, or rather loud pop, followed by silence.

FS (full scale) as it's referenced to absolute maximum of LPCM waveform. dBFS values are not translatable into "real" dB, or dBA/dBm/dBW (if you prefer) without another point of reference (usually dBu/dBV or Vrms). Not sure where is the disagreement then?
I think that's what @mozee meant.

 

Last edited: Dec 11, 2017

Yes and no.
While you pretty much described the sound of clipping (and square waves), the output from a DAC is, indeed, a straight line -- DC current.

In all cap coupled/tranny amps. Though not in direct-coupled ones. But you're right, there are some lofi amps out there, and we can always make them worse if we tried.

If you wish to discuss hardware, I don't mind, but it has nothing to do with the [seemingly simple & self-evident] notion I'm having to defend here, namely that [over]compressed things are, indeed, loud; so much & so unpleasantly so [vs. uncompressed material] that streaming services felt the need to introduce normalization as a countermeasure.

Do you feel that YT introduced normalization to make compressed stuff sound LOUDER than uncompressed material?

I'm unsure which part of
OdBFS, when rendering a 16-bit int file, translates to the highest number that fits into 16 bits (or highest positive/lowest negative number, if you prefer). Exceeding 0dBFS in this scenario will cause clipping.
you wish me to clarify.

0dB is as real (or unreal) as the ceiling when you jump (or, rather, the floor & ceiling, since you prefer signed & exactitude) -- trying to jump any higher than the ceiling will give you a headache (cause [potentially unwanted] clipping)

Edit: If anyone cares about YT normalization, right-click => Stats for nerds =>

 

Last edited: Dec 11, 2017