[LiquidAcid]

Originally Posted by Xenofan 29A

Do I know the formula? No. The algorithm has not been released. If you want, you can try to judge the results against those of other similar measuring tools, and you'll get a pretty accurate picture of how it is calculated.

I doubt that anyone is going to suceed with this kind of reverse engineering. It's a lot more complicated than you might think (see below).

Originally Posted by Xenofan 29A

I know the general method by which it works, and I explained this above.

With all due respect, I don't think you have any clue about how it's computed. If you indeed do, then at least me give some pseudo-code here. This so called "explanation" is worth nothing.

Remember the thing about why a plane is flying in physics lesson? The stuff about the air traveling with different speeds above and below the airplane wing? Anyone tried to build a real plane just based on this information? Hopefully not, because you will probably kill its passengers instantly. The "explanation" is overly simplified and furthermore incomplete. It might give someone a very coarse idea about how things work, but it's worthless if you actually want to fly / build a plane / evaluate it's flying capabilities.
So if this is the kids explanation to airplanes, then your explanation Xenofan is the kids explanation for dynamic range computation.


Need to make it more clear?

Originally Posted by Xenofan 29A

It's primarily a reading of the average difference between the highest and lowest levels at a given point.

1) Level is not defined. What should that be? Sample value?
2) What kind of average? Linear? Weighted?

There are infinitely many ways to compute momentary volume/loudness levels. You just need to take a look at the RG specs about this part:
RG1 specs - loudness measurement

It's sophisticated, it's complicated and it matters how you do it. You seem to believe that the loudness level is some natural property that just comes along with the audio data. This is simply _not_ the case.
And therefore it's of utter importance to know how a tool, which you ultimately want to use to compare different data, computes it.

Swap out this part of the scanner and a track with previous result of 12 might now produce 10, while another track changes from 10 to 12. What is "better" now?

--------------------

Originally Posted by Xenofan 29A

Any natural sound has several "layers" to it that are at different volumes, so a collection of natural sounds
will have a wide range in dB in any given instant.

It's not clear if the algorithm knows of these so-called "layers". I highly doubt that, since this would involve some way to automatically "extract" single instruments from the data. And there's a limit what you can do with a FFT / DCT.

Originally Posted by Xenofan 29A

It is, on the other hand, useful for comparisons.

And that's what I doubt the most. How do you define a proper metric for comparison if you don't know what these values are anyway.

This really reminds me of all this bullshit about radiation in the media when hell was breaking loose in Fukushima. Bombardement with information about radiation in Sievert, Becquerel, Gray and what not. And everybody had their opinion on that matter, even though these units didn't tell them shit (sry for cursing). 10mSv here, 20mSv there -- people just knew "more = bad". It was just numbers for them. But everybody suddenly felt like an expert...

Originally Posted by Xenofan 29A

Certainly a far cry from "guessing stuff out of thin air".

It's unscientific in every conceivable way. I'm honestly amazed how you guys can trust something so undocumented (if it really this, maybe I just haven't found the right FAQ yet) like this.


And to prevent any misunderstandings: I'm only talking about the technical aspects of the scanner here. The interpretation itself is a totally different matter and as you pointed out numerous times Xenofan, it's best to let the ear decide in the end. And I've got no doubt that you can do this perfectly, probably better than most of the people here. I just doubt you understanding the technical subtleties that come with such a scanner.

To quote from the EBU site:

Originally Posted by EBU

EBU R128 is the result of two years of intense work by the audio experts in the EBU PLOUD Group , led by Florian Camerer (ORF).

This _is_ rocket science and not something you think up in your 5 minute coffee break. People are vastly underestimating this subject.



EDIT: I already spotted some comments on HydrogenAudio that somewhat mirror my whole critique here (Link).
Quoting from the thread:

Originally Posted by Northpack

I don't see anything professional about all of this. They don't update their website, forgetting their own announcements as you've said. They don't properly document the algorithm used by the Dynamic Range Meter - but if you actually figure it out, you'll notice that it's pretty coarse and far less apt to match subjective perception than those developed by the EBU or by David Robinson. Than there is the strange case that the foobar plugin still gives DR values hughely different from those reported by the offline DR meter (and that noone seems to care about this fundamental flaw, even 3 months after it has been reported).

OK, they gave that tool a neat design and made up a coherent CI. Maybe that's the real defition of beeing professional: to enshrine technically deficient products within a big marketing bubble, made out of fancy terms and labels. Some people would certainly agree.

Aha...


Another funny thing is that I found a claim about the scanner being open-source, which would pretty much destroy my critique here. But the source itself is nowhere to be found, so I don't really believe that.
More funny stuff: The author even mentions EBU R 128 in his ChangeLog, so probably he's aware of the issues with the scanner.