Theoretically (and when things were always done as could) you are correct.
Ok, I only now see that I never put up the XX-Analysis data for the HiRes version (I just wanted to quote it all), and the HiRes version really isn't much different from my version to that aspect. The Hires has a little higher SPL (so, it's a little louder on average) and the available headrom I forgot (and I'm not at location right now), but was somewhat less and not worrying (say 31000 remapped to 16 bits. But what does that all mean ?
Ok, here's that quote after all :
I suppose you bought the same tracklisting;
The 01 you send me is not of the same length (but within a second), and the average SPL of yours is close to 6dB louder. Yours is close to the digital limit (for output peaks), while mine has loads of headroom. Notice this latter is related to the total album (well, if all is put on there with the same relative level), so t doesn't say much at all. But might you want to compare for the whole album, this is from the XXAnalysis file for mine :
MaxVolume : 32752
SPL : 1546
--------------------------------------------------------------------------------
For my 01 track :
MaxVolume : 23877
SPL : 1176
For your 01 track :
MaxVolume : 31514
SPL : 2013
My 23877 has spades of headroom (limit is 32768 as ABS() value), but it also clearly did not use all the space there is. So, less dynamical than could have been. By SPL of 1176 though, is so low, that the dynamic space is from very low up to that also low figure of the maximum level used. Compare this with Gerard's track;
Average SPL is almost twice as high (twice = 6dB), meaning that the base level is already lifted significantly. But, his also has a higher maximum level (which indeed could be about one peak only); If you plainly make percentages of this, you'll see that my version uses more of the space, hence must be more dynamical (while the S/N relation on Gerard's version theoretically will be better -> nonsense, because we would FIRST be talking about tape hiss, and that's amplified along with liftiing up the level).
So this is how it works, and the Hires (with IIRC SPL of in the 1200 or something and max level of in the 30000 range) will be the most dynamical. Well, I do not perceive that at all, which can be explained by that one peak being in there, which in both cases is just that peak which is not limited or anything and if you'd calculate the percentages I'm sure both will be rather equal. Not Gerard's version, because that *must* be compressed (that's the (IIRC) 35% vs. 45% calculation I did in that post, while also there's hardly headroom left, and *what* is there, is there for technical reaons (like the engineer didn't want to reach real maximum which is a bad thing for players because of C++ anomalies so to speak).
It is more difficult, because in either case (16 or 24 bits) the dynamical peak to peak range turns out to be the same (hey, what about that 96 vs. 144 dB (without dither) then eh ?). However, within 16 bits there are 256 times less steps for the volume to (gradually) vary, so from sample to sample it would be more rough steps. Whether this is too rough ? we may not know, because first your DAC must be able to follow those rougher steps, and usually it can't, which makes it a nice analogue "filter" (think about this !). This is also while it is so difficult to make a "better" DAC, because all what happens is that the digital steps are followed better, which may not be what you want at all. Now, assumed that this analogue filter is nicely at work, you could even say that what's normally contained in the 16 bits, just nicely stays there, and now ADDITIONAL peaks can be added in that 256 times more level available (48dB). 1812 really would fit in there (I don't know, 120-125 dB maybe ?).
But to perceive something of the music, you'd now first need to turn up the volume with 48dB more than you're used to today. And *now* you just don't want to wait for those guns anymore ...
So you see, your idea wasn't so wrong at all, and actually it's how it could have been done really. But it doesn't make much sense, because everything would break.
Also to keep in mind : where we might think that the HiRes could at least have 6dB *less* of SPL (that allowing for 6dB more dynamic space), it must have a reason. I mean, when no 96+6 dynamic space is in the recording anyway, what to do it for, and it only would make the music softer, the SNR worse, and more (digital stuff) which works out for the worse.
Disclaimer :
When I talk about these things with mastering engineers, they have complete other ideas about this all, because they don't incorporate the real merits of digital. They can't - it's not their job. SNR always has been an important thing though, and not to compress or limit is a known phenomenon / good idea. But there's also practice, and when we (these days) record in digital, how to prevent
a. the level of jumping singers etc. not to exceed digital maximum;
b. the level getting *really* low when they for one time don't jump (think of a level of 18dB less easily);
c. it not being allowed to put out an album which is waaaayy too soft.
And so limiters *are* used, and so compression *is* used. There is no other way.
Yeah, record in analogue. But mind you, attennuating that is a means of compression in my book.
It again is more complicate when we look at the filtering which happens normally in our DACs. I mean, if you think of the too rough stepping (from sample to sample and the level changes), all is totally moot when you see what that filtering makes of it. It is so completely unrecognizeable that head nor tail can be made of it. Therefore we can really wonder how important the above really is, when we first use DACs that devistate anyway.
That for many it will be so that Arc Prediction -let loose on an NOS DAC (of 24/at least 192 !)- results in out of band other anomalies is yet another thing, and that many problems are incurred for when the impulse response is able to follow those (too rough) steps - again another (nothing analoguely filters what's wrong in the base). But the latter makes this whole story of vast importance, and no mastering engineer will get *that* (hey, they don't have the DACs).
And this is still a nutshell ...
Peter
(sorry for typos; I didn't read back)
