Upsample hi-res files?

[doublelife]

Does anybody here upsample hi-res files (eg 24/192) in XXHighEnd? Is there any good reason to routinely upsample such files (or to routinely not upsample them), or is just down to personal taste?

Thanks
Paul

[PeterSt]

Hi Paul,

I always do because it is more hard to not let it happen (requires a settings change) then let it go.
But at comparing I'd also choose for it (but the last time I did that was with Arc Prediction and not with the Custom Filter).

To be honest, I actually don't play Hires or it must pass by accidentally without me explicitly selecting it.

Regards,
Peter

[pedal]

To be honest, I actually don't play Hires or it must pass by accidentally without me explicitly selecting it.

It's an enigma to my why you don't acknowledge the benefit of 24bit resolution above 16bit; (better micro dynamics, more natural and fluid sound, less dryness/mechanical character).

[PeterSt]

Today I count 740 albums I denoted as "nice stuff". This is out of the 40K albums I have here. This is 1.85%.

I also have 1091 so-called Hires, of which about 950 will be fake (and I can't help that really). Anyway, that not counted, I would have 1.85% x 1091 = 20 Hires albums I'd theoretically like. This is not enough to be bothered.
In practice I count 2 (two) albums within my Nice Stuff.

I know, once in the x years the story recurs. Your question remains the same and my response remains the same.
My next question to you could also be the same again. But we better skip that this time.

Peter

[PeterSt]

It's an enigma to my why you don't acknowledge the benefit of 24bit resolution above 16bit;

Let's do it the other way around this time;

Yesterday I ended with The Doors at selecting the best version I had of Strange Days, which in itself happened because of a track I had of it in my Demo Gallery gave me a stuffed ears feeling. And these days this can happen because of the crazy highs output I achieve these days (all is relative, but have that not on par and stuffed ears are your unavoidable share). Btw this was a DCC Gold etc. version and I ended with a HDCD version. Miles better (superb, actually).

And so, also thinking of this topic, I searched for what I have of The Softparare, and ended up at the Perception Box (6 volumes); 24-96.

And there we go again :

- There is no power in it no matter how loud you play it;
- It is super flat;
- It's almost lacking stereo image all together;
- It does nothing - it is nothing.
I also tried without any upsampling/filtering, but that seemed to make it worse. At least somewhere underway I thought that the highs where "just highs" as in from my young years, and not cymbals.

I also compared with the version I normally play (orginal Redbook) and this Hires compresses 6dB more.

Halfway at the track The Softparade itself (one of the best pieces ever made in my view) I turned it off. This was because I was sort of scoffed with this :
"The Softparade has now begun !" ... and then some idiot who I think thinks to mimic Jesus or something is humming along almost louder as the main lyrics.
So yes, it is a remaster at the same time. In 99% of cases it is (you can see it by the compression alone), and if it is not, something else is going on (like fake/upsampled); there is just no other way you know - or go to Justin Peeper and the like.
So new stuff ? what new stuff.

The main point you can learn from this is that the lust to ever try a next will only be there when I have forgotten about this sheer debacle for the so many'st time.. I mean, I can have a 1000+ Hires albums, but why waste time on something that most probably (!!) fails anyway and again. So I don't even really try them. I have better things to do ... (play).

See you the next time !
Peter

[doublelife]

Of course it is very true that quite a bit of so-called 'hi-res' files avaiable for download are nothing of the kind, just repackaged and compressed sh***y versions of Redbook. However, that's not always the case and you can find some real gems in the 24/96, 24/192 or 24/384 filesets if you are choosy. I recommend this website, you probably have seen it already, but I always use it to try to see which version is worth purchasing:

http://dr.loudness-war.info/album/list?artist=doors&album=strange+days

Interestingly, for the specific album Peter mentions (Strange Days), the 'best' i.e. highest dynamic range, version appears to be the DVD-A from Perception box set (the version Peter did not like at all). However that rating appears to be skewed by a single track with DR=21, the others are all much lower.

[manisandher]

I'd be very careful equating DR with SQ. Sure, a DR of 6 or so suggests massive compression, from which one might surmise that the mastering may also be pretty poor.

As an example, I have two digital versions of Brubeck's 'Time Out', one with a DR of 13, the other with 12. The former sounds absolutely terrible. I mean, really flat and lifeless. The latter sounds soooooo much better. (But then again, I still prefer the sound of my 45RPM vinyl version over all the digital versions I have.)

Recording/mastering is waaaaaay more important than anything else, and I don't think this can be 'measured' in any way... other than by listening.

Mani.

[doublelife]

Mani,

Yes, agreed. I suppose I use the DR database as a first pass of which versions to avoid. As you say, a very low DR value will likely sound terrible.

Paul

[PeterSt]

Interestingly, for the specific album Peter mentions (Strange Days), the 'best' i.e. highest dynamic range, version appears to be the DVD-A from Perception box set (the version Peter did not like at all).

Haha, with all the up-and down-mixing I'm afraid something mixed out.
I talked about the Softparade being lousy from the Perception box set, but I'm afraid this is not rated at all : http://dr.loudness-war.info/album/list?artist=the+doors&album=the+soft+parade
What you anyway can see is that the normal CD hardly can be improved, if it were about DR. And for anyone's fun : when I started comparing, I did that with Wild Child (DR of 16). But mind you, when I heard it from the Box Set and then (thus) to the sense of : wait a minute, I really know this one with better quality (while we see now that the DR of the CD really is the very best). So all is relative of course, but the mind/memory is hard to fool.

Otherwise, for Strange Days, I did not listen to the Box Set version, but the HDCD version came out as the best by far. And this now is rated the worst of all (http://dr.loudness-war.info/album/list?artist=the+doors&album=strange+days).
For fun, here my (XXHighEnd's) interpretation of matters :

The (Strange Days) one which bugged me has an average SPL of 3254 (DCC Gold). The one I selected in the end reads 2348 (HDCD and this is after correction of HDCD content being 6dB less to begin with). In dB "normalisation difference" this compares as the HDCD one being 3dB less compressed
Just a minute ago I also examined the Perception Box Set version and that reeds 2272 and theoretically is a little bit better than my HDCD version (at 0.5 dB rounding both are the same).

I have no idea where my HDCD version comes from and nothing about that is on the cover. It just says "1985" which IMO is the original year of release (and HDCD did not exist I'd say).

Then I did something strange for my own fun : I got the album from Tidal (only one is in there) and saw the Release Date of 2006 (which is the date of the Perception Remasters). Side note : The Release Dates I see are different from what you see  (in XXHighEnd) because I changed a few things (now it is for real, and what you out there still see is the date Tidal obtained it (that is, often it shows that date and it is useless of course)).
The result : 2350 which is quite equal to my (2348) HDCD version.
The worst part : that one too is HDCD.

Getting the hang of things, I now checked out The Softparade, and that too is HDCD from the Box Set (the 16/44.1 one). Aha, so probably all 6 of them are. The rating is 2272 vs 2533 of the Hires one. Aha again (in this case a 1.5dB difference to the advantage of the Redbook).

Continuing ...
This is in the comment of the Strange Days DVDA DR rating :
DVDA multichannel tracks downmixed to stereo in audacity (mix&render; amplify); source files 96/24 from Perception box set (incl. the 2 bonus tracks)
As you can see, someone is actually doing this. As in "this is me, and I am going to determine the downmix settings". Here my remark "it can't be done" weighs in once more; what is/was this about ? any 6 channel recording will show 12dB too much of level to begin with after downmixing (6dB per 2 channels beyond the 2 of the target). With 6 channels we need to lose 12dB somewhere on to the (downmixed) 2 channels. Thus, someone is manipulating this, in this case in AudaCity. "Mix & Render & Amplify" my a**.
Point is : this is not official (as how I see it). Ah, see ? another fake (a whole box of them). It is fine for the DVD-A version and machine, but not for Redbook.
Btw, dropping 4 channels is another means to do it (there's a Dire Straits - Brothers in Arms version around which does this - sounds superb but is of course flawed miserably again).

Then what about my HDCD versions ?
Good question. But they sound superb (no, I must be honest, I listened to one only - Strange Days) and this is not from any 1999 HDCD studio whatever complete recordings.

On to Tidal again ... what about these Complete Studio Recordings ...
OK, found fast. Let's Prepare Strange Days (CD 02) ...
Rates 2135 ... did we have that number ? No. 2348 was by at first selected HDCD one. One problem with this : it does not utilize the full headroom. So compression is fine, but headroom is wasted (by something like 9%).
Now let's play it ...
D*mn, not HDCD.

Lastly, I did the same with the Softparade; With 1815 it rates the best and the full headroom is used. Again no HDCD.
The best ? nah ... I also have my own one (the one I talked about yesterday and which I compared with the Hires), rated at 1501 and *again* 1.5dB better. Sadly, this one too does not use the complete headroom (waste of ~10%).
Btw, yesterday I said 6dB worse, but I now see it is 4.5dB.


Guys, this may look as wasting my time. Perhaps to you it is indeed, but possibly you have no clue about to what degree "versions" matter. And trust me, the lower that SPL number (in the XXAnalysis.dat file of your album folder, after Volume Normalisation), the better the sound WILL be. Do notice that this combines with the MaxVolume number in there, which should be as high as possible and won't be more than 32768. But you can bet that if you see 32768 (or 32767) the music softclips (was clipped in order to let it fit) or was dynamically compressed otherwise (again to let it fit in the digital space available). In the above I mentioned two times a "waste", but you can bet that these two were the least manipulated and again WILL sound the best.

Regards,
Peter

PS:
I found another HDCD of The Softparade in my library. Rated at 1854 and 1.5dB better than the HDCD from Tidal. At least this one says HDCD on the cover (A Steve Hoffman one). But you know what (I never saw before) ? It says "HDCD copyright Microsoft" (2009). Oh yeah ... So MS really did something with HDCD (I just did not know that).