Valuable Resource

[Jud]

Regarding all the discussions about the sound of RedBook vs. Hi-Res: Have a look at DR (Dynamic Range) Database, which provides dynamic range numbers for tracks/albums through a Foobar plugin.  Look particularly at the recordings for a given artist in date order, to see what the Loudness Wars have done over the years.

For instance, Peter likes to point to Rolling Stones CDs as exemplars of good RedBook sound.  Here's a URL that should take you to a page with the Stones' recordings in date order:

http://www.dr.loudness-war.info/index.php?search_artist=rolling+stones&sort=year&order=asc

Just look at the colors - should tell you all you need to know about the quality of the recordings.  Or here's R.E.M., even more dramatic if anything:

http://www.dr.loudness-war.info/index.php?search_artist=R.E.M.&sort=year&order=asc

So no real puzzle about the sound of hi-res downloads - they're being sourced from the New! Improved! reissues the labels are doing, with all the compression the market supposedly demands these days.  Thus we are able to hear in pristine hi-res the squashed dynamics and consequent loss of detail and emotion in the reissues.

[JohanZ]

DR (Dynamic Range) Database, which provides dynamic range numbers for tracks/albums through a Foobar plugin.

Hi Peter,
Last week i found for the first time this Dynamic Range information on www. What do you think about adopting this DR-meter / functionality in XXHE?
Regards johan

[PeterSt]

Johan,

Just apply the Normalized Volume and look what the net volume is going to be. The lower the more the compresion ...

FYI: I use a Reference Album with a result of 4914 SPL; it is my experience that this SPL number gives the "fairly average" hence input and (Normalized) output volume are about the same for these "fairly average" compressed albums. In relation to these the better albums albums are 10-12 dB less compressed (on average !).

Judging like this is not always easy; for instance for classical it won't work much (too much variation). Drums only - same story.

I'm sure the above is tough to understand. It needs some experience ...

Peter

[Jud]

Peter -

You may be interested in the album with the second-widest dynamic range in the entire database:

http://www.dr.loudness-war.info/index.php?sort=dr&order=desc

[PeterSt]

You mean Tord Gustavson Jud ?
Haha, welll, I guess you knew that already from me, right ?

Not that I knew about this superb dynamic range, but most often that comes together with "good recordings" ... obviously.

But watch Charly Antolini who also is high-ranked. That's drums. And so I guess (besides I just talked about that) that it is difficult to capture the *real* DR.

Interesting to see is Crime of The Century. I *almost* mentioned it in my previous post as *the* example of something to count out. So notice please :
This album has a digital headroom of more than 6dB. So, totally wasted space and a poor (or fast) job of the engineer involved. The price ? it's, sadly, one of the most poor sounding albums I own - on CD that is !. This, while it is one of the best musically. Such a waste.
The other day I found a "vinyl rip" of it and finally I heard some power coming from it. Sadly it's chaos just the same which is how vinyl comes across these days. Notice that this is "NOS1" comparison which is not totally honest to others.

So guys, besides the poor Crime of The Century example, notice that Foobar's Dynamic Range can go sky high by a just one sample which maxes way above the average. It tells nothing. Well, sometimes it does not. Same with how to observe it through XXHighEnd although there the maxing out of one sample won't do much (it's just average SPL in there - still more reliable in my view, because with the experience it really tells something).
Btw, look at your Analysis.dat file in the album folder, once you apply the Normalized Volume. So, there you can see the digital headroom as well. Half of 32767 would be 6dB ...
But also : see 32767 in there and be sure the album has been compressed. Combine with the Average SPL and know it all ...

Regards and thanks Jud,
Peter

[Robert]

A very interesting site this one. After recently purchasing Supertramp on Blueray live and watching I was inspired to repurchase Crime of the Century. I'd only ever owned the vinyl which I sold years ago after giving up on vinyl.

I ended up downloading the MFSL version taken from the gold CD of Crime of the Century. It certainly sounds great and is the only remaster that is ok so far of this recording.

[Jud]

Though it's fun to see who's at the top (and bottom!) of the dynamic range numbers in the DR Database, and very generally I recognize some artists at the top end whose recordings have good reputations for sound quality (Tord Gustavsen, Rickie Lee Jones, the late Michael Hedges), I think comparisons among artists are subject to all the problems Peter has pointed out.

Where I think the database may be of more value is comparing between recordings/masterings of the same album.  There, the consistent march over the last 30 years (or more, since there is vinyl in the database also) to squeeze everything toward the top of the loudness scale is clear.

[PeterSt]

It looks like in 1994 something new has been invented ...

[JohanZ]

FYI: I use a Reference Album with a result of 4914 SPL

The reference album i use has a vaulue of something like 1000 SPL. Do I make a mistake somewhere? What is the title of your reference Album?

Isn't it possible to use in XXHE the sliding scale from 1 to 20 from the Dynamic Range Database? (http://www.dr.loudness-war.info/)  To make comparisons easier.

http://www.hydrogenaudio.org/forums/index.php?showtopic=88373

[Jud]

It looks like in 1994 something new has been invented ...

Danzig happened!    Hee hee, they apparently slot albums from lowest to highest DR range within each year when putting them in date order.  So there was good stuff (at least from the DR Database viewpoint) in 1994 as well:

http://www.dr.loudness-war.info/index.php?sort=year&order=asc&page=87

(Some surprises in there - Jack Bruce, Kool and the Gang, Burning Spear, Prince....)

[Jud]

Isn't it possible to use in XXHE the sliding scale from 1 to 20 from the Dynamic Range Database? (http://www.dr.loudness-war.info/)  To make comparisons easier.

This I believe is the same DR "meter" that has been configured for Foobar as a plugin:

http://www.pleasurizemusic.com/ja/ja/download#menu1

[PeterSt]

Isn't it possible to use in XXHE the sliding scale from 1 to 20 from the Dynamic Range Database? (http://www.dr.loudness-war.info/)  To make comparisons easier.

Johan,

Maybe this is not a bad idea. The data is all in there (also per track if you Normalize the albums per track) and all it takes is some "remapping".
In the mean time I would like to improve on it. So, right now it will work the same (or very similar otherwise) as DR does, but this is a kind of useless at times. So, when one drum hit in a track reaches the maximum value while the remainder of the album has peaks on half of the headroom as their max, this should not imply the 6dB extra which would come out in this case. So, the 6dB will be true all right, but it's only for this one drum hit. What to do with that. So, this should be counted out. It would imply few compression though, so that's a figure by itself.

Also, supposed there's a choir the whole track. That would imply a dynamic range of close to nothing if it's only singing at the same level all the time. And that can happen. So, a poor figure while there's nothing wrong with it ?
So this requires the "lowest dips" to be incorporated into the equation.

Peter

[PeterSt]

For those who read it already : I had two posts in here about Crime of The Century Analysis. But something was wrong in the data I used for that, so I have to re-do it ...

Peter

[PeterSt]

I am sorry to say it, but DR is completely worthless. Look below. I squeezed out all the dynamics in the track; only at -21.5dB (where only close 1/3 of the original Dynamic Range is left) the original figure of 16 dropped to 15. This, while the 16 already could not be true in the first place IMO.
(this is the original Crime of The Century track).

I can't tell where DR goes off, but generally the more *headroom* left (like it is here 2/3 of the total space) the better it reports. So, possibly the designer thought that this was key to a good recording hence good Dynamic Range, but it doesn't make any sense.

I know, I am more often the bringer of bad messages; this is another one.

Peter

[PeterSt]

And since I made this screenshot anyway yesterday, plus by now I listened to it :

This is the original (MFSL is exactly the same) at the left, and the 2002 remaster at the right. This remaster is not compressed at all - only limited at smart points (to the benefit of the remainder of the track). This one (finally) sounds good to me; the original just lacks resolution which is obvious at utilizing close to half of the digital headroom only.

DR reports 10 for this track. That's just wrong ...


PS: PeterSt judges a remaster as better than the original ? hey, that's news !

[Robert]

I did wonder about all of this when I checked out Led Zeppelin remasters which didn't do very well on the DR site.

I owned a German pressing of LZ 1 CD and when the remasters came along in the 90's I got hold of a German pressing and after extensive listening comparing. I felt the remaster was definitely better.

I mention German pressings as these were always superior to our local pressings we got mostly made in Australia.

Regarding Crime of the century I only ever owned the vinyl not the original CD so was not able to compare original with the MFSL download I currently have. I was put off downloading the remastered version due to adverse comment on download sites re the so called 'loudness wars'.

I thought 'loudness war' was created for radio and small sound system playback Ipods etc to attempt to create a big open dynamic sound. Its practised in radio stations and studio's. 

I have always been wary of remasters being better and listened before buying. Although, I so often like the music, I would buy it anyway even if its only marginally better.

I also note no Golden Earring albums have been tested for DR.

[Jud]

I think it still pays to be careful.  As you said when you were thinking of how to implement something like a dynamic range measurement in XXHE earlier in the thread, it isn't necessarily as simple as determining what the difference is between the very loudest and very softest milliseconds of the track.  Headroom and lack of it, leading to clipping or limiting of the widest swings, is a reasonable factor to consider. 

On this Supertramp recording it may give anomalous results.  But on the R.E.M. recording I bought recently ("Part Lies..."), it's quite evident that it's just been made too damned loud.  And the highly touted Nirvana remasters turn out to be compressed as well, for me removing much of the drama of the groundbreaking "Nevermind" album.  The database shows what I hear for these recordings.  It's one factor that may be considered, and it's a factor that in some cases at least can be considered (along with other information) *before* purchase.  Not that it can't be or shouldn't be improved, though.

[PeterSt]

[brainstorm !]

I'll try to explain a few things :

I am not sure whether it is really a known phenomenon, but at least I recognized quite some years back that when we have a CD which doesn't use the headroom by far, it sounds more dynamical. More open. More air. More depth. More everything. I reported so on this forum as well (somewhere in between and not much as an explicit subject).
This is about the headroom utilized at 66% or so (EAC reports on this during ripping, so you might recognize it).

When you look at this at first glance, this makes no sense. So, less headroom utilized means less dynamical range in order. Notice though that this 66% would imply something like 4.5dB only. So, this is not *that* much difference and therefore fairly harmless. Still 4.5dB less, but harmless in my view. BUT :

It would kind of guarantee that no compression has been in order. How ? well, if I were this mixing engineer and finalize the album or track, and I would use compression, then I would do that with as much speace as possible. So, 100%. Not 66%. That would be the stupiest thing to do. So ... when we would take it as a "rule" that an album which doesn't utitlize the 100% but far under that (which would be this 66%) and put out a "DR meter" figure on that, the 66% would receive a say, 20. Now, to me it looks like DR operates on just that, and nothing more of intelligence. But now think :

When the recording is calibrated precisely and no single peak extends the 100% but just under that (impossible for a live recording) then the DR should be higher than the same recording reasing 66% only. But DR doesn't accept this "rule" and the more of the headroom is used, the worse it's figure is going to be. This, while it should be better.
My little Crime of The Century test emphasizes it; The at first utilized Dynamic Range was squeezed and squeezed (just by attenuating digitally, loosing more and more bits) but the DR figure remained. Only when almost nothing was left (see picture from before previous post) it dropped by one number.
And so I say "worthless".

A remaster like that 2002 from Crime of The Century : the level is taken up so the theoretically available Dynamic Range is utilized better, but it gets even more better when the handful of occurring peaks above the 100% are limited (read : almost cut-off hard) so the remainder has more dynamic space. This is not bad; only those handful of peaks are out and you may never notice it; and when you do, it's still about those handful of peaks only.

A remaster like "The best Disco Ever" (just making up one) : The level is taken up way more, and now each second will contain peaks above the 100%. This can't be limited anymore because it will be audible as distortion. So now it's "compressed". The louder the level the more it's compressed and this goes by various many settable parameters. It changes the music and what happened to the wave - which looks flattened - happens to the sound just the same (sound flat). Here, really the Dynamic Range has been narrowed, but now by different means than not utilizing the space;

In the case the original take utilized 66% it is not said that you will perceive that roughly 96 - (96 - 4.5) = 91.5dB minus noise of 30dB = 61.5dB. It will when it's a drum track (from silence to max-hit) but it won't when it's this choir singing nicely all the time; the choir may sing at 30dB above the noise anyway. So, perceived dynamic range of the choir is 31.5dB (61.5 - 30).

When this is carefully recorded, 4.5dB is added and peaks are just at 100% (not over it). Choir is now 31.5dB + 4.5dB = 35dB of dynamic range.
Drums would be 61.5dB + 4.5dB = 66dB of dynamic range. Notice that both have an absolute value of 66dB max, but drums start more down because they start in silence.

Now we make Disco of the combination of choir and drums;
Drums don't start in silence because the choir sings at 30dB (above noise) to begin with (softest). Besides, drums are way louder and should sound louder. So say that the silence level of the drums now is to be the level of the choir singing the softest. So, dynamic range of the choir was 35 and dynamic range of the drums was 66;
35 + 66 = 101, but technically there was also the noise of 30, thus 131.

131 does not fit in the 96dB which is technically available in digital, so something has to happen.

Did you follow so far ? the latter is how I described the recording process of today, which is the same as "remastering". It starts upside down. And earlier ? earlier, virtually (!) the volume was taken down with 35dB (131 - 96) and the fun is : now the choir has gone. Virtually, because when the recording was analogue, all what happened is that the drums were louder for excursions - hence dynamic range was larger. Now, in digital (16 bits that is) this can't be. So, solution ?

Cut off the tops ! haha
hmm

So, the top parts are a kind of gently compressed, but mind you, in my example (for what that's worth) with 35 dB. The gentleness causes much more to be compressed because otherwise it would be cut-offs. And so everything is squeezed and the more loud the more. This makes the choir still to be audible, but the dynamics of it all to vanish. And, overall the sound (SPL) becomes larger.
The perceived (but it's literal) dynamic range has become less.

DR reports on this.

When I lower the utilized dynamic range like I did it with Crime of The Century (remember, just attenuate it), that perceived dynamic range can sort of stay the same. All the lost bits were in the noise anyway (that first 30dB of acoustic noise) and maybe not so much has been lost.
Hasn't it ?

It has, because when I expand that later (just louder volume) the resolution in the level(s) has gone. Before (in the case of Crime of The Century) some 27000 decimal steps were available (so, 27000 utilized of the 32767 available) while now there's only 3000 or so steps available. It is clear that this implies a huge roughness (and sheer distortion), never mind the theoretical dynamic range still being the same (say, when it is remapped to the full 100% again).

Notice that this Crime of The Century already exhibits this distortion (to me) with this 27000 utilized.

What DR does, is holding the RMS value (which is the average level) against the peaks. The difference of that (with some proper math) gives the dynamic range (like I tried to describe with the choir singing at softest). Should be around 0 for a choir because no peaks are there. Anyway, when making Disco out of it, peaks will be there because of the drums;
RMS will be lower when attenuated like I did, but drums will peak as much above the RMS value as before. Dynamic range stage the same for that (and now I think of it, lost bits at the bottom (is lost lowest levels) will hardly matter -> only contributes a little to RMS which goes up now).

My conclusion :
DR may show the dynamic range all right. But it says nothing because :

1. It depends on the music itself (drums-only will show best figures);
2. It is wrong to hold the RMS value against the peaks without incorporating the absolute levels.

Ad 2 (a)
When this is done, a subjective quality factor could be incorporated. This should rate my attenuated Crime of The Century to 0. Ok, 1 for the trouble.

Ad 2 (b)
This is more complex than it seems at first glance, because when on one track is a choir and on the other is those drums, the choir can only ("over"-)improve on resolution when it's made (recorded) louder. This is not the intention and so maximum levels are just OK here. Now, at this time I really would not know how to discover whether it's just a poorly utilized headroom (implying unnecessary level distortion plus unnesessary noise (because the volume has to be cranked up)) or that it's just an in-balance (choir) level compared to the next track which is drums.

Sorry for the brainstorm,
Peter


PS: I should add explicitly that although XXHighEnd doesn't provide any DR number, it's problems actually come down to the same. Regarding this I should notice that I'm not bashing on DR, but it may have looked like it. But also : in my posts which I scratched again because they were no good, I already said so (XXHE being as "bad") but this context had disappeared of course.

[Jud]

Sorry for the brainstorm

Not at all, it was terrific stuff, very understandable and informative.

But (always one of those!) I think, even with the very evident problems with this DR measurement, there are circumstances where it is informative. 

The one I always come back to is Nirvana.  If you recall their biggest hit, "Smells Like Teen Spirit," it juxtaposes quiet sections with Cobain singing fairly softly, playing individual notes on the guitar slowly and rather softly with a bit of vibrato, with sudden transitions into Cobain almost screaming (still in tune, though), power chord guitar, bass and drums wailing away at top volume.  Those transitions give this song and many other Nirvana songs a lot of their drama.

What the remaster did is raise the volume level of those relatively quiet passages.  The transitions aren't as big a change, causing the song to lose a great deal: there you are anticipating the big entry of the screaming vocals, guitar, bass and drums, but the entry is not so big any more.  So when DR Database shows the 1991 CDs with a DR of 11 or 12, while the "Super Deluxe" 2011 version and HDT download have a DR of 7, that's what's happening. 

I do understand with the DR methodology there will be situations like Crime of the Century (which I'm very familiar with - one year in college it was a couple of my housemates' favorite record) where the initial recording was in effect too quiet.  But that's not what is happening with albums like Nevermind.

So is there any way to say whether the Crime of the Century scenario or the Nevermind scenario applies in a given situation?  Perhaps there actually may be.

One reason a recording may have been made without utilizing the available headroom is because of the physical limitations of vinyl.  It was quite normal before the advent of CDs to limit dynamic range to what the vinyl/stylus system would track at, and what the possible groove width was, for the loudest bass notes of a given track.  (Something I noticed on the Beatles remasters is that McCartney's bass is definitely louder.)  So there we may have the Crime of the Century scenario, where the CD format has more available headroom.  But where previous CD issues have substantially higher DR numbers, especially where (1) the earlier DR numbers are themselves middling, and/or (2) the recording was first issued on CD, then I think it is quite likely we have the Nevermind scenario, where it is not the limitations of the LP that dictate, but the fact that even the softest parts of the song must be made loud enough to be clearly audible over standard issue iPod headphones on a crowded city street or through a car radio.