***EDIT*** OK these two samples are not good at all.
In order to jump right to the interesting uploads, go down to 2BDecided posts : http://www.hydrogenaudio.org/forums/index....ndpost&p=205696
...and following ones.
***EDIT***


Here are two samples illustrating how dynamics compression may artistically damage a song. There are already several threads about the biggest replaygain alues. This one is different. The replaygain value of the track is "only" -7.23 db, but the negative effect on the song is maximal.

They are two different performances of the same song : Qntal - Ecce Gratum

Live.mpc is taken from the Video DVD "Qntal - Live". Technically, the recording is bad, because it was directly taken from the live performance.

Studio.mpc is the same part, but from the studio album Qntal III.

When I bought the CD, I found this song very boring. For me, the worst of the album by far. I really disliked it.
But watching the DVD, the exact same song was completely different ! In fact, I ended finding it not bad at all.

Technically, there are only 3 db more dynamics in the live version. That is not much, but it seems to me that the "loudness race" has terribly damaged the studio version. Every instrument tries to make as much noise as possible.

The live sample can be listened to several times in a row... the voice emphasizes every syllab of the text.
In the studio sample, it seems to me that the emphasis has been crunched. In fact the voice seems lost in a stack of noise. I find it very unpleasant to listen to.

What do you think of them ?

Second sample

with most music i find the bootlegs/live recordings to be more interesting anyways...

it isn't specific to a particular record (i.e. more like a general rule for me) so i'm curious ... is this your argument?


later

It would have been more interesting if the two performances were the same, because here, we can argue that the live sample is more listenable because the singer and performers did a better job.
It's difficult to tell what comes from the performers, and what comes from the mixing/mastering.

It seems to me that dynamics compression was applied to the studio voice, and that it sucked all life from it. The studio sample also reminds me of what a french humorist, Coluche, once said in a sketch :
"Not everyone can be soloist !"

So my argument is very different. I was concentrating on the mixing, trying to ignore differences between the performances.

The ambience / acoustics are also significantly different on the live recording.

Too many factors come into play when comparing studio vs live recordings, so I for one do not think that you can conclude anything about the impact of the loudness race from these samples.. just my $0.02

perhaps not these samples in particular

i guess the best method for this would be to focus on material that has gone through a remaster, etc.

i.e.

Led Zeppelin Original CD Release vs Led Zeppelin Box Set Remaster Release, etc.

Pio,

i do understand some of your intentions but i guess doing live vs. studio isn't the best way to handle this situation


later

Maybe it would be easier to compare an old release vs. a new, remastered release?

Like Pink Floyd from the 1992 remasters (maybe Shine On, considered by many the best Pink Floyd remaster) - which are from before the loudness race started - vs. recent remasters.

I have a more general question concerning this:

I've heard -7db CD's that sound fairly decent (not often). But I've also heard -2 db (or whatever) CD's that sound flat. So the replaygain value is obviously not the only meaningful piece of information. Is this where the peak value comes into play? Is there some other piece of info that comes into play? Is there a layman's way to explain this to me? Am I asking a dumb question?

Aerosmith - "Greatest Hits" vs. "O Yeah! Ultimate Aerosmith Hits"

"Greatest Hits" is horribly compressed, while "O Yeah! Ultimate Aerosmith Hits" seems to have much larger dynamics on the tracks - I don't know which album was released first. Listen to Sweet Emotion on both CDs, the difference is just too large to imagine.

I'm not into that kind of music in the slighest, however, i do agree with you that the live version is more enjoyable to listen to.

not sure if this is an example of the sonic difference, but it points out the stupidity of the loudness race:

i recently borrowed a friend's vinyl edition of Beastie Boys - Ill Communication, and compared replaygain values. these cannot be completely accurate, as my cartridge, stylus and tonearm are all sub-optimal, however i took care in the recording (going so far as to record with 10% speed reduction and then re-EQ'ing and speeding up).

the song "sabotage" had these values:

CD: -7.14 dB
vinyl: -2.38 dB

now, my recording was not a quiet one - peaks were pushing -0.5dB while i recorded it, and after re-equalising with foobar's convolver and advanced limiter, it tickled fullscale all through the song (i might want to do it again to avoid this, but it really doesn't bother me for now).

aesthetic effect is difficult to observe... there's the "warmer" vinyl sound (ie. more distortion, and exaggerated bass caused by record wear and RIAA EQ curve), which could account for maybe 1 or 1.5dB of the replaygain value, but generally the sound is far more effective with the more dynamic recording. the beats kick you more when the volume goes up, the vocals aren't smothered by the bassline.

this strikes me as unbelievably stupid considering the noise floor of this LP (after meticulous cleaning, etc) was a high -41.9dB, where the noise floor of an undithered CD is -96dB.

why is it that the superior medium is being abused?

I have a strong desire to do just this with several of my discs, but I only have a single version of each unfortunately. In particular, I would be most intrigued to compare some of my older "audiophile" Mobile Fidelity Ultradisc masterings to more modern mass market remasterings. If anyone is interested I'd be more than willing to upload 30 second segments of songs from any of these discs to compare with someone elses modern remaster.

I would be especially interested in comparing segments of Rush's Moving Pictures (2112 as well) or the Blind Faith album to the recent remastered versions since these are two of the oldest Ultradiscs I own, before the advent of their ominously named "GAIN system".

Kuuenbu (edit: not Dan!) mentioned the former album as a good example of properly done, modern remastering in the loudness thread...if anyone wants to compare "Tom Sawyer" or "Limelight" or "YYZ" or "Witch Hunt" (or Camera Eye, Barchetta, Vital Signs) samples I would be extremely gratified.


Do I know what rhetorical means?

As a preventitive measure in case I get bogged down with work, here are the first thirty seconds of "Tom Sawyer" from the MFSL Ultradisc II remastering of Moving Pictures in case someone wants to upload the companion segment from the new remaster. In this case we would be potentially comparing two different mastering techniques from different years on the same source material, which might be instructive.

Track is not wavgained of course, peak level for the track was about 90% with 100% extraction quality

I don't have two different versions of the same track from CD to upload.

But I have played around with using multi-band compression and hard limiting to make some of the best recordings I own sound like some of the worst!

This is the original...

...and this is my compressed version...

I used some manual multi-band compression in Cool Edit Pro (i.e. I filtered it and then compressed each band separately). I then EQed it to make it sound closer to the original, used 6dB of hard limiting, and then tried to match the level to that of the original. The result is that the compressed version peaks at -9.5dB, while the original peaks at -0.5dB.


ReplayGain thinks the compressed version should be 2dB louder than this to match the original, but I disagree! It could be 9.5dB louder before clipping than what I've uploaded here, and certainly would be on a commercial release. At 9.5dB louder, it would just blow away the much quieter original.

Even with the levels matched, the compressed version sounds "punchier" - however, the hard limiting is overdone, so it also sounds quite distorted.

What's most marked is that it doesn't sound like a real trumpet in a real acoustic anymore! It's just noise.


I'd invite anyone with cheap speakers (or better still, both cheap and expensive speakers) to compare these two, and say what they hear.


(btw, the first second of the original track isn't transparent with lame 3.90.3 aps or lame 3.96 ps - the trumpet sounds rougher - ABX 8/8 both versions - will upload a lossless version if anyone's interested.)

Cheers,
David.

Here is an extract from the passage Ego sum abbas in Carmina Burana by Carl Orff.
It was performed by the London Symphony Orchestra, conducted by Richard Hickox.
The CD is an all digital (DDD) production

file details:
FLAC compressed
27s
1.95MB
Ripped using CDex

The extract has a wide dynamic range and clear vocal which seems to stress codecs at lower bitrates.

Luke

Ok, my samples were pathetic. Now, since people seems interested in this, here are more dynamics examples.

First, the most dynamic CD in my collection.
Edvard Grieg - Peer Gynt - I Dovregubbens Hall (Gothenburg Symphony Orchestra and Chorus - Neeme Jarvi, 1987)

Here is the original

And here is a version that I compressed myself.
Please forgive the awful artifacts, I'm not good at all in compression.


This one illustrate how and why music is compressed. If you don't pay attention to the artifacts that my process introduced, the voices are much louder and clearer than in the original.

And now, the same as above, but leveled 15 db below, so that the voices are as loud as in the original.

Load this one first in your playlist, then the original, and play them one after another.

These two illustrate the lack of punch in compressed recordings.

Finally, in addition to the above sample, I'd like to show another example. I know that I'm uploading much data. I hope that it is OK. They are not lossless, so they don't take very much space.


These two are not from the same original, nor even from the same artist, but their musical genre is similar.

This is Kovenant - Mirrors Paradise.

It is overcompressed. I'd like to turn up the volume to enjoy it, but I can't ! It makes too much noise, and has no punch at all. I find it very frustrating.

And for the comparison, this is Artrosis - Gora Przeznaczenia.

The instruments used sound similar, but this one is much more dynamic. And I can turn up the volume louder than Kovenant. It has much punch, and is very enjoyable.

Thankies, these are exactly what I was looking for in my thread. Nice work, Pio ^^

I tried listening on both PC speakers (Altec Lansing ACS48; nice, but definately far from high end) and my headphones (Beyerdynamic DT880).

But I can't say the headphones were more "revealing". Or in other words, the speakers didn't mask the flaws in the compressed sample.

Things I noticed in the compressed version:

The attacks of the trumpet are killed.

Especially the piano has more reverb. This had an interesting effect on me. Think of an ADSR envelope of a single piano note. When I was listening to the "R" part, it was so loud, that my brain expected the "A" part to be much louder than that and my ears prepared for the loud attack, but the "A" part actually wasn't that much louder. This could explain listening fatigue. Your ears are allways on the edge, ready to face a loud part. For short periods of time the ear can dampen loud sounds. Probably for no particular reason, this mechanism is also activated when yawning.

It's hard to focus attention on an individual instrument - they are all right in your face, screaming: "ME! ME! ME!"

Exactly - that's my theory - the equipment has to be complete junk for the compressed version to sound better (when level matched with the dynamic version).

You could say my attempt at compression was quite poor, but it sounds similar to many ILRs (independent local radio stations - i.e. commercial radio) in the UK.

Your idea about listener fatigue may have a lot of truth in it - I wonder how you could test it?

Cheers,
David.

here's an uncompressed (from CD) rock sample with relatively high dynamics

edit: matched the volume

and here's the compressed version (-9db hard limited)
listen especially to the drums, they loose lots of their original dynamics
it nearly sounds like an echo was added

totally overcompressed, sounds like crap to me
you can easily hear how the compressor cuts the louest parts out (the bassdrum especially)

Here's part of my Redistributor program: it'll input a wave file and show you the distribution of the pcm samples in that wave file.

It's pretty quick at processing, but wait for it to finish before you try another file; I haven't coded it so you can stop in the middle really.

The Y-axis is non-linear to boost the weaker signal at the low end. The X-axis corresponds to zero signal on the left and a full-scale signal on the right. It will automagically adjust to the scale of the sample with crappy "nearest neighbor" interpolation.

As usual, if this program steals your girlfriend, eats your dog, begins nuclear holocaust in southern Mongolia, doesn't function at all, or damages anything, I'm not responsible.

I don't have the latest remaster, but here's the 30 seconds from a very old CD version (mid 80s).

edit: second try to attach sample...

Not only easier, but more accurate. We all know that today's mastering engineers can achieve staggering RMS volumes while still preserving some semblance of the music... my experiments with multiband compression and limiting fall far short of what professionals are capable of.

Luckily, I have access to The Smashing Pumpkins' original albums, along with their newer Greatest Hits compilation. I downloaded an MPC-compressed version of the Greatest Hits CD (I own all the original albums anyway), and some of the songs on the GH album have track Replaygains of 4 or 5 dB more than the same songs on the original albums. I don't think anything was done to the songs other than some compression and equalization (i.e. no re-recorded tracks).

I'll listen to the songs tomorrow and find a couple good comparisons to highlight. Since there's no MPC file-trimmer that I know of, I'll either post uncompressed clips or recompress the GH audio (Q7-->Q7 shouldn't sound bad at all). Someone who owns the GH CD can come in and supply samples from the CD itself at a later time, if they want.

A quick listen makes me think this comparison will do a nice job of highlighting the pitfalls of overcompression.

Edit: School's just crushing me right now, so the samples will have to wait until the weekend, at the earliest.

I actually have an example of just that on my website. The songs on Depeche Mode's The Singles 86>98 are absolutely slaughtered sonically and dynamically, with "A Question of Lust" being a particularly bad offender.

Depeche Mode - "A Question of Lust" (Original version)
Depeche Mode - "A Question of Lust" (Remaster)
Depeche Mode - "A Question of Lust" (Remaster, attenuated)

EDIT: Dug 30 seconds out of my original two samples.


Actually, judging from my wavefomrm views of the tracks they don't even look like there was much, if any, compression at all; rather it seemed that Howie Weinberg simply just hard clipped everything, something that Mr. Weingberg is very ought to do as I have observed from much of his work.

We shouldn't necessarily equate hot albums with compression as sometimes engineers go the distortion route instead.

I'm not sure if that's a good example. Anyway, taken from LOTR soundtrack, not supposed to be compressed.

That's the normal CD version.
» http://www.lsi.usp.br/~agbueno/temp/Enya%2...-%20normal.flac

And this is my compressed version of it. I think this is probably how it would sound on a rock compressed CD. Sorry if the quality is not so great, it is just me playing around.
» http://www.lsi.usp.br/~agbueno/temp/Enya%2...%20boosted.flac

Now, seriously, I'll tell you guys one thing: I don't know one single person that prefers the original sound over my compressed. And I've showed it to quite a lot of people. I think now you guys know why overcompressed CDs sell so much. Just for fun, here are the screenshots of both versions

CD » http://www.lsi.usp.br/~agbueno/temp/enya-normal.png
Compressed » http://www.lsi.usp.br/~agbueno/temp/enya-boosted.png

[EDIT] I removed the samples from my server, if you want them, ask me.

It's not suprising that everybody prefers the compressed version. It's louder. In an test between two samples, people regularly prefer the louder one. That's why you need to match levels during subjective listening because otherwise the louder source will almost always sound more agreeable than the quieter source, making the comparison unreliable.

Yes, it is true. But they apply compression to make it louder, doesn't them? Which is why I haven't adjusted any levels on my samples. I've let people choose whether to use RG or not when listening.

Indeed, the puming effect in your sample seems to show that they applied a general dynamics compression on the whole song, instead or remixing it. It needed compression, that's sure, the original Black Celebration album peaks at -6 db in average ! (with a peak album at 0 db ) But the result is very unprofessional. It sounds like a newbie playing with CoolEdit. They should have only compressed the highest peaks, leaving most of the content dynamically unchanged.

Except people usually just turn down the volume when swtiching to a louder CD anyway. You could simply play the original very loud during the demonstration which will then result in them request in the boosted version being turned down when it's played.

Hocico - Signos de Aberracion - 02 Instincts of Perversion (2001)

With this mexican CD, a great step towards white noise was taken, but...

Into the darkness DVD compilation - Live shots - 01 Hocico - Instincts of Perversion (live)

Once more, the DVD video, even in MP2, provides the correct dynamics for the track.

It's obvious looking at them and listening that the CD is seriously messed up, but it actually has the lower min RMS and higher peaks, giving it, statistically, the greater dynamic range.

It is not obvious. In terms of objective measuring, these samples have almost no difference. CEP showed a only 1 or 2 dB quieter average RMS for the DVD version, while I was very surprised about wavgain reporting that the DVD version is almost 1 dB louder. These differences are nothing, from my experience I know that they can hardly be audible (regardless that it's not obvious which of the two is the quieter one) When looking at the waveform views, it also seemed to me that the DVD version has a bit more dynamics, and, subjectively, the DVD version sounded much better. Conclusion: It's the mastering/mixing that makes the difference here, not the clipression.

I've got the Greatest Hits album, too, and all the original releases (pitfall of being a completist?). I'm no professional engineer (though I'm training to be one), but from what I can tell, EQ-wise it hasn't been touched, and to make it louder, they simply took the originals and amplified them a few dB. No compression, no limiting, just amplifying. Now, naturally, some of them were already loud enough (Machina, need I say more?), but certainly Gish and Siamese Dream originally had much to boast about as far as dynamics went.

Anyway, judge for yourself with these 15-second clips from Disarm. First the Siamese Dream version (Replaygain -3.99):

EDIT: I didn't know that the HA forums saved ALL of the files I've ever uploaded and count it against me if I continue uploading other files. Lesson learned. I dumped a couple older ones. So anyway, here's the Rotten Apples (Greatest Hits) version (Replaygain -6.69):

SECOND EDIT: The replaygain values refer to the clips only, not the entire songs.

Does this mean that personal upload capacity becomes less and less with each new upload ?

Quite possible. Now all is see is "Global Space Left: 0bytes", and I can't upload anything anymore.

I have two good examples that show as the over-compression destroys the natural sound of the musical instruments.

In order to illustrate this I take 30 seconds from a song, rich in drums transients.

these 30 seconds are sounds of many keyboards and drums.
this example does not have any processing.

download here:Click to view attachment

I compressed (hard limiter) this sample (that originally had 92,4 dB SPL) up to 97,4 dB SPL, in full scale without clipping.
in order to be able ABXable the sample compressed I reduced his value up to 92,4 db SPL; that it corresponds to the value of normal_drums sample.

download here:Click to view attachment

normal_drums sample versus compressed_drums sample

results: the drums transients was totally destroyed.

if anyone of you unloads these two samples and compares in ABXtest; you could be surprised with the results...

EDIT: Figured out the index.php stuff. Gotta rename the file.

Anyway, Acid Orange Juice inspired me. I've got a great example to illustrate the loss of drum transients with overcompression. Metallica's Black Album, while not everyone's favorite, is probably the best overall sounding album they ever put out. The drums are big and authoritative, the guitars roar throughout, etc. Best of all, it's a loud-but-not-overcompressed album.

Here's a 15-sec clip from Sad But True (Replaygain of clip: -3.93)

And here's the same clip after I hard limited it. Notice how the snare sounds kinda saggy when compared to the good solid crack of the original. (Replaygain of clip: -9.22)