I know this question is kinda dumb. But it not meant for a big discussion thread. A simple yes or no is enough and the moderators can close this thread.

My question is this: Is ReplayGain info a reasonable information on compressed CDs? I mean, can I use RGs gain values to say "this CD is very compressed, this other CD is not"?

Sorry if that's not the best way to ask these kind of questions. Please PM me if that's the case.

Thanks a lot.

This is going to spawn some discussion and I'm glad this question/discussion is coming up.

No, it's not in my oppinion. Some music just is really fuckin loud, which will lead to high ReplayGain values. Of course it's an indication and this indication will be right most of the time, but after all it's more important what it actually sounds like than what some algorithm says.

I have some old Rolling Stones discs with a ReplayGain value of -8 and -9, but there is no dynamic compression. Like dev0 said, it's just loud.

No, some genres tend to have higher RG values, compressed or not, for example, compare an old Death metal song with a pianissimo move. The metal tune will have a higher RG adjustment, regardless of compression.

So, what would you guys suggest as nice indication of a CD's compression? A screenshot of its wave form?

Adobe Audition has a nice tool for analysing the dynamic range (/analyse/statistics)

Exemple:

http://membres.lycos.fr/guruboolez/AUDIO/l...to_original.png

You can see here that this song have a wide dynamic range (from -70 to -15 dB = 55 dB).

the way it sounds. over-compressed CDs sound like shit. they sound constipated, sterile.

RG takes RMS values from the entire song and uses the 5th percentile as the replaygain value. (Correct me if I'm wrong, that's just what I remember) This means if a song has an extremely loud part that takes up at least 5% of the song, even if the rest of the song is K-20, the RG value will still end up as high as the really loud part.

However, in the production of most modern genres, levels are usually kept consistent throughout an entire song, so a high replaygain value will indicate a severely compressed song. A low replayain value, however, can mean a compressed song that was attenuated. If a mix CD was burned with replaygain and someone rips it, the rips will have a replaygain of 0, even if the original tracks were pushing -8 RMS the whole time.

in my experience it goes like this:
if an album has a high replaygain value, and the separate track rg values are similar (no differences over say 1db), there is high possibility that it's overcompressed
if an album has a high replaygain value (-7, -9) but the tracks have track rg values like one track -3, one -10, one +1 etc, it's very unprobable that the album was squashed in mastering.

if i am really wrong, please correct me, but this is the impression i've got from my collection.

The Rolling Stones disc I mentioned has an album gain of -8 and almost all tracks have a track gain of -7 to -9. Anyways, like I said above, the disc was not compressed.

ReplayGain isn't entirely reliable because the scores are also based on an equal loudness contour. Basically, ReplayGain reports how loud the album SOUNDS, not how loud it's actually cut at. A CD dominant in upper-mid frequences can have values of up to -9 and still sound dynamic, wheras a bass-heavy CD with those kinds of values will most likely be very squashed since you'd have to get a higher raw RMS average to get the same ReplayGain score.

This is outside my area of expertise but I think that (almost) any recording of (almost) any "real" signal picked up by a microphone will have a much larger peak to average ratio than is found on typical pop CDs.

ReplayGain isn't designed to measure this peak to average ratio, though various assumptions/coincidences/common practices mean that a very negative ReplayGain value typically indicates a more squashed disc.


If the peak to average ratio is small, then either the original source didn't have a very large dynamic range, or compression has been used.

Overdriven guitar amps have very little dynamic range. They're always saturated/limited, which itself is a form of compression. So, the sound is compressed, but it's compressed at source. If they dominate the mix, then the track may have a small peak to average ratio without the use of subsequent compression. (I give this as an example, not something that happens in practice!)

However, vocals and drum kits (unless the mics are driven into saturation) can both have huge peak to average ratios. If these are prominent in a mix, but the resulting CDs only have a tiny peak to average ratio, then compression has been used.


I'm not claiming that's necessarily a bad thing, but I am claiming that most CDs which "are just loud" (dev0 and Sebastian Mares have suggested that's because the music is loud) are in truth only "loud" because they've been compressed. If the peaks weren't squashed, the CDs would sound comparatively quiet (unless you turned the volume up, in which case they would seriously kick!).

There are exceptions, and I think dev0 has suggested some genuine ones in other threads, but the Rolling Stones aren't one of them! The analogue tape recording itself will have introduced "tape compression", and I would guess there were several other intentional compression stages along the way.

Cheers,
David.

Another important point was made by dev0 and me:

http://www.hydrogenaudio.org/forums/index....ndpost&p=212746

Compression isn't = compression. Some do a "better" job at it than others. So, it can happen that you do have two albums, both same genre, same band, both compressed - but the first one sounds "okay", while the second one sucks!

This is just another reason why no matter what method is used to detect a compressed CD, it is always just an indication of the quality of the CD. Your ears are the final judge.

For listening, it doesn't matter if a cd is compressed or not - it only matters how it SOUNDS. While aggressive compression in the mastering-process tends to degrade the sound, the result can sometimes be less worse than one would think.

- Lyx

Well said. Both Pink Floyd's Dark Side of the Moon and The Beatles' Sgt. Pepper's Lonely Hearts Club Band use compression (not a very high amount though). Both are regarded as two of rock/pop's better recordings available.

Well, the Dark Side Of The Moon SACD sounds pretty bad to my ears. I am waiting for a CD pressed in the 80s to do an ABX test.

As you know I have the MFSL Dark Side of the Moon. If it interests you, I have access to the first CD pressing: Capitol CDP 7 46001 2 of Dark Side of the Moon.

Humm, very interesting. Although I couldn't see your picture I can have an idea this would be a nice way to measure DRC. Then this CD isn't overcompressed, right?

I had this question basically because I wanted to know whether a CD was compressed or not. Heavy Metal CDs, for example, are really hard to tell due to their nature, aren't them? Still though, I was wondering something could tell me if it could be a little more dynamic, that's the reason why I was asking this.

Here are the statistics of a song by Opeth called Bleak, from their album Blackwater Park (2001). This is a Sweedish death metal band. This is what horribly overcompressed and very bad dynamic range look like:

Opeth - Bleak_Stats

Here's a pic of the wav:

Opeth - Bleak_wav

Damn! That wave screenshot does look horrible.

About the histogram, you mean bad because most samples are concentrated in a very low dB? Like around -10?

Having the peak at (high) -10db just shows high overall loudness, concerning compression the fact that loudness isn't spread over a greater dynamic range is more relevant. The narrow base of that graph is an indication of dynamic compression.

A capture of a waveform of a song that's 9 minutes long, zoomed out all the way, in a rather small display nonetheless. How very reliable!

Here's a screenshot of "Black No. 1" from Type O Negative's Bloody Kisses album, released in 1993:



An album sounding like that, released in 1993? That can't be right. But wait! Here's a screenshot of the same song, zoomed in to an approximately 60 second area:



Waveform editors like CoolEdit and SoundForge view waveforms on a highest sample value-per pixel basis. Which means that ANYTHING will look like a brick if you zoom out enough.

Blackwater Park does sound like crap, though. But I found this out through listening and careful analysis, not looking at a full zoom-out of a 9-minute waveform.

Humm, that's interesting. I've got to install this program. Thanks a lot guys.

...


Here's a zoomed in pic of the wav:

BleakZoomed

I was trying to show that this album's material was mastered so loud and compressed that the whole song has little dynamic range, but alas, zoomed in it is still a brick.