Hello, I have downloaded The Last Samurai OST on the internet and then scanned it with replaygain to find an album gain of -2.32dB (and many tracks are near 0dB modification). This is really surprising, I decided to buy the CD as it was very good, I ripped it (as the person that ripped it may have normalised it or anything) and the cd masterization has been properly made!

Feel free to move the thread to proper forum.
Hans Zimmer - The Last Samurai - 01 - A Way of Life
replaygain_track_gain = +2.90 dB
replaygain_track_peak = 0.753692
replaygain_album_gain = -2.32 dB
replaygain_album_peak = 0.988647

Hans Zimmer - The Last Samurai - 02 - Spectres in the Fog
replaygain_track_gain = -1.03 dB
replaygain_track_peak = 0.988586
replaygain_album_gain = -2.32 dB
replaygain_album_peak = 0.988647

Hans Zimmer - The Last Samurai - 03 - Taken
replaygain_track_gain = -3.55 dB
replaygain_track_peak = 0.988586
replaygain_album_gain = -2.32 dB
replaygain_album_peak = 0.988647

Hans Zimmer - The Last Samurai - 04 - A Hard Teacher
replaygain_track_gain = +6.37 dB
replaygain_track_peak = 0.453857
replaygain_album_gain = -2.32 dB
replaygain_album_peak = 0.988647

Hans Zimmer - The Last Samurai - 05 - To Know My Enemy
replaygain_track_gain = -0.11 dB
replaygain_track_peak = 0.988586
replaygain_album_gain = -2.32 dB
replaygain_album_peak = 0.988647

Hans Zimmer - The Last Samurai - 06 - Idyll's End
replaygain_track_gain = -0.79 dB
replaygain_track_peak = 0.988586
replaygain_album_gain = -2.32 dB
replaygain_album_peak = 0.988647

Hans Zimmer - The Last Samurai - 07 - Safe Passage
replaygain_track_gain = -1.90 dB
replaygain_track_peak = 0.988586
replaygain_album_gain = -2.32 dB
replaygain_album_peak = 0.988647

Hans Zimmer - The Last Samurai - 08 - Ronin
replaygain_track_gain = -4.92 dB
replaygain_track_peak = 0.988586
replaygain_album_gain = -2.32 dB
replaygain_album_peak = 0.988647

Hans Zimmer - The Last Samurai - 09 - Red Warrior
replaygain_track_gain = -6.92 dB
replaygain_track_peak = 0.988647
replaygain_album_gain = -2.32 dB
replaygain_album_peak = 0.988647

Hans Zimmer - The Last Samurai - 10 - The Way of the Sword
replaygain_track_gain = -3.26 dB
replaygain_track_peak = 0.988647
replaygain_album_gain = -2.32 dB
replaygain_album_peak = 0.988647

Hans Zimmer - The Last Samurai - 11 - A Small Measure of Peace
replaygain_track_gain = -1.98 dB
replaygain_track_peak = 0.988586
replaygain_album_gain = -2.32 dB
replaygain_album_peak = 0.988647

Sorry for duplicated data (RG albumgain).
Mathieu

I dunno... I don't come across a lot of orchestral scores that are "hot mastered".

Hmmm just listen to "Songs for the Deaf" by Queens of the stone age and you will see what I mean (album gain of -10.7dB).

Yeah but Queens of the Stone Age is rock, not orchestral. Generally music that will get radio airplay is hot mastered meaning the music is compressed to death to get it as loud as possible. Orchestrial and classical music aren't usually hot mastered because the majority of people that listen to that music want dynamic range and record labels don't see a need to make it as loud as possible. Record labels got it in their heads that louder is better and the louder music is the more people will notice it on the radio over other music which is total crap in my opinion. But whatever, until they get it out of their heads we're stuck with heavily compressed rock and pop music until we stop buying it.

also, sountracks are mixed to the movie industry's strict mastering practices. (-31db dialog level).

this means a person talking normally would be at -31dB, leaving plenty of room for explosions and orchestral crescendos.

it's good to see they didn't play with it much when they put it on CD though (-2dB says "mild compression" to me)

Don't forget that ReplayGain is not (and was not intended to be) an universal indicator of good/bad mastering.
Some albums are just loud and some are not. Of course it helps detecting hot-mastered material, but I've heard excellently mastered albums with over -9db album gain.

Teqnilogik, the even crazier thing about the radio is that radio stations have huge amouts of compression on their output so that can can sound as loud as possible without over-modulalting the transmitter. So it makes no difference if the CD track was compressed to death or has the dynamic range intact they all sound the same on the radio.

One radio station here in New Zealand (The Rock) even has obvious pumping and breathing. (level compression artifacts)

A clever trick used in the old days (the 80's) by the production team of Stock, Aitken, and Waterman was to mix tracks with less bass so that the radio station compressors wouldn't reduce the gain as much and therefore the track would sound louder than the others. It's unlikey to work today because radio stations have much more sofisticated processing available to ruin musicality (i mean compress the signal).

-Iain

I think most actual movie sound tracks are mixed about 4dB louder than this, and brought down by 4dB by the AC-3 (Dolby Digital) encoding and decoding. In the AC-3 encoder, there's a parameter you can set to adjust the loudness to match the reference -31dB dialogue level - it works rather like the ReplayGain values.

However, there's no reason for this level to be maintained on the soundtrack CD. In fact the ~0dB ReplayGain on this discs suggests 89dB average loudness on a calibrated system. This means it's (very approximately) a K-14 recording, whereas I believe a typical movie soundtrack would be approximately K-20.

http://www.digido.com/portal/pmodule_id=11...dder_page_id=59

Cheers,
David.

that probably makes sense.

in SoftEncode, the default Dailog Normalization level is actually -27.

unfortunately, most if not all DVD authors don't actually know what this parameter is for - they leave it on default, and if the result is annoyingly pumpy, they simply turn DRC off.

it's hella annoying to watch Fight Club - the DRC is hard-coded into the sound, and can't be turned off in the DVD player. so when Jack is talking at the start, and the camera whooshes down to the basement, there's some serious pumpage going on. it brings a tear to the eye.

Songs for the Deaf has a hidden track that's orchestral. Replaygain -8.5 dB. Sounds a lot better than the rest of the album though. =D

I have heard a lot of tracks that *sound* good at -9 dB replaygain, but the difference between those tracks and quieter uncompressed tracks is something you can feel. When the transients are left intact, everything sounds punchier, and you can even feel the vibration of the drums without a subwoofer. When I listen to tracks from the 80s in my headphones, I can feel a kick in my head with the bass drum that you just don't get from modern albums.

On loud and compressed albums it becomes harder to hear the drums actually. Everything mastered is at a more similar volume level and drum dynamics seem to disapear. You may get more "kick" but personally, I'd like to be able to hear drums...

You seem to have missed what I said. I said that you feel the kick more out of the quieter uncompressed tracks.

The fun part is that people buy CD players and headphones with bass boost functions to feel the bass more. They wouldn't have to do that if the CD was mastered properly in the first place.

I like to see that nearly every week there's a discussion about loudness war here on HA. That is appropriate to the importance of this issue. Hopefully many people that still don't care about loudness race will read this.

soundtrack CDs: Most of them seem to take also part in the loudness war. I remember at least some soundtracks that were dynamically compressed/limited (in a still acceptable way).

I fully agree. This becomes very noticeable if you listen to a -9db track and then switch to say a K14 track. The quiter one has much more punch and grace while the louder one gives a feeling of 'looking through glass' -It might be mastered properly and sound good, but all the instruments are on the same level in the foreground, all in a mush they sound weaker.

Its hard to put into words, can't really abx it.. In the last 12 years I've grown attached to albums that sounded really good (sound wise). Back then I had no internet, didn't know HA, replaygain or the loudnessrace. Recently I analyzed the gain of those albums and they fall roughly around K14.

Speaking of rock, get a copy of 'jump' by Van Halen, replaygain it together with a compressed modern track. Turn replaygain on, play the compressed one then play 'jump'

Oh yeah. Maybe I should read posts when I'm more coherent. To much exam studying lately...

Maybe. Bass boost is also useful to compensate for headphone/speaker shortcomings, listening environment, and... personal preference.

89db is K-20, not K-14.

EDIT: Also, soundtrack CD != the film itself. 83db may be the film standard, but it applies to movies and movies only. Audio CDs are exempt from this standard no matter what the content.

You do realize that 2Bdecided is the progenitor of the ReplayGain standard, right?

I'm pretty sure That 83dB is the reference pink-noise volume for a K-20 recording, hence when MP3Gain first came out, and its default volume was 83.0dB, it was trying to match all MP3's to the loudness equivalent of a K-20 recording. Snelg soon changed it to 89.0dB though, corresponding to K-14, because almost everyone's MP3's were being made way too quiet.

I read through the link that 2Bdecided provided (to the digido.com article), and it was a bit hard for me to make sense of it, so perhaps my interpretation is incorrect. But I think it's right...

Very true, and this becomes more clear in electronic music. The artist has so much control that the album will come out sounding precisely as he desires it. To compare, mu-ziq's Tango 'n' Vectif album, released in the mid to early 90s has an album gain of around -10, with one track at -15 or -16. I don't have my computer running properly so I can't give an exact figure, unfortunately. The extreme RG values appear to come from an extremely overdriven bass drum and bass line.

OTOH, there are some slovenly mastering jobs done by some electronic artists as well. For example, check BT's Emotional Technology CD. This is really weird, because BT is an incredible producer. May have something to do with the push of that CD towards pop music. It's intentional, for certain, but it doesn't sound as good as it could.

Just for info, the soundtrack for the movie, “The Mission” (1986) by Ennio Morricone has a replaygain value of -0.70 dB. This CD was mastered in 1986 and remains to be remastered. This soundtrack is roughly a K-14 master.

A very good example:

Sirenia - At Sixes & Sevens
Albumgain: -9,25db
Comment: this album sounds plain flat, boring and with absolutely no punch, its horrible! It also sounds "hot" and "dry"

Tristania - Beyond The Veil
Albumgain: -9,64db
Comment: this album has lots of punch and sounds awesome. I was really surprised when i saw the albumgain-value, because i expected something like -2db. The "sound" of this album also is much more natural and sounds less "dry and hot".

Both albums are gothic-metal and although the bandname is different, its partially the same band! (Tristania disbanded and some band-members formed Sirenia)

- Lyx

It is. Like 2Bdecided said in this thread,



I think we have ample evidence that a Replaygain value of 0 = 89 db SPL = about K-14.

How exactly does SPL factor into a volume adjustment being done inside a computer? SPL is defined as actual amount of sound hitting your physical ears, and depends on how loud your speakers are. Until a signal hits the air molecules, SPL does not apply.

Yes, and Bob Katz is the progenitor of the K System, and he says that 89db is K-20. Either 2Bdecided is wrong or Bob Katz is wrong.

From the CD honor roll page:


So basically, 0db is the reference for K-20 on the honor roll. Here are the WaveGain results of the CDs from the list: Ænema by Tool:


Bob Katz designated this CD as having a monitor gain of -9, which means it is 9db above the K-20 standard. If you look at the WaveGain results, you'll see that the highest score is -8.47, and most of the more normally-arranged tracks tend to be very close to -8. And that's taking into account Tool's highly dynamic style (lots of soft portions in the loud rockers); had they had more conventional pop/rock arrangements that are less macrodynamic-oriented these scores might be even higher. If a ReplayGain value of 0 were actually K-14, then according to Bob Katz the louder of these WaveGain results would be far closer to -3 than -9, and we all know that the latter is the case in this scenario.

Now I don't know about orchestral arrangements, but we can definitely conclude from this is example that a rock song with a ReplayGain value of 0 is far closer to K-20 than K-14.

Kuuenbu, I just saw the other threads where you've brought this up. My interpretation (and pretty much everyone else's that I've read here) is that K-20 corresponds to playback of the standard -20dB pink-noise at 83dBC. I mean, here's the quote from Bob Katz:

Basically, to get K-14, you can either play -20dB pink noise at 77dBC, or (presumably) -14dB pink noise at 83dBC. Please, show me a quote where Katz says "89dB is K-20" the next time you want to bring this up. I'll be happy to eat my words if you can prove me wrong with a Katz quote.

Anyway, I noted your comment on Ænema, so I looked at a couple of my own CD's that are on the Katz honor roll. Here's Pink Floyd's Dark Side of the Moon, which Katz says sounds about right with a monitor gain of -1dB (corresponding to a K-19 recording).

Replaygain scores it at -3.71dB for album gain. It's possible that I have a different remaster than Katz's reference, though.

Another album with monitor gain of -1dB is Talking Heads' Stop Making Sense.

I'm pretty sure this is the same release Katz has. I checked the barcode and copyright date.

A few more albums (which, once again, may be from different masters than Katz's): Beatles' Sgt. Pepper scores -3.69dB album gain, and Katz gives it -6dB monitor gain. Rage Against the Machine's self-titled album scores either -6.63dB or -7.25dB album gain, depending on whether I use the MPC replaygain program or Foobar's, and Katz says -8dB monitor gain gives the appropriate listening level. Ænema you've already addressed: I get either -7.23dB or -7.60dB album RG's, and Bob says it's a -9dB monitor-gain album.

So, what does all this information say? I'm not sure, I'm still doing mental math with the numbers. What you said makes sense, though: that RG values should typically be 6dB less negative than Katz's numbers, if RG is using K-14 as a reference. These RG numbers are all too negative, though. So, while I think your argument that "89dB is K-20" is heading in the wrong direction, you may have a point when you say "RG of zero is K-20, not K-14".

One problem may be that we are only measuring rock albums. I don't have anything else in my CD collection, but maybe someone who owns some other genres that Bob lists can report their RG values. Bob may have high (less negative) monitor gains for rock music because it's supposed to ROCK! but, as mentioned by Guruboolez, if you take a harpsichord album, you might see a trend in the opposite direction.

Sorry Kuuenbu, but you've totally misunderstood the quote you've posted from the digido site. However, your results are very interesting!

As SometimesWarrior said, a zero 0dB monitor gain is the monitor gain at which a -20dB FS RMS pink noise signal causes a level of 83dB SPL within the room (SMPTE RP 200).


As I've said before, a ReplayGain of 0 (ref 89dB SPL) should be roughly equivalent to a K-14 mix (monitor gain ref SPMTE RP 200 at -6dB if you prefer).

This is what makes the above results interesting. There are plenty of reasons why the human preferred listening loudness for a given track might not match the ReplayGain value, and there might even be a skew overall due to inadequacies in the ReplayGain calculation or human preferences.

If anyone can trace a consistent problem, I would be interested. However, I think SometimesWarrior may have found it already - replay gain tries to make everything a similar loudness, whereas a human listener will decide that ROCK music should be louder than a harpsichord album.

Someone suggested a genre based correction, which would help, but you'd have to pick the genre carefully: "classical" could mean a harpsichord work or a 100 piece orchestra at full belt! It's more likely that, to always maintain "intended" loudness, you need human input to the loudness matching process, which is why I repeatedly suggest the need for a "Real Life" type loudness measure.

Cheers,
David.

This K/replaygain issue confuses me!


(David)

I assume wavgain uses the same reference level as replaygain: Why does wavgain then 'RMS-normalize' the average RMS volume of the differently loud titels to about -20dB? This doesn't depend on RMS window, I've checked this several times and with different examples. I thought the same as Kuuenbu:
replaygain/wavgain value = 0dB means K-20 !??

Is Wavegain still using the 83dB reference level by default?

If true, that would explain this, and the results in previous posts.

Cheers,
David.

I don't think so. I've used MP3Gain for monitoring levels very and it scores ReplayGain values on a pure SPL basis rather than based on a reference number. Comparing the same scores with WaveGain suggests that WaveGain does, in fact, use 89db as the reference level.

I don't want to interfere with your REPLAY-GAIN orgies but i find them a little pathetic. Let me explain...

Upon recording of a song, instruments usually get recorded SEPARATELY (each instrument/group of instruments has its sensor/mic). At this stage a song is composed of, let's say, 8 separate tracks each containing some instrument(s). It would be quite foolish to think that these don't get processed before mix. Actually, many studio engineers use compressors even on separate instrument tracks, thus making them sound louder and, unfortunately, flatter. This helps mixing alot as there are no big dynamic steps in the tracks and the overall volume of the song is much more easilly predictable.
But quickly to the point...
When mixing such tracks, an engineer can easilly meet your K-14 criterion (14db headroom) on the complete track. Then a REPLAYGAIN could indicate 0db (what a perfectly engineered song! ) but the tracks used to create such song could have been over-volumed since the beginning.
Beginning to feel embarassed? No wonder, as that BAD studio engineer easily fooled your precious REPLAYGAIN.

I meant no offense by this post but you have to realize that nothing is black or white!

While considering an engineering job, you should use your EARS in the first place, not some computer algorithm that completely lacks intelligence and judgement. Do not have programs make the conclusions for you (instead of you)!

Of course replaygain/wavegain can say only something about sound quality, not everything.
David's original intention (in principle) was to make ABX tests for dynamic and compressed versions of the same song possible (the loudnesses aren't different), so people would hear how bad the compressed version sounds.

My my - how educational - I never knew that.





I've never suggested that ReplayGain should be used for judging the quality of recordings.

However, I'm very interested in the data that SometimesWarrior and Kuuenbu showed up. I don't see it as a fault or a problem, it's just interesting - that's all.

Cheers,
David.

(David)

But I remember reading it somewhere in your posts here on HA. Unfortunately, I've forgotten where it was.
by the way: Judging the sound quality in this way is almost the only (and a good) sense for replaygain/wavgain that I can think of.

You could be right. It probably wasn't meant as a recommendation (for all the issues well covered in this thread), rather an observation.



and to think I liked you before you made that comment

You were on better ground, in your previous post, noting that ReplayGain is great for showing how bad over-compressed albums sound in comparison to other material. My hope was that maybe something like ReplayGain would become common enough that there would be little point in over compressing albums. I've suggested a similar thing for radio stations in the UK for exactly the same reason, but there's little interest.

Cheers,
David.

I ripped and analysed the soundtrack CD from above.
I find it not as good as the replaygain values might suggest. There's one track (track 9, the lowest gain, -6,...) which has a loud part with shouting men and blade sounds. The music is heavily compressed there and there's also some clipping, hidden in the compression. It is audible. The clipping probably occured during the downmix process and the master has been compressed then.
The average and total RMS levels in cool edit are rather close to -20dB than to -14 what suggests a K20 recording. But that's not enough for it. I consider the music style to be much more like orchestral clasical music than pop or rock. It has a lot of "long-time" RMS dynamics, that's what causes the low average volumes/replaygain values. But at the loudest parts it's as compressed as a recording from the end -90s (about 7 decibels of the peaks clipressed away)
This CD is a lot worse than it should be.
I wonder how well the CD's clipression status matches what we hear from the DVD or in the cinema:
Is it a original stereo master, that has been mixed into the multichannel movie sound, or is film music made as a multichannel master, which has been separately downmixed and mastered for CD ?

I think ReplayGain makes the assumption that the mixing engineer does a good job, and that most of the final damage, if any, is made by the mastering engineer.

This is probably a very valid assumption, because most mixing engineers just want to make a good recording/mix and care as much for dynamics as we do.

I highly doubt that.
Records get clipressed on each step - clipped on recording the single tracks, clipped during downmixing and finally compressed during mastering.
Often there can be found e.g. clipping on the vocals while the master is quiet and visually unclipressed, this must originate from the single track.

Hm, I have done some recordings in different studio's and in all cases clipping was prevented completely by the engineers. In fact, the use a compressor during recording with a max 1 - 2 dB reduction is to avoid clipping on instruments with great dynamics (vocals, electric guitar..).

Currently, I'm recording in an own studio, and every time a single track gets clipped by accident, it is re-recorded (doesn't matter how good the performance was). As far as I know, almost every studio works that way.

I mean, we should not think we're the ones with "golden ears" and mixing engineers are deaf fools who try to destroy our listening pleasure. Good engineers know what they're doing.

I think, when a record is made in a good studio, all clipping is introduced in the mastering stage. Mastering engineers probably get pushed by record companies to make records as loud/hot as possible.

For the record, all of Autechre's releases are uniformly excellently mastered. Untilted was released in May and RGs to -3.69dB. Extremely good dynamics and sound quality, although some releases are a little dry.

I agree on the Autechre part. Still, ‘Draft 7.30’ sounds terribly monophonic, while still maintaining good dynamic range, so nothing is perfect, heh.

Oh by the way, my ‘Dark Side of the Moon’ albumgains at -2.94 dB.
How many different masterings of that album is there in existance?!

Too many... The original, The SACD Remaster, the anniversary remasters, and maybe a few others...

I have the SACD remaster and the CD layer (the only one I can play) has an album gain of -6.12 dB using foobar 0.8.3's scanner. Certain tracks like "Time" and "Us and Them" have very obvious problems with dynamic range, and I have never heard any other version of Dark Side of the Moon.

The Batman Begins OST (composed by Hans Zimmer and James Newton Howard) is also pretty well mastered. Average RG value is about +2 db

regarding Tango 'n' Vectif cd1 has a value of -4.75 cd2 has a value of -4.28 on my rip, im not sure how to make foobar output the gain for all tracks so there could be one with a very high value. i dont know if i can make foobar sort my stuff by the gain values but the highest i have seen for an album has to be Venetian Snares + Speedranch - Making Orange Things -17.54 and still almost unlistenable on headphones at my normal listening level.

I've never gotten a gain value higher than -12.5 and at that level it sounds really bad. I can't imagine what -17.54 would be like.

(Re: Dark Side Of The Moon)
I have four different copies:
the 1983 original issue (album gain: +0.47db)
the 1988 MoFi gold (album gain: +0.78db)
the 1994 remaster (album gain: -3.68db)
the 2003 SACD (album gain: -6.12db)

Subjectively, I'd have to say that the best sounding copy is the 1988 MoFi, followed closely by the 1983 original issue. The others, aside from their "clipression" issues, simply do not have the clarity (ie, the change rattling at the beginning of Money) or the punch (ie, Nick's toms; Roger's bass) that the older issues have, to my ears.


Just for visual comparison...

Here's the waveform of Money from the 1983 original issue:

No clipping.

Compare that with the 2003 reissue (SACD (CD layer))

362 instances of clipping.


...and for anyone who hasn't seen this yet, here's an article examining the SACD version (Stereophile.com, June 2003).

I bought the hybrid DSOTM a couple of years ago for the 16/44 layer which I assumed would be exemplary, I'd never owned it on CD.

How wrong I was. It got played maybe twice (ripped to audio PC) and then forgotton about, because there is simply no drama there.

2nd crescendo/guitar-solo in 'Money' used to send me, but on this version it's just 'more of the same'.

R.

The Ace of Base album "The Sign" has an album gain of -0.11dB

The Pearl Harbor OST was done by Hans Zimmer as well. It's a -3.3dB album gain.

I always wondered why listening to the music at the end of a DVD sounded so good when the credits were rolling. It also definately explains why i have to crank up the volume when watching a movie.

-Joe

But from the look of those .wavs it's not like the 1983 track had tons of dynamic range either -- the track appears to be mostly a solid block of green with short bits of grass growing out of it

If you normalize it so that the peak is at 0 dB, I suspect it will look a fair bit like the 2003 CD layer... minus the gnarly clipped peaks of course.

Maybe I'm just seeing an artifact of the shrunken image...I'll have to go check the wavs I amde of the SACD and CD layers of the hybrid.




My own personal comparisons of the two layers of the disc, using A-->D conversion, bear their results out. Any buyers who thought the CD layer might have the same mastering philosophy behind it as the SACD layer , were duped.