Hello Guys!

I have a question and I hope I place it in the right sub-section. Today I bought the score for the movie G.I. Joe: Rise of the Cobra. I don´t want to talk about the movie, I´m not interested in seeing it, I just want to hear the music which was composed by Alan Silvestri (Back to the Future, Forrest Gump, The Abyss etc.). The label which released the score is called Varèse Sarabande, they reside in California. They have a long history in releasing music from movies and they are an independent label.

There german subsidary is called Colosseum Schallplatten which subsequently was responsible for the release I´d like to ask about. This release was bought by me today at a Saturn market, I have the bill to prove it. When doing the things I usually do when I prepare a CD for listening (grabbing and upsampling and store the results with WavPack lossy) I took a look at the original files.

What can I say? The files on this pressed, official and brand new CD appear to be compressed with a lossy codec. They have the typical frequency response of music compressed with MP3 or AAC, I´d say if MP3 was used it had a bitrate of 192 kBit/s. I tested the disc with TauAnalyzer - the result was "UNKOWN".

Here is a screenshot I did with Wavelab:


I´ve already written a very angry mail to Varèse Sarabande! I can´t believe this: I´m going into a store, buy a brand new CD and what do I get? Lossy music! Don´t get me wrong, the whole CD sounds really good apart from being mastered 6 dB too loud but nevertheless I didn´t expect lossy music on a regular CD for that much money.

I´ll upload an excerpt from the score for you that you can judge yourself.

EDIT: Here you can download an excerpt from the score. I think you can use whatever analyzing tool you want. I see the missing frequencies in WaveLab and in iZotope RX. Transients occasionally show up but the rest doesn´t. Very strange.

This is not a proof of lossy compression being used.



What's the problem then?

A lowpass in and of itself is not proof of lossy encoding, but a spectrogram view of your upload clearly shows gating of low amplitude high frequency content, which is a patently clear sign of lossiness.

If that isn´t proof enough, download and look at the sample. There you can see yourself.

And the answer to the rest is very easy: first and foremost when I buy a CD I expect the full frequency, not just something that was lossy before. If I´d want lossy quality I would buy iTunes stuff. I buy CDs for exactly the opposite reason: full quality (and a printed cover, a real booklet etc.).

The second and more severe reason: if I decide to make some MP3s out of this CD myself for listening on my portable MP3 player, I would re-compress an already lossy compressed source. In other words I´d do a form of lossy transcoding (via converting to .wav throught the CD). Listening to artifacts should be easy then, don´t you think?

Still a possible sign not a proof. Band-limited gate or compression may be used without noisy compression.


Though it sounds to you very goog with the frequency range it has.




Did you try this? Could you hear artifacts?

Do you actually have principles? I don´t want to be harsh but I believe it´s understandable that I can expect full quality for ca. 19,- Euro (don´t know what it is in $).

To answer your questions: with good I meant I don´t hear typical artifacts like pre-echo. I only hear distortions and pumping effects - maybe due to brickwall limiting. You can hear that also in the excerpt. This discussion is not about the fact if I´m able to hear lossy compression or not, it is about the fact that a manufactured CD is pre-compressed. For me, that´s a first and also for me, quite unaccetable.

I did a test with 160 kBit/s MP3 (Lame). Result: compression artifacts. When I usually use that bitrate I sometimes have slight compression artifacts. Now I have many. Increasing the bitrate causes my MP3 player to loose its battery power much sooner. And you don´t expect me to buy a new one with a longer battery life so that I can play 256 kBit/s files, do you?

I am very certain, that this is not lossy compressed. The spectrogram doesn't show any sign of blocking at any zoom level, that AFAIK all common lossy encoders share. It's perfectly smooth! Many mastering engineers don't go by the schoolbook. They often use a specific hardware (or software) equalizer just because they like the "sound" of it and that may have been the case here.

BTW, have you designed those arrows especially for these plots?

PS: I just got my hands on a large pile of DCC and MFSL gold releases, that I'm importing into iTunes right now. Pure diamonds! If those ever cross your sight, drop everything and grab'em!

You don´t? I can clearly see low pass filtering in the spectrum analyisis in iZotope, which is sometimes expanded by transient sounds. Otherwise I agree with you. Mastering engineers and composers alike love to play around with several tools. But then I wonder... for years Alan Silvestri used HDCD. Now he uses something that applies a low pass filtering at around 16 kHz to violins, brass, flutes, synth and sometimes percussion. Above that stays some tiny little noise pattern that doesn´t move with the music

And the image was done in 1 minute with photoshop.

But they are so expensive. Well... not one bit of compressed sound If they would release more classical music (not just dvorak or stokowski as they did many years ago) I´d buy them immediately. Do they still master with half speed?

I think MP3, AAC, and even MPC and MP2 can be ruled out. They usually leave a clearly blocked pattern in the spectrogram. It's not very probable, that anything else has found its way into this studio.

My first thought was maybe this was a try to sabotage MP3 creation, but the brickwall filter alone doesn't sabotage anything and other artifacts don't seem to be there.



Yes, the classical lineup was always poor. But they have released a lot of fantastic remasters from the 60's-80's: Miles Davis, Coltrane, Sonny Rollins, Aretha Franklin, Velvet Underground, The Police, etc. and later even Nirvana, Guns N' Roses and the Pixies. And yes, completely without the "hot" sound of compression that they tell us we want.

Most of those releases aren't available anymore, so a large sharing scene of perfect EAC rips has evolved over time. Therefore you don't necessarily have to pay $100 for every of those 24K gold CDs to dive into this.

Yes, the lowpass appearance is damning evidence of some lossy scheme, but what I want to know is if all the tracks on the album look that way, if in fact they are all the same composer, performer, etc.

Couldn't anyone cite a lossy codec, that doesn't leave blocks in a spectrogram, for a change? Maybe I'm missing something. Even the old MP3 tries to avoid this by overlapping, but usually there are always some traces left.

I know of this releases. But you know, I´m very old-fashioned. I actually want to hold a CD in my hands, want to look at it. A booklet printed by my printer doesn´t look real (maybe I need a better printer). That´s why I still tend to buy a lot. To be completely honest, the CDs I want, for example from TELARC, are simply not existing on the file-sharing networks. But then, I´d buy them anyway...

I think I can answer that. Most of the tracks have this evidence of some lossy scheme, but not all. There are 21 tracks on this score album and roughly 30 % look lossless. And all the tracks are composed and conducted and produced by Alan Silvestri with executive album production by Stephen Sommers and Robert Townson (head of Varèse).

A friend of mine was so nice to point out that someone before me already has discovered this, I didn´t know:

1. Intrada Soundtrack forum
2. FSM Board

On there one can find a reaction from Varèse Sarabande:



It really is strange to see that a company who released several SACDs in the past and used people like Mike Matessino for remastering jobs would behave that way. If they speak the truth it would mean, that Alan Silvestri himself approved of the filtering/compressing/whatever. Some filmmusic composers tinker a lot with sound apart from composition or orchestration to achieve the desired reaction from the audience, maybe it is the same here.

Hi,


He never denies that they used actually compressed material, they just stated that they gave their ok after listening.
Or am I just missing the point?

Regards, Clobon

This original CD is made from a lossy source, look at the spectrogram.

Not every CD made has to be full-range, especially if the company releases the material from distance

No, I think it's really possible that the source sound files could be lossy. It is analogous (pun intended, I guess) to the limitations of cassette, which nobody really really noticed in the case of the lack of another medium to compare it to.

If you think about it, now there might be some HD version of this released later on in its un-capped glory

Just see what the a DVD-A of this is release performs. $$$

.halverhahn's spectrogram looks blocky enough to me.

If it was something exceptional (which I don't think it is), it could be AC-3 (Dolby Digital) - I can't remember what that looks like.

The 2000 single of Toca's Miracle...
http://en.wikipedia.org/wiki/Toca's_Miracle
...is clearly (and very audibly) from a lossy source - I think it's multi-generation minidisc, but that's just a guess based on years of listening to MD and mp3. Only the 7" mix is affected - the 12" mix is lossless (but you can't quite get the 7" mix by editing the 12" mix - some parts aren't duplicated in both). Probably the worst obvious audio fault on a UK number one single since Jive Bunny's first single (mono mix transferred off a dodgy cassette with incorrect azimuth!).

Cheers,
David.

Exactly. That´s why I bought only CDs almost from the beginning. Wait, I even had one cassette from TELARC (!) - fantastic sound (for a cassette). Recorded in 1x speed onto a TDK-SA Chrome Cassette. And they even used HXPro. Maybe TELARC did use a home tape deck for it - but on the cassette it states that it was produced at Terre Haute (DADC). It was back in the 80s...

Very cynical but you´re right. Then hopefully with full frequency content AND without loudness-war-artifacts. But that won´t come. Varèse is notorious for being economical.

AC-3 stops from my experience at precisely 19 kHz but I´m not sure about this.

What? You´re kidding, right? I have Jive Bunny´s first single (I was young and without taste once), it always sounded good to me (and in stereo). Did Germany have a different pressing?

Did anyone think of it as an attempt to force users to abandon CD media? At least that seems to be a common strategy of the music business lately...

Varèse wouldn´t get a commercial benefit out of it if they used compressed music for CD releases. Reading over at the soundtrack forums, there is the information that the german Varèse daughter Colosseum admitted that some mistake happened with the master.

The american mother Varèse on the other hand denies this. Douglas Fake, head of Intrada (another independent score-only label), expressed his concern about the sound quality of G.I. Joe but he doubts that Varèse will remaster the release.

I guess I should have used the term "blocking" when I said "gating". halverhahn and I are talking about the same thing.



There are really funny things going on above the ostensible lowpass, as you can see. Based on the length of the blips I can discount AAC from contention because some blips are not clear multiples of 1024. They could be multiples of 576, so MP3 is a contender.

My money is on the pressing plant receiving an MP3 by accident supplied by the mastering house for proofing. I'm likely to incline with both Cavaille and VS here - this sucks but it is unlikely to cause an audible problem unless additional information becomes available as to the specific encoder used.

Where high frequency blocks are maintained but lower frequencies are completely removed, that gives rise to an annoying sound some listener refer to as "ringing". (Not the ringing associated with filters).

My ears don't have enough high frequency extension to hear it (never have had, within the lifetime of mp3!) - but a few other gifted listeners report it for other samples with spectrograms like that.

So maybe someone somewhere will hear a problem with this.

...and who can say what effect it would have on transcoding?

Cheers,
David.

I must correct myself on the issue blocking. My spectrogram was left in a "look nice" setting and added visual blur that is not an representation of the actual signal.

Ah, thanks for the information. At least this maybe points to MP3. But what a shame for the mastering house (Bernie Grundman Mastering).


How does this "ringing" effect sound?

BTW, here is the technical information about the disc:


Mastered by Patricia Sullivan Fourstar - quite a household name, don´t you think?

It sounds like it's ringing.

Here are comments from people who have actually heard it(!)
http://ff123.net/ringing.html

and pictures...
http://ff123.net/ringing_graph.html

Cheers,
David.

Ok, thanks. Now I think I know what you mean. I always thought it would belong to what I call "flanging" - though I´ve never heard it that sharp and so high. I didn´t observe these effects with G.I. Joe so far... but then I listened with loudspeakers and not with my headphone. If an MP3 encoder was used, does Lame still produce these artifacts anyway?

I agree, this track is clearly compressed. Very very sad, in my opinion. This would be a no-go for me, especially if I'd have to pay 15 Euros for it.

Indeed, the track could originate from a 44.1- or 48-kHz MP3. Looking at .halverhahn's image upload and my Audition plot, I can see blocking with a length of around 1100 samples (i.e. roughly 2x576 or 2x576x(44.1/48)) and band-wise gating during the quiet passage around second 19 - 22. The borders of these bands are at around (as measured visually in Audition) 6.5ish, 8.0ish?, 9.6, 11.5, 13.8, and 16 kHz. Does anyone have the scale factor band borders for MP3? Do they coincide?

Actually, exporting to 192-kb CBR MP3 in Audition 1.x gives a spectrogram very similar to the upload.

Update: Found the scale factor band tables here. Sorry, couldn't wait


If we take the 48000 table, we get:

24 kHz * 156 / 576 = 6.50 kHz
24 kHz * 190 / 576 = 7.92 kHz
24 kHz * 230 / 576 = 9.58 kHz
24 kHz * 276 / 576 = 11.50 kHz
24 kHz * 330 / 576 = 13.75 kHz
24 kHz * 384 / 576 = 16.00 kHz

Bingo! It's a 48-kHz MP3 decoded to 44.1-kHz. Probably 192 kbps. Ouch...

Chris

Wow! That was really impressive work. You guys are great!! Then we know for sure it is MP3 - and I´m also quite sure that the codec used is the original one from the Fraunhofer Institute. Look at the scan I provided of the cover - ProTools was used and Digidesign offers an MP3-option for it. This option uses the Fraunhofer encoder. It´s only a guess and not really proof though...

But how stupid to decode a 48 kHz-file to 44.1. But when decoding happens like that, is there any sampling rate conversion involved or is it something different which converts the sample rates? Or does it mean that the music simply is too fast?

Oh, I forgot... I did some proper listening with my headphones. Besides some pumping and several distortions I´m pretty sure I´ve heard some artifacts which could be described as "gating". Does something like this exists with MP3? I did not hear some "ringing" so far - but then there´s a lot of synthesized parts in the music and the string sections of Alan Silvestri always sound like they are "ringing" or "grainy".

When I´m back at home I´ll provide some further samples.

Thanks! Well, I'm pretty sure they resampled to 44.1 kHz when decoding the 48-kHz MP3. Haven't heard the uncompressed version, but this one doesn't sound too fast to me.

Now that you know it's an MP3, you shouldn't trust your hearing any more Assuming it's 192 kbps, it probably sounds just fine, and compression artifacts, if any, are probably only audible in direct comparison to the uncompressed version... which unfortunately we don't have. I don't hear any artifacts. But I agree with 2Bdecided, some people might find the sound dull due to the bandwidth limitation.

Chris

That´s exactly the problem. How can someone know how the original sound was? No one (aside from the composer, producer and several other people) don´t know the original. So we can´t compare. And I don´t know that much about MP3 that I could safely assume about artifacts. I could some years ago when I started with computer audio but since then the codecs have developed and they are much better now.

At least that´s the case with LAME. I don´t know about the codec from Fraunhofer institute if that one is still developed besides adding capabilites like surround or else.

As I said, there is something which sounds like artifacts but I´m not so sure. It could be "ringing" but it also could be the for Alan Silvestri typical high frequency strings. When I´ll upload more sample I´ll also provide a sample from "Van Helsing", recorded in 2004 with similar dynamics and from the same team. Maybe one can draw conclusions from that.

another visual representation:

"Economical"... so that's the expression for asking the full bundle of euros for compressed content.

Since when do we hear music with our eyes? I listened to the clip and it didn't sound bad. As for what you see in the lowpass--I've seen an album in the past with a similar cutoff from an original pressed cd, and it didn't sound any worse than if it had the info all the way to 22khz. I think the engineers know what they're doing...and Varese is a good label.

Well I also doubt that I'd be able to notice any artefacts (I mostly can't ABX lame -V5 from the original) but thats not really the point. I'd still be very peeved if I bought a pressed CD that was mastered from 192kbps mp3's. You think the "engineers knew what they were doing" well I very much doubt it, this is almost certainly a stuff up. Regardless of whether problems are immediately audible or not this is still a bad situation. What about ripping this CD to mp3's, you're going to have to transcode lossy to lossy just to do that - it's just plain wrong! If I had a CD like that I could see myself using lossless compression for situations where I'd otherwise use lossy - just to avoid transcoding - and then to add insult I'd know that my 800kpbs lossless files were really only 192kbps mp3 quality. Personally I'd be very unhappy with that purchase.

How do you know that? Do you have the version with 22-kHz bandwidth available?

A young colleague at work recently gave low grades to an item in a blind test, while the other participants didn't. When I asked him why, he said the item sounded dull compared to the original. Well, the test item had a 16-kHz lowpass applied to it. That was basically the only artifact. So, again, even if something sounds "not any worse" to you, it might actually sound worse to someone else.

I agree with uart, when I buy a CD, I expect the highest quality possible via that medium given the original recording. Nothing less.

Chris

That was a goof on my part when I wrote that. No of course I don't have anything else to compare it against. What I'm saying is, the only reason I even knew there was a cut in the frequency is I ran a spectro on a track, and had I not done that, I would've been none the wiser about the cutoff. Its not that LeAnn Rimes' early work cd sounds bad. It could've been the recording equipment that had limits because these came before she made it big in 1995.

You're right though, you should be able to get the whole thing, and any modern equipment would not have blocking like you see. I don't understand it either.

So, finally the promised additional excerpts. You can find them here.

I don´t know if you can hear the same things as me but I´m under the impression that the high violins (given the fact that they are on the left side (as it is custom with the standard orchestra)) seem to wander around sometimes. Would that be typical for MP3? Or is it more a sign of the Dolby Surround processor used for this recording? Then there is the percussion (not the synthesized percussion, the one from "real" instruments) which sounds strange (for the lack of a better word). In the excerpt "The JOEs mobilize 3" there is a strange warbling artifact which not necessarily has to be a compression artifact.

When the percussion kicks in on some parts the violins seem to lose their "air" and their brilliance.

I have to admit that some of these things could very well come from processing via some DSP used in the production and not necessarily from the compression. Without a reference this is very hard but I´m under the impression that the whole score sounds kind of bright, a bit thin (apart from the overpowering deep bass) without the air that normally comes with a bright sound.

For comparison I also offered an excerpt from "Van Helsing", released in 2004. The style is quite similar (even the tonality) with less synthesizers but the same overall frequency response.

What software was the CD ripped to WAV? Is the CD copy-protected? If ever, please don't throw TOS #9 on me, but if the CD is copy-protected, and was ripped altogether, one may get a resulting rip with 16kHz lowpass even if the CD has tracks with full frequency spectrum. The culprit: another set of compressed audio files hidden within the CD, which surfaces when one rips a particular CD, instead of the original uncompressed tracks, when ripped improperly.

Just my 2 cents worth though.

EDIT: Just browsed the CD on Amazon, its playtime is beyond 70 minutes, so I guess there won't be enough space for a data part on the CD, and therefore, may not be copy-controlled.

Thats not how a copy protected CD works. If you try to rip one of those, either you get the actual PCM, or else it errors out.

Exactly. No copy protection on the disc. Extracting was done with EAC.

Cavaille, how many tracks on the Joe CD seem to be MP3 compressed? All of them?

Chris

While some say, only half of the tracks are compressed I can say that the whole CD has a frequency cutoff at 16 kHz. There is a difference though that has nothing to do with the lossy compression: the first 9 tracks are mastered differently. They are dynamically compressed with brickwall filtering with 0.2 dB to 0 dbFS left. All 12 tracks coming after that leave 0.5 dB to 0 dBFS.

But the MP3 compression is applied to the whole CD. Why do you ask?

EDIT: Here is the frequency response for the whole CD:


I seem to be mistaken. The last tracks seem to be coming from a lossless source.

Indeed, starting with track 15 to the last track 21 everything is lossless with fully frequency content. From approximately 70 minutes of music this still makes 50 minutes of lossy compressed music. Thank you for the question because if you wouldn´t have asked it I would have continued to say that the whole CD is lossy.

Maybe Patricia Sullivan Fourstar got so used to the sound of her iPod that she applied this on purpose as it sounds "just right" to her ears. Kind of like Portishead did in the 90's, when they pressed their studio masters on vinyl first and then sold the needle drop on CD. Next thing will be producers re-recording music from little mobile phone speakers before releasing, because that's supposed to sound as kids are used it.

I know of a vinyl release that's cut from mp3s because the producer lost the masters.

And some Aphex Twin CD releases are mastered from cassette tape.

Kung Fu Records' CD reissue of Alkaline Trio's Maybe I'll Catch Fire shows clear signs of being MP3-sourced.

Thanks! Interesting. What is also interesting: In the "Mobilize" track excerpts you uploaded, I can see that every single brickwall-limited waveform peak extends to precisely -0.50 dB. This shows that the mastering house was using MP3s as their source for most of the tracks (i.e. lossy compression was there before the limiting). Now let's recap.

The audiophile community tells us that even a 192-kbps MP3 is far from being transparent (source). However, a supposedly acclaimed mastering engineer obviously did not get suspicious about the sound of her mastering product, partially made from MP3s in the range of 192 kbps. Now, either

- said mastering engineer did not have a choice (had to use MP3s for half of the tracks, which I seriously doubt), or
- said mastering engineer has (or at least had) no clue what a spectrogram is and what its benefits are, and does not have the ears to distinguish an MP3 from an uncompressed source, and therefore made a serious mistake (I'm pretty sure of that), or
- contrary to the beliefs of some (or most?) of the audiophile community, an MP3 can be transparent even if it is almost one hour long (I'm pretty sure of that as well)

Chris

The first possibility is the one I´d expect the most from Varèse Sarabande. I truly believe that the mastering engineer did recognize the problem (almost every DAW has a spectogram or at least a frequency meter and a mastering engineer would literally have to be blind) and she did inform Varèse about that. They in turn said "We don´t care. Our buyers won´t hear it anyway because we doesn´t. And providing them with lossless tracks would be more costly". So the mastering house had to use this source. Remember, many mastering engineers are destroying material on purpose (Loudness-War) because the money givers say so. Same case here IMO.

And please be careful. You can´t say that audiophiles are generally stupid because of this CD. I´d describe myself as a parttime audiophile. And it is true, I wouldn´t have recognized the compression on this disc if I wouldn´t rip and upsample every little tidbit that goes inside my DVD-ROM. But this fact alone is no proof that MP3 is transparent. It is merely a staunch proof of the fact how conditioned we are. I expect pristine sound on CD and from years of buying CDs I grew accustomed to the fact that there is lossless music on them. Up until now, this was true for 99.999999 %. So the simple experience of having lossless CDs for around 20 years created a placebo effect. Yes, the same placebo effect that normally creates sound differences with some dubious little thingis. A placebo effect of listening to pristine sound.

Truth to be told, the sound is perfectly allright. But I still wouldn´t say that it is transparent. It is really good, but not transparent. Once you know about the possibility of an imperfect CD you can search for artifacts and that´s what I did. Another truth to be told, I don´t know many artifacts when I hear them. During researching this CD I found that MP3 actually had evolved. Music sounds really good even with 160 kbps and for the live of me, I can´t hear a difference from 16/44.1 to 320 kbps MP3. Now I´m using WavPack lossy for two years. And WavPack lossy does have other artifacts (in fact, only one). I had to listen to some selfmade MP3s with a low bitrate to find out again how MP3-artifacts sound.

So, since I´m not experienced in this matter I came here to find out where the artifacts are. But as one person said, it is difficult without a reference.

I for my part have given up. Varèse will not remaster this CD, they will not answer my mails. Instead I created with the help of WaveLabs Spectralizer additional frequencies above 16 kHz which were mixed afterwards into the lossy stuff. I think it sounds better but I didn´t bother to actually compare that (besides, I did use a bit of equalizing to hide possible effects).

Forgive me for the bashing of your conclusions but I couldn´t let the comment about audiophiles be written there without reply. In my mind it was what happens here all the time, only this time the other way round.

Cavaille, I never said or even implied that audiophiles are stupid! Sorry if I was unclear. What I wanted to express is that lossly compressed audio can be transparent, or in other words, of more than sufficient quality, even to a mastering engineer which is used to high-quality recordings (as the "G.I. Joe issue" showed). It seems that some (of course not all) audiophiles don't want us to believe that. Of course that's no proof that it always is transparent.

The thing with Varèse: Why would providing the mastering house with lossless tracks instead of MP3s be be more costly? Transmission over the Internet has never been faster. If it's the producer's fault, they could easily have re-sent the tracks at no extra cost. Unless the mastering engineer already mastered the MP3s... which is what I think happened.

But I give you that, a good mastering engineer would probably never apply loudness-war compression levels from personal belief, but only because the money-givers want it that way.

Chris

I'd like to believe that, but in a thread at some hi-fi forum (linked from here some time ago and I don't remember either forum, exact comments or anything), users/testers(/other mastering engineers?) were praising a mastering engineers work high above the sky for it's "hotness" - Which are mastering-engineers general term for "compression/limiter-abuse".


It seems to me (from the spectrogram) that the lossless-part recording has picked up a 16khz sine probably from a tube-tv... Which leads me to think that more than said "mastering "engineer"" both have no idea of what a spectrogram is (and should look like) and lost hearing ability at these frequencies.

A great example of this is Celine Dion's "A new day has come (radio edit)", which has a VERY LOUD 16khz sine in most chorus which are driving ME crazy, but surely didn't do anything to the "mastering "engineer"".