Fraunhofer

BigD-Bluez

Someone brought up the topic on the eMu MB of What Happened To LAME? This led me to do a little checking and it seems that as far as I can tell every release I have downloaded from from the UMG since the drop has been in Fraunhofer. These include all the following titles - The Wild Tchoupitoulas, The London Bo Diddley Sessions, A Decade of Chicago's Finest:Chess Soul, Afro-Cuban Jazz Suite, Tell Mama-The Complete Muscle Shoals Session (Etta James), Yule Be Miserable, Below The Bassline -Ernest Ranglin, In Pas(s)ing-Mick Goodrick, Fiddler on the Rock-Don Harris, Gene Ammons-Boss Tenors In Orbit & The Gene Ammons Story-The 78 Era.
In other words every UMG release. Since I'm not that versed in bitrate conventions, why exactly should this be a concern? I can see from comparision to song files of similar length encoded in LAME 3.98 that the Fraunhofer files are considerably thinner - 3.7MB for LAME vs. 2.7MB for Fraunhofer for 2:28 songs. Is this a big old stinking rip-off in progress?

thomnottom

I always try to caution people when looking at mp3 files about comparing bitrates and file sizes, etc. Unless you're talking about lossless, more is not always better!

Files that are the same length and same bitrate should be about the same size. But a properly encoded VBR file can significantly save in file size over the annoying CBR that so many sites offer (I really do get annoyed by all of the 256 and 320 CBR files I've bought over the years). So those LAME files - while using a better codec - probably have a lot of wasted space because they are CBR. The Fraunhofer files are probably VBR encodes and save some space.

This is all just my guess. And most likely UMG paid to use Fraunhofer to make sure they can't be sued in case there ever a falling out over LAME.

thomnottom

For a better understanding of why VBR is better, think of John Cage's 4'22". Say it's encoded at 256kbps CBR (Constant Bit Rate). That means the file size would be:

(256 kbps * (4*60 +22) seconds) / 8 bits = 8384 kB / 1024 bytes = 8.1875 MB

Now instead lets encode that using LAME's VBR0 setting - which aims for the best quality at around 245kbps. I forget the actual range used, but I believe the lowest bitrate will be 32kbps. Even though it aims for the 200s, if you're basically encoding silence with a little bit of ambient noise it's largely going to be in the 30s and 40. We'll say it averaged out to 48kbps. So the file size would be:

(48 kbps * (4*60 +22) seconds) / 8 bits = 1572 kB / 1024 bytes = 1.535 MB

And guess what - those files would sound the same. Obviously I'm taking an extreme example, but that's the best way to show you how much things can vary.

elwoodicious

And guess what - those files would sound the same.

Tut-tut. It all depends on the performer and the sound engineer! :-P

/dbag

kargatron

thom, why are you talking about CBR in this context, since eMusic doesn't offer CBR files? If you want to defend UMG's Fraunhofer decision, talk about the likelihood that a ~25% decrease in filesize between two VBR encodings won't have much affect on quality.

thomnottom

A quick scan of my collection shows that eMusic does, in fact, offer some CBR files. Less than 1%, but they are there.

While I went off topic, I was simply trying to educate the masses on how quality, bitrate, and filesie interact. Without knowing the settings used, I merely made a guess that he picked a file that had been poorly encoded to begin with.

thomnottom

Another point to add, the VBR settings can have their ranges modified. Because most users do not understand that bitrate does not equal quality, it appears that some sites will use VBR V0 from LAME but limit the range to make sure that the bitrate comes out in the 200s.

I got the Complete Stax/Volt Singles this year during the screwed up Warner drop or something. 244 files encoded at VBR V0. But the bitrate comes out to an average of 160kbps. This is pretty common with older material that was not mastered after the Loudness Wars. Now if you force the codec to keep the bitrate up over 200 even though the source material doesn't require it, you'll increase the file size by about 50% without increasing the quality. Suddenly the directory goes from under 800 MB to around 1.2 GB.

Maybe I put too much trust in the codecs, but my guess is that without the number being shown most people would have no clue that they weren't getting all of those bits.

amclark2

Does anybody have any comment on the actual sound quality of this stuff?

Doofy

Fraunhofer, hmm? Sprechen Sie Englisch, bitte! These tunes are or are not going to sound crappier than the standard eMu files?

I'm hoping not, because I am looking to score me some Wild Tchoupitoulas.

BigD-Bluez

Does anybody have any comment on the actual sound quality of this stuff?

I haven't had the opportunity to do any quality listening since this came up - this long weekend will provide - but I was listening to the Chess Soul comp last weekend, just on the desktop, not that loud, and I was a little underwhelmed, but it's old stuff and I wasn't paying a world of attention.

jackedUPjazz

Since most peeps are probably listening to their downloaded music through dime sized computer speakers you may not be able to tell the diff.

I am using old but fairly decent speakers and I cannot tell the diff

However as soon as I get my CD burner up to snuff I plan on burning a bunch of my music to CD so that I roll up bumpin in my chariot with the Bose sound system.

I don't want to find out later that I have crappy rips and they don't sound good.

I hadn't noticed until you mentioned it but even some of my pre Universal DLs are fairly low bitrates. Some are fairly high bitrates but they all used to be high bitrates back before the prior price increase.

So I hope you keep up the pressure on the other place until they splain theyselves

What tha

thomnottom

I'm worried about how everything will sound once I get an awesome stereo hookup, too. But here's the thing - more bits ain't gonna do much. Unless you switch to lossless, you're not going to hear much difference.

I am by no means an expert, but those threads always make me cringe because they mistake bitrate for quality. Those are not synonyms.

kargatron

I don't think this Fraunhofer thing is necessarily about bitrates. BDB's off the cuff comparison of a 2.5 minute track (I did assume musically comparable, though) and seeing 25% lower size is worrisome. But more to the point, LAME is very well-tested over the years: I searched on LAME/Fraunhofer comparisons and boy, there's very little out there that isn't years old. Fraunhofer seems to have held a low profile (implying less experiential testing to me) for years now. Yes, these are worries based on conjecture, but not based on misconceptions afaict.

BigD-Bluez

For the record the comparison was between tracks from Chess Soul (UMG) and Ain't Nobody Here But Us Chickens, a compilation of even older material of a bluesy/jump R&B vein, so genre wise they were pretty close, besides being contiguous in my eMusic playlist. I have to say when I listened to the chicken songs the sound was happening!

More to the point also in Fraunhofer - the bonus tracks I downloaded from Motorhead's Iron Fist. How the bleedin' 'ell am I supposed to listen to Lemmy in anything less than the finest flipping hi-fidelity?

Dr. Mutex

Fraunhofer is a perfectly good codec and for years the LAME folks worked to get as good as Fraunhofer. I'm suspicious of files that are significantly smaller than LAME VBR0 files. It suggests that the highest quality setting wasn't used. The best comparison would to get some of the tracks from CD and compress them with LAME and see what size it comes up with. The bitrate requirements of a piece of music can be surprising. You might think solo piano would compress very well but it's just the opposite. The piano has a fast attack and there's nothing to mask any artifacts.

BigD-Bluez

Well FWIW I've hunted up an actual comparison from my eMusic downloads - I'd Rather Go Blind by Etta James - both versions at 2:37 - from the album Raw Blues Volume 1 downloaded in October: 3.5MB in LAME 3.96, 183kbps(VBR); from last week's UMG drop from Tell Mama-The Complete Muscle Shoals Session we have 2.8MB in Fraunhofer , 143kbps(VBR). On the desktop and through the sennheiser headphones they sound pretty much the same.
The song Mama Don't Lie by Jan Bradley from the A Decade of Chicago's Finest:Chess Soul, a UMG drop selection, came in at 2.3MB at 149kbps(VBR). I burned this song into the computer from the Chess Rhythm & Roll box set for comparison - being a Mac it's an AAC file, at 3.5MB at 256(VBR). Again to be honest, the CD rip may have a slight edge but if so it's barely perceptible. So there you have it. I'd just like a response from eMusic which has not yet been forthcoming either on the MB or in response to my e-mail to CS.

BigD-Bluez

Five days in and finally a response from the Ministry Of Truth -

All:
We are following this thread and are looking into the issue. We will get back to you soon with more information on your concerns. Sorry I didn't post sooner.

Thanks,
Cathy
eMusic

A near identical response was posted to the thread about low classical bit-rates. Wow, that's customer service. Let me go back to holding my breath.

Doofy

No kidding, what are the odds they are going to re-encode and re-up all that stuff?

It would be interesting to see if there is any difference in UMG stuff DL'd from eMu versus other mp3 sources. For example, I have ECM tracks from both Amazon and Guvera that are 256 kbps. Amazon is encoded in LAME, Guvera in "unknown" (make that, "Gawd only knows.")

Yes, I know bitrate isn't everything, just wondering if some UMG mp3s are more equal than others.

xtrev

being a Mac it's an AAC file

Why? I assume you're using iTunes, but you can change the default format to mp3 or a few other things in the preferences.
Although I've been an Apple fan (ish) for a looong time, there's no need to help them lock you in :-)

Although I think technically AAC is an open format, pretty much nobody else uses/supports it as far as I know.

thomnottom

I'd be curious to run a comparison on some ECM albums. For starters, because they are just about the only thing from this drop that truly interests me. But even bigger is the fact that I have 22 albums from ECM via eMusic UK. They were encoded with LAME 3.96 set to VBR V0 with an average bitrate of 229kbps.

If anyone has grabbed any Pat Metheny, Keith Jarret, Gary Burton, or Chick Corea from this drop it's possibly that we could at least compare the nitty gritty of the encoding process. I'd be plenty willing to set up a spot for us to share and compare (no, I'm not suggesting we actually share albums with each other, rather those of us who own the album could share the different version strictly for comparison).

thomnottom

Side note with AAC - it is slightly more open than MP3. Probably a better format - but just like VHS, the better format doesn't always win.

Bad Thoughts

Vice President of Lackiness Cathy H-N has given a response:

Please refrain from the personal insults while I keep my own sarcasm in check. Also note that eMusic was closed for the US holiday weekend beginning on Wednesday afternoon.

As for the bit rate concerns, we had our tech team look into this. UMG's tracks are indeed encoded with Fraunhofer encoder. The decision to use Fraunhofer was made on our end from the encoding options offered to us by UMG. For FLAC files, which are delivered to us by most labels, we typically we use the LAME encoder. After your post, we did a small scale analysis of bit rates with tracks from a variety of labels and found only a small difference among them, which we will be repeating within the next few days for confirmation. We understand your frustration on this issue; we continue to explore it and are looking into ways we can improve audio quality in the future.

Two things to note. First, she's responding to the snark and ire. That's not good. Second: the encoding is done by eMusic, not the labels, in most cases? Shouldn't eMu be able to up the quality of older tracks at their own leisure?

kargatron

I infer that for whatever bizarre reason, UMG disallowed LAME as a choice. Can't imagine why emusic would choose otherwise if they were free to. I was also surprised to learn emu encoded things. Am I wrong to mistrust that claim? I suppose it's not unrealistic that my impression that distributors supplied the mp3s was wrong.

Bad Thoughts

Kargatron, I don't think you are wrong. The removal of Fantasy/Prestige/Contemporary albums was presented as the label's (Concord) decision to re-encode at a higher rate. Perhaps eMusic was not originally in posession of the FLAC masters, but it seems to be based on Concord's decision, not eMusic's.

BigD-Bluez

AAC vs. MP3 is another grey area for me - yes, it is the default on the Mac, and I just continue to use it because I burn my CDs to Music CDRs which will generally play on any machine I've encountered, and they could be copied for a friend on a dubbing deck, not that that is anything I would ever do.

BigD-Bluez

Well, here's the latest from the Ministry of Truth - it raises as many questions as it answers:

As for the bit rate concerns, we had our tech team look into this. UMG's tracks are indeed encoded with Fraunhofer encoder. The decision to use Fraunhofer was made on our end from the encoding options offered to us by UMG. For FLAC files, which are delivered to us by most labels, we typically we use the LAME encoder. After your post, we did a small scale analysis of bit rates with tracks from a variety of labels and found only a small difference among them, which we will be repeating within the next few days for confirmation. We understand your frustration on this issue; we continue to explore it and are looking into ways we can improve audio quality in the future.

For clarification - most of you probably know that the VBR (variable bit rate) as displayed is the average bit rate for the tune. Some parts of the song may use a very high bit rate. Other parts of the song – which may not be particularly human-distinguishable -- may use a very low bit rate. This is different from CBR (constant bit rate) in which the entire song is encoded using the same bit rate. We do target 256kbps, but that doesn't mean that every song will have a VBR of precisely 256kbps.

I hope this information is helpful.

Cathy
eMusic

Any thoughts, opinions?

thomnottom

Their incompetence knows no bounds. "Hmm, there's some complaints about the quality of our recent and much publicized UMG catalog acquisition. Should I respond promptly before our long holiday or assume that no one will care that we ignore it for almost a week?"

WWCHND is my new motto.

BigD-Bluez

I am wrong in thinking that eMusic's defense for low bit-rates in the past was basically that the labels not eMusic were responsible for how the material was delivered to them? Did Cathy just let the cat out of the bag whilst trying to deflect this more recent question?

Bad Thoughts

I am wrong in thinking that eMusic's defense for low bit-rates in the past was basically that the labels not eMusic were responsible for how the material was delivered to them?

As I wrote above, eMu deliberately said that Concord was responsible for pulling its catalog and delivering at a higher bitrate.

thomnottom

I am SHOCKED. She actually apologized when I called her out for not posting before the holiday weekend. Of course she still hasn't apologized for lying to us before.

BigD-Bluez

One week later I finally got a response from Customer Service:

We’ve looked into this and yes, UMG's tracks are indeed encoded with Fraunhofer encoder. The decision to use Fraunhofer was made on our end from the encoding options offered to us by UMG. For FLAC files, which are delivered to us by most labels, we typically we use the LAME encoder.

We have performed a small scale analysis of bit rates with tracks from a variety of labels and found only a small difference among them, which we will be repeating within the next few days for confirmation. We understand your frustration on this issue; we continue to explore it and are looking into ways we can improve audio quality in the future.

For clarification – you may already know that the VBR (variable bit rate) as displayed is the average bit rate for the tune. Some parts of the song may use a very high bit rate. Other parts of the song – which may not be particularly human-distinguishable -- may use a very low bit rate. This is different from CBR (constant bit rate) in which the entire song is encoded using the same bit rate. We do target 256kbps, but that doesn't mean that every song will have a VBR of precisely 256kbps.

Wow, that's so much clearer than what Cathy posted on the MB on Monday - hey, wait a minute it is what Cathy posted. Stayed up all night writing that one they did. Enlightenment is a hard and strenuous task.