hardcore audiophiles can not reliably distinguish a 128K-bit MP3 file from the master recording

Don't believe everything you read, Corby . Most people can easily hear the difference between a CD master recording and any MP3 copy if they compare on a decent stereo system. The MP3 theory that it can eliminate certain frequency bands without anyone noticing the difference is grossly exaggerated: the difference is significant: It is as if the leading tunes eat into the sounds that are not in the foreground - I find that there are 'dips' in the background music lines, whereas on decent playback systems, all music lines are autonomous; not affected by other sounds played at the same time. MP3s also combines lower frequency left and right channels on the assumption that the lower tones don't contain directional info - the same theory that Bose uses to market small speakers with a single bass box that's supposed to be hidden behind a couch. Well, the stereo image of MP3 is affected, for which the mono lower frequencies contribute. But with 90% reduction in file size, the results are remarkably good and Emma does a great job in reproducing vast CD collections from such a small box. If you don't believe that the differences between CD and MP3 are significant, you're invited and I'll demonstrate on my home system: It 'll tell you know how well CD's can sound. MP3 are no comparison for serious listening.

I agree with you that the differences are only distinguishable at home: the car is such a bad listening environment (shape / size / speaker placement / speaker quality / electrical noise) that the differences are probably hardly noticeable even if one were able to eliminate engine and traffic noise (even tanstaaf may agree . . . do you Doug?). Also, the amps and the speakers are much less refined than at home. So if you'd play Emma in the car only, normal compression may well suffice. But as a somewhat of an audiophile I believe that higher quality should be up the chain in any audio chain: amp should be better than speakers - pre-amp better than amp etc. so I don't mind to have 10% larger MP3s that sound a tiny bit better. Moreover, I also use Emma at home: it provides very decent background music and when played over the home stereo system, the better quality of normal/high encoding clearly shows.

But for close listening, nothing beats a properly mastered CD played on a good system (yet?) - There is so much potential left, that I doubt if DVD (or whatever) can improve on it, as longs as the basic technology (high sampled wav) remains unchanged. I can recommend to carefully listen to a good stereo set-up and then try to find a system that you can afford that sounds like it. Well recorded MP3s then prove worthwhile.

Henno
# 00120 (6GB+18)

Thanks for the intro to the 'guys'. Nice to 'meet' you.

And in the interest of reciprocity, why not fill out your BBS user profile so we can "meet" you, too?

Tony Fabris
Empeg #144

I suspect we'll use a bit of flash to store IP parameters for those who want static addresses: our DHCP client on the empeg does uPnP (universal plug & play) like win98, so that you can just slam an ethernet card into your unit and not even open the network control panel - but still have it work.

Hmm, time to read man dhcpd etc....

How about another (admitedly primitive) protocol: put manually empeg's desired MAC to IP mapping to your Linux arp table, than access empeg. It will grab the target IP address from the first packet whose ethernet frame has its MAC as target and assign it as its own. (Those cigarette-pack sized print servers often use this scheme - OK, I admit it is a relic from times before DHCP, but I still find it convenient.)

Cheers!

Dragi "Bonzi" Raos
Zagreb, Croatia
#5196

Since you put it so nicely, how can I refuse? ;-)

Hey Tony,

My friend Ash and I (check out his two car computers ) have had an on-going debate regarding the quality of 128kb mp3's vs. higher.. I forwarded your post to him, check out his response:

---

This makes sense to me. Because higher frequencies have more waves per
second, it seems that the repetitiveness of these high frequencies are
harder for the compression algorithm to recognize and therefore you have
some degree of alternating sources of the repeat which will cause you to
hear a flanger-type effect in cheering, and most other high frequency
sound sources. But what this document doesnt explain is why some come
out sounding different whereas other cd's compress fine. 95% of music
is going to have low, mid, and high range frequencies. The high range
is the one in question because according to this, those high
frequencies are what create these anomolies but, all songs have high
frequencies. It may have to do with the CD itself and how purely those
high frequencies are recorded. Our ears may hear a perfect 8th octave
C note but the digital transcript on the CD would show some
inconsistencies in the wave pattern that the MP3 encoder will see and
translate that small inconsistent section as another repetitive cycle,
but unlike the moment before as a perfect 8th C. Therefore creating a
flanging effect by switching from one repetitive clip over to this short
inconsistent clip that almost sounds the same, but different, and then
back to the first clip once more. Imagine this happening at say, 20
times per second on one note. This will create a 20Hz flange that will
be audible. Of course, what I have said to be the source of this
(inconsistencies on the CD) is only a theory. I dont know if this is
actually the cause, but I think I'm on to something. BTW: 8th octave C
is actually the highest note on a piano at the far right. It is at
16KHz, pretty damn high.
Ash

---

-mark

...proud to have one of the first Mark I units

Tell your friend his car computers look sweet. The fabrication of the replacement dash plate on the Z is particularly nice.

As far as the quality discussion goes... I actually commented on this in another thread with Mike... When I talk about high-frequency artifacts, they always seem to manifest themselves in high-frequency noise. In other words, the very thing that's hardest to compress. And the kinds of artifacts I notice happen in parts of the music that are particularly saturated with high-frequency noise. It seems that a 128kbps encoder can handle some high-frequency noise reasonably well, but not too much of it. So you only end up noticing it on certain songs.

Your friend's theory of the possible inconsistencies in a digital sample is known as jitter correction and it's already taken into account at each stage of the MP3 creation process: By the digital master of the original CD, by the drive mechanism you ripped the CD with, by the encoder, and by the playback firmware. Note that "jitter correction" is used two different ways: The first (and the correct usage), refers to the time-base correction of digital samples and their relative accuracy to a given wave. The second (and technically incorrect usage) refers to the ability of a CD ripper to sector-synchronise and correct errors when performing Digital Audio Extraction from a CD. Bad jitter correction can, as he predicted, cause artifacts that sound similar to compression artifacts. But that's not what's happening with a 128kbps MP3.

Here is a very well-informed discussion about jitter correction as it relates to digtally-sampled waveforms, and here is a discussion about jitter correction as it relates to ripping audio data blocks from a CD.

I have heard that data-compressed music fails to capture certain subtleties that audiophiles can notice. For example, the phrase "not enough air around the instruments" keeps coming up. I can't hear this myself-- perhaps it's the nature of the music I listen to. But that same phrase seems to get used in reference to jitter correction problems, too, so there's some sort of parallel going on. In the end, both jitter problems and data compression can make the final output sound slighty different from the original source, and therefore might induce the same sorts of audible artifacts.

Tony Fabris
Empeg #144

MP3s also combines lower frequency left and right channels on the assumption that the lower tones don't contain directional info

I'm going to pick a small nit here -- this is misleading. Lower channel frequencies can be combined in Layer III for better compression, but there's no loss of stereo imaging here; essentially the channels are encoded as L+R and L-R. This is called middle/side or M/S joint stereo. Both channels can be reconstructed perfectly.

Layer III can also optionally use something called intensity joint stereo, in which some high frequencies are encoded as mono plus some stereo imaging information. In this case you can't always reconstruct both channels perfectly. This might be what you're thinking of, as it's the only joint stereo option for Layer I and Layer II. However, it's not very common in Layer III.

For Layer III, either of these joint stereo options, both, or neither are possible. M/S is probably most common, as there's usually little reason not to use it. I have found inconsistencies, however, in the way some decoders handle a combination of M/S and intensity joint stereo, so I don't recommend anyone use both.

-v

errr.... mainly because we have had experience with bad RIPs carried out on music we are familiar with?

Plus, the open-top argument carries weight, yes - but don't forget you can use the box at home when you definitely will hear cracks, pops, and compression artifacts on high frequency stuff.I don't know about you, but this sets my teeth on edge....

Can, but in all honesty, probably won't. I have a top of the line a/v setup (52" projection TV, DVD, LD, CD, AC3, external DA converter, etc) and the 40 CDs are in a rack right next to it. I might use it at the office, though, but that'll be through small ear-plug headphones, which combined with office noise, and given I can't have it so loud that I can't hear the phone ring, it'll probably do.

I started ripping at fixed 128, then went to VBR 75% and am now down to VBR 50% and still can't tell the difference. The file sizes are good, and looking at the actual rates on VBR, I see everything from 192 (electronic trance with lots of precise high frequencies - little harmonics) down to 96 (acoustic guitar with background vocals).

---

You're saying that now:) but you'll use it indoors... I have a 200disc changer that I never use anymore; the empeg's just alot more effecient at swapping cd's (instant - in place of the 4-5 second pause most changers have..), and it's alot easier to use/control then disc changers; it's very easy to make playlists with the emplode software, something that's typically a pain to do on disk changers...
-mark

...proud to have one of the first Mark I units

Tony,

Thanks for the tip on www.allmusic.com! I just checked it out, and wow! ... Great resource! Now i can fix my year tags, and learn more about my favorite artists! I didn't know some of these albums exsisted!

thanks again.


-CHiP

Yeah, allmusic.com is one of the best resources on the net, if not THE best. Try allmovie and allgame.com too. allgame isn't that great, but allmovie is really good. I've used allmusic for years, and even used it in a school report :) They've got books in print but for some reason they list free, updated information on the internet. Who knows...

Anyway, I've got an interesting link about bitrates. It's mostly a review/comparison of about 4 encoders, but there's alot about different bitrates.
http://arstechnica.com/wankerdesk/1q00/mp3/mp3-1.html

I don't know who said that even audiophiles could not tell the difference between 128 and wav. I'm no audiophile but if I listen to a file encoded at 128 and an original CD. There's no question. I even did one of my CDs at 320 once just to check, and the difference is unmistakeable. I haven't done alot of testing at only slightly higher bitrates, but I imagine it could make some difference.

By the way, what do you mean your Variable Bitrate is set at 6%? All my AudioCatalyst has is 5 settings from low to high. what exactly is the VBR?

Thanks for the tip on www.allmusic.com!

Yeah, it's a great resource for people making MP3s. I use it when I rip soundtrack/greatest-hits/compilation albums. I can tag the MP3s with accurate year and album info. I mean, how silly is it to tag The Beach Boys' "Don't Worry Baby" with the date of 1999 and the album name of "Never Been Kissed Soundtrack"? With a few minutes of research, I can pull the correct date and album.

Although, for some pieces of music, it's a little difficult to look up those particulars. The way their search feature is laid out, it takes a little work on the users' part. Especially if the song is published by more than one artist, or was published by the same artist on more than one album. Still, great site.

Tony Fabris
Empeg #144

...definitely will hear cracks, pops, and compression artifacts on high frequency stuff.

Just to be painfully clear here (since Jens is new to this black art of MP3 creation)...

If you encounter cracks and pops (or instrument sounds shifting suddenly between the left or right channels), you're encountering issues with the Digital Audio Extraction from the CD-ROM drive, and these have to be fixed by activating the "sector synchronization" feature of your ripping software.

Compression artifacts are unrelated to those things, and are a lot more subtle.

Remember that ripping a CD is two distinct steps: 1) The digital extraction of the audio from the CD, and 2) the data-compression of that full-bandwidth audio into MP3 format. Things can sometimes go wrong in both sides of that equation, but the problems manifest themselves differently.

Tony Fabris
Empeg #144

Dionysus is right, Jens. I've scrounged up three extra power supplies for the Empeg, and I now have little pre-wired "stations" that I plug the Empeg in at: My desk at work, my desk at home, and the stereo in my living room at home. I understand that you've already got a high-end stereo system in your living room, but if there's a free set of inputs on your tuner, you'll really enjoy plugging the Empeg in there and having it shuffle-play. I haven't put in a real CD in months (except to do A/B comparisons or something).

Tony Fabris
Empeg #144

I would agree that 128K files can lack a lot of detail with certain genres however I seriously doubt whether the difference is "unmistakable" at 320K. If this is the case then I suspect your PC audio system is the cause. The empeg sounds an order of magnitude better than most PC systems.

I haven't heard of anyone consistantly identifying an MP3 track at 256K/sec in a controlled blind test against the CD original. A number of tests have yielded similar results at lower bitrates but I suspect that these are very genre specific.

In practice the majority of our clients report that 128K is satisfactory for them and a significant minority (myself included) encode at 160K. My experience has been that whenever I note a track with poor quality, I later discover that the original CD was dirty or worn. In other words, ripping was the problem, not encoding.

Rob

There seems to be some confusion about the style of compression mp3 uses. It isn't based on repeating clips or any such time-domain analysis. All the work gets done in the frequency domain, in a sequence (1 per frame) of spectra.

My understanding of how it works is it splits the wave into its frequency components by a DCT (discrete cosine transform). A DCT is like an FFT (fast fourier transform - which gives you a spectral display) but with the nice property that there are no imaginary numbers floating about. It then takes into account a property of the ear: you cannot hear quiet frequencies which are close to loud ones. This effect is called 'psychoacoustic masking'. For these masked frequencies, it can use less accuracy. All the coefficients are then quantised (have bit accuracy reduced) depending on their required accuracy, and huffman coded. Quantisation means that there are less 'symbols' for the huffman code to deal with, so the data compresses. It's quantised by an amount which makes the compressed data the correct size.

Ok, that description was properly full of technical inaccuracies, seeing as I'm not completely clued up on all aspects, just the basics :)

At any rate, high frequencies are no more difficult to compress than low frequencies. It's the quantity of frequencies which are loud in relation to each other that determine how well it compresses. The absolute worst case, as somebody else here pointed out, is white noise, which contains all frequencies in equal measure - which is remarkably similar to a cymbal crash.

The effect you hear when mp3 compression doesn't work is 'noise modulation'. Imagine standing a distance off from a waterfall. After a while, your ear will filter out the sound and you won't be concentrating on it. If however the waterfall was turning itself on and off, you would be most alarmed. Quantisation gives this effect, but mp3 attempts to get around it by quantising masked frequencies more than those which aren't. But if there's white noise all it can do is quantise everything - so you get that weird 'swishing' noise (very technical term).


- John.

(The above may not represent the views of empeg :)

that's exactly what it does. it's also just what the link I posted says:

http://arstechnica.com/wankerdesk/1q00/mp3/mp3-1.html

I'm still wondering about that VBR thing everybody here's been talking about. It appears we have either different encoding programs or different Audiocatalyst versions.

I'm still wondering about that VBR thing everybody here's been talking about. It appears we have either different encoding programs or different Audiocatalyst versions.

If you're referring to that "6%" thing, it's because that's an option found in a completely different piece of software (not Audiocatalyst). I believe it was MusicMatch, although I'm not sure.

That link to the Ars-technica article is good (especially in terms of explaining the nature of compression and describing the compression artifacts well), but the article makes one big mistake: It makes its final conclusions based on only fixed-bitrate compression.

You see, the article praises the Fraunhofer encoder for its ability to reproduce the music well at fixed bitrates, and complains about Xing because it doesn't sound quite as good as the Fraunhofer encoder at the same fixed bitrates.

While I agree with the article's conclusions on that level (and agree that it's necessary to compare apples to apples in such tests for accurate results), it doesn't take into account that the Fraunhofer encoder (as far as I know) currently only offers fixed bitrate encoding, whereas Xing will do VBR.

His complaints about the Xing encoder seemed to only apply at low bitrates. But the space-cost of cranking the bitrate on a VBR file is a lot less than the cost of cranking the bitrate on a fixed-rate file. What we really need is a comparison of the sound at a given file size rather than a given bit rate. For instance, if you encode something at a fixed bitrate with Fraunhofer, and compare it to a Xing VBR file that's the same size, I'd bet that the VBR file will probably sound a tiny bit better. This would be a more accurate "real-world" comparison, since the end-user mainly cares about the size/quality tradeoff.

Tony Fabris
Empeg #144

Cool, sounds interesting.

So just some clarifying questions:
a) Did we all decide that that 6% rate was good?
b) If so, what is the comparable rate in AudioCatalyst (in terms of low->low/medium->medium->medium/high->high)?
c) How does the 6%/comparable bitrate compare to the 128 bitrate?

I dont use music match or what ever it is, so no, i dont know what 6% means. Since I have no idea what it means, I cant tell you if its good or not. I use hycd to encode my cds - it uses the xing encoder. I use the high quality vbr setting on the songs that i listen to frequently. Although this is an interesting subject, i think that which setting is the best is entirely dependent on each individual listener and what quality they expect in their music. I suggest you try some different settings with your encoder and figure out what sounds the best for you based on how much hard disk space you have to give. Good luck

Term

Perhaps I'd better explain the 6% thing!

According to the help file, "Choose a VBR setting from 1 to 100, low end being the lowest quality/highest compression and high end being the highest quality/lowest compression, to customize your recording." So 6% is decidedly bad! I ended up choosing VBR at 65%.

I've had varying results (all sound good) with this setting, some files are encoded at 190 (mostly electronic trance and techo), but most are around the 150 mark. I've actually had some files encoded as low as 34 on this setting and they still sound fine (some tracks from Bryan Ferry's 'Boys and Girls').

Remember I'm listening through the PC speakers (middle of the line, with subwoofer) so not the same quality as the empeg or home setup. I've encoded over 2Gb now, which will be enough for me to start with when I get the empeg and not too many that it'll p*ss me off to have to do them all again!

What I'll do (when I get time) is to encode the same track at different rates and make them available from my FTP site - perhaps you guys could then do me the favour of testing them 'in the field' as it were?

J.

What I'll do (when I get time) is to encode the same track at different rates and make them available from my FTP site - perhaps you guys could then do me the favour of testing them 'in the field' as it were?

Um, don't do that, please. We just got done talking about the RIAA in another thread.

Besides, it's totally up to you as to whether the stuff sounds good or not. Just encode them so that they sound OK to you. If you ever reach a point where you're listening to the Empeg, and you dislike the compression on one particular song, it's not so tough to re-rip that song from your original CD (now that you've seen how easy it is with the proper software).

Tony Fabris
Empeg #144

I'll second that study - I read it too and it seemed on the money.

I'll concede that 128kbit might have some artifacts in certain parts of certain songs. I can hear it quite clearly as a digital processed sound on the snares on New Order's Blue Monday. High frequencies, white noise and other near random sound, and sharp attacks such as kick drums or explosions all suffer the most under MP3.

However, firstly, the encoder you use makes a big difference. A recent article on Ars Technica studied the frequency response of various players. While the Fraunhofer encoder reproduced the sound all the way up to 20 or 22KHz (I can't remember exactly), AudioCatalyst and BladeEnc (and another I can't remember) all dropped off at around 16KHz in standard 128kbit mode. People with good hearing can hear these tones directly, but there's plenty of evidence to suggest that the rest of us can still notice if a cymbal is lacking the really high frequencies. Most of this frequency loss went away as you increased the bit rate - the spectrum would return to normal at 160kbit or 192kbit.

Secondly, double blind is the only way to accurately determine real results. I've seen too many 'experts' able to tell the difference between a well encoded JPEG and a TIFF ten times its size when they know which is which, but very few can identify them when they don't know. And hearing is much more subjective - the memory is not very good at remembering precise tones or sound characteristics over time, and by the time you've loaded up that next track you've probably forgotten the exact details of the comparison part you wanted to look at.

Try things out before you commit to someone else's ideas.

Save the whales. Feed the hungry. Free the mallocs.

Good, now I have a justification for ripping all my music at 192 via Audio Cataylist. I'll have to dig through some book marks, as I know a site had the same song in 96, 128, 160 and 192. All of them I herd differences in, and would never encode that particular song any lower then 192.

My empeg site is:http://24.236.3.131/empeg/

Okay, I think I'm going against what I've said before, but I'll say it anyway.

I've decided that 128 is fine (for me). My reasoning is that I recently got the Klipsch ProMedia's, and now I can actually HEAR my music. Considering I'm listening to Candlebox's "You" at 112 right now and it sounds fine, I've decided I'll be fine in a car setting with 128.

It seems to me from reading this thread that pretty much everybody finds 128k indistinguishable from a CD for most of the time. Doesn't this mean that a good VBR is the way to go? That way you get much higher bitrates for the (apparently) problematic cymbal crashes and so forth, without blowing your whole file size out.

Personally, I've been encoding my stuff with lame, with VBR, at a nominal bitrate of 64k. I think in general the file sizes come out to an equivalent bitrate of around 140k. I haven't checked individual frame bitrates yet. So far, listening in the car, I haven't found any audible artifacts. If the psychoacoustic model in your VBR encoder is good, then that's how it should be. I really like the idea of being able to fine tune your psychoacoustic model, so that it knows just what YOUR thresholds for distortion at various frequencies are. Once you've got that, then you can be pretty much assured of getting the smallest possible file size that won't give you any audible problems.

Richard.

Jens --

Listen to Dionysus and tfabris.

All you have to lose by encoding at a higher bitrate is disk space, and you have everything to gain. From the reviews I've been reading, 160 is good, but 192 seems to be the sweet spot beyond which diminishing returns takes effect. Disk space is cheap. The difference in cost of an empeg with enough space to hold all your music at 128 Kbps compared to one big enough to hold the same music at 192Kbps is only going to be a few hundred dollars. Believe me, after you've had the empeg for a year you'll never miss that couple hundred bucks.

Anybody who can afford a Mustang Cobra can afford enough hard drive space to encode his music for optimal playback!

tanstaafl.

"There Ain't No Such Thing As A Free Lunch"

I agree with you that the differences are only distinguishable at home: the car is such a bad listening environment (shape / size / speaker placement / speaker quality / electrical noise) that the differences are probably hardly noticeable even if one were able to eliminate engine and traffic noise (even tanstaafl. may agree . . . do you Doug?)

Not really, Henno. Yes, the typical slap/dash car installation, put the CD player in, play it through the miserable factory speakers, then crank the bass up as high as it will go so the subwoofers in the trunk can annoy the neighbors.... you're right.

But I have seen cars in competition with incredible sound systems -- one Lexus that I know of has a system that we onlookers estimate at over $40,000 (the owner won't say one way or another) and while my own system is pretty good, the Lexus puts it to shame. You've told me a little bit about your own home system, and yes, it probably is better than even the Lexus... but from your description, there probably aren't very many stereos on the whole planet to equal yours.

A car can be an excellent listening environment IF PROPERLY CONFIGURED for the very reasons you cite as making it a poor environment. The fact the the environment is so.... finite means that you can do some very specific tuning without having to deal with the variables involved in a large, unsealed room. It just takes a bit more work to find the right speakers to work with the amount of air volume in your car; the right locations for the speakers to work with the shape of the interior and the rake of the windshield and the different materials and reflective surfaces; some of the best quality speakers available are made for in-car audio, such as MB Quart; the same could be said about amplifiers. One of the things that is driving the push towards quality car audio is the audio competition environment; many cars do a quite respectable job of filtering out road, engine, traffic, and electrical noise, and by the time you have the volume cranked up to 90 decibels or so (for some reason the very ambience of a moving car seems to encourage playing the music louder than you would at home, irrespective of any external noise) all you'll hear is the music, you won't hear anything else, and that music can sound very good.

Whether you'd pick up on the subtleties of 192Kbps vs 128 Kbps.... I have no first hand experience on that. I think I would. Little things like the brightness of the treble might give it away. Toni (oops, pardon me, Tony) might be able to verify an analogy: is a comparison between 128 and 192 Kbps similar to a comparison between the original Rush CDs and the re-masters? If so, then I can absolutely promise and guarantee that I will be able to tell the difference, at least in my car.

tanstaafl.

ps: sorry to be so slow in responding to this -- I've been "off-line" for the past two weeks...

"There Ain't No Such Thing As A Free Lunch"

Especially anyone who can afford the parking prices in NY. They're INSANE!